Healthcare 4.0: Next Generation Processes with the Latest Technologies [1st ed.] 978-981-13-8113-3;978-981-13-8114-0

Table of contents :
Front Matter ....Pages i-xxi
An Introduction to Healthcare 4.0 (Janya Chanchaichujit, Albert Tan, Fanwen Meng, Sarayoot Eaimkhong)....Pages 1-15
Internet of Things (IoT) and Big Data Analytics in Healthcare (Janya Chanchaichujit, Albert Tan, Fanwen Meng, Sarayoot Eaimkhong)....Pages 17-36
Blockchain Technology in Healthcare (Janya Chanchaichujit, Albert Tan, Fanwen Meng, Sarayoot Eaimkhong)....Pages 37-62
Application of Artificial Intelligence in Healthcare (Janya Chanchaichujit, Albert Tan, Fanwen Meng, Sarayoot Eaimkhong)....Pages 63-93
Optimization, Simulation and Predictive Analytics in Healthcare (Janya Chanchaichujit, Albert Tan, Fanwen Meng, Sarayoot Eaimkhong)....Pages 95-121
Innovative Health Technologies and Start-Ups Process in Healthcare Industry (Janya Chanchaichujit, Albert Tan, Fanwen Meng, Sarayoot Eaimkhong)....Pages 123-159
Transforming and Managing Healthcare Projects (Janya Chanchaichujit, Albert Tan, Fanwen Meng, Sarayoot Eaimkhong)....Pages 161-194
Conclusion (Janya Chanchaichujit, Albert Tan, Fanwen Meng, Sarayoot Eaimkhong)....Pages 195-197
Back Matter ....Pages 199-202

Citation preview

Healthcare 4.0 Next Generation Processes with the Latest Technologies Janya Chanchaichujit Albert Tan Fanwen Meng Sarayoot Eaimkhong

Healthcare 4.0 “In a complex world that is advancing rapidly at an exponential pace where nearly every sphere of human existence seemingly finds the internet indispensable, therein lies profound benefits to be gained at the convergence of virtual reality, simulations, big data analytics and Artificial Intelligence. Healthcare is no exception. So physicians, health economists and healthcare administrators of this generation and beyond ultimately have to thrive in such a technological environment to remain relevant in the medical profession which provides care for the population in the modern age of digitalisation. This book elegantly shows how the paradigm of healthcare has gradually morphed into the current Healthcare 4.0 version that encapsulates the core principles of Industry 4.0. The authors should be highly commended for assembling a fascinating book which is provocative, edifying and exceedingly readable to a wide audience from all walks of life.” —Associate Professor Melvin Khee-Shing Leow, Clinician Scientist and Senior Consultant Endocrinologist, President of Endocrine and Metabolic Society of Singapore; Duke-NUS Medical School, Singapore; Tan Tock Seng Hospital, Singapore “This book stands out as a timely key to open the new pathway to the future of healthcare where digital technology intertwines with patient care and healthy living of tomorrow. It is a great reading both from the technologist’s point of view– looking to enter and disrupt the healthcare market, or from the health providers’ perspective preparing oneself for an unforeseen revolution.” —Dr. Nares Damrongchai, CEO, Thailand Center of Excellence for Life Sciences, Thailand “This book is a timely contribution demonstrating the benefits of adopting Industry 4.0 principles in healthcare. The authors develop a framework for implementation, providing real examples on Healthcare 4.0 projects, along with guidelines on how to overcome the challenges. It is essential reading for healthcare policy makers, managers, and academicians interested in understanding Healthcare 4.0, which is supported through emerging technologies, such as the Internet of Things, Big Data, Optimisation and Predictive Analytics.” —Associate Professor Tugba Cayirli, Ozyegin University, Istanbul, Turkey

Janya Chanchaichujit • Albert Tan  Fanwen Meng • Sarayoot Eaimkhong

Healthcare 4.0 Next Generation Processes with the Latest Technologies

Janya Chanchaichujit School of Management Walailak University Thasala, Nakhon Si Thammarat, Thailand Fanwen Meng Department of Health Services & Outcomes Research National Healthcare Group Singapore, Singapore

Albert Tan Malaysia Institute for Supply Chain Innovation Shah Alam, Selangor, Malaysia Sarayoot Eaimkhong National Science and Technology Development Agency Pathum Thani, Thailand

ISBN 978-981-13-8113-3    ISBN 978-981-13-8114-0 (eBook) https://doi.org/10.1007/978-981-13-8114-0 © The Editor(s) (if applicable) and The Author(s), under exclusive licence to Springer Nature Singapore Pte Ltd. 2019 This work is subject to copyright. All rights are solely and exclusively licensed by the Publisher, whether the whole or part of the material is concerned, specifically the rights of translation, reprinting, reuse of illustrations, recitation, broadcasting, reproduction on microfilms or in any other physical way, and transmission or information storage and retrieval, electronic adaptation, computer software, or by similar or dissimilar methodology now known or hereafter developed. The use of general descriptive names, registered names, trademarks, service marks, etc. in this publication does not imply, even in the absence of a specific statement, that such names are exempt from the relevant protective laws and regulations and therefore free for general use. The publisher, the authors and the editors are safe to assume that the advice and information in this book are believed to be true and accurate at the date of publication. Neither the publisher nor the authors or the editors give a warranty, express or implied, with respect to the material contained herein or for any errors or omissions that may have been made. The publisher remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. Cover illustration: Pattern © John Rawsterne/patternhead.com This Palgrave Pivot imprint is published by the registered company Springer Nature Singapore Pte Ltd. The registered company address is: 152 Beach Road, #21-­01/04 Gateway East, Singapore 189721, Singapore

Preface

The internet has been a revolutionary technology for the healthcare sector and it has helped in optimizing the entire supply chain and providing more detailed patient outcomes. There has been a great deal of research into Industry 4.0 and its benefits, challenges and opportunities with regard to the healthcare industry. However, incorporating Industry 4.0’s core principles into healthcare practices is still not widespread enough to have created the transformation that is possible. The main motive of this book is to demonstrate the benefits of implementing Industry 4.0 in healthcare services and to recommend a framework to support this implementation. Key topics in this book include: • Discovering emerging technologies and techniques to support Healthcare 4.0. This includes the Internet of Things, big data analytics, blockchain, Artificial Intelligence, optimization and predictive analytics; • Illustrating some examples of such advanced implementation in Healthcare 4.0; • Recommending a development process to develop health technology start-ups and entrepreneurial activities; and • Discussing the transformation methodology used to redesign healthcare processes in order to overcome the challenges of implementing a Healthcare 4.0 project.

v

vi 

PREFACE

This book consists of eight chapters. Chapter 1 presents the background and core principles of Industry 4.0 in healthcare services. It comprises six core principles which demonstrate Industry 4.0’s capabilities and which can be applied to the healthcare industry. In this chapter, the authors summarize the evolution of IT implementation over the decades. In Chap. 2, the Internet of Things (IoT) and big data analytics in healthcare are presented. This chapter aims to introduce the reader to the IoT and big data analytics as elements of Industry 4.0 and the healthcare industry. The techniques and technologies used, the main advantages, the challenges of how to apply IoT and big data analytics in healthcare industry, and strategies to overcome these challenges are presented. In this chapter a case study of how big data analytics is being applied in Singapore is presented. The chapter concludes by highlighting lessons learnt from the use of the IoT and big data analytics case studies in the healthcare delivery system. Chapter 3 is concerned with blockchain technology in healthcare. The main objective of this chapter is to show the reader how blockchain technology is an effective system for managing patient records and tracking medical drugs along the pharmaceutical supply chain. The application of blockchain technology is illustrated with two case studies: using blockchain technology for an Electronic Health Record system in a developing country and using blockchain in the pharmaceutical industry. Chapter 4 presents the role and significance of Artificial Intelligence, commonly known as AI, in the Control and Management of Tuberculosis (TB). The complexity of the disease and problems in TB diagnosis are introduced. Following this, initiatives and opportunities for using AI in TB diagnosis in Thailand are shown as a case study. The chapter concludes by discussing the current limitations of AI improvement, and alternative models and key success factors in the implementation of AI in treating TB. Chapter 5 discusses the use of operations research techniques like optimization, simulations and predictive analytics in healthcare. The chapter introduces optimization problems in healthcare, from strategic resources and capacity planning to operational and clinical issues such as resource scheduling and treatment planning. Case studies using operations research in healthcare in Singapore are presented, followed by some insights into improved healthcare delivery. Chapter 6 discusses an innovative health technology and development process in the healthcare industry, challenges and trends in the healthcare system, and health technologies for the future. The chapter concludes by recommending methodologies for developing health technology to start-ups and entrepreneurs. Chapter 7

 PREFACE 

vii

presents the process transformation methodology used to redesign healthcare processes in order to overcome challenges of implementing Healthcare 4.0 projects. In addition to the methodology, agile project management for implementing such projects is introduced to cope with changes in IT project. The chapter concludes by highlighting key success factors for implementing a Healthcare 4.0 project. Chapter 8 concludes the book, with summaries and recommendations for implementing Healthcare 4.0. Malaysia, Thailand and Singapore  2019 March  

Janya Chanchaichujit Albert Tan Fanwen Meng Sarayoot Eaimkhong

Acknowledgements

The authors would like to express their sincere gratitude to Dr Krit Pongpirul for his generous support and time in sharing his knowledge and information about the use of AI in his work. The authors worked closely with him in the writing of Chap. 4. His experience with the Thai government in implementing a uniquely designed AI algorithm specifically for diagnosing TB gave us valuable insight into employing AI in real practice for our book. Special appreciation is extended to Ms Malai Williams. She contributed to the book regarding the development of a healthcare project and change management by using Waterfall and Agile methodology, as can be seen in Chap. 7. Her expertise in IT project management, especially in healthcare, has helped us to share key success factors with the reader on transforming and managing healthcare projects. We would also like to thank to Dr Albert Tan’s students at Curtin University, Singapore, who have contributed their work to this book as follows: Ms. Vithya Laxme Samiappan and Ms. Vu My Linh for their research on developing a concept framework for blockchain technology and case studies in drug traceability and Electronic Health Records in Vietnam (Chap. 3), Mr. Sinarta Wirawan for conducting a literature review of Chap. 4 and Do Tran Phuong Uyen for conducting a literature review on IoT for Chap. 2. Finally, the authors wish to express special thanks to their families, who supported them wholeheartedly during the preparation and writing of this book. This book was partially supported by the Institute of Research and Innovation, Walailak University (Grant number WU61202), and the ix

x 

ACKNOWLEDGEMENTS

Research Institute for Health Science, Walailak University (Grant number WU60-010). Malaysia, Thailand and Singapore Janya Chanchaichujit, Albert Tan, March 2019 Fanwen Meng and Sarayoot Eaimkhong

Contents

1 An Introduction to Healthcare 4.0  1 1.1 Introduction  1 1.2 Industry 4.0  3 1.2.1 Key Components of Industry 4.0  3 1.2.2 Core Principles of Industry 4.0  5 1.3 Challenges in Implementing Information Technologies in Healthcare  8 1.4 Stages in Healthcare IT Transformation  9 1.5 Drivers for Healthcare 4.0  11 1.6 Recommendations 12 1.7 Conclusion 13 References 13 2 Internet of Things (IoT) and Big Data Analytics in Healthcare 17 2.1 Overview of the Healthcare Supply Chain 17 2.2 Interoperability Issues in the Healthcare Supply Chain 18 2.3 Internet of Things (IoT) 21 2.3.1 Internet of Things (IoT) in the Healthcare Supply Chain 22 2.3.2 Challenges of Implementing the Internet of Things 23 2.4 Overview of Big Data Analytics in Healthcare 25 2.5 How to Implement Big Data in Healthcare 26 2.5.1 Type and Source of Healthcare Big Data 26 xi

xii 

CONTENTS

2.5.2 Techniques and Technologies Used in Healthcare Big Data Analytics 27 2.5.3 Application of Big Data Analytics in Healthcare 28 2.6 Case Study 29 2.6.1 Big Data Analytics in Managing the Future Health of Singapore 29 2.6.2 Diabetes in Singapore—A Nation’s Big Challenge 29 2.6.3 Challenges to Implementing Big Data in Healthcare 32 2.7 Conclusion 32 References 33 3 Blockchain Technology in Healthcare 37 3.1 Blockchain Technology 37 3.1.1 Smart Contract in Blockchain 40 3.1.2 Blockchain in the Healthcare Industry 41 3.1.3 Challenges of Implementing Blockchain 42 3.2 Case Studies 44 3.2.1 Case Study 1: Using Blockchain Technology for EHR Systems in Developing Countries 44 3.2.2 Case Study 2: Using Blockchain in the Pharmaceutical Industry 54 3.3 Conclusion 61 References 61 4 Application of Artificial Intelligence in Healthcare 63 4.1 Introduction 63 4.1.1 Machine Learning 64 4.1.2 Neuro Learning 65 4.1.3 Deep Learning 66 4.2 Case Study 66 4.2.1 Case Study 1—The Role and Significance of Artificial Intelligence in the Control and Management of Tuberculosis 66 4.3 Innovation Management of Tuberculosis in Thailand Using Artificial Intelligence 82 4.3.1 TB Situation in Thailand and Opportunities for AI in TB Diagnosis 82 4.3.2 Shortfalls of Current AI Platforms in TB Diagnosis 84

 CONTENTS 

xiii

4.3.3 Alternative Models and Key Success Contributors for the Implementation of AI in TB 85 4.4 Conclusion 87 References 88 5 Optimization, Simulation and Predictive Analytics in Healthcare 95 5.1 Overview of Operations Research in the Healthcare System 95 5.2 How to Apply Operations Research in Healthcare 97 5.2.1 Healthcare Service Planning 97 5.2.2 Healthcare Management and Logistics 98 5.2.3 Clinical Practice 99 5.3 Case Studies 99 5.3.1 Case Study 1. Managing Elective Admissions Using the Robust Optimization Approach100 5.3.2 Case Study 2. Shift Capacity Planning for Nursing Staff Using Mixed Integer Programming107 5.3.3 Case Study 3. Analysis of Patient Waiting Time Governed by a Generic Maximum Waiting Time Policy112 5.4 Conclusion118 References118 6 Innovative Health Technologies and Start-­Ups Process in Healthcare Industry123 6.1 Introduction to the Health Technology123 6.2 Challenges and Trends in the Healthcare Systems125 6.2.1 Rising Healthcare Expenditure126 6.2.2 Life Expectancy Increase127 6.2.3 Rise of Technology and High Expectations128 6.2.4 Shortage of Healthcare Personnel129 6.3 Health Technology for the Future131 6.3.1 Digital Healthcare131 6.3.2 Personalized Healthcare136 6.3.3 Nanorobotics and Nanomedicine142 6.4 Development of Health Technology and Healthcare Start-­Ups143 6.4.1 Smart Healthcare Industry: Future Value Chain in Healthcare144

xiv 

CONTENTS

6.4.2 Technology and Innovation Development Processes in Health Technology145 6.4.3 Regulatory and Legal Requirements149 6.4.4 Start-Ups and Entrepreneurship in Health Technology152 6.5 Conclusion155 References156 7 Transforming and Managing Healthcare Projects161 7.1 Process Transformation162 7.1.1 Step 1: Assumption Surfacing162 7.1.2 Step 2: Challenging Assumptions164 7.1.3 Step 3: Idea Generation164 7.1.4 Step 4: Idea Grouping170 7.1.5 Step 5: Idea Evaluation174 7.1.6 Step 6: Ideas Integration and Digitization176 7.1.7 Step 7: Process Validation and Implementation178 7.1.8 Step 8: Project Management and Change Management181 7.2 Challenges Faced in Managing Healthcare IT Projects188 7.2.1 Patient Safety188 7.2.2 IT Integration Difficulties188 7.2.3 Lack of Commitment189 7.2.4 Resources Constraints189 7.2.5 Changes in Regulations189 7.2.6 Fear of Job Loss189 7.2.7 Change of Organizational Structure and Culture190 7.3 Change Management Is an Important Part of Healthcare Transformation Projects190 7.3.1 Political Skills191 7.3.2 Analytical Skills191 7.3.3 People Skills192 7.3.4 Leadership Skills192 7.4 Key Success Factors for Implementing Large-­Scale IT Systems192 References193 8 Conclusion195 8.1 Implementing Healthcare 4.0196 I ndex199

About the Authors

Janya  Chanchaichujit  is an assistant professor in logistics and supply chain management at the School of Management at Walailak University in Thailand. Chanchaichujit holds a PhD in Logistics and Supply Chain Management from Curtin University in Australia, an MSc in Operational Research from the University of Hertfordshire in the UK and a BSc in Mathematics from Mahidol University in Thailand. Albert Tan  is an adjunct professor at Malaysia Institute for Supply Chain Innovation. He has worked as a consultant for the Center for Corporate Learning at the Singapore Manufacturers’ Federation (SMa) and as Director in a government agency at Singapore. He has held teaching positions in Dubai, Vietnam and Indonesia as well as in Singapore. Fanwen Meng  is working as an operations research specialist at National Healthcare Group, Singapore. Meng holds a PhD in operations research from National University of Singapore (NUS). He previously worked at NUS and University of Southampton in the UK.  His research work has been published in leading journals including Operations Research, Mathematics of Operations Research, Scientific Reports, Journal of the Royal Society Interface. Sarayoot  Eaimkhong  received his PhD in chemistry focusing on bio-­ nanotechnology from UCLA. He worked with two start-ups to develop novel biosensing platforms before joining a life-science consulting firm in California. He has worked closely with many large pharmaceutical and xv

xvi 

ABOUT THE AUTHORS

medical companies to develop business, market access and regulatory strategies. He specializes in promoting healthcare innovation, commercialization of health technology, increasing the Thai industrial capability, and R&D within Thailand and South East Asia.

List of Figures

General process for transferring medical information into blockchain46 Fig. 3.2 Patient registration and setting up personal profile 46 Collecting medical data utilizing IoT 49 Fig. 3.3 Fig. 3.4 Encrypting data into blockchain for future reference 51 Data recorded in each block stored in the blockchain platform 59 Fig. 3.5 Fig. 3.6 Flow of medical products and data captured along the supply chain60 Fig. 4.1 Global trends in the estimated number of incident TB cases and the number of TB deaths between 2000 and 2016 (WHO, 2017) 67 Fig. 4.2 The development of Tuberculosis from the point of introduction 69 to active TB (WHO, 2018a) Fig. 4.3 Diagnosis and ruling of TB (CDC, 2016) 70 Fig. 4.4 Implementation of digital health products to different 74 components of TB (WHO, 2015a) Fig. 4.5 Schematic diagram of digital transformation of TB (WHO, 2015a) 75 Fig. 4.6 Input variables for training of early AI in TB (Farrugia et al., 1993) 76 Fig. 4.7 (a) Posteroanterior chest radiograph shows upper lobe opacities with pathologic analysis-proven active TB. (b) Same posteroanterior chest radiograph, with augmentation with 79 convolution layers (Lakhani & Sundaram, 2017) Fig. 4.8 Example of anomalies selected by the developers for AI training 81 on chest X-rays (Qure.AI: http://qure.ai/qxr/) Fig. 4.9 TB screening during the “World TB Day 2018” campaign 83 (Sonrueng, 2018) Fig. 4.10 TB screening in prisons in Thailand (The Nation, 2018b) 83 Fig. 3.1

xvii

xviii 

List of Figures

Fig. 4.11 Fig. 5.1 Fig. 5.2 Fig. 5.3 Fig. 5.4 Fig. 5.5 Fig. 5.6 Fig. 5.7 Fig. 5.8 Fig. 5.9 Fig. 6.1 Fig. 6.2 Fig. 6.3 Fig. 6.4 Fig. 6.5 Fig. 6.6 Fig. 6.7

Fig. 6.8

Fig. 6.9

Schematic diagram of the process and algorithm of DAC4TB to 86 diagnose a CXR An illustrative diagram of inpatient flow in hospital 101 An example of different bed allocation policies 102 Average emergency admissions by day of the week 105 Comparisons of bed shortages of Model 4 with Models 1 and 2 106 with additional beds Fraction of days with bed shortfalls over the evaluation period 107 Half-hourly demand at emergency department by time of day 110 Mean deviations of workload and their reductions compared 114 with the current mean deviation An illustration of original waiting distribution and maximum 115 waiting time policy An illustration of transformed waiting distribution 116 Graduates of allopathic medical schools in the USA, 1980– 130 2005 (Salsberg & Grover, 2006) Classification of technologies (Weiss et al., 2018) 133 Old and new healthcare paradigm with digital health technology. 139 Source: Deloitte Healthcare Solutions (Taylor, 2015) A personalized treatment approach tailored by the integration of exome sequencing and Avatar mouse models (Garralda et al., 2014)140 Outline of precision medicine in Oncology (Morash et al., 2018) 141 Physical value chains are disrupted and transformed by digital technologies (digital value chain: https://slideplayer.com/ slide/10496307/)146 A typical health technology (medical device and pharmaceutical products) development pathway (https://www.ttopstart.com/ news/the-occurrence-of-a-second-valley-of-death-duringmedical-device-development; https://www.slideshare.net/ AmberHolHoreman/design-for-reimbursement-in-medicaldevice-development-50344101)147 Development of a medical device from fundamental research to marketing with the two “valleys of death” (https://www. ttopstart.com/news/the-occurrence-of-a-second-valley-ofdeath-during-medical-device-development; https://www. slideshar e.net/AmberHolHor eman/design-forreimbursement-in-medical-device-development-50344101)148 Development pathway of pharmaceutical products (https:// steveblank.com/2013/08/19/reinventing-life-sciencestartups-evidence-based-entrepreneurship/)150

  List of Figures 

xix

Fig. 6.10 The regulatory path and approach of the USFDA to regulate medical devices (https://www.slideshare.net/Healthegy/ breakout-session-cybersecurity-in-medical-devices)151 Fig. 6.11 Regulatory standards for an electromedical device with software (https://blog.cm-dm.com/post/2013/04/12/MD-andIVD-standards%3A-IEC-60601-1-and-IEC-61010-1%2Cversus-IEC-62304-Part-2)152 Fig. 6.12 Health technology innovation quadrant when considering average funding and average age of the technology (https:// www.slideshare.net/NathanPacer/venture-scanner-healthtech-report-q3-2017)153 Fig. 6.13 The venture capital investment profile in the heath technology sector (Pacer, 2017) 154 Fig. 7.1 Current process flow for a specialist private clinic 163 New process flow for specialist private clinic 177 Fig. 7.2 Waterfall methodology 184 Fig. 7.3 Fig. 7.4 Agile methodology 185

List of Tables

10 Table 1.1 Transition from Healthcare 1.0 to Healthcare 4.0 12 Table 1.2 Emerging technologies to support Industry 4.0 Table 4.1 Main shortcomings of commonly used TB diagnostic tests 72 (Vikas K. Saket, 2017; Dande & Samant, 2018) Table 4.2 Development of TB detection methods augmented by and 77 integrated with AI Table 5.1 Descriptive statistics of patients 104 105 Table 5.2 Different scenarios in numerical study Table 5.3 Difference of bed shortage compared with Model 1 106 112 Table 5.4 Different scenarios in numerical study Table 5.5 Number of nurses per shift by the model under different scenarios113 117 Table 5.6 Descriptive statistics of patient waiting time data Table 5.7 Mean and standard deviation of new waiting time under 117 various scenarios Table 7.1 Templates for assumption surfacing 164 Table 7.2 Assumptions and challenges regarding a private clinic 165 Table 7.3 Ideas generated for a private clinic 168 Table 7.4 Format to evaluate each option 175 176 Table 7.5 Scoring of each option for a private clinic Table 7.6 Staffing requirements to support new process 180 Table 7.7 Migration plan for new IT system 182 186 Table 7.8 Agile versus Waterfall methodology Table 8.1 Emerging technologies to support Industry 4.0 196

xxi

CHAPTER 1

An Introduction to Healthcare 4.0

Abstract  The Industry 4.0 revolution is already redefining how companies manufacture “things” today. It sets out the concepts for how companies can achieve faster innovation and increase efficiencies across the value chain. But, in the world of healthcare devices and services, which is burdened with regulatory compliance and is still largely dependent on paper-­ based processes, what does Industry 4.0 really mean? If healthcare services are to incorporate Industry 4.0 core principles, they require proper guidelines or a framework within which to incorporate the core principles. Based on the key factors determined from our research and based on the knowledge acquired from literature reviews, a set of emerging technologies is proposed for implementation in the healthcare sector. Keywords  Industry 4.0 • Healthcare 4.0 • Internet of Things (IoT) • Internet of Services (IoS) • Interoperability • Cyber-physical systems (CPS)

1.1   Introduction New opportunities and challenges emerge in an industry as a result of customer demand and the desire for advanced technology to further enhance sophisticated technological services. Such transformation promotes new context configurations, new environments and new motivations which will

© The Author(s) 2019 J. Chanchaichujit et al., Healthcare 4.0, https://doi.org/10.1007/978-981-13-8114-0_1

1

2 

J. CHANCHAICHUJIT ET AL.

eventually impact upon a company’s performance. Today, technical advancements and innovations are gaining serious importance in several industries such as the biotech industry, IT industry and automotive industry. These industries are incorporating new technologies which make use of automation and provide intelligent solutions. These transformations are the precursors to the rapid changes which are leading us towards a new industrial revolution, or Industry 4.0 as it is known. This “revolution” will have an industry-wide impact (Tjahjono, Esplugues, Ares, & Pelaez, 2017). Society itself is being influenced by this revolution. Our general operational structure, man–machine interaction optimization, economic views and patterns, and other significant circumstances are all affected. Industry 4.0 enhances the ability to accurately interpret and recognize progressions, along with an awareness and understanding of market trends. The organizational structure of a firm can be streamlined and made more cohesive by utilizing new discoveries in Industry 4.0. This revolution is promoting new models such as social networks, additive manufacturing, collaborative innovation, digital platforms and a shared economy that initiates change in organizations. Production in the manufacturing industry has been completely modernized and automated. Industry 4.0, with its technical advancement, has not only promoted improvements but in itself is a concept that promotes transformation (Lu, 2017). There has been a great deal of research into Industry 4.0 and its benefits, challenges and opportunities with regard to the healthcare industry. The internet has been a revolutionary technology for the healthcare sector and it has helped in optimizing the entire supply chain and provided more detailed patient outcomes. However, incorporating Industry 4.0’s core principles into healthcare practices is still not widespread enough to have created the transformation that is possible. Industry 4.0 is not just a technical advancement; it is a profound concept that can enhance the performance of any industry. The six core principles of virtualization, modularity, interoperability, decentralization, service orientation and real-time capabilities constitute the concepts contained within Industry 4.0. Industry 4.0 core principles are driven by emerging technologies such as blockchain, Internet of Things (IoT), big data and Artificial Intelligence (AI) (Manogaran, Thota, Lopez, & Sundarasekar, 2017). The main aim of this book is to demonstrate the benefits of implementing Industry 4.0 in healthcare services and to recommend a framework to support this implementation.

1  AN INTRODUCTION TO HEALTHCARE 4.0 

3

1.2   Industry 4.0 In 2011, German manufacturing industries were strengthening their competitiveness in the sector by promoting a new concept, with the help of pioneers from different fields such as politics, academia and business. This concept was proposed at the Hanover Trade Fair 2011 as Industry 4.0. The concept was widely supported by the German government with the view that Industry 4.0 would develop into a high-level competitive strategy in the future. The “virtual marketplace” is expected to influence the connection of the physical world (people, products, machines and systems) to a constant virtual world. In this way, service platforms and software-based systems will play a significant part in future manufacturing processes. These virtual connections are the optimal way of analysing and providing data that supports the communication between product and machine. In other words, virtual connection is the process of connecting physical and digital processes to “smart” products. Key Components of Industry 4.0 1.2.1   The fundamental components of Industry 4.0 are cyber-physical systems (CPS) and Internet of Services (IoS). These components are equipped with actuators and sensors which help in effective communication and support factories to work autonomously and in a decentralized way (Zezulka, Marcon, Vesely, & Sajdl, 2016).  yber-Physical Systems (CPS) C CPS are fundamental elements of Industry 4.0 that connect the physical and virtual world. CPS are characterized by remarkable coordination of distributed ledgers and internet services. In other words, the systems that connect physical processes with computations are called cyber-physical systems, or CPS. Computation is influenced by physical processes such as feedback loops, embedded networks and computers, and physical process controls (Poovendran, 2010). The physical process is also influenced by computation. High-level transparency, efficiency, surveillance and control in the operations process are a few of the noteworthy advantages of CPS.  CPS consist of two parallel networks such as cyber networks and physical networks, and constitute communication linkages between these networks. The cyber network includes intelligent controllers while a

4 

J. CHANCHAICHUJIT ET AL.

­ hysical network incorporates infrastructure of manufacturing. Using sevp eral controls, actuators, communication devices and sensors, CPS connect these two networks (Garibaldo & Rebecchi, 2018). I nternet of Things (IoT) The internet has allowed technical advancements and had an unprecedented impact on both communication systems and data-sharing systems. Moreover, it has facilitated the exchange of and access to real-time data from any place in the world and at any time. It also enhances the coordination required between the customer, the supplier and the company, along with the interaction of man and machine. IoT has served as an Industry 4.0 initiator since the 1990s (Gubbi, Buyya, Marusic, & Palaniswami, 2013). Smart products can allow us to overcome or cut across the traditional boundaries of a product, providing greater reliability, new functionality, expanded opportunities and high-level product utilization. IoT promotes an environment where every individual can be connected to web service provided by smart technology, which is self-managing, self-­ aligning and self-organizing anywhere and at any time. IoT constitutes the most dominant and conservative technology that can stimulate numerous opportunities for economic growth (Wortmann & Flüchter, 2015). I nternet of Services (IoS) Concerning the idea of a “service society” and IoT, web-orientated services have also been developed. These services are called IoS and they allow private users and companies to connect, develop and promote advanced value-added services. Future industries are expected to rely upon internet value-added services. From a technical point of view, ideas such as BPO (business process outsourcing), SOA (service-oriented architecture) and SaaS (software as a service) are more similar to IoS. In essence, IoS is a business service transaction between two parties. The aim is to perform the required activities and the result is to gain advantage from that performance. To perform such activities, one party temporarily uses the resources of the other (Drăgoicea, Pătraşcu, & Bucur, 2012). Smart Factory According to the literature, Industry 4.0’s core components are IoS, CPS and IoT. CPS connects to IoS and IoT through cyber and physical networks and communication linkages, thus promoting a Smart factory

1  AN INTRODUCTION TO HEALTHCARE 4.0 

5

set-­up. This points to these core components as the initiators of a Smart factory. Smart factories are being developed with the aim of building a social network set-up in which resources, human beings and machines communicate with each other easily. A decentralized manufacturing system can also be built with the help of this social network. Existing complex production processes and production logic can be altered by integrating human beings with transport systems, machinery and products at sophisticated levels, and this goal can be effectively achieved by a Smart factory (Zuehlke, 2010). A cost-effective manufacturing Smart factory provides a highly flexible, distinct and individualized production process along with availability, positioning and locating of the product. A Smart factory reduces complexity in production and provides effective tracking of both process and product. It not only promotes changes in the production process, it also minimizes the duties and responsibilities of employees. The advantages of a decentralized manufacturing system encourage employees to act independently and instantly (Radziwon, Bilberg, Bogers, & Madsen, 2014). 1.2.2  Core Principles of Industry 4.0 Industry 4.0 is not just a technological development; it is a concept that could disrupt many industries. The concept is developed by integrating six core principles, namely interoperability, decentralization, virtualization, modularity, service orientation and real-time capabilities (Qin, Liu, & Grosvenor, 2016). Interoperability Interoperability is the primary principle or the key initiator of Industry 4.0. The system’s capacity to communicate with various other systems to coordinate distinct functions and to exchange data is termed interoperability. Interoperability provides man and machine the ability to obtain real-­ time data, which enables faster and more effective decision-making. Without Interoperability, immense amounts of data gathered in warehouses every day remain unused, or are not exchanged with other systems for processing. To develop opportunities and to increase the presence of man and machine integration, facilities should be linked with IoT. Interoperability enables the integration of software such as Enterprise Resource Planning system (ERP), Electronic Medical Records (EMRs), Laboratory Information Management system (LIMS) and other software,

6 

J. CHANCHAICHUJIT ET AL.

thereby minimizing the transaction cost between the software systems in analysing and consolidating the data. The data gathered from distinct systems and devices is processed and consolidated into knowledge which can assist in and improve decision-making (Lu, 2017). There are a few procedures which enhance the ability of interoperability. The standard and proprietary protocols of current processes and systems require analysis and evaluation. Protocols that require costly custom-coding or that promote a single supplier strategy should be eliminated. Where there are several business issues, the issue that requires more consideration should be prioritized. Uptime, productivity, cost, speed and accuracy are some significant business issues. Business issues can be prioritized using real-time business case research, and this prioritization enhances the ability of interoperability. The aim is to develop longterm guidelines for real-time decision-making and man and machine integration in the future. Long-term guidelines also promote improvements in interoperability and the ability to adopt other core principles. Interoperability is also used to allow IoT to intersect with other Industry 4.0 components (Wollschlaeger, Sauter, & Jasperneite, 2017). Modularity Today’s software companies would benefit greatly from anticipating future risks and being able to “disable” uncertainty. The goals of enhanced productivity and profitability competitiveness may be achieved by the adoption of new technologies. Yet the companies which adopt this technology still face difficulties in upgrading, since various disruptive changes cannot necessarily be anticipated, nor the changes incorporated into the new technology. A system which intrinsically adapts to change and new advancements is termed a “modular system”. Modular systems enable a company to respond quickly to demand fluctuation and ensure the security of initial investments during fluctuation. Virtualization Some functions that cannot be executed in the physical world may be performed in the digital world. Data obtained from facilities, along with their equipment and processes, is simulated with virtual simulation models to develop a digital view of operations. This digital view is termed Virtualization and it provides the ability to minimize equipment downtime, enhance processes and handle complex situations. The virtualized view is helpful in coordinating and monitoring the physical and digital world.

1  AN INTRODUCTION TO HEALTHCARE 4.0 

7

Remote services are performed using Augmented Reality. Augmented Reality is one of the key components used to incorporate virtualization capabilities. Providing remote services and monitoring the condition and location of the product are just some of the tangible benefits of Virtualization. Many organizations face challenges in understanding the benefits and impact of incorporating new technology into their processes. Virtualization provides an exact view of activities performed by “human and machine”, along with the capability of optimizing processes and using preventative measures to mitigate risk. The combined benefits of mobile robots, virtual reality and Augmented Reality equipment may provide us with great opportunities in the future (Klement, 2017). R  eal-Time Capabilities To obtain ongoing or real-time information on equipment and its processes is the ultimate aim of the core principles. Thus the Virtualization and Interoperability principles of Industry 4.0 promote real-time capabilities. CPS is used to collect real-time data across an entire supply chain. Robots, Auto Guiding Vehicles (AGVs) and equipment which interfaces with computerized devices such as scanners, sensors and Radio Frequency Identification (RFID) tags and which connects with IoT provide visibility and real-time data. In such cases, man and machine can make real-time decisions with the help of real-time data. Adequate data can be collected to enhance current operations. In summary, real-time data is collected to optimize operations and to enable real-time decision-making (Wittenberg, 2016). D  ecentralization In the traditional manufacturing process, several subsystems in each stage of the process were supported by a centralized system. In a centralized structure, a central computer embedded with business logic is used to provide solutions to other subsystems. With Industry 4.0, there are certain restrictions around having a centralized structure. A centralized structure limits scalability. It is also difficult to adapt to upcoming advancements or respond to fluctuations, as the structure cannot be altered once it reaches its maximum capacity. In a distributed structure, logic nodes can be used to help or handle the subsystems or remote components. To enhance intelligence and functionally in a distributed structure, the data collected is shared with every node and the capabilities of every node are combined. Components or ­subsystems are programmed with business logic in a completely decentralized structure. This capability enhances the intelligence required to execute neces-

8 

J. CHANCHAICHUJIT ET AL.

sary functions and allows coordination with other subsystems to manage more complicated tasks. From the Industry 4.0 perspective of decentralization, more robots and AGVs can be added to enhance the easy scalability of the operation, and decentralized decision-making enhances speedier execution of operations. The subsystems and workers are coordinated with the help of CPS. Improvement in intelligence and functionality can only be fully realized when the subsystems and their operations are distributed or decentralized. Service Orientation The activities or services carried out by machines and humans are optimized by connecting to the internet. IoS is used to optimize the service and this is performed in order to enhance service orientation. From the initial stage of movement of merchandise to the final stage of data analytics, every service involved is overseen via the internet to mitigate specific business issues (Jiang, Ding, & Leng, 2016). To illustrate the preceding point, if a modular assembly station equipped with AGVs is subjected to a service-oriented approach, IoS serves as a platform for the AGVs and modular stations to perform necessary services. The RFID tags on merchandise contain design procedures, and the services required with respect to the design are decided autonomously by the machines. At that point, the machine formulates the required procedure and directs the services to be performed through IoS. In spite of collecting and storing large amounts of data, the exchange of information between various systems becomes too complex. Yet, service orientation empowers more liberated data streams between and within systems. Software used by a company serves as a tool to manage internal services, which in turn maximizes the benefits of external functionality. The supporting software serves as a well-grounded platform to optimize and execute business processes. Finally, a greater capacity to alter processes and give higher scalability is provided by service orientation (Stock & Seliger, 2016).

1.3   Challenges in Implementing Information Technologies in Healthcare Despite healthcare delivery services having much to gain from implementing information technologies, they have been the slowest of all industries in the adoption of such technology. There are many reasons for IT failures in the healthcare environment, but the single most important cause is a

1  AN INTRODUCTION TO HEALTHCARE 4.0 

9

technology capability mismatch in addressing work processes within healthcare service organizations (Yusof, Stergioulas, & Zugic, 2007). For over 20 years Information and Communication Technology (ICT) and healthcare service organizations have been unable to find a comprehensive solution. It may be necessary to research less into design and implementation and further into how an end user reacts to already implemented IT solutions (Morrow, Robert, Maben, & Griffiths, 2012). IT investment will only be successful if the fit between IT and clinical processes is a comfortable one, reflected best by the acceptance or rejection by end users (Yusof, Kuljis, Papazafeiropoulou, & Stergioulas, 2008). In the short history of IT, the emergence of new, disruptive technologies plays a crucial role in closing the capability gap and in gaining more acceptance from the main users. The latest innovations are changing or disrupting how medical care is organized, practised and delivered. They are also redefining a host of other aspects such as changing the patient-physician model and facilitating the emergence of new industry players within the value chain. It is hoped that these innovations will be successful in delivering better, smarter care.

1.4   Stages in Healthcare IT Transformation Between 1970 and 1990, we saw the emergence of modular or silo IT systems in the healthcare industry. This period could safely be called Healthcare 1.0. Throughout the next decade and a half, most IT systems commenced networking, and Electronic Health Records (EHRs) that were being generated started integrating with clinical imaging, giving doctors a better perspective. This was Healthcare 2.0. From the year 2000 onwards, we saw the development of genomic information, along with the emergence of wearables and implantables. The integration of all the resultant data, along with networked electronic patient records, saw the emergence of Healthcare 3.0. However, due to data incompatibility and resistance from healthcare providers, the a­ doption of IT in Healthcare 3.0 did not produce significant improvements for the community. What we are seeing today is the emergence of Healthcare 4.0. Its intention is to apply some of the principles of Industry 4.0 by integrating technologies with IoT for data collection, increasing the use of AI for analysis and using the overlay of a blockchain for patient medical records. The focus on collaboration, coherence and convergence should make healthcare more predictive and personalized.

10 

J. CHANCHAICHUJIT ET AL.

The enhanced amount of data available to doctors should be of great benefit; however, the critical advantage may be found in the ability to extract insights from the data being captured, and the portability of this data using blockchain. Data portability and interoperability would allow patients and their physicians to access information anytime, anywhere. Enhanced analytics would allow for differential diagnoses and medical responses that can be predictive, timely and innovative (Chawla & Davis, 2013). Healthcare 4.0 allows valuable data to be used more consistently and effectively. It can pinpoint areas of improvement and enable people to make more informed decisions. What it also does is help move the entire healthcare industry from a system that is reactive and focused on fee-for-service to a system that is value-based, which measures outcomes and encourages proactive prevention. Table  1.1 shows the transition from Healthcare 1.0 to Healthcare 4.0.

Table 1.1  Transition from Healthcare 1.0 to Healthcare 4.0

Main objective

Focus

Information sharing

Healthcare 1.0 Healthcare 2.0

Healthcare 3.0

Healthcare 4.0

Improve efficiency and reduce paper work Simple automation

Improve data sharing and productivity

Provide patient-­ centred solutions

Provide real-time tracking and response solutions

Connectivity with other organizations

Interactivity with patients

Within an organization

Within a cluster of healthcare providers EDI and cloud computing with HL7 messages for exchange Sharing of critical information only but not interacting with patients

Within a country

Integrated real-time monitoring, diagnostics with AI support Global healthcare supply chain

Key LIMS and technologies administrative used systems Limitations

Stand-alone systems with limited functionality

EMR, Big data, wearable devices, optimization system Different standards used within the community with limited interoperability

IoT, Blockchain, AI, Data analytics

New and untested technologies with concerns about data privacy

1  AN INTRODUCTION TO HEALTHCARE 4.0 

11

1.5   Drivers for Healthcare 4.0 There are several factors influencing the drive towards improved healthcare. First, in some developed countries such as Singapore, the USA and the UK, central governments have proposed a national IT “backbone” that will help in integrating EMRs and making them portable. For governments, initiatives like this are key in meeting such societal objectives as enhanced access to healthcare and improved patient outcomes (Qin et al., 2016; Sligo, Gauld, Roberts, & Villa, 2017). The second factor is the rise of an IT-savvy population that is more informed and as such demands better service from its healthcare providers. The healthcare providers themselves—diagnosticians, physicians, surgeons and hospitals as a whole—are realizing that with the increased use of Healthcare 4.0-enabled tools, their efficiency is enhanced and outcomes are becoming more effective (Eysenbach et al., 2013). The third factor is the emergence of integrated care. The move from patient-centred to integrated care will shift the primary focus from disease and treatment to wellness and prevention. Truly integrated care will take into consideration not just the individual but also factors like medical history of their family, the patient’s lifestyle, demographics and their ability to access healthcare. Taking all these factors into account, healthcare plans can be developed with a focus on providing care that is personalized, enabling and coordinated, and that treats people with compassion and respect. The final factor that is driving this change is the data revolution that is currently taking place in many countries (Chawla & Davis, 2013). Access to affordable, high-speed data connectivity—as a result of some governments’ initiatives and private sector competition—makes it possible for ­ enefits both doctors and patients from smaller towns to access some of the b of Healthcare 4.0. To sum up, the focus is shifting to integration of capabilities, integrated care and ownership of clinical outcomes. One key issue that requires tackling is the need expressed by patients and even doctors for a physical presence. This need calls for a hybrid model that combines hi-tech with hi-touch. What may provide a solution is a platform that connects the entire ecosystem (integrated and connected devices), one that extracts valuable patient information, mines the derived intelligence, innovates with predictions and delivers blended (hi-tech and hi-touch) services across the entire spectrum of wellness, prevention, cure and care.

12 

J. CHANCHAICHUJIT ET AL.

1.6   Recommendations If healthcare services are to incorporate Industry 4.0 core principles, they require proper guidelines or a framework within which to incorporate the core principles. Based on the key factors determined from research and the knowledge acquired from literature reviews, a set of emerging technologies is proposed for implementation in the healthcare sector. Table  1.2 demonstrates how these emerging technologies are fulfilling the core principles of Industry 4.0. The benefits of adopting these emerging technologies for Healthcare 4.0 include: 1. IoT and wearable devices—Empowering patients to perform self-­ management of medical needs, and provide channels for more interactive communication with healthcare professionals; 2. Blockchain technology—Providing real-time capturing of patient clinical records; 3. Artificial Intelligence—Providing more accurate predictive models of a patient’s condition; and 4. Big data and mobile applications—Maximizing healthcare resources, and increasing the preventive and predictive components of care with the expectation of keeping individuals as healthy as possible and less dependent on curative care. Table 1.2  Emerging technologies to support Industry 4.0 Core principles of Industry 4.0

Internet of Things/ wearable devices

Interoperability Decentralization Virtualization Modularity Service orientation Real-time capabilities

Yes Yes

Big data and mobile Apps

Blockchain Yes Yes

Yes Yes Yes Yes

Artificial Intelligence

Yes

Yes Yes Yes Yes

1  AN INTRODUCTION TO HEALTHCARE 4.0 

13

1.7   Conclusion Industry 4.0 is not just a technological advancement; it is a concept that may be used to enhance the “intelligence” and functionality of any industry. This concept comprises six core principles which demonstrate Industry 4.0’s capabilities and which can be applied to the healthcare industry. The author has summarized the evolution of IT implementation over the decades and how Industry 4.0 can help to transform the healthcare industry. The following chapters of this book will explain in more detail the use of IoT and big data analytics, blockchain, Artificial Intelligence and optimization techniques in healthcare. The book concludes by highlighting the innovative health technologies and start-up process in the healthcare industry and recommending processes to transform and manage healthcare. We hope that the adoption of these emerging technologies will raise the level of the healthcare industry towards Industry 4.0.

References Chawla, N., & Davis, D. (2013). Bringing big data to personalized healthcare: A patient-centered framework. Journal of General Internal Medicine, 28(Suppl. 3), 660–665. https://doi.org/10.1007/s11606-013-2455-8 Drăgoicea, M., Pătraşcu, M., & Bucur, L. (2012). Service orientation for intelligent building management: An IOT and IOS perspective. Paper presented at the UNITE 2nd Doctoral Symposium, R & D in Future Internet and Enterprise Interoperability, Sofia, Bulgaria. Eysenbach, G., Siadat, H., Solvoll, T., Keeling, D., Li, J., Talaei-Khoei, A., … MacIntyre, C. R. (2013). Health care provider adoption of eHealth: Systematic literature review. Interactive Journal of Medical Research, 2(1). https://doi. org/10.2196/ijmr.2468 Garibaldo, F., & Rebecchi, E. (2018). Cyber-physical system. Journal of Knowledge, Culture and Communication, 33(3), 299–311. https://doi.org/10.1007/ s00146-018-0802-3 Gubbi, J., Buyya, R., Marusic, S., & Palaniswami, M. (2013). Internet of Things (IoT): A vision, architectural elements, and future directions. Future Generation Computer Systems, 29(7), 1645–1660. https://doi.org/10.1016/j. future.2013.01.010 Jiang, P., Ding, K., & Leng, J. (2016). Towards a cyber-physical-social-connected and service-oriented manufacturing paradigm: Social manufacturing. Manufacturing Letters, 7(C), 15–21. https://doi.org/10.1016/j. mfglet.2015.12.002

14 

J. CHANCHAICHUJIT ET AL.

Klement, M. (2017). Models of integration of virtualization in education: Virtualization technology and possibilities of its use in education. Computers & Education, 105(C), 31–43. https://doi.org/10.1016/j.compedu.2016.11.006 Lu, Y. (2017). Industry 4.0: A survey on technologies, applications and open research issues. Journal of Industrial Information Integration, 6, 1–10. https:// doi.org/10.1016/j.jii.2017.04.005 Manogaran, G., Thota, C., Lopez, D., & Sundarasekar, R. (2017). Big Data Security Intelligence for Healthcare Industry 4.0. In L. Thames & D. Schaefer (Eds.), Cybersecurity for Industry 4.0: Analysis for design and manufacturing (pp. 103–126). Cham: Springer International Publishing. Morrow, E., Robert, G., Maben, J., & Griffiths, P. (2012). Implementing large scale quality improvement: Lessons from The Productive Ward: Releasing Time to Care™. International Journal of Health Care Quality Assurance, 25(4), 237–253. https://doi.org/10.1108/09526861211221464 Poovendran, R. (2010). Physical systems: Close encounters between two parallel worlds [Point of View]. Proceedings of the IEEE, 98(8), 1363–1366. https:// doi.org/10.1109/JPROC.2010.2050377. Cyber–. Qin, J., Liu, Y., & Grosvenor, R. (2016). A categorical framework of manufacturing for Industry 4.0 and beyond. Procedia CIRP, 52(C), 173–178. https:// doi.org/10.1016/j.procir.2016.08.005 Radziwon, A., Bilberg, A., Bogers, M., & Madsen, E. S. (2014). The smart factory: Exploring adaptive and flexible manufacturing solutions. Procedia Engineering, 69, 1184–1190. https://doi.org/10.1016/j.proeng.2014.03.108 Sligo, J., Gauld, R., Roberts, V., & Villa, L. (2017). A literature review for large-­ scale health information system project planning, implementation and evaluation. International Journal of Medical Informatics, 97, 86–97. https://doi. org/10.1016/j.ijmedinf.2016.09.007 Stock, T., & Seliger, G. (2016). Opportunities of sustainable manufacturing in Industry 4.0. Procedia CIRP, 40, 536–541. https://doi.org/10.1016/j. procir.2016.01.129 Tjahjono, B., Esplugues, C., Ares, E., & Pelaez, G. (2017). What does Industry 4.0 mean to supply chain? Procedia Manufacturing, 13, 1175–1182. https:// doi.org/10.1016/j.promfg.2017.09.191 Wittenberg, C. (2016). Human-CPS Interaction—Requirements and human-­ machine interaction methods for the Industry 4.0. IFAC PapersOnLine, 49(19), 420–425. https://doi.org/10.1016/j.ifacol.2016.10.602 Wollschlaeger, M., Sauter, T., & Jasperneite, J. (2017). The future of industrial communication: Automation networks in the era of the internet of things and industry 4.0. Industrial Electronics Magazine, IEEE, 11(1), 17–27. https:// doi.org/10.1109/MIE.2017.2649104

1  AN INTRODUCTION TO HEALTHCARE 4.0 

15

Wortmann, F., & Flüchter, K. (2015). Internet of things. The International Journal of WIRTSCHAFTSINFORMATIK, 57(3), 221–224. https://doi. org/10.1007/s12599-015-0383-3 Yusof, M. M., Kuljis, J., Papazafeiropoulou, A., & Stergioulas, L. (2008). An evaluation framework for Health Information Systems: Human, Organization and Technology-fit factors (HOT-fit). International Journal of Medical Informatics, 77(6), 386–398. https://doi.org/10.1016/j.ijmedinf.2007.08.011 Yusof, M.  M., Stergioulas, L., & Zugic, J. (2007). Health information systems adoption: Findings from a systematic review. Studies in Health Technology and Informatics, 129, 262–266. Zezulka, F., Marcon, P., Vesely, I., & Sajdl, O. (2016). Industry 4.0—An introduction in the phenomenon. IFAC PapersOnLine, 49(25), 8–12. https://doi. org/10.1016/j.ifacol.2016.12.002 Zuehlke, D. (2010). SmartFactory—Towards a factory-of-things. Annual Reviews in Control, 34(1), 129–138. https://doi.org/10.1016/j.arcontrol.2010.02.008

CHAPTER 2

Internet of Things (IoT) and Big Data Analytics in Healthcare

Abstract  This chapter presents an overview of the Internet of Things (IoT) and big data analytics in healthcare. The healthcare supply chain is introduced in order to understand the material and information flow and how healthcare data is utilized. The concept of IoT as an element of Industry 4.0 and the healthcare industry is explained. IoT aims to identify, track and authenticate objects and people, in particular medical devices and patient data, for further analysis, which is where big data analytics plays a major role. The chapter is followed by a discussion of how to use big data analytics in healthcare and the challenges of and strategies for implementing big data. It aims to bridge existing knowledge in the literature and draws upon a prior big data analytics project in Singapore as a case study. The chapter concludes by highlighting lessons learnt from the use of IoT and big data analytics case studies in the healthcare delivery system. Keywords  Internet of Things (IoT) • Big data analytics • Preventive • Intervention • Diabetes • Singapore

2.1   Overview of the Healthcare Supply Chain The current healthcare supply chain is much more complex than it was 20 years ago. In addition to hospitals and patients, the supply chain includes various stakeholders. Members of the healthcare supply chain can be © The Author(s) 2019 J. Chanchaichujit et al., Healthcare 4.0, https://doi.org/10.1007/978-981-13-8114-0_2

17

18 

J. CHANCHAICHUJIT ET AL.

divided into four major groups, namely producers, purchasers, providers and patients. Healthcare producers are responsible for manufacturing medical supplies, surgical supplies, medical devices and pharmaceuticals. These medical products are provided to purchasers include wholesalers, distributors and Group Purchasing Organizations (GPOs). Purchasers are intermediaries who keep inventory and deliver the right products in a timely fashion to the downstream customers. The majority of healthcare products are allocated through wholesalers and distribution centres. Shah (2004) stated that 80% of products are delivered by wholesalers. Another method of product distribution is carried out via GPOs, which may provide considerable savings to providers where large-volume purchases attract discounts from manufacturers. Healthcare providers refer to those working directly with patients. Providers comprise hospitals, clinics, physicians, pharmacies, integrated delivery networks and nursing homes. In addition to the producers, purchasers and providers, other elements such as insurance companies, governmental policies and regulatory agencies also contribute to the complexity of the chain (Mathew, John, & Kumar, 2013).

2.2   Interoperability Issues in the Healthcare Supply Chain Supply chain management is about managing the three flows in the chain: material flow, information flow and money flow. In the manufacturing sector, the flow of product starts at the manufacturer and reaches the end users through several channels, depending on the product type or characteristics. With health data, the flow of information must run smoothly from the upstream supplier to the downstream customer and back and forth. However, due to various factors such as privacy and confidentiality in the healthcare network, the reliability and validity of health data is difficult to maintain. Each party in the network has different ways of utilizing health data. Producers need health data to analyse the efficiency of medicines in treating symptoms, as well as data on the frequency of side effects from an active ingredient. Healthcare providers can use historical health data in health examination processes, while patients who wish to manage their care better use data. In order to provide all parties with real-time health information, EHRs are widely adopted due to the diverse clinical benefits they bring. An EHR system refers to a historical electronic records

2  INTERNET OF THINGS (IOT) AND BIG DATA ANALYTICS IN HEALTHCARE 

19

system which contains the health information on a patient including demographics, health issues, medication, health examination reports, recovery progress and past medical history. The EHR system allows information to be exchanged electronically among relevant parties, in order to provide timely treatment. Because of its automation, the EHR system facilitates reduced paperwork and allows streamlining of patient data. As Menachemi and Collum (2011) argued, the most fundamental advantage in adopting the EHR system is in the generation of improved clinical outcomes in terms of fewer medical errors, more accuracy in measurement of patient health and better quality of care. Studies have shown that the application of EHRs contributed 78% to the improvement of patient care (Kulkarni & Sathe, 2014). In addition to improved care, EHRs can assist in eliminating unnecessary clinical tests, which in turn reduces related costs and enhances patient satisfaction. The majority of patients felt respected when clinical practitioners appeared to listen to their problems and dedicated themselves to treating the patient after implementing EHRs (Mysen, Penprase, & Piscotty, 2016). Additionally, the availability and timeliness of data in EHRs supports more thorough medical research. According to Bowles et al. (2013), studies for the purpose of continuous improvement of healthcare services can contain data on patient characteristics, such as health status, doctor’s assessments, medication process, functional parameters and depression risk score. Based on the analysis of these figures, doctors and clinicians may determine the types and length of treatment required. Even though EHRs have made a great contribution to the recording and storage of health data, there exists a big challenge for every EHR system: interoperability. The general concept of interoperability refers to the interconnection and ability to exchange information between two or more components (Bhartiya, Mehrotra, & Girdhar, 2016). Interoperability in healthcare is defined as the ability to exchange, communicate and make use of health information between one organization and another, for the purpose of enhancing the quality of healthcare delivery to individuals and the public. The movement of health data between systems and organizations needs to conform to several interoperability norms: • Preservation of the meaning and purpose of data; • Consistent presentation of data regarding different information systems;

20 

J. CHANCHAICHUJIT ET AL.

• Consistent system of controls supporting similar actions across all applications; • Secured and integrated data that grants authorization to specific people and programmes; • Protection of patient’s private information; and • Consistency in degree of quality, in terms of availability, reliability and responsiveness. This is the first step that distinct systems need to take to achieve the goal of collaboration and effective networking (Maheshwari & Janssen, 2014). Interoperable EHRs need to satisfy these requirements when sharing information between different units of a hospital or different hospitals. Three elements need to be taken into consideration: the content of data exchanged, the tools used to exchange data and the amount of data (Bhartiya et  al., 2016). However, there are several barriers to complete interoperability between distinctive EHRs. EHRs are designed to fulfil the requirements of the user. That is, each health provider implements a specific version of EHRs depending on the existing infrastructure and their budget, even though the EHR system is provided by the same vendor, which poses a challenge in collecting data and presenting data to the user in the earlier version and the latest version. In addition, each site may require specific customization in terms of importance of the presented information, the wording or the order of a standard list of options in response to a specific question (Bowles et al., 2013). Additionally, a doctor can play multiple roles in the same hospital and thereby request different information. For example, they may be a primary care physician of a patient and a secondary specialist for another at the same time. The complexity of job assignment and responsibility makes it difficult to authorize and control the access to health records (Bhartiya et al., 2016). A lack of interoperability prevents effective data sharing in the healthcare environment. It not only affects healthcare providers in advancing their healthcare services, but patients are also limited in their interaction with and access to their medical records. The healthcare industry is highly fragmented in nature. As health problems can occur suddenly over one’s lifetime, and there is now greater movement from one provider to another, patients may find it extremely difficult to access their past data and review their medical records. For example, the Health Insurance Portability and Accountability Act in the USA stipulates that patients must be allowed to see and make amendment to their records routinely. In the age of automa-

2  INTERNET OF THINGS (IOT) AND BIG DATA ANALYTICS IN HEALTHCARE 

21

tion and technology advances, patients are desirous of controlling their information. Therefore, it is appropriate to bring about a new system which prioritizes the benefit to the patient. A new system would make data available and accessible to relevant parties, while resolving the interoperability issue. Without interoperability, and with scattered data, it is impossible to create a comprehensive record and gain insights.

2.3   Internet of Things (IoT) Being an element of Industry 4.0, IoT has received increased attention from private users, businessmen and researchers. The term Internet of Things is not as simple as its fundamental idea, which is to connect a variety of things or objects in a network at any time and any place (Lu, Papagiannidis, & Alamanos, 2018). Different users and researchers with unique purposes have distinct views on what IoT really means, and its implications for the future. It is therefore necessary to thoroughly understand some of the views of IoT, before considering the achievements and challenges brought about by its presence. A common definition, which is given by Atzori, Iera, and Morabito (2010), characterizes the IoT paradigm as a combination of three “visions”, namely “Things-oriented”, “Internet-oriented” and “Semantic-­ oriented” visions. According to Atzori, Iera and Morabito, the concept of IoT is confusing because two distinctive terms have been put together in a new framework. From the Things-oriented or Internet-oriented viewpoint, IoT is a network of heterogeneous devices which are interconnected and addressable through a mutual communication protocol. On the other hand, Gubbi, Buyya, Marusic, and Palaniswami (2013) explained IoT as a “smart” environment comprising sensing devices which are connected in a mutual digital structure and able to share data over different platforms to create a complete integration between applications and be open to future innovations. Data sharing is supported by sensing techniques, data analytics and cloud computing. This interpretation encourages the utilization of any available protocol to develop durable and compatible applications without depending on a standard means of communication. In addition to the general idea of connecting physical things in an intelligent network, Patel and Patel (2016) argued that IoT involves the ­interaction through the internet in three ways: machine to machine, people to machine, and people to people. The authors have visions or concepts of a variety of things which are linked using wireless or wired

22 

J. CHANCHAICHUJIT ET AL.

connections and given unique addresses to cooperate and converge with each other to create a smart world. The paradigm refers not only to electronic devices, but also to non-electronic materials such as food and clothing. Despite the discrepancies among interpretations, the ultimate goal of IoT is unchanged, that is, allowing any device to connect at any time, in any place to anybody, deploying any kind of network and any service. The future of such a ubiquitous network is enormous, and it is predicted to be gradually achieved through the enabling technologies that IoT currently facilitates in several industrial areas. 2.3.1  Internet of Things (IoT) in the Healthcare Supply Chain IoT has been broadly utilized in the field of medical care for a long time. Most IoT-based health applications are designed to identify, track and authenticate objects and people, collect patient and staff data, and use sensors for specific purposes (temperature, smoke, etc.). These applications can be categorized into two sections: clinical care and remote control. Clinical Care IoT provides an infrastructure for paying close attention to patients. Many hospital devices are now attached to sensors to constantly capture patients’ health indicators, including blood pressure, level of oxygen in blood, heart rate, cholesterol level and other data. Real-time data is then transmitted to a central device such as a computer or mobile in a wireless network where it is classified and analysed. IoT helps save the time and effort of medical staff, who can monitor patients through an autonomous continuous information flow instead of performing the repetitious tasks of collecting information. In this way, IoT can enhance the quality of healthcare services and lower related costs. Many smart health device ideas using IoT have been proposed and many of them are in progress. For example, Withings devices are smart connected devices for healthcare purposes. Two popular Withings devices include the Wi-Fi body scale to measure body fat, and the Withings Blood Pressure Monitor to estimate the blood pressure and heart rate of a user (Yin, Zeng, Chen, & Fan, 2016). Data is transmitted through Wi-Fi or Bluetooth and synchronized in a Google Device Health service application, which can be accessed via multiple mobile platforms such as Android and iOS. Further developments in wearable devices for tracking weight, temperature and daily calories have received much attention, for instance AliveCor Heart Monitor and Viatom Checkme.

2  INTERNET OF THINGS (IOT) AND BIG DATA ANALYTICS IN HEALTHCARE 

23

Remote Control Wireless IoT solutions make it possible to access patients’ data at any time and remotely monitor it. A network of sensing devices and healthcare wearables facilitates the collection of the complete health profile of a patient, which can be a guideline towards recommendations for treatment and appropriate medicine. Doctors and nurses can take care of every vital sign and use records to avoid any wrong diagnosis or misuse of medicine (Fang, Dan, & Shaowu, 2013). Real-time tracking systems using RFID tags or Infrared Data Association (IrDA) technology can keep updating the real-time location and conditions of patients as well as hospital staff. The tags can be attached to medical equipment or a patient’s wristband, identifying the location of the tagged objects. Under emergency circumstances, the system can help to identify the exact location of a patient where immediate treatment is needed or it can alert caregivers if a patient leaves hospital without permission (Fang et al., 2013). Another application of RFID technology is in the tracing of prescription drugs. An RFID tag contains information on the lot number, expiry date and every point of contact from the manufacturer to the dispensing pharmacy, allowing quality and inventory control. Challenges of Implementing the Internet of Things 2.3.2   Interoperability Interoperability refers to the ability of two systems to exchange information and make use of information. This is the core operating basis in a traditional computer environment, as computers are connected to the internet on a peer-to-peer basis (Elkhodr, Shahrestani, & Cheung, 2016). However, achieving interoperability between IoT-based devices is more difficult. Each device is designed with a different bandwidth, level of security and amount of power consumption, thereby requiring a distinctive standard and communication layer. An IoT network combines a vast number of connected things, so a common configuration that meets the requirements of all components poses a challenge to all IT programmers. On the other hand, IoT is developing rapidly as an indispensable ­consequence of the digital technology era, which expands the opportunities for more and more devices to communicate with each other. For the smooth and effective operation of the wireless network, the interoperability issue needs to be addressed.

24 

J. CHANCHAICHUJIT ET AL.

Data Management IoT changes the way data is collected, stored and exchanged. Proper data management that ensures complete, accurate and consistent data across various devices and systems is a key success factor in an IoT network. The possibility of human error has been recognized with regard to the generating and gathering of data, thus leading to lower reliability of the system (Taylor-Adams & Kirwan, 1995). IoT adoption should help optimize the process of collecting information and avoid subjective errors. The interconnection between heterogeneous devices calls for a vast amount of data from diverse sources to be shared in a short time, and this appears to be a burden for IoT applications. How to store, analyse and control the volume of generated information for better business performance becomes critical (Abbasi, Memon, Memon, Syed, & Alshboul, 2017). Privacy and Security Whenever a computer connects to the internet, there is the chance that it will be attacked by anonymous individuals or malware and data may be stolen, altered or destroyed if an inadequate method of security protection is employed. IoT breaks the barriers between devices for data transmission, thus creating conditions for cyber-attacks. Atzori et  al. (2010) pointed out several reasons why an IoT network is vulnerable to cybercrimes, including shortage of supervision of IoT components, lack of complex secured methods due to IoT components’ characteristics, and wireless environments leading to easy physical attack. Some vital problems regarding the security of IoT networks concern data access management, data integrity, authorization and authentication of exchanged data, physical attacks and denial of service attacks (Elkhodr et al., 2016). An attack occurring in one node of the network can easily bridge sequential disruption to the entire network (Abbasi et al., 2017). Furthermore, large amounts of data exchange, including personal and confidential information of users, raise more concerns in terms of privacy assurance. Because the features of IoT facilitate the automatic collection, storage and sharing of information, there is a higher risk that attackers could gain access to an IoT application and interrupt personal data. Numerous cases of information leakage on mobile platforms have been reported (Elkhodr et al., 2016). Secure mechanisms to control the disclosure of personal information are required for IoT to function effectively.

2  INTERNET OF THINGS (IOT) AND BIG DATA ANALYTICS IN HEALTHCARE 

25

2.4   Overview of Big Data Analytics in Healthcare A survey by IDC Health Insights found that the top priority for investment of 50% of hospitals and healthcare insurers over the next 10 years was that of increasing their data analytics capabilities and Industry 4.0 approach. The benefits of IoT, big data analytics, blockchain and predicting analytics range from predicting epidemics to curing cancer and making hospital stays a more pleasant experience. In addition, Marr (2015) highlighted that big data analytics will be a major force of change in the healthcare industry. The majority of data in healthcare was previously stored in the form of hard copy records or single source data storage. But today, given an increasing amount of real-time data exchange between various sources such as handheld devices, wearable devices, smart devices and the use of digital health data such as EHRs, the trend of data management in healthcare is moving towards digitization. All this data is termed big data. Big data is not a new term, being coined in 2001 by Doug Laney. Laney (2001) identified big data according to the significant characteristics of Volume, Velocity and Variety (3Vs). Volume refers to the massive amount of data that is unable to be managed with traditional tools and techniques. Today, health-related data from various sources has created exponential growth. Dell EMC (2014) reported that in 2020 big data around the world is expected to reach 44 zettabytes (1021 gigabytes), and in the healthcare system it is estimated to increase up to 35 zettabytes (33). Velocity is the speed of data generation. Due to current advances in the technology of data management, health data is generated with greater speed and accumulated in real time (Raghupathi & Raghupathi, 2014). Real-time data analytics will be of benefit in detecting disease as early as possible. The last V, Variety, is data diversity comprising structured, unstructured and semi-structured data. Following this, Fieldman, Martin, and Skotnes (2013) proposed the fourth V, Veracity, as a big data key characteristic in healthcare. Veracity refers to the accuracy and quality of healthcare data. This characteristic is vital, as poor-quality data may have very harmful consequences in patient care. Later, Manyika et al. (2011) introduced the fifth V, Value, as an additional key dimension of big data. Big data in healthcare is not only big in size and number but it varies in the different types of data collected from different sources. The speed, quality and reliability of data and, importantly, value received from data analytics allow meaningful actions to cure and save patient’s lives. The last

26 

J. CHANCHAICHUJIT ET AL.

characteristic of big data, “Value”, has set a new standard for healthcare big data analytics and other industries in providing value for customers and patients in health services. In general, big data analytics in healthcare refers to health data that requires advanced analytical techniques and technologies to retrieve, store, distribute and analyse. This analysis allows the discovery and understanding of data trends, patterns and meanings. Through the analysis, insights may be obtained, allowing better decision-making, improving healthcare systems, saving more lives and preventing diseases.

2.5   How to Implement Big Data in Healthcare 2.5.1  Type and Source of Healthcare Big Data Healthcare data is expected to grow dramatically due to the availability of digital data. Previously, healthcare data was kept from physician’s notes and prescriptions and mostly stored in hard copy files or on a local computer network. Hospitals today are collecting a multitude of data from different sources and devices such as sensors and other smart devices through the Healthcare Information System in EMRs. In addition, healthcare data generally has a different format and structure (Rouse, Serban, & Moses, 2014) and is generated from internal and external sources. Healthcare data can be categorized into two main groups: health data (biomedical) and nonhealth data (administrative). Biomedical data comprises the omics group of data (genomics, microbiomics, proteomics and metabolic data) and patient-generated data. This can be used to analyse disease mechanisms of each individual patient for more effective treatment. This data can be derived from various sources such as EMRs, wearable devices or sensorgenerated, computerized physician order entries and health reports. Administrative data includes EMRs, clinical data, insurance claims and pharmaceuticals. This type of data provides benefits to physicians in understanding the patient’s health background and providing better healthcare treatment. Another group of data is the non-­health data from social media, external databases, wearables and sensor-generated devices. These types of data are used to provide ­information on patient behaviour and lifestyle. Non-health data becomes more useful in healthcare analytics when combined with other information. For example, Saeb et al. (2015) found that the use of mobile phone data was correlated to patient depression rates. Thus, it can be seen that the right amount of health and non-health data can benefit healthcare analytics.

2  INTERNET OF THINGS (IOT) AND BIG DATA ANALYTICS IN HEALTHCARE 

27

2.5.2  Techniques and Technologies Used in Healthcare Big Data Analytics Previously, healthcare data analytics could easily be performed by using business intelligence tools and techniques. However, today’s healthcare data has become big data, which is more sizeable, complex and dynamic, and with different formats and sources. Therefore, a sophisticated computing infrastructure and efficient data analytic tools are necessary to manage and analyse the data. There are four types of healthcare data analytics: descriptive, diagnostic, predictive and prescriptive (Thammasudjarit, Pattanateepon, & Pattanaprateep, 2018). In Thammasudjarit et al.’s work (2018), heart failure patients admitted to hospital demonstrated how different types of data analytics are applied. In this work, descriptive analytics are used to explore the data related to the current condition of patients to find out the possible occurrences. Diagnostic analytics are then applied to factors associated with those occurrences. The next step is predictive analytics. Machine Learning, data mining and advanced statistics can be used to develop prediction models regarding heart failure conditions, given several risk factors. Finally, prescriptive analytics are used to implement a treatment programme for the patient. Thammasudjarit et al. (2018) suggested chatbot as one technology for facilitating a patient’s self-care after discharge from hospital. Moreover, Asante-Korang and Jacobs (2016) emphasized that incorporating various types of data analytics can help to improve healthcare service quality. The authors also highlighted predictive analytics as promising analytic tools in healthcare data analytics, especially for chronic disease. In order to outline and delineate techniques on healthcare big data analytics, Mehta and Pandit (2018)’s research summarized healthcare analytical techniques and highlighted how these techniques can be applied in healthcare. The list of big data analytical techniques includes cluster analysis, data mining, graph analytics, Machine Learning, Natural Language Processing (NLP), neural networks, pattern recognition and spatial analysis. Machine Learning can be used to predict risk of disease. Likewise, NLP has been used to predict future disease in patients. A wide range of healthcare applications has been presented to support each analytical technique, such as cluster analysis and data mining, and these can be used to explore data to identify chronic disease risks. Furthermore, Raghupathi and Raghupathi (2014) introduced the conceptual architecture of big data analytics, along with platforms and tools. Since healthcare data comes from multiple formats, locations and sources, transformation tools and techniques such as middleware and data warehouse are applied

28 

J. CHANCHAICHUJIT ET AL.

to convert raw data and transfer it to a big data platform for further analysis. At this stage several platforms and tools (Hadoop, MapReduce, Pig, Hive and others) for big data analytics can be used to present big data analytics in different applications such as reports, queries or in data mining. Joshi, Meng, and Yan (2018) added that numerous case studies highlighted that descriptive analysis is a useful tool for analysing healthcare data. The authors have also added that data interpretation from visual systems and report systems is also important in the use of big data analytics tools in order to visualize and interpret data to support a physician’s decision-making. 2.5.3  Application of Big Data Analytics in Healthcare One distinct benefit of big data analytics is the ability to enable healthcare organizations to explore new insights and find optimal solutions from complicated variables for a better health service. In terms of data analytics capability and benefits, they found that there are five potential benefits for healthcare organizations: IT infrastructure benefits, operational benefits, organizational benefits, managerial benefits and strategic benefits. Joshi et al. (2018)’s study shows many healthcare organizations using big data analytics to improve clinical workflow efficiency and operational management. Moreover, IT infrastructure benefits from reducing healthcare system redundancies as well as showing improvements in quality, safety and speed of data transfer between hospitals and healthcare providers. However, Joshi et al. (2018)’s study revealed that organizational, managerial and strategic benefits are still limited at this early stage of healthcare big data evolution transformation. Moreover, big data analytics has the capacity to reform the process of decision-making by providing greater clarity and transparency in the approaches adopted in functional operations, and achievement in performance. Researchers have agreed that big data analytics in healthcare is being applied in several areas ranging across clinical, biomedical, public health and administration. These areas include genomics (Maia, Sammut, Jacinta-­ Fernandes, & Chin, 2017), elderly care (Jiang et al., 2016), cardiovascular disease (Rumsfeld, Joynt, & Maddox, 2016), diabetes (Bellazzi, Dagliati, Sacchi, & Segagni, 2015) and heart failure (Thammasudjarit et  al., 2018)

2  INTERNET OF THINGS (IOT) AND BIG DATA ANALYTICS IN HEALTHCARE 

29

2.6   Case Study 2.6.1  Big Data Analytics in Managing the Future Health of Singapore Singapore is one of the key leaders to adopt big data analytics in healthcare. The initial project started in 2013 in major hospitals such as Changi General Hospital, Khoo Teck Puat Hospital, the National Heart Centre Singapore and Singapore General Hospital. The results show that the use of big data brings possible insights for better decision-making. It is also improving hospital operations efficiency and, to a certain extent, effectiveness. The existing research in Singapore healthcare, as well as that of other scholars, reveals that big data analytics will enable Singapore healthcare to progress from treatment care to proactive healthcare, shown by each patient referred to in patient profile analytics (Raghupathi & Raghupathi, 2014). By applying advanced analytics, predictive modelling can indicate which patients have the potential for developing a specific disease, especially a chronic disease such as diabetes. From this knowledge, the patient may benefit by implementing preventative care such as changes in lifestyle, including diet, to prevent those specific diseases. The following section presents a big data analytics case study applied to manage diabetes in Singapore. The case study analysis was obtained from secondary sources by reviewing publicly available materials, ranging from reports and research articles to white papers, case studies and company reports. This chapter intends to bridge the existing knowledge gaps in the literature. It draws upon prior Singaporean big data analytics projects implemented to manage diabetes. It looks at this as a case study as well as providing an in-depth analysis of the efficiencies and challenges in adopting big data analytics. 2.6.2  Diabetes in Singapore—A Nation’s Big Challenge In Singapore, more than 48% of disease is related to chronic diseases. The biggest challenge of these is diabetes. In 2017, the proportion of diabetics in Singapore was 10.99%, making it number 41 in the world diabetes atlas ranking (International Diabetes Federation, 2017). In addition, 46.9% of Singaporeans are undiagnosed diabetics; the estimate is that by 2040 the total percentage will increase to 16.3% (International Diabetes Federation, 2017). Diabetes is a major public health issue and is one of the most

30 

J. CHANCHAICHUJIT ET AL.

expensive to manage in terms of medication expenditure. In addition, a study has forecast that the total economic loss from diabetes will be US$1.9 billion by 2050. Moreover, the average medical expenditure on diabetes patients was twice as high as the cost of prevention. There are two types of diabetes, type I and type II. Type I diabetes is caused by the pancreas not producing enough insulin. Diabetes I is genetic and unpredictable. On the other hand, type II diabetes is related to lifestyle and can be prevented by weight management and exercise. Over 90% of diabetics suffer from type II diabetes (American Diabetes Association, 2017). As type II diabetes is related to lifestyle and diet, some changes in these behaviours are interventions which can prevent an advance in the disease. It has been confirmed that diabetes is one of the most preventable diseases in that it can be cured by a change in behaviour (American Diabetes Association, 2017). However, lifestyle changes are not easy. Constant support and encouragement are required to motivate the patient and assess the progress that is being made. The Singapore Minister of Health, Gan Kim Yong, announced a “war on diabetes” and a new diabetes task force for Singaporeans. He said the key to managing the challenge was to identify the disease as early as possible for those at risk or undiagnosed and to control the progression of the disease (Khalik, 2018). Holmusk (2018) pointed out that the intervention strategies for diabetic management should be initiated from a diabetic population equation as follows: Diabetic Population ( D ) = Early Stage ( ES)

+ LateStage ( LS) + Complex Stage ( CS)



(2.1)

The importance of this equation is that it shows a diabetic population broken down into different stages of disease. Since all diabetic patients have different types of healthcare support, this equation can lead to a bespoke approach of the stages for each patient. From this, the strategy for each patient stage can be developed. In addition, this equation can lead to data analytics production for each stage of a diabetes population segment. The focus of the strategy is on the Early Stage (ES) patient population, where prevention may stop the development of the disease to further harmful stages. Effective intervention prevention strategies and guidelines should be deployed to reduce the conversion from this patient stage to the Late Stage (LS) or Complex (CS) stage. At this point, digital intervention, in the form of big data analytics, is expected to play a major role in reducing

2  INTERNET OF THINGS (IOT) AND BIG DATA ANALYTICS IN HEALTHCARE 

31

disease progression. Smart technologies such as wearable devices and mobile applications allow real-time data monitoring. The patient can be monitored closely by tracking glucose levels, assessing lifestyle and examining wellness data, allowing medical staff to intervene and adjust treatment if needed. Moreover, researchers (L. Wang & Alexander, 2016) have developed predictive models and named risk factors based on healthcare information for the onset of type II diabetes. The aim is to gain insight, in order to predict which factors trigger disease development in the early stage patient. The benefit of big data analytics is that is provides a preventative care approach to diabetes management for patients but also allows healthcare budget management for governments. In 2011, Singapore established the National Electronic Health Record (NEHR) system which is owned by the Ministry of Health and managed by Integrated Health Information Systems to keep track of all patients’ health records from different hospitals and health service providers. This initiative has made Singapore one of the few countries that has digital records of its patients. These records provide an opportunity for building further health analytics specifically to manage chronic diseases. Many researchers emphasized that the use of EHRs can enhance the quality of the delivery of care and that they could be particularly useful in early detection and intervention. In addition, there are several projects being established in Singapore to support the “war on diabetes” including research and seeking solutions for pre-diabetes interventions. A current project at SingHealth’s Diabetes and Metabolism Center at Singapore General Hospital collects data by using smart technology such as wearable devices and mobile applications to track and monitor a patient’s ­behaviour. This information is then sent to the doctor to provide assistance and allow personalized treatment in order to improve patient care. The project’s ultimate aim is to improve diabetes care from risk prediction through to monitoring and treatment with a view to developing a long-term treatment for this disease (Choo, 2018). The collaboration between the University of Oxford and Holmusk, a digital health and data analytics company based in Singapore, is another attempt in the nation’s war on diabetes. This project could prove successful in leveraging skillsets between academia and industry in solving chronic disease, particularly diabetes. Oxford University has provided expertise in investigating the role of abnormal cardiac metabolism that correlates to heart failure and diastolic dysfunction, and Holmusk’s advancement in healthcare analytics techniques and tools will be used to optimize multi-omic data analytics

32 

J. CHANCHAICHUJIT ET AL.

research. The collaboration aims to obtain and construct models based on diabetic heart metabolics in type II patients to predict changes in metabolism. 2.6.3   Challenges to Implementing Big Data in Healthcare Although the majority are agreed on the enormous potential benefits of big data analytics in healthcare, concerns and challenges arise in three main areas, namely privacy and confidentiality, data management knowledge, and technical challenges and organization change management through IT-enabled transformation challenges. Of these challenges, privacy and confidentiality challenges are mentioned as being of most concern in healthcare (Y.  Wang, Kung, Wang, & Cegielski, 2017). Thus, government control and tight procedures around privacy and consent are important to address this concern. Data management knowledge and technical challenges including appropriate IT infrastructure, data integration and knowledge of data analytics techniques are the next most pressing concern. Finally, organizational change management such as how to transform digital healthcare into practice is another challenge. This change management must be prepared for by both management team and staff. Mehta and Pandit (2018) reviewed a recent study in big data analytics in healthcare between 2013 and 2018 and summarized five strategies to overcome big data analytics challenges in healthcare: (1) implement big data governance, (2) develop and share information, (3) employ security measures, (4) train key personnel to use big data analytics and (5) incorporate cloud computing into an organization’s big data analytics. Furthermore, Raghupathi and Raghupathi (2014) call for a simple, convenient and transparent big data analytics system. Thus, applying the above strategies and overcoming the current challenges by developing an action plan to design a simple, convenient and transparent system will lead to the implementation of successful big data analytics.

2.7   Conclusion Today, the Internet of Things and big data analytics appear to be an important aspect in the transformation of the healthcare system. With the advance of IoT device technologies such as sensors and wearable devices’ capability to capture and transfer data, healthcare data has become big data. Healthcare services are expected to gain benefits from this revolution

2  INTERNET OF THINGS (IOT) AND BIG DATA ANALYTICS IN HEALTHCARE 

33

in various areas ranging from clinical care to biomedical information and general administration. A case study of how big data analytics is being applied in Singapore to manage diabetes has been presented. In the future, one can foresee more applications of big data analytics to healthcare services. However, several challenges still exist. Healthcare organizations must prepare policies and strategies to address these challenges. Acknowledgement  The authors would like to express their sincere gratitude to Dr Albert Tan’s students at Curtin University, Singapore, Mr. Do Tran Phuang Uyen for conducting a literature review on IoT for this chapter.

References Abbasi, A., Memon, Z.  A., Memon, J., Syed, T.  Q., & Alshboul, R. (2017). Addressing the future data management challenges in IOT: A proposed framework. International Journal of Advanced Computer Science and Applications, 8(5), 197–207. American Diabetes Association. (2017). Diabetes myths. From American Diabetes Association. Retrieved from http://www.diabetes.org/diabetes-basics/myths/ ?loc=db-slabnav. Asante-Korang, A., & Jacobs, J. P. (2016). Big data and paediatric cardiovascular disease in the era of transparency in healthcare. Cambridge University Press, 26(8), 1597–1602. Atzori, L., Iera, A., & Morabito, G. (2010). The Internet of Things: A survey. Computer Networks, 54(15), 2787–2805. https://doi.org/10.1016/j.comnet. 2010.05.010 Bellazzi, R., Dagliati, A., Sacchi, L., & Segagni, D. (2015). Big data technologies: New opportunities for diabetes management. Journal of Diabetes Science and Technology, 9(5), 1119–1125. Bhartiya, S., Mehrotra, D., & Girdhar, A. (2016). Issues in achieving complete interoperability while sharing electronic health records. Procedia Computer Science, 78(C), 192–198. https://doi.org/10.1016/j.procs.2016.02.033 Bowles, H. K., Potashnik, J. S., Ratcliffe, H. S., Rosenberg, D. M., Shih, D. N.-W., Topaz, D. M., … Naylor, D. M. (2013). Conducting research using the electronic health record across multi–hospital systems: Semantic harmonization implications for administrators. JONA: The Journal of Nursing Administration, 43(6), 355–360. https://doi.org/10.1097/NNA.0b013e3182942c3c Choo, F. (2018, December 9). SingHealth, A∗Star fight diabetes with smart tech. The Straits Times. Dell EMC. (2014). The digital universe driving data growth in healthcare: Challenges opportunities for IT. Retrieved from https://www.delltechnologies. com/en-us/perspectives/categories/transformative-technology/.

34 

J. CHANCHAICHUJIT ET AL.

Elkhodr, M., Shahrestani, S., & Cheung, H. (2016). Wireless enabling technologies for the Internet of things. In Innovative research and applications in next generation high performance computing, IGI Global (pp.  368–396). https:// doi.org/10.4018/978-1-5225-0287-6.ch015. Fang, H., Dan, X., & Shaowu, S. (2013). On the application of the Internet of things in the field of medical and health care. In 2013 IEEE International Conference on Green Computing and Communications and IEEE Internet of Things and IEEE Cyber, Physical and Social Computing (pp. 2053–2058). Fieldman, B., Martin, E. M., & Skotnes, T. (2013). Big data in healthcare—Hype and hope. Dr Bonnie, 360, 122–125 [Press release]. Gubbi, J., Buyya, R., Marusic, S., & Palaniswami, M. (2013). Internet of Things (IoT): A vision, architectural elements, and future directions. Future Generation Computer Systems, 29(7), 1645–1660. https://doi.org/10.1016/j.future. 2013.01.010 Holmusk. (2018). Multi-omic data analytics collaboration between the University of Oxford and Holmusk. Retrieved from https://www.holmusk.com/ research/#publications. International Diabetes Federation. (2017). IDF Diabetes Atlas, 8th Edition. Retrieved from https://diabetesatlas.org/resources/2017-atlas.html. Jiang, P., Winkley, J., Zhao, C., Munnoch, R., Min, G., & Yang, L. T. (2016). An intelligent information forwarder for healthcare big data systems with distributed wearable sensors. IEEE Systems Journal, 10(3), 1147–1159. Joshi, A. P., Meng, H., & Yan, W. (2018). A survey on security and privacy issues of blockchain technology. Mathematical Foundations of Computing, 1(2), 121–147. Khalik, S. (2018). Singapore health declare war on diabetes. Retrieved from http://www.straitstimes.com/singapore/health/moh-declares-waragainst-diabetes. Kulkarni, A., & Sathe, S. (2014). Healthcare applications of the Internet of Things: A review. International Journal of Computer Science and Information Technologies, 5(5), 6229–6232. Laney, D. (2001). 3D data management: Controlling data volume, velocity and variety. META Group, TechRep. Lu, Y., Papagiannidis, S., & Alamanos, E. (2018). Internet of Things: A systematic review of the business literature from the user and organisational perspectives. Technological Forecasting & Social Change, 136, 285–297. https://doi. org/10.1016/j.techfore.2018.01.022 Maheshwari, D., & Janssen, M. (2014). Reconceptualizing measuring, benchmarking for improving interoperability in smart ecosystems: The effect of ubiquitous data and crowdsourcing. Government Information Quarterly, 31(1), S84–S92. https://doi.org/10.1016/j.giq.2014.01.009 Maia, A.-T., Sammut, S.-J., Jacinta-Fernandes, A., & Chin, S.-F. (2017). Big data in cancer genomics. Current Opinion in Systems Biology, 4, 78–84.

2  INTERNET OF THINGS (IOT) AND BIG DATA ANALYTICS IN HEALTHCARE 

35

Manyika, J., Chui, M., Brown, B., Bughin, J., Dobbs, R., Roxburgh, C., & Byers, A.  H. (2011). Big data: The next frontier for innovation, competition, and productivity. Marr, B. (2015). Big data: Using smart big data, analytics and metrics to make better decisions and improve performance/Bernard Marr. Chichester, UK: Wiley. Mathew, J., John, J., & Kumar, S. (2013). New trends in healthcare supply chain. Paper presented at the POMS 24th Annual Conference “Integrating Practice in POM Research and Teaching”, POMS, Denver, USA, 2013. Mehta, N., & Pandit, A. (2018). Concurrence of big data analytics and healthcare: A systematic review. International Journal of Medical Informatics, 114, 57–65. https://doi.org/10.1016/j.ijmedinf.2018.03.013 Menachemi, N., & Collum, T. H. (2011). Benefits and drawbacks of electronic health record systems. Risk Management and Healthcare Policy, 4, 47–55. https://doi.org/10.2147/RMHP.S12985 Mysen, L. K., Penprase, L. B., & Piscotty, L. R. (2016). Patient satisfaction with electronic health record use by primary care nurse practitioners. CIN: Computers, Informatics, Nursing, 34(3), 116–121. https://doi.org/10.1097/ CIN.0000000000000218 Patel, K. K., & Patel, S. M. (2016). Internet of Things-IOT: Definition, characteristics, architecture, enabling technologies, application & future challenges. International Journal of Engineering Science and Computing, 6(5), 6122–6131. Raghupathi, W., & Raghupathi, V. (2014). Big data analytics in healthcare: Promise and potential. Health Information Science and Systems, 2(1), 3. Rouse, W. B., Serban, N., & Moses, J. (2014). Understanding and managing the complexity of healthcare. MIT Press. Rumsfeld, J.  S., Joynt, K.  E., & Maddox, T.  M. (2016). Big data analytics to improve cardiovascular care: Promise and challenges. Nature Reviews Cardiology, 13(6), 350. Saeb, S., Zhang, M., Karr, C. J., Schueller, S. M., Corden, M. E., Kording, K. P., & Mohr, D. C. (2015). Mobile phone sensor correlates of depressive symptom severity in daily-life behavior: An exploratory study. Journal of Medical Internet Research, 17(7), e175. Shah, N. (2004). Oral health care system for elderly in India. Geriatrics & Gerontology International, 4(s1), S162–S164. https://doi.org/10.1111/j. 1447-0594.2004.00187.x Taylor-Adams, S., & Kirwan, B. (1995). Human reliability data requirements. International Journal of Quality & Reliability Management, 12(1), 24–46. https://doi.org/10.1108/02656719510076221 Thammasudjarit, R., Pattanateepon, A., & Pattanaprateep, O. (2018, April–June). Big data analytics in healthcare. Ramathibodi Medical Journal, 41(2), 116–122. Wang, L., & Alexander, C.  A. (2016). Big data analytics as applied to diabetes management. European Journal of Clinical and Biomedical Science, 2(5), 29–38.

36 

J. CHANCHAICHUJIT ET AL.

Wang, Y., Kung, L., Wang, W. Y. C., & Cegielski, C. G. (2017). An integrated big data analytics-enabled transformation model: Application to health care. Information & Management, 55(1). https://doi.org/10.1016/j.im.2017.04.001 Yin, Y., Zeng, Y., Chen, X., & Fan, Y. (2016). The Internet of Things in healthcare: An overview. Journal of Industrial Information Integration, 1, 3–13. https://doi.org/10.1016/j.jii.2016.03.004

CHAPTER 3

Blockchain Technology in Healthcare

Abstract  In this chapter, we introduce blockchain as an effective system to manage patient records and track medical drugs along the pharmaceutical supply chain. Blockchain technology, in short, is a digitalized and decentralized ledger that could be made public or private depending on the user where transactional data is being recorded and unable to be tampered with in the system. Blockchain offers a robust tracking solution for healthcare patients and medical drugs. The application of blockchain technology to healthcare is in its infancy, and there are still some challenges in deploying it in the healthcare sector. Keywords  Blockchain • Decentralized • Peer-to-peer • Distributed ledger • Smart contract • Pharmaceutical supply chain

3.1   Blockchain Technology People have paid attention to blockchain since the emergence of Bitcoin and other digital currencies such as Ethereum, Litecoin and Ripple. Blockchain has become a trend and a phenomenon with the introduction of a peer-to-peer electronic system of payment without depending on financial institutions (Nakamoto, 2008). However, blockchain technology is not limited to cryptocurrencies; its unique technology has potential in a wide range of applications which need a trust-free system. The term block-

© The Author(s) 2019 J. Chanchaichujit et al., Healthcare 4.0, https://doi.org/10.1007/978-981-13-8114-0_3

37

38 

J. CHANCHAICHUJIT ET AL.

chain refers to a chain of blocks which contain transaction information. Risius and Spohrer (2017) described blockchain as a “fully distributed system” which captures and stores the information on transactions by encoding. Stored data expands over time and is assured of consistency and immutability. This system functions as a distributed ledger in recording and verifying transactions. Each block includes transaction data and a timestamp which is linked to the previous block. It is extremely difficult to alter information once chained, as a change in a transaction forces a change in the preceding transaction. A block can contain a large amount of transaction data. The more transactions involved, the greater the increase in the size of the chain and the number of blocks added in the chain through a hashing algorithm, which is the essential function of blockchain management (Tama, Kweka, Park, & Rhee, 2017). Thus, the number of interconnected blocks is infinite. Blockchain can be summarized as embodying the following characteristics: • Decentralization Unlike a centralized transaction management system in which every transaction is conducted by a central institution or an intermediary, blockchain ignores the presence of the third party (Zheng, Xie, Dai, Chen, & Wang, 2017). Instead, the access, validation and audit of transactions are implemented equally by participants in the system, or nodes. This means that a consensus mechanism by networked users (nodes) is required to maintain the coherence of data and prevent systems from failing due to malicious information. • Distributed ledger Blockchain is a database which is shared on a peer-to-peer basis among network nodes. A list of transactions is grouped in blocks, which is linked to the preceding and following block. Each node keeps a copy of the entire ledger and administrates the chain independently. Thus, even if one node fails, there is no disruption of the system as the remaining nodes continue operating. • Cryptographically sealed Blocks are created using a cryptographic hashing algorithm, which is a mathematical function to convert the input data of any length into a fixed-length output. Each block contains a root hash which is created from internal transactions and a hash pointer, including the address and the hash of the content in the previous block, which links together to create a chain. Any modification in a transaction in

3  BLOCKCHAIN TECHNOLOGY IN HEALTHCARE 

39

a block results in a new root hash and invalidates the consecutive block (Dinh et al., 2018). Due to the property of the one-way function, it is difficult to completely alter or delete a well-formed block once it is recorded in the chain, without altering subsequent blocks (Conte de Leon, Stalick, Jillepalli, Haney, & Sheldon, 2017). The hashing function creates a high degree of secured protection of networked assets, as all the nodes in the chain can detect the change when it occurs. • Chronologically-based Blocks are connected together to form a chronological chain which permits users to trace events or transactions back to the initial block. This feature enforces data integrity and transparency in the blockchain. • Consensus Protocols As mentioned previously, a consensus mechanism is vital for the successful operation of a blockchain. There are two typical approaches to achieve mutual consensus: • Proof-of-Work This consensus strategy requires all nodes to “work” to verify a new block. All nodes have to solve the encrypted function to find the hash in the next block, which demands an enormous amount of time and computing energy. This process is referred to as “mining”, and nodes solving the hash function are “miners”. The new block is “solved” if the calculated hash is equal to or lower than a given value (Zheng et al., 2017). • Digital Information is digitalized in blocks, which excludes the manual documenting process. In addition, every participant is entitled to a digital signature, which is a kept record of a transaction by a specific person. A digital signature is based on asymmetric encryption which generates a pair of keys: a public key and a private key. The private key encrypts data in a message sent to another user within the network, while the public key is given to the recipient to verify the message (Zheng et al., 2017). A digital signature grants a user the right to make transactions in a given account. Possessing such features, blockchain provides a robust solution for the dynamic, efficient and transparent operation of transactions (Risius & Spohrer, 2017). It also strengthens cyber security and privacy (Kshetri, 2017).

40 

J. CHANCHAICHUJIT ET AL.

3.1.1  Smart Contract in Blockchain One of the outstanding implementations of blockchain technology is in a programmable computer protocol which is able to verify and execute terms based on pre-determined factors. The protocol is known as the smart contract (Giancaspro, 2017). By using blockchain platform, the smart contract naturally contains a code in every node of the blockchain which defines the conditions of a transaction. Once the transaction meets the pre-determined requirements, a smart contract is automatically activated to execute the transaction. The smart contract may be likened to a vending machine, where performance is automatically achieved if there is the personal involvement of a party, such as someone inserting a coin into the machine (Christidis & Devetsikiotis, 2016). In addition, a smart contract goes beyond traditional contracts as it triggers every transaction and enforces the obligation of involved parties without the help of a court (Savelyev, 2017). Several obvious advantages of smart contracts can be noted. First, as they are run upon a blockchain platform, transactions are validated by the consensus of the network’s users. Contracting parties are only required to decide the content of the contract, and smart contracts take over all of the subsequent work. The process of approval is system-wide and instant, since there are less frequent delays in obtaining the authorization of an intermediary institution. Less human involvement in fulfilling an agreement is likely to improve efficiency (Giancaspro, 2017). In addition, the lack of a central authority ensures the transparent nature of the transaction and preserves the privacy of the personal information of users, which is normally provided to verify commercial transactions. Another benefit also comes from the absence of a “middleman”, which reduces transaction and legal costs. Such costs are incurred through the preparation, implementation and supervision of conventional contracts, as well as contractual enforcements, with the help of a legal entity or service (Giancaspro, 2017). Additionally, the utilization of smart contracts in place of an intermediary offers the promise of cutting down on operational and infrastructure costs for businesses and the government (Giancaspro, 2017). Nevertheless, smart contracts are far too complex to completely replace traditional contracts. Their compatibility with the law of commercial contracts seems to be the most significant issue with smart contracts. Contract law requires several key elements in an enforceable contract to activate the application of a digital smart contract, such as the presence of an offer and acceptance,

3  BLOCKCHAIN TECHNOLOGY IN HEALTHCARE 

41

the intent to be legally binding, consent to the terms, the will of all parties to make it effective, and applicable remedies for breach of contract (Savelyev, 2017). Moreover, many scholars are concerned about how fundamental factors in conventional contracts are represented in code in smart contracts. Examples include the duty of acting in good faith, mutual mistakes in contract terms which require correction, and rules or implied policies that apply to commercial contracts. These issues seem to be straightforward in written contracts but they challenge the smart contract framework in both technical and legal contexts. Being built upon cutting-­ edge technology, there is a lack of a regulatory environment to manage the use of smart contracts, thereby raising concerns over security and cybercrime; this is being discussed in open debates (Christidis & Devetsikiotis, 2016). 3.1.2  Blockchain in the Healthcare Industry S mart Healthcare Management The benefits of a distributed database hold the promise of managing data in the healthcare industry. Due to the complexity in terms of the stakeholders involved in the healthcare network, a massive amount of information must be transferred backwards and forwards. The application of blockchain becomes important in dealing with the need of a variety of involved parties to gain access to the same type of information. Medical treatment processes are prioritized when considering the added value created by blockchain. A pilot project, based on the Ethereum platform, which is called the Gem Health Network, gives different healthcare providers full access to treatment information. Such an ecosystem tackles the issues of concurrent accessibility, thereby limiting the possibilities for medical negligence caused by outdated data. In addition, operational costs incurred from maintaining past diverse databases can be reduced. The system permits caregivers to track patients’ medical information on a frequent basis, as well as to review the historic interaction between medical experts and patients, which subsequently enhances the transparency and quality of the entire medical environment (Mettler, 2016). In addition, blockchain empowers patients by addressing the medical data disruption issue. MedRec is an interesting example of how blockchain facilitates putting data in patients’ hands. MedRec utilizes smart contracts to integrate data based on the creation of patient-provider relationships

42 

J. CHANCHAICHUJIT ET AL.

(PPRs) (Bell, Buchanan, Cameron, & Lo, 2018). Rather than storing data directly, smart contracts contain references to scattered health data in existing storage systems. This means that different fields of data can be gathered into a single record (Azaria, Ekblaw, Vieira, & Lippman, 2016). This provides a continuous and comprehensive log of medical records which is made available for patients and doctors to view, keep track of and audit their own information. It also enables patients to accept or reject relationships with healthcare providers (Bell et al., 2018). By combining data into a consistent database without direct storage, MedRec supports the use of an existing infrastructure (Azaria et al., 2016). Another promising contribution to the healthcare industry relates to using blockchain against counterfeit medical products. Counterfeit medicine has been a significant issue which directly endangers the lives of consumers. Through the ability to trace the original drug data in blockchain, the origin of counterfeit drugs can be detected and the drugs removed before reaching the hospital or pharmacy. The initiator in this field is the Counterfeit Medicines Project, which targets forged or poor-quality products by using a timestamp attached to each medicine. This approach allows identification of the exact time when, and location where, drugs were produced, as well as detecting any ownership transfer from the manufacturer to any party along the destination before reaching the patient. Using this method, blockchain attempts to authenticate products and protect the health and safety of consumers, which should lead to reduced healthcare follow-up costs (Mettler, 2016). 3.1.3  Challenges of Implementing Blockchain Security Security poses a greater challenge for a distributed network like blockchain, compared to other technologies. According to Mendling et  al. (2018), the issues of security involve the confidentiality, availability and integrity of the network. Since data is intimated over the network and all users possess the same copy of every transaction, confidentiality appears to be low. Wide intimation also assists in readability in the network rather than writing availability. Despite the challenges, the system has considerable integrity (Mendling et al., 2018). The most important case regarding blockchain’s security is known as the 51% attack. In the case of cryptocurrencies, this attack can occur if a

3  BLOCKCHAIN TECHNOLOGY IN HEALTHCARE 

43

single node (an individual or group of “miners”) takes control of 51% of the hashing power or of the total number of Bitcoins in the network, which results in the distortion of the entire network (Efanov & Roschin, 2018). Apart from manipulating the network, an attacker can conduct subsequent attacks such as modifying and reversing transactions, launching double-spending attacks, interfering in the validation process of transactions and impeding other miners’ operations. Even though this type of attack is considered theoretical, the vulnerability of blockchain at this early stage cannot be underestimated. Scalability Transactions are stored in blocks and new transactions wait to be added into the chain any time. New transactions require confirmation of the source of transaction before achieving system-wide consensus. Limited block capacity is not compatible with the requirements to simultaneously process a vast number of transactions, and the creation of a new block takes time. Consequently, there exists an issue of transaction delay or rejection when a block becomes full (Joshi, Han, & Wang, 2018). As Zheng et al. (2017) argued, two possible movements can address the scalability issue: storage optimization and redesigning. With the first option, a new cryptocurrency scheme is proposed which allows old transaction data to be removed from the network while maintaining transactions from non-empty addresses. In redesigning blockchain, a conventional block is divided into a key block and a micro block, in which the key block handles leader election and the micro block is responsible for recording transactions. The key block is generated through the mining process, and the leader is in charge of generating micro blocks. Legality As blockchain is a novel technology, there is a shortage of legal approaches to protect network users. Various Bitcoin scandals associated with money laundering have influenced the reliability and true nature of blockchain. In addition, the enforceability of smart contracts becomes difficult when considering the written laws of contract. According to Yeoh (2017), the adoption of blockchain is restricted by an absence of governing mechanisms to provide rules and engender trust beyond the financial market. Once clear policies and procedures are set up, the real benefits of blockchain may be fully realized.

44 

J. CHANCHAICHUJIT ET AL.

3.2   Case Studies 3.2.1   Case Study 1: Using Blockchain Technology for EHR Systems in Developing Countries Even though developing countries such as Vietnam have improved their public healthcare environments, there are some differences in medical procedures which limit the goals of equivalent quality across different public hospitals in the network. In the first scenario, a patient registers with the hospital’s reception, provides their health insurance card to the nurse and takes a booklet which gives the patient a queue number and is used to record medical results. The patient then meets the specialist and may undergo medical tests such as blood tests, X-rays, ultrasound scans, magnetic resonance imaging scans or an endoscopy. Each test is implemented by a different physician in a different department, and it takes time to complete all the tests and obtain the recommended treatment. Following this, the patient comes back to reception and verifies their insurance card. Depending on the type of disease and medical tests performed, the patient will be reimbursed partially or fully by the public health insurance organization. Finally, the patient receives medicines at an appointed pharmacy, based on the treatment and prescription given by their specialists. In the second scenario, with no financial support from the public insurance provider, a patient is obliged to pay an initial amount of the medical fee before making an appointment with the specialists. The amount remaining to be paid depends on the tests required. Following consultation and treatment, the patient may purchase medicines from the hospital pharmacy. Both cases share common drawbacks. Patients often take a whole day to complete medical examinations because of the many steps involved in diverse departments in the hospital and the waiting intervals between tests or after finishing a test to obtain the results. Some serious diseases with complicated symptoms require patients to take a number of tests in order to prescribe accurate treatment. Additionally, where there is inconsistent development between central and local hospitals in terms of facilities and technology, hospitals and clinics at local levels are not capable of providing specialized treatment for patients with chronic diseases. Such patients must then be transferred to central hospitals where they have to retrace every step from the beginning, which is obviously time-consuming and costly.

3  BLOCKCHAIN TECHNOLOGY IN HEALTHCARE 

45

Another existing issue is related to how patient information and medical data is stored and managed. A large number of hospitals and clinics rely on manual methods to record data, including X-ray films, booklets and papers. Whenever a medical device is used to carry out a test, the result obtained is written on paper and handed to the patient. Results which are printed from computers connected to medical devices, such as ultrasound scanning results, are physically stapled together with papers. The patient’s medical examination booklet eventually contains all the diagnostic results as well as the physician’s recommendations for treatment. Additionally, each medical facility provides a separate examination booklet, which can cause confusion for the patient when they have to carry various booklets if they have medical examinations in different hospitals or clinics at different times. Thus, the paper-based recording method can make it difficult for patients to manage their own data, as damaged, forgotten or missing papers are inevitable. Further difficulties may be encountered where clinicians are unable to retrieve medical data from a patient’s previous examination. Patients are required to retake medical examinations, including blood and urine tests, even if they are transferred from district hospitals to municipal or central hospitals for specialized treatment. Because of the decentralization of the medical records system, a patient may have multiple medical records from different healthcare providers and these records may be either inconsistent or duplicated. It is clear that the seamlessness of patients’ medical data across a nationwide system is hard to achieve. As the number of patients keeps growing, paper-based storage methods may be deemed inefficient in controlling such a massive amount of data. Even though such a means of information management seems to be easy to implement, it is a waste of resources where data cannot be shared among medical facilities, or used for research purposes in other healthcare organizations. P  roposed Architecture Given the current challenges facing the healthcare systems in developing countries, the following blockchain model is proposed. Its purpose is to assist all stakeholders in overcoming existing issues with medical data interoperability and to enhance healthcare outcomes. The complete process for streamlining health information records utilizing IoT, blockchain and smart contracts is shown in Fig. 3.1. The new process can be applied to typical medical procedures in midand high-level hospitals in any country with an existing infrastructure. Whenever patients have medical checkups, their personal information is

46 

J. CHANCHAICHUJIT ET AL.

Step 1: Patient Register and setup smart contract

Step 2: Medical diagnostic perfomed

Step 3: Clinical data recorded in EHR sytem

Step 4: Data is encrypted into blockchain

Step 5: Data block is accessible by authorised user

Fig. 3.1  General process for transferring medical information into blockchain

Fig. 3.2  Patient registration and setting up personal profile

registered at reception, at the same time the smart contract is established. Medical diagnostics are then performed. The test results are collected through IoT-based devices and recorded in the EHR system, which is connected to the smart contract in blockchain where data is encrypted. Once it is added into blockchain it is not possible to alter data, but it can be made available and accessible to authorized parties only. Specific steps on how to apply both technologies to address the stated problem are represented in Fig. 3.2.

3  BLOCKCHAIN TECHNOLOGY IN HEALTHCARE 

47

This is the first step in booting the EHR system. In order to maintain a clear, explicit and continuous medical record, each patient is provided an EHR account when they register at reception. This account is unique and similar in every public and private hospital, clinic and health centre at all levels which implement the system, thereby preventing the discontinuity of health information even if the patient has not had a medical examination for a long time. The EHR account contains the patient ID number and their personal information. As most people participate in the social health insurance programme, their patient ID can be the number on their social health insurance card, which is also the number of their EHR account. In case a patient does not have a health insurance number, the account number can be generated randomly from the computer. The patient needs to provide their fundamental personal information, including name, age, contact, health insurer (if any), medical history, chronic diseases, allergies and adverse drug reactions. Following this, the EHR account is added into blockchain awaiting further updates following medical results. A smart contract is activated in the blockchain in which the EHR account is stored. A smart contract comprises coded regulations that govern new registrations and it orients the performance of transactions added into the chain, for example enforcing transactions to obtain consensus from both healthcare providers and patients before giving viewership to a third party. Three types of policies are recommended for inclusion in the smart contract in terms of personal data privacy, health data records and third-party involvement. Personal data privacy policies are related to how healthcare institutions and other stakeholders collect, use, manage and disclose the personal information of patients. They also inform patients about which specific information is gathered, and whether it is kept confidential and protected from misuse or abuse through sharing with partners. Health data record policies regulate which type of medical data is collected and recorded, standards for healthcare documentation and management, and the obligations of healthcare personnel in treating data. These policies must be consistent with common law and current best-practice requirements. They also include standards for health indicators such as blood pressure, cholesterol, protein levels, heart rate and glucose levels in blood, all predetermined for a smart contract to automatically execute. ­ Third-­ party involvement policies concern the rights and obligations of pharmaceutical companies, insurance companies and the government regarding the collection, use and disclosure of patients’ personal information, as well as

48 

J. CHANCHAICHUJIT ET AL.

health results. After registering an EHR account, patients provide their digital signature to activate a smart contract. Any breach of contract can be detected immediately and an alert sent to the patient and caregivers. A smart contract is structured to serve several purposes. First, the smart contract controls new identity registrations. Hence, identity registration is only performed in certified institutions. Second, it contains the PPR contract, which is issued when there is a data interaction derived from both parties. It identifies the data ownership and accesses permissions associated with the health records kept by the physicians. A PPR contract defines the data pointers attached with the hash of the patient data subset that prevent data from being changed at the source. It also allows for patients to have full access to and control of their records, along with the option of sharing their data with other viewers. Finally, it carries a list of references to the PPR contract, which indicates all the previous and current interactions between two nodes in the system, for instance all the patients served by a doctor, or all healthcare providers who have provided services to a patient, or an authorized third party with whom a patient shares data. Blockchain and smart contract maintains all medical logs indefinitely, so a single participant can always gain access to the logs by updating the latest blockchain from the network. Steps 2 and 3 concern medical test performance and the collection of medical data through IoT-based devices. A typical medical checkup comprises a series of tests, such as a height and weight check, blood pressure check, blood sugar test, urine test, cholesterol level check, eye check, ear check, throat check, electrocardiogram test and chest X-ray. In order to cut down the time spent gathering and analysing medical data in typical procedures, a wireless sensor network is adopted. IoT-based devices facilitate data collection and analysis through four phases, collection, transmission, processing and monitoring, as shown in Fig. 3.3. In the data collection phase, a patient’s vital signs such as temperature, heart rate, blood glucose, blood pressure and cholesterol level are collected, along with other measures, using multiple wireless sensors attached to the body during medical examinations. The sensors resemble the wired sensors used in an electrocardiogram test, but they are of a smaller size, thereby minimizing any discomfort to patients. The type and number of sensor nodes used depend on the purpose of the test, the health indicators that are being measured and the corresponding treatment. These sensors capture all the changes in a patient’s health status in real time. They are then transmitted to the server, where the medical specialist can view the

3  BLOCKCHAIN TECHNOLOGY IN HEALTHCARE 

49

Fig. 3.3  Collecting medical data utilizing IoT

results and detect any abnormal signs/danger to the patient. Additionally, sensors are used for inpatients and elderly patients to monitor their daily diet and drug consumption. In this case, sensors are embedded in wearable devices to avoid being removed by patients who want to move around the hospital and may trigger disconnection with the network and errors. Imaging data and text data such as diagnostic transcripts, treatment and doctors’ prescriptions is prepared on the computer by the doctor or nurse after finishing the tests, which are submitted directly to the system. In the transmission phase, gathered data is conveyed to the server through wireless communication. The medical information of an individual must be transferred and stored securely due to its sensitivity. Medical data is transmitted, without human intervention, in a timely manner to the centralized server. In this phase, the server investigates the data through a group of technical layers to grant access permission to different users, including doctors, nurses and patients, at any time. As all components are connected through the internet, the system is vulnerable to cyber-attack from attackers and malicious software. Therefore, authentication is a requirement for ensuring system security. Monitoring is the last phase, in which recorded data is accessible via intelligent monitoring interfaces provided for end users. The interfaces combine a mobile application and web service which are used for analysing

50 

J. CHANCHAICHUJIT ET AL.

and monitoring health data in a real-time fashion. They facilitate different enquiries from different end users such as medical centres, emergency centres or healthcare practitioners, depending on the particular purpose of the service. End users are kept informed about recent medical data as well as their health situation since information is updated frequently and in real time. Because EHR connects with a smart contract in the blockchain, recorded medical data is encrypted and verified by the smart contract before giving authorization to different participants in the network to access and utilize the data. Blockchain contains various provider nodes and patient nodes, so every time the healthcare provider adds a new record, or the patient shares a portion of data with an authorized third party, an automated notification is made in the network that informs the receiving party to verify the proposed information before accepting or rejecting it. In this way, participants are kept updated with new data which is transmitted continuously and participants are engaged in the process of data evaluation. Let us consider a case where new information on a particular patient is added to the system, either a test result which is transferred through IoT devices or a transcript that is added into the computer by a doctor or nurse. The smart contract on the blockchain first matches recent information with the patient’s Ethereum address and locates their corresponding blockchain contract. Following this, the doctor uploads a new PPR indicating the ownership of data to the patient’s address. The provider node then updates the PPR accordingly and links it to the patient’s contract, which is later located in the patient node. From the patient node perspective, after the PPR is updated, the system sends a signal across the network to inform the user about the change. The user, or patient, is able to accept or reject the record, which results in updating their contract correspondingly. If the change is acknowledged, the system automatically obtains the new information and updates the PPR status. At the same time, the provider related to the new PPR is addressed in the network. In addition, the system implementation allows a patient to recover personal data from the provider node or authorize a third party to access their shared data. The patient node sends a query request including the reference to the PPR that is being asked for permission. Next, the system certifies the cryptographic signature of the requester along with the blockchain contract to ensure permission. If the verification finishes and the user address satisfies the conditions, the system runs the query over the pro-

3  BLOCKCHAIN TECHNOLOGY IN HEALTHCARE 

51

vider node’s database and returns appropriate results to the user. The process is similar when a patient selects a part of their data and shares it with a third party. In that circumstance, the patient node links the PPR of the patient with the third-party’s blockchain contract, which automatically verifies access to the system. Once accepted, the third party is informed that they may access the required information, as shown in Fig. 3.4. A smart contract also facilitates notification to the caregiver about the occurrence of any health signals that may endanger the patient’s health. As mentioned, a smart contract contains limits for most of the health indicators. When these indicators are measured, the results are transferred to the smart contract for verification. The smart contract then detects any non-­ conforming factors; for example, if there is a higher level of blood glucose than expected, this may be a symptom of diabetes. A digital notification will be sent to the provider node, which is acknowledged by the provider, and they may then provide appropriate treatment. The user interfaces mentioned in the previous steps are integrated into the blockchain. The provider can manage the upload of data as well as system implementation via the web service. Patients are permitted to view their data, to receive notifications of all data updating and sharing and retrieval options, and to take action to accept or reject them. Data presented on the mobile application is synchronized with that on the web interface, and thus patients are able to obtain access to their medical

Fig. 3.4  Encrypting data into blockchain for future reference

52 

J. CHANCHAICHUJIT ET AL.

records through one of them. The application is designed to be compatible with common operating systems in mobile devices.  enefits of the Proposed Architecture B This proposed system coordinating IoT and blockchain brings a digital solution, which is a comprehensive, credible and accessible medical log for healthcare providers and patients. There are several key features which prove that the prototype is targeting patient agency, autonomy and control. First, medical records are stored in the separate decentralized databases of healthcare providers and patients, instead of in a traditional centralized database. Because blockchain allows each node to have a copy of the entire ledger, all information is transparent. Patients have full control of their data through an intelligent communication tool. A distributed log is helpful to avoid a central attack attempt to alter or leak sensitive personal data. Second, patients are kept informed regarding all changes in their historical and current medical records. This prevents the providers from varying the results of medical tests without notice, on purpose, or returning wrong results to the patient. Finally, the system enables data exchange between patients and other participants in the network once verification is given from the system. Opportunities are therefore limited regarding the misuse or abuse of medical data from unauthorized parties. More importantly, this prototype addresses one of the significant issues in healthcare, which is interoperability. Both state-of-the-art technologies are integrated into the existing data storage infrastructure in the majority of hospitals in developed countries, which physicians and patients have been familiar with. This allows savings on the building of a totally new infrastructure. The system is able to receive medical data from any point of origin, including patient health centres, physicians’ offices and hospital servers. Patients’ personal information as well as their medical history is processed immediately, then maintained infinitely and immutably in the blockchain, so the meaning of the data is preserved. Moreover, blockchain implementation does not require frequent system upgrades. Participants need only download the latest copy of the blockchain ledger onto a mobile application or web interface, where data is presented consistently. Specifically, the system is developed to support data exchange from one point of care to another. In a common scenario, a patient makes a medical appointment at hospital A, but the doctor needs reference to historical data which was created during the previous medical checkup at institution B. This can be a lower-level hospital or a private health centre. Previous

3  BLOCKCHAIN TECHNOLOGY IN HEALTHCARE 

53

data is kept separate by a professional in institution B. This scenario covers an important question of how the physician at hospital A communicates with institution B to receive the needed data right away. Making a phone call or writing an email may not be possible or timely enough. Institution B may not be permitted to breach patient confidentiality by exchanging data. In addition, privacy restrictions control who is granted permission to access to the data. Other questions might be: How reliable is the data and how professional is the institution from which it came? When adopting the proposed technological pattern, institutional medical records may be accessed by other healthcare providers. An individual’s medical history should then be always reliable and available for health specialists to track back at any time. The issue of data privacy is also controlled by cryptographic algorithm implementation. In addition, other healthcare institutions such as health insurance companies, pharmaceutical companies, healthcare providers and governmental organizations benefit from the availability and transparency of medical data, provided the patient permits them to access their data.  otential Risks in Using Blockchain P One potential risk of applying this framework relates to identifying the governing rules of the system. As the right to exploit medical data in the network is unlimited, every party who is authorized can have access and use the data of a particular person. However, data exploitation is not free. Who will pay for each piece of data exchanged or received from the patient? How much will the receiver pay for it? Which payment method will be used? How can the patient value the data they share with a third party? The incorporation of a cryptocurrency into the network can be a solution. It is a digital wealth portion that the miner has to give in for some requested data. However, many scandals have occurred around cryptocurrencies such as Bitcoin. Another disadvantage is the lack of a legal system to control the circulation of these currencies in developing countries. At present, it does not appear possible to control the currencies or the circulation. It is obvious that this architecture brings great benefits to patients, healthcare professionals, hospitals and health centres. However, some health institutions and governmental organizations already manage a large amount of data from different sources. Installing one more data storage system may be a risk to their data and impinge upon their profits.

54 

J. CHANCHAICHUJIT ET AL.

3.2.2   Case Study 2: Using Blockchain in the Pharmaceutical Industry The pharmaceutical supply chain links all the stages involved in the production and distribution of medical products to end consumers. It is common for the supply chain to have numerous stages involving dozens of geographical locations. However, due to lack of transparency along the supply chain, fraudulent events such as counterfeiting of medical products that could seriously affect their quality, safety and efficacy cannot be traced effectively throughout the supply chain. The inability to trace counterfeit or low-quality medical products along the supply chain also hinders the drug recall process, which affects the safety of end consumers.  ata Standardization for Traceability Systems D Lack of uniformity in data standards in the pharmaceutical industry has led to problems in attaining global visibility. The use and sharing of traceability data allows the pharmaceutical industry to develop solutions that can enhance the safety and security of the supply chain. Supply chain traceability in today’s global economy involves complying with each country’s and region’s laws. This could burden an organization with multiple traceability requirements. Traceability data powers traceability systems. This data is generated through a variety of business process executions that are carried out by each organization. When an organization executes a traceability-relevant process, traceability data is generated. This data includes: • Who: The parties involved; • What: The object being traced; • Where: The location of the object; • When: The date and time a specific event occurred; and • Why: What business process took place at the time of the event. End-to-end supply chain traceability is achieved through accessing and combining data from different organizations. A traceability system needs to be able to support many applications and use cases. In addition, it should also be adoptable. As many parties are involved in the pharmaceutical supply chain, one party’s traceability system may not be able to interoperate with another party’s system in the supply chain. Therefore, organizations need to build their systems on a common set of standards to

3  BLOCKCHAIN TECHNOLOGY IN HEALTHCARE 

55

ensure interoperability. The systems used need not be the same. However, the systems need to be able to support standardized data. Interoperability can be provided by the GS1 system of standards. This system provides standards to identify, capture and share information regarding an object throughout its lifecycle. • Identify: Objects and locations are identified by supply chain partners using standardized identifiers. • Capture: An object’s identity and attributes are captured and encoded in a standard manner in a data carrier such as RFID or barcode. This allows the object to be read consistently throughout the supply chain. • Share: Gathered data can then be shared using semantics that are standardized and in a format that is standardized. Traceability data needs to be aligned, exchanged and recorded between partners in the supply chain. There are three categories of data traceability: • Master data: Permanent or constant data that provides descriptive attributes to identify products, locations and parties; and • Transaction data: Data that is created from the transactions. This transaction triggers or confirms a function that is executed. • Visibility event data: Capturing of physical activity details of a product such as what object, when the process took place, where the objects are and why the process took place. This data is often captured by RFID or barcodes. Therefore, to achieve end-to-end visibility along the supply chain, a standardized common language of the three categories of data traceability needs to be developed first in order to achieve interoperability between applications throughout the supply chain. In the pharmaceutical industry, smart contracts can be used to improve the visibility of the medical products along the supply chain along with IoT devices, such as the RFID technology. As mentioned earlier, IoT technology attached to a product, pallet or a package has the ability to trace the product’s movement on a real-time basis from the manufacturer, shipping by sea or air, to the end consumer. In addition, IoT smart sensors can track the temperature and humidity of the products. Any temperature deviations can be captured through the

56 

J. CHANCHAICHUJIT ET AL.

sensors and this data will be recorded via the smart contract in the blockchain. The smart contract rules can then execute notifications and the affected stakeholders can then perform the necessary actions. The IoT device which is attached to each product can communicate via the blockchain at any particular location and time, while the smart contract will be executed to indicate that the products have been delivered. The stakeholders can then trace the product information as the data will be stored in the smart contract. The ownership of the product can be recorded in the smart contract, as well as other data such as the product’s location, time and condition, which is broadcast to the specific participants in the blockchain network. The smart contract functionality of blockchain technology along with IoT devices can provide pharmaceutical stakeholders an effective drug tracing and tracking system whereby a product’s full provenance and its condition are accessible at any point and time. A complete audit trail of each product along the supply chain is thus available from hospital to wholesaler, distributor and finally to the drug manufacturer.  aw Material and Manufacturing Process R Raw materials used for the production of medical products can be a source of contamination due to mistakes in storage and handling of the materials. This could lead to a mix-up of materials or selection errors. Moreover, inappropriate exposure to harsh environmental conditions can cause degradation. Lastly, the use of materials that do not meet the accepted specifications can lead to contamination. During the manufacturing process, contamination could occur due to lack of cleaning in between batches and on equipment and materials used in manufacturing facilities. In addition, usage of “open” manufacturing systems that expose products to room environments can cause contamination. Lastly, failure to implement sound manufacturing processes and/or adhere to best-practice can lead to substandard drugs. Wholesaler In a legitimate supply chain, pharmaceutical companies sell medical products directly to authorized distributors. These authorized distributors ­supply the medical products to pharmacies and hospitals. However, the drugs can flow from authorized distributors to secondary wholesalers. These wholesalers sell drugs to one another. Counterfeit drugs from different countries that are dangerous can enter the supply chain at this point.

3  BLOCKCHAIN TECHNOLOGY IN HEALTHCARE 

57

These counterfeit drugs are then distributed from wholesalers to hospitals and pharmacies and finally to end consumers. The pharmaceutical industry is new to blockchain technology. The following steps provided by an international consultant for the healthcare industry can aid the pharmaceutical industry in making decisions on the implementation of blockchain technology. The first step involves initiating the blockchain project. A pharmaceutical company can initiate the project if certain preconditions are met. These preconditions include the following: (1) transactions are generated by multiple parties that change the information shared in the repository, (2) trust is needed on the validity of the transactions by parties, (3) intermediaries are not trusted and (4) there is a need for enhanced security to ensure the system’s integrity. The next step involves designing applications that are relevant to facilitate the process of tracing the medical products along the supply chain by integrating the blockchain system with tracking enablers, such as RFID chips. This would enable real-time visibility of the medical products. The third step involves strengthening the system through smart contracts. The smart contract can trigger events when certain conditions are not met. This would ensure that the medical products that are exposed to harsh conditions can be tracked and recalled immediately before reaching the end consumer. Finally, the implementation step involves the selection of the type of blockchain and its platform. This could be a permissioned or permissionless blockchain. Blockchain technology can be applied in the following supply chain areas to improve the traceability of medical products, along with the use of RFID technology, thus enabling the identification and elimination of products that are counterfeit and products that are not within the predefined conditions: • Raw material ingredients; • Manufacturing processes; and • Distribution channels A permissioned networked supply chain monitoring system is first created. All stakeholders having an interest in the authenticity, purity and integrity of the pharmaceutical product would have to be identified. Later, appropriate roles that define access rights to the system by each of them would have to be granted. The fundamental aim is to provide for a tracing facility that keeps counterfeit products at bay and eliminates

58 

J. CHANCHAICHUJIT ET AL.

products that do not meet predefined conditions. At the same time, it must be ensured that the digital ledger underpinning the blockchain, that is accessible in the network, is maintained in a secure and protected form, by only allowing permitted access levels to those who have been authorized. As there are close parallels in how both blockchain and RFID technologies might be applied to the various supply chain areas identified above, a detailed explanation on a likely implementation process that might be adopted at Point C is provided. It is further assumed that a permissioned blockchain platform that uses smart contracts is in place. Each new product that has to be traced would be subjected to the following sequence of RFID and blockchain steps. . Register the RFID chip in the supply chain. 1 2. Manufacturer takes possession of chip, embeds product details on the chip and attaches the chip to the product. This data is stored on the blockchain. 3. Products are transported to the port, for example, and shipped by a shipping company. 4. RFIDs inside the shipping company container update the RFID chip relating to the storage conditions, such as temperature and humidity, at periodic intervals until the product reaches the destination port. In addition, shipping details might be added if desired. This data is then updated on the blockchain. 5. From the destination port, the goods are delivered to the wholesaler. RFIDs at the wholesaler’s storage facilities would continue registering the data pertaining to the storage conditions, temperature, humidity, and so on, on the product’s embedded chip. In addition, the wholesaler’s details might be added if desired. This data is then updated on the blockchain. 6. The wholesaler delivers the product to the end consumer, such as hospitals and pharmaceutical outlets. 7. Verification as to the product’s authenticity might be carried out in Stage 6, or any other prior stage, by any of the stakeholders by retrieving information from the blockchain and checking for any breaches indicated by missing or failed links. An unbroken chain is symptomatic of a product that has maintained its uncompromised integrity throughout the entire process.

3  BLOCKCHAIN TECHNOLOGY IN HEALTHCARE 

59

Figure 3.5 shows a sample of transactions carried out for a product at different stages of the pharmaceutical supply chain. These transactions are recorded in the blockchain platform with the following format: This data recorded in the blockchain platform makes the supply chain fully traceable. The product has a unique identification number. At every production stage or site change, the product’s RFID chip is scanned and the data is recorded in the blockchain platform. The transaction data is stored in chronological order. Therefore, it is possible to obtain a fully traceable history of a product that has extensive safety features built into it to prevent fraudulent tampering, from the point of exit from the manufacturer’s production facilities to that of final consumption by an end consumer. In this example, the medical product was manufactured in Cambodia and a unique country ISO (KH-12) code was deposited in the blockchain. The product arrived at the shipping company, in the province of Koh Kong (KH-9) and was then shipped to the wholesaler facility in Hamburg (DE-HH). The wholesalers then distributed the drugs to hospitals and pharmacies in Berlin (DE-BE) and Saxony (DE-SN). This efficient and secured system that is in place for tracing medical products can prevent counterfeit drugs from entering the supply chain. Each item would have a registered RFID chip with a unique ID attached to it and the data would be constantly recorded in the blockchain platform when the chip is scanned. Therefore, all the stages from the drug manufacturing process to the distribution process until the drug reaches the end

Fig. 3.5  Data recorded in each block stored in the blockchain platform

60 

J. CHANCHAICHUJIT ET AL.

consumer would be recorded in the blockchain platform. A counterfeit medical product which has the exact same packaging as the authentic product and with a RFID chip attached to it can enter the supply chain from different countries through the wholesaler. However, this chip’s historical data would not be recorded in the blockchain platform as the RFID chip attached to the counterfeit drug would not be registered in the particular blockchain system. When the unregistered chip is scanned and recorded in the blockchain as it enters the supply chain, the participants in the blockchain network would be able to determine that the product is not legitimate, as the previous transaction data for that item was not recorded in the blockchain system. This would prevent the counterfeit drug from entering the market and it could be recalled immediately. In addition, at each stage of the product’s transportation from the producer’s facilities to the end user, the environmental conditions that ensure the integrity of the medical product, such as temperature and humidity, are also captured by the RFID tag, and this data can be tracked via the smart contract in the blockchain. The smart contract rules will execute notifications if there are any deviations in the temperature or humidity, which will thus allow the stakeholders in charge to perform the necessary actions. This allows drugs that do not meet predefined conditions to be recalled on time. Figure 3.6 illustrates a simple flow of medical products along the supply chain in a blockchain ecosystem, and shows how each transaction is recorded in blocks in the blockchain platform. Stage 1

Stage 2

Stage 3

Stage 4

Stage 5

Manufacturer attaches RFID to product

Manufacturer ships products to distributor

Distributor scans RFID

Distributor ships products to pharmacy

Pharmacist’s scans RFID

Smart Contract

Blockchain Platform

Fig. 3.6  Flow of medical products and data captured along the supply chain

3  BLOCKCHAIN TECHNOLOGY IN HEALTHCARE 

61

3.3   Conclusion The application of blockchain technology to healthcare is in its infancy, and there are important challenges to face and big decisions to make going forward. While data privacy remains a challenge for patients, blockchain technology still has potential benefits to patients if embraced carefully. If the challenges of interoperability continue to be overcome, dependable privacy established, good protocols developed and consensus achieved among the healthcare community, then blockchain technology is going to revolutionize the healthcare industry. Acknowledgement  The authors would like to express their sincere gratitude to Dr Albert Tan’s students at Curtin University, Singapore, who have contributed their work to this chapter: Ms. Vithya Laxme Samiappan and Ms. Vu My Linh for their research on developing a concept framework for blockchain technology and case studies in drug traceability and Electronic Health Records in Vietnam.

References Azaria, A., Ekblaw, A., Vieira, T., & Lippman, A. (2016). MedRec: Using blockchain for medical data access and permission management. Paper presented at the Open and Big Data (OBD), International Conference. Bell, L., Buchanan, W. J., Cameron, J., & Lo, O. (2018). Applications of blockchain within healthcare. Blockchain in Healthcare Today, 1. Retrieved from https://blockchainhealthcaretoday.com/index.php/journal/article/view/8 Christidis, K., & Devetsikiotis, M. (2016). Blockchains and smart contracts for the Internet of Things. IEEE Access, 4, 2292–2303. https://doi.org/10.1109/ ACCESS.2016.2566339 Conte de Leon, D., Stalick, A. Q., Jillepalli, A. A., Haney, M. A., & Sheldon, F. T. (2017). Blockchain: Properties and misconceptions. Asia Pacific Journal of Innovation and Entrepreneurship, 11(3), 286–300. https://doi.org/10.1108/ APJIE-12-2017-034 Dinh, T. T. A., Liu, R., Zhang, M., Chen, G., Ooi, B. C., & Wang, J. (2018). Untangling blockchain: A data processing view of blockchain systems. IEEE Transactions on Knowledge and Data Engineering, 30(7), 1366–1385. https:// doi.org/10.1109/TKDE.2017.2781227 Efanov, D., & Roschin, P. (2018). The all-pervasiveness of the blockchain technology. Procedia Computer Science, 123, 116–121. https://doi.org/10.1016/j. procs.2018.01.019 Giancaspro, M. (2017). Is a ‘smart contract’ really a smart idea? Insights from a legal perspective. Computer Law & Security Review: The International Journal

62 

J. CHANCHAICHUJIT ET AL.

of Technology Law and Practice, 33(6), 825–835. https://doi.org/10.1016/j. clsr.2017.05.007 Joshi, A. P., Han, M., & Wang, Y. (2018). A survey on security and privacy issues of blockchain technology. Mathematical Foundations of Computing, 1(2), 121–147. https://doi.org/10.3934/mfc.2018007 Kshetri, N. (2017). Blockchain’s roles in strengthening cybersecurity and protecting privacy. Telecommunications Policy, 41(10), 1027–1038. https://doi. org/10.1016/j.telpol.2017.09.003 Mendling, J., Weber, I., Aalst, W., Brocke, J., Cabanillas, C., Daniel, F., … Zhu, L. (2018). Blockchains for business process management—Challenges and opportunities. ACM Transactions on Management Information Systems (TMIS), 9(1), 1–16. https://doi.org/10.1145/3183367 Mettler, M. (2016). Blockchain technology in healthcare: The revolution starts here. Paper presented at the e-Health Networking, Applications and Services (Healthcom), 2016 IEEE 18th International Conference. Nakamoto, S. (2008). Bitcoin: A peer-to-peer electronic cash system. Retrieved from http://bitcoin.org/bitcoin.pdf Risius, M., & Spohrer, K. (2017). A blockchain research framework: What we (don’t) know, where we go from here, and how we will get there. Business and Information Systems Engineering, 59(6), 385–409. https://doi.org/10.1007/ s12599-017-0506-0 Savelyev, A. (2017). Contract Law 2.0: ‘Smart’ contracts as the beginning of the end of classic contract law. Information & Communications Technology Law, 26(2), 116–134. https://doi.org/10.1080/13600834.2017.1301036 Tama, B. A., Kweka, B. J., Park, Y., & Rhee, K.-H. (2017). A critical review of blockchain and its current applications. Paper presented at the Electrical Engineering and Computer Science (ICECOS), 2017 International Conference. Yeoh, P. (2017). Regulatory issues in blockchain technology. Journal of Financial Regulation and Compliance, 25(2), 196–208. https://doi.org/10.1108/ JFRC-08-2016-0068 Zheng, Z., Xie, S., Dai, H., Chen, X., & Wang, H. (2017). An overview of blockchain technology: Architecture, consensus, and future trends. Paper presented at the Big Data (BigData Congress), 2017 IEEE International Congress.

CHAPTER 4

Application of Artificial Intelligence in Healthcare

Abstract  This chapter presents the role and significance of Artificial Intelligence, commonly known as AI, in the control and management of Tuberculosis (TB). The complexity of the disease and problems in TB diagnosis are introduced. Following this, initiatives and opportunities for using AI in TB diagnosis in Thailand are shown as a case study. The chapter concludes by discussing the current limitations of AI improvement, alternative models and key success factors in the implementation of AI in TB. Keywords  Artificial Intelligence (AI) • Tuberculosis • Neuro Learning • Deep Learning • Innovation management • Chest X-ray

4.1   Introduction Artificial Intelligence, commonly known as AI, is one area of computer science that focuses on the creation of intelligent machines that work and react in a very similar way to humans. AI systems carry out processes usually associated with human intelligence and human behaviour. These include adaptations, learning processes, planning and even problem solving. AI systems incorporate an “intelligent agent” which is able to access objects, categories and properties and even “understand” or use the concept of common sense, the idea of the sound judgement that the majority © The Author(s) 2019 J. Chanchaichujit et al., Healthcare 4.0, https://doi.org/10.1007/978-981-13-8114-0_4

63

64 

J. CHANCHAICHUJIT ET AL.

of people might use. This intelligence is known as an “agent”. AI a­ lgorithms are designed to make artificial machines capture, store and analyse massive amounts of data, for the benefit of human beings. Furthermore, Gheorghe discussed the fact that as business industries are growing, the “agent” will be able to analyse massive amounts of information, thus increasing the demand for and development of AI. The agent can adapt to the environment and use algorithms and regression to analyse the inputs, making an acceptable output which is suitable for the required environment. Early adoption of AI attempted to reproduce human capabilities such as memory, in an attempt to create a mechanical “brain”. With the development of technology, AI will attempt to further understand how the human mind works. At present it is beginning to mimic the human decision-­making process, and in the future it is hoped that it will carry out tasks in an even more “human” way. According to Gesing, Peterson, and Michelsen (2018), AI was first discovered in 1956 when it had very limited processing power due to limited information and processes. However, as technology has advanced, AI has improved significantly in information analysis, which helps it to recognize data and better mimic the human process. The AI platform is built based on business intelligence and industry-specific consulting intelligence. This actionable intelligence allows for a more real-time decision-making process based on a situation that is as human as possible. The fundamental purpose of AI is to focus on digitization that will revolutionize the area of decision-making (Syam & Sharma, 2018). As AI grows, there are three types of learning that will impact upon its development: Machine Learning, Neuro Learning and Deep Learning. 4.1.1  Machine Learning According to Syam and Sharma (2018), Machine Learning is a core part of AI that deals with the simulation of intelligent behaviour in computers. The ability to mimic human behaviour has led to AI “learning”, or accumulating information, and solving problems. By finding and learning a pattern from multiple inputs, the machine thus learns further and goes on to classify and use regression to determine suitable outputs. Machine Learning focuses on using the data to establish correlations and make predictions. The data analysed comes from a substantial pool of data and requires high-level processing power, which will allow learning from the model, based on the parameters set.

4  APPLICATION OF ARTIFICIAL INTELLIGENCE IN HEALTHCARE 

65

Although Machine Learning is sophisticated in that it allows learning and has the facility to make predictions, it is one of the most basic practices of AI. By using algorithms, Machine Learning analyses the data and creates a predictive model. When any new relevant data is incorporated into the current predictive model, Machine Learning combines the information and creates predictions based on regression. With a significant increase in the digital information available, and the development of machines that can do many things far more efficiently than a human, improvements in the analysis of information are desirable. Ultimately, the primary aim of Machine Learning is to allow a machine to understand information, without any human intervention. Machine Learning can be broadly categorized into supervised learning and unsupervised learning. Furthermore, supervised learning is classified as using either classification or regression methods, while unsupervised learning uses the clustering method. Syam and Sharma (2018) define supervised learning as a statistical model that will predict and estimate outputs based on one or more inputs. On the other hand, unsupervised learning is defined as output which is not supervised, which results in a set of unexplained variables due to unstructured and unlabelled data. Linear regression is commonly used within regression models in determining the model and actual values, which reduces the training data (Barga, Fontama, Tok, & Cabrera-Cordon, 2015). Neuro Learning 4.1.2   Neuro Learning is the next step in AI after Machine Learning. According to Schmidhuber (2015), Neuro Learning is developed due to the possibility of both supervised and non-supervised learning creating errors due to unseen sequences of input events. Neuro Learning uses the basic approach of Machine Learning, combined with artificial neural networks to create algorithms that work in a similar way to a human brain. As this neural system uses multiple layers, it connects and directs data, and then links it all together in the manner of a human brain. With active networks, the nodes in the neuro system receive different data. Neuro Learning is tasked with connecting this data and creating a single value to assist in determining the outcomes. Similar to the human nervous system, which allows continuous learning, Neuro Learning can expand and extract information more efficiently the more it learns. Neuro Learning is able to sense or notice if information is wrong. It can process and mimic how humans communicate, and communicate back using similar natural language.

66 

J. CHANCHAICHUJIT ET AL.

According to Syam and Sharma (2018), neural networks solve problems differently compared to traditional computer algorithms. This is because neural networks are better suited to analysing disordered and complicated data, in a more efficient manner than human methods and traditional computer systems, which may not be able to detect and solve such issues. 4.1.3  Deep Learning Deep Learning is the next step in the development of AI after Neuro Learning. Deep Learning analyses how Machine Learning captures data, and predicts the outcomes. At the same time, with Neuro Learning in place, decision-making can be more accurate. Deep Learning breaks down all the layers developed within Neuro Learning and leads the Machine Learning process towards better outcomes. In essence, Deep Learning involves utilizing systems with a great deal of data. Based on the information, the system creates a decision-making function. It begins by going through the Machine Learning process, transforming binary data into multi-layer processes within the neural networks. All data that passes through is classified into predicted outcomes. As the neural network grows, the development of deep neural networks (DNNs) becomes the logical network. This makes it easy for the computer, or in this case an “agent”, to create a high degree of prediction, representative of human prediction. According to Lemley, Bazrafkan, and Corcoran (2017), generic DNNs will enhance the accuracy of outputs.

4.2   Case Study 4.2.1  Case Study 1—The Role and Significance of Artificial Intelligence in the Control and Management of Tuberculosis Tuberculosis, Worldwide Health Issues Tuberculosis (TB) is still a major urgent issue (WHO, 2014) and a life-­ threatening infectious disease. The WHO reported in 2016 that around 10 millions people fell ill with the disease resulting in around 1.7 million deaths, with 95% of the deaths occurring in low- and middle-income countries. This makes TB among the top 10 causes of death in the world. India, Indonesia, China, the Philippines, Pakistan, Nigeria and South Africa account for around 64% of new TB cases, and Thailand is among the 30 highest TB-burdened countries (WHO, 2017) (Fig. 4.1).

4  APPLICATION OF ARTIFICIAL INTELLIGENCE IN HEALTHCARE 

TB deaths

10

All TB cases 5

Notifications of new and relapse cases HIV-positive TB cases

0 2000

2008

2016

Millions per year

Millions per year

TB incidence

67

1.5 1.0 0.5 0

TB deaths among HIV-negative people

TB deaths among HIV-positive people 2000

2008

2016

Shaded areas represent uncertainty intervals.

Fig. 4.1  Global trends in the estimated number of incident TB cases and the number of TB deaths between 2000 and 2016 (WHO, 2017)

TB diseases and infection are caused by M. tuberculosis bacteria (MTB). These bacteria mainly affect the lungs and results in pulmonary TB, which accounts for over 85% of TB cases. However, MTB can also affect other parts of the body, and this is referred to as extra-pulmonary TB (Fanning, 1999; Singh, Kant, Gaur, Tripathi, & Pandey, 2018). TB is treatable with a standard six-month course of four antimicrobial drugs (i.e. Isoniazid, Rifampicin, Pyrazinamide and Ethambutol) under proper supervision and management, following WHO guidelines (WHO, 2010). Early diagnosis is crucial and significantly increases the chances of survival in people infected with MTB. Between 2000 and 2016, the WHO estimated that around 53 million lives were saved through TB diagnosis and treatment (WHO, 2017). TB is an airborne disease which can be spread through the air when a person with MTB in the lung or throat coughs, sneezes or even speaks. People nearby inhale the bacteria and may become infected, transferring the infection from one person to another (CDC, 2016). As MTB is easily transmissible, an outbreak of the disease is highly likely to infect large populations. In the past decade, the WHO approximated that one new case was occurring every second (WHO, 2006b). This prompted action to end TB, and its prevalence was recognized by the worldwide community. A strategy and campaign to end the TB epidemic by 2030 was embraced and included in the WHO’s Sustainable Development Goals (UN). In 2014, the World Health Assembly adopted the WHO End TB Strategy as a blueprint for countries to end TB epidemics and set a goal to drive down

68 

J. CHANCHAICHUJIT ET AL.

deaths by 90% and reduce new cases by 80% between 2015 and 2030. The WHO itself has taken the further step of setting a higher target to reduce deaths by 95% and the incidence of TB by 90% by 2035 (Uplekar et al., 2015)  omplexity of the Disease, and Problems in Tuberculosis Diagnosis C Modern molecular techniques attest to TB being one of the oldest diseases in the world; evidence of the M. tuberculosis complex has been found in ancient skeletal remains (Zink, Grabner, & Nerlich, 2005). Factors contributing to the complexity of managing TB include the easy transmission and spread of MTB, slow growth of the bacteria colony and difficulty collecting sputum samples of sufficient strength for diagnosis. MTB can be easily passed around through inhalation, and can infect large populations in a relatively short amount of time. In many cases, standard diagnoses to confirm the presence of the bacteria have failed because of the difficulty in obtaining potent enough sputum samples, along with the slow growth of the MTB culture (WHO, 2006a). Consequently, alternative approaches were needed to prevent the spread of the bacteria. The Mantoux tuberculin skin test (TST) was one common standard TB test that was 30% effective in providing evidence of infection (Jereb, Etkind, Joglar, Moore, & Taylor, 2003) and was proven to be useful in early diagnosis. Infection with Tuberculosis bacilli (TB) manifests as different stages of the diseases: latent TB, active TB and multi-drug-resistant TB (MDR-TB). The variance in stages of TB infection adds to the challenge and complexity of detection as well as the difficulty of managing the disease with promptness and precision. Once infected with MTB, only 5–10% of the population will develop a clinical manifestation of active TB within two years (Lin & Flynn, 2010). The active TB condition is infectious to others and makes the carrier sick. It can develop quickly in the first few weeks after infection (Fanning, 1999; Singh et al., 2018; WHO, 2010). The majority of people infected with TB will have the bacteria remain in the inactive state, known as latent TB. The condition is not contagious; however, it may turn into active TB when the immune system is weakened and can no longer resist the bacteria, as shown in Fig. 4.2 Consequently, those who have conditions that weaken the immune system or those who have “high risk factors” such as HIV/AIDS, diabetes, severe kidney disease and certain cancers are among the people recommended for treatment for TB if infected (WHO, 2018a; MayoClinic). The WHO reported in 2015 that there are

4  APPLICATION OF ARTIFICIAL INTELLIGENCE IN HEALTHCARE 

69

Fig. 4.2  The development of Tuberculosis from the point of introduction to active TB (WHO, 2018a)

currently 2–3 billion people living with latent TB, and therefore it is crucial to prevent the development of active TB with preventative treatments (CDC, 2016; WHO, 2018a).  urrent Diagnosis of Tuberculosis and Intervention C Effective management and efforts to reduce TB infection require robust surveillance, with diagnosis playing a pivotal role (Uplekar et al., 2015). As recommended by the WHO and the Centers for Disease Control and Prevention, USA (CDC), people at high risk, once infected, should be monitored and tested as to the progression of TB (WHO, 2017; CDC, 2016). Management and intervention in different geographical areas should be customized to fit the situation. For instance, it is more appropriate to identify active TB in areas with outbreaks or high-burden countries, while identifying infections of MTB is more strategically important in other areas for latent tuberculosis bacteria infection (LTBI) treatment. Therefore, the schema of diagnosis plans should be flexible and also align with the strategy and implementation of TB management. Overall, it is the prevention of LTBI that is the goal, with steps being taken to stop the development of active TB and to control TB in general as a strategy. If LTBI cases can be controlled and managed, the spread of TB can be significantly reduced. In order to do this, testing, follow-up medical evaluation and treatment should be planned and integrated together for maximum benefit and outcomes (CDC, 2016). In TB infection diagnosis, pathology is commonly used to confirm bacterial infection. However, challenges in accu-

70 

J. CHANCHAICHUJIT ET AL.

rately identifying the bacilli still remain due to their very small size (< 1 μm in diameter). Visualization of TB bacilli under a microscope is time-consuming and may require trained personnel, along with other resources. Many test methods have been developed to address these issues, such as TST, Interferon-gamma release assays (IGRAs), PCR and RNA scopes. Despite this, none of these tests have been proven to be reliable or accepted widely (Xiong et al., 2018; Little, 2004), and researchers are still working together to develop new tests that will provide efficiency as well as costeffectiveness. Due to its convenience and relatively low cost, the CDC has adopted TST and IGRAs in their guidelines. Figure 4.3 shows a decision

Fig. 4.3  Diagnosis and ruling of TB (CDC, 2016)

4  APPLICATION OF ARTIFICIAL INTELLIGENCE IN HEALTHCARE 

71

tree for LTBI surveillance in the USA (where the spread of TB is relatively rare (CDC, 2016; Kanabus, 2018)), in order to distinguish LTBI and active TB disease for appropriate care management. While the USA is among low-burden countries for TB, the disease appears to be brought in by immigrants and foreign visitors (Cohen & Murray, 2005). This means that the monitoring and surveillance of the incidence of TB are of equal importance outside high-burden countries. Although TST and IGRAs may be cost-effective (Tasillo et  al., 2017; Nienhaus, Schablon, Costa, & Diel, 2011) and used widely in the USA, many other countries’ pay-per-test ability is lower and they would benefit from more robust and cheaper test methods to allow regular monitoring. Furthermore, other issues have not been addressed by either method, such as the potential to develop boosters and the reduction of waiting times. In addition, there is a lack of equipment which might assist researchers and scientists to develop newer tests to address these gaps in low- and middle-income countries. Many recommended standard TB diagnostic techniques commonly work around three principles: identification of symptoms, confirmation of pathogens, and clinical outcomes, such as lesions on or anomalies in the lungs (CDC, 2016; Lindsay McKenna & Lessem, 2014; Vikas K. Saket, 2017). In previous years, numerous tests were developed to improve test accuracy and speed to confirm the presence of TB bacteria. However, every test has its limitations in sensitivity, specificity or practicality of implementation (see Table 4.1) (Vikas K. Saket, 2017; Tsara, Serasli, & Christaki, 2009). Thus, WHO screening guidelines recommend compiling the results from multiple inputs and different techniques. With adequate information, care providers should confidently be able to diagnose and distinguish different types of infection (e.g. multiple drug-resistant TB, latent TB infection, TB diseases) (WHO, 2015b) and manage the cases correctly. Other than TST and IGRA, common test methods for TB usually include physical examinations (e.g. chest X-ray), smear microscopy (dye microscopy) and genetic tests. Though these tests also provide important additional data, the guidelines still recommend administration of the “gold standard” to grow the culture and the reviewing of medical records in order to distinguish and confirm the infection. These inclusion guidelines were implemented due to the complexity of TB infection that exists in many stages. Consequently, having the confirmation of bacilli in the body alone is not enough to determine the appropriate intervention and management for many cases. Therefore, an ideal and complete diagnosis

72 

J. CHANCHAICHUJIT ET AL.

Table 4.1  Main shortcomings of commonly used TB diagnostic tests (Vikas K. Saket, 2017; Dande & Samant, 2018) Test

Methodology

Chest X-ray

X-ray of chest recorded to detect inflammation in the lungs

Interpretations

Shortcomings

Abnormal Cannot exclude shadow visible on extra pulmonary X-ray TB, cannot differentiate stages of TB TB skin test Injecting small The bigger the May give false amount of raised area of the results if person tuberculin into the swelling, the was infected by lower arm and more chance of some other observing the being infected by bacteria. Cannot swelling TB differentiate between latent TB and active TB TB Mix blood sample Blood sample Interferon-­ with special must be instantly gamma substances to examined, release assays identify interferon laboratory (IGRAs) gamma cytokine. required, test is for detecting latent TB Morphological In cases of HIV Sputum smear A series of special characteristics and TB test stains are applied co-infection, TB to a thin smear of identification to detect presence of cannot be patient’s sputum and it is examined M.tuberculosis detected due to under microscope low levels of TB for signs of TB bacteria bacteria Fluorescent Illumination of Morphological Expensive and microscopy patient’s sputum characteristics time-consuming smear with quartz identification to high-pressure detect presence of mercury lamp M.tuberculosis Culturing bacteria to test

Culture the bacteria from a biological sample of patient on M.tuberculosis selective media

Detection of presence of bacteria by observing colony characteristics

Time-consuming

References Health

CDC (2018) and Healthwise (2017)

WHO (2011b)

WebMD, Center (2016)

Dowsland and Thompson (2000) and WHO (2011a) WHO (2008)

(continued)

4  APPLICATION OF ARTIFICIAL INTELLIGENCE IN HEALTHCARE 

73

Table 4.1 (continued) Test

Methodology

Polymerase The assay targets chain reaction the KatG gene having a unique sequence in TB bacterium GeneXpert Identification of test DNA present in TB bacteria Nucleic acid Amplification of amplification nucleic acids from test biological specimens of suspected patient

Interpretations

Shortcomings

References

Presence of the M.tuberculosis complex will give a positive test result If DNA found, patient is TB positive If nucleic acids found, patient is TB positive

Expensive

Research

Expensive

Scott et al. (2011)

Lower sensitivity for respiratory tract specimens

Hans and Marwaha (2014)

of TB may require care providers to collect and review data on medical history, carry out physical examinations (e.g. chest X-ray), smear microscopy and TB culture, and gather genetic test information. However, waiting time and limited resources are critical factors for this epidemic, and they need to be balanced if complete diagnosis may cause longer waiting times or other complications in its management (Lindsay McKenna & Lessem, 2014; WHO, 2015b; Dande & Samant, 2018). More data on patients requires comprehensive analysis, and experienced personnel are needed, which may cause a bottleneck in the delivery of results in a timely manner. With the rise of numerous technologies, AI, big data science, communication and blockchain may provide efficient solutions in solving the many issues in TB diagnosis. Thus, a complete digital transformation of TB diagnosis and management could offer the ultimate potential in bringing an end to TB globally in the near future, as proposed by the WHO. I nitiative to Use Technologies in Solving Unmet Needs As mobile devices become more affordable, people in low- and middle-­ income countries are gaining increased access to mobile devices and digital platforms. Electronic healthcare through mobile devices was developed to leverage the connectivity that mobile devices provide and it has become a rapid and inexpensive solution in many low-income countries (Bastawrous & Armstrong, 2013). Many digital platforms have been deployed and enabled collection of health data, which is extremely important in devel-

74 

J. CHANCHAICHUJIT ET AL.

oping practical and useful policies. Digital technologies were also adopted and have helped the management of many public healthcare issues, including TB, to become more effective and efficient. The WHO is one of the first organizations to recognize the potential of digital technologies, and to implement many electronic and mobile health products in TB care and prevention (WHO, 2015a; Falzon et  al., 2016). In many TB control efforts, the WHO has implemented digital technologies and strategized to deploy digital technologies to support TB control in four categories: patient care, surveillance, programme management and e-learning (Doshi et al., 2017; WHO, 2015a). The goals for digital transformation of TB control can be classified into three stages, as shown in Fig. 4.4: diagnostic connectivity, data repository

Pillars and components

Electronic tools to help stock management and procurement SMS communication

Electronic notification of TB cases Mobile telephone credit as enabler

1) Integrated, patient-centred care and prevention a) Early diagnosis of TB including universal drugsusceptibility testing, and systematic screening of contacts and high-risk groups b) Treatment of all people with TB including drugresistant TB, and patient support c) Collaborative TB/HIV activites and management of comorbidities d) Preventative treatment of persons at high risk and vaccination against TB 2) Bold polcies and supportive systems a) Political commitment with adequate resoures for tuberculosis b) Engagement of communities, civil society organisation, and pubulic and private care providers c) Universal health coverage policy, and regulatory frameworks for case notification, vital registration, quality and rational use of medicines, and infection control d) Social protection, poverty allviation and actions on other determinants of tuberculosis 3) Intensified research and innovation a) Discovery, development and rapid uptake of new tools, intervention and strategies b) Research to optimise implementation and impact, and promote innovations

Automated laboratory results VOT eLearning for staff

eLearning for patients

Digital uniqua identifier Add-on hardware to smartphones to permit clinical measurement Mobile devices as resources for data collection

Fig. 4.4  Implementation of digital health products to different components of TB (WHO, 2015a)

4  APPLICATION OF ARTIFICIAL INTELLIGENCE IN HEALTHCARE 

Diagnostic connectivity Connecting medical diagnostic devices for digitisation and transmission of data Current focus

Data repository and gateway Providing secure data storage and routing

75

Data presentation and applications Supporting timely and Actionable data use e.g. patient results notification and medical record system Future focus

Fig. 4.5  Schematic diagram of digital transformation of TB (WHO, 2015a)

and gateway, and data presentation and application. The transformation can start with the use of multiple digital products to increase diagnostic connectivity and assist in the acquisition of large amounts of health data from medical diagnostic tests and devices. This data is then digitized and stored in a secure data repository to await further processing and analysis. In order to gain useful insights, it is strategically important that the data is handled and managed suitably, ready for presentation or application. Moving into the data presentation and application phase proposed by the WHO, big data analysis and AI are among the solutions needed for extracting insights from large health data repositories (Fig. 4.5).  sing Artificial Intelligence in TB Diagnosis U AI is a promising solution to the handling of large amounts of data and extracting reasonable outcomes according to the set of rules provided to the algorithm. This branch of computer science focuses on the automation of intelligent behaviour (George, 1992) and allows Machine Learning, which gives computers the ability to learn and respond without re-­ programming (Samuel, 1959). This technology has been used and embedded in many healthcare applications such as speech transcription software and computer-aided diagnostic platforms, to recognize visual, voice and text input (Marr, 2016). With regard to TB, AI has been utilized in multiple platforms of diagnosis to help assess medical cases. Since many ­countries of high TB prevalence are mostly low or middle income, the implementation of AI in surveillance and monitoring could offer more

76 

J. CHANCHAICHUJIT ET AL.

rapid and accurate diagnostic tools with a higher sensitivity and specificity, compared to the current techniques such as microscopic sputum examination, TB chest X-rays, skin tests and cultures (Dande & Samant, 2018). Extensive research has been conducted on the development of new rapid and accurate TB diagnostic tools that include AI in their platforms. The history of AI in TB diagnosis dates back to 1999 (Farrugia, Yee, & Nickolls, 1993) where 21 input variables on radiographic findings, constitution symptoms and geographic variations (Fig. 4.6) were applied to an artificial neural network (ANN) by El-Solh et  al. The results revealed a sensitivity of 100% and a specificity of 72%, and suggested a significant role and the usefulness of radiographs in the prediction of TB. After this work, many researchers followed up on improving AI and ANN platforms for evaluation of chest radiographs. Other methods of detection such as pathology, clinical outcomes, blood samples and biosensors were also investigated for potential augmentation and integration with AI, as shown in Table 4.2. Since pulmonary issues are the major manifestation of the disease, TB chest X-rays (CXR) are one of the main methods suggested in WHO guidelines for monitoring and surveillance because of their c­ ost-­effectiveness and high sensitivity. Digital chest radiography, however, still requires trained professionals or radiologists to read and confirm the infection. The

Variation due to demography • Age • CD4 count • Diabetes melitus • Purified protein derivative

Constitutional symptoms • Chest pain • Weight loss • Cough • Night sweats • Fever • Shortness of breath

Radiographic findings • Upper and lower lobe infiltrate

• Upper and lower lobe cavity

• Adenopathy • Unilateral and

Bilateral pleural effusion • Pleural thickening • Miliary pattern

Fig. 4.6  Input variables for training of early AI in TB (Farrugia et  al., 1993)

39

Blood sample

Patient epicrisis N/A reports Pathology/ N/A Microscopy Plasmonic ELISA, N/A biosensing

Number of parameters

Test method

China Malaysia

201 N/A

Digital scanning image, slides Chromaticity analysis

Iran

France

97.94 83.65 >97

99.14

93.93

Demographic Accuracy (%)

175

175

Sample size

Haemato chemistry

Haemato chemistry

Input types

Table 4.2  Development of TB detection methods augmented by and integrated with AI

AbuHassan et al. (2017) and Tania et al. (2017)

Xiong et al. (2018)

Er, Temurtas, and Tannkulu (2010) Shamshirband et al. (2014)

Reference

4  APPLICATION OF ARTIFICIAL INTELLIGENCE IN HEALTHCARE 

77

78 

J. CHANCHAICHUJIT ET AL.

lack of experts in this crucial area has been one of the major challenges to performing regular and rapid chest X-ray evaluations, and has resulted in many patients being diagnosed too late. To address this issue, researchers and developers have been deployed and trained in AI using previously diagnosed CXR data. Numerous studies have shown an improvement in accuracy, both in sensitivity and specificity, using various computational techniques and algorithms, that is, fuzzy logic, ANNs, genetic algorithms and deep convolution neuron networks (DCNNs). Currently, AI is commercially introduced and incorporated into computer aid detection, which helps pre-identify abnormalities of the lungs in chest X-rays before further reviews and confirmation by experts. This is of great benefit and reduces the workload for the experts, and allows faster diagnosis for patients. Nowadays, remarkable advancements in Machine Learning and AI have driven many to believe a completely automatic detection platform for TB is possible in the near future (Yahiaoui, Er, & Yumusak, 2017; Satheeshkumar & Raj, 2006; Chui et  al., 2017), using Deep Learning techniques. One of the key success factors in developing an AI platform for TB detection is the training of the algorithm with a great deal of data, which will later enable it to differentiate the infection from normal cases. For high accuracy and reliability, a large database is required for Machine Learning. In recent years, many researchers and scientists have illustrated affirmative results with selected CXR databases. In order to analyse and represent data based on more complex features that can lead to learning processes, predictions and classification of information, more complex and advanced artificial neural networks such as DNNs have been considered. The latest DNNs comprise more complex hidden layers (e.g. a higher number of hidden layers and/or number of neurons per hidden layer) than older generations of DNNs. They have therefore been gaining attention as practical solutions in healthcare and preliminarily have proven to be appropriate systems to analyse complex data, for example images, CXR and genetic information (Dande & Samant, 2018). A relatively large research and development project using a DCNN that included over 1,007 X-rays of patients with and without active TB was completed by Dr Lakhani and his team. The type of AI used enables computers to finish tasks after relationships of data have been established (Lakhani & Sundaram, 2017). Dr. Lakhani’s work, later commercialized by Semantic MD, was based on data from the National Institutes of Health, the Belarus Tuberculosis Portal and Thomas Jefferson University Hospital, and sug-

4  APPLICATION OF ARTIFICIAL INTELLIGENCE IN HEALTHCARE 

79

Fig. 4.7  (a) Posteroanterior chest radiograph shows upper lobe opacities with pathologic analysis-proven active TB. (b) Same posteroanterior chest radiograph, with augmentation with convolution layers (Lakhani & Sundaram, 2017)

gested a net accuracy of >90% compared to other AI for TB which had around 80% accuracy. Several other commercial platforms using AI have been implemented and proved highly competent in TB detection. Computer-Sided Detection for Tuberculosis (CAD4TB) by Delft Imaging Systems is among the most widely-adopted platforms by many professionals for CXR evaluation. The CAD4TB platform was trained by applying Machine Learning to thousands of X-rays from over 15 countries, of healthy and diseased lungs. It also received expert feedback from lung specialists, making it extremely accurate, and it has been incorporated into the WHO e-Health Compendium (Fig. 4.7).  urrent Limitations of AI and Improvement C Promising and accurate results from many studies and investigations have encouraged practitioners and policymakers to adopt and implement AI technology. However, these technologies still need to be thoroughly proven and further investigation is required in clinical environments, where manifestations of the diseases may vary from subtle to severe. The retrospective images and data were a good starting point for Deep Learning and the development of the initial AI working platform, but

80 

J. CHANCHAICHUJIT ET AL.

further training with real clinical data is required in order to enhance the performance and reliability of the platform. Similar to the learning acquired by human doctors, experience and familiarity with clinical data provides confidence that leads to accurate decision-making. However, clinical data on the same disease presented in different countries could be vastly different, and the diagnostic decision tree could be different from one country to another. In the development of AI, this implies that Deep Learning with data from one geographic region may give accurate results only when applied to similar groups of populations, but it might not be enough to train AI for diagnosis in another region. This assumption was brought to light by two studies that used two de-identified Health Insurance Portability and Accountability Act (HIPAA)-compliant datasets including the National Institute of Health Chest X-ray dataset and the National Library of Medicine Shenzhen Hospital X-ray. They illustrated that when an AI platform using a DCNN was developed from one database and applied to analyse the other dataset, the accuracy was shown to decrease. The findings suggest different training data sets are likely to engage in variations of diagnostic accuracy of Deep Learning in different populations, highlighting the utility of a context-specific approach in the use of AI for automated CXR classification (Pongpirul, Sathitratanacheewin, Sunanta, & Kampa, 2018; Sathitratanacheewin & Pongpirul, 2018). Therefore, the reliability of AI platforms developed in one region may not be directly applicable to populations in another country before necessary justifications and further training using local clinical data. For instance, in a country with a high prevalence of TB and vast land size such as India, implementation of its own locally-trained AI might be a more beneficial solution than adopting foreign-trained AI to provide greater accuracy in detection. The design and architecture of AI to handle, extract, learn and analyse selected features of CXRs is critical and important. In CXR evaluations, abnormalities are among the critical features (Fig. 4.8) typically chosen for AI training by developers such as AiQure, an Indian start-up company (Qure.AI). These abnormalities include opacities (e.g. consolidation, fibrosis, calcification and atelectasis), blunted costophrenic angle, cavity, tracheal shift, cardiomegaly, hilar prominence and pleural effusion. The weights of each input and the relationship between these parameters are optimized and computed in the algorithm created. The more nodes and layers there are, the greater the accuracy and the better the Deep Learning ability in general terms. For CXR evaluation, DCNNs are among the best AI options for processing clinical images, and the resolu-

4  APPLICATION OF ARTIFICIAL INTELLIGENCE IN HEALTHCARE 

81

Fig. 4.8  Example of anomalies selected by the developers for AI training on chest X-rays (Qure.AI: http://qure.ai/qxr/)

tion of these images determines the quality of the inputs that enter the AI algorithm. Many platforms re-process these images to reduce the dimensions and resolution and to change the format in order to accelerate ­analysis and save on storage. This could potentially reduce the accuracy of the algorithm in detection, and hence it is best to balance all the attributes of AI in order to achieve the goals of TB detection. Although CXR detection is the most advanced method, with AI integration the limitations of CXRs in TB detection still remain for many cases such as non-pulmonary TB (see Fig. 4.8). Furthermore, CXR confirma-

82 

J. CHANCHAICHUJIT ET AL.

tion does not provide information that leads to a decision on the correct intervention, which is indeed important for TB disease control. Incorporation of supplementary inputs such as genetic information, pathology, sputum tests, biosensors, geographic information and symptoms was experimented with in AI (Dande & Samant, 2018; Er et  al., 2010), and could provide additional references to ensure the outcome and accuracy of AI algorithms in many cases. Additionally, more comprehensive AI that processes multiple types of data could offer meaningful results and solutions that may suggest appropriate intervention once various clinical test data is incorporated into the same algorithm. This kind of comprehensive model would be extremely helpful and bring AI technology many steps closer to achieving diagnosis in a similar way to a human expert who would call for multiple test results before reaching a final conclusion.

4.3   Innovation Management of Tuberculosis in Thailand Using Artificial Intelligence 4.3.1  TB Situation in Thailand and Opportunities for AI in TB Diagnosis The World Health Organization placed Thailand among the top 22 countries with the highest prevalence of TB, including drug-resistant TB. Each year, there are about 120,000 new TB cases, which is 1.3 times above the global average. The Public Health Ministry aims to reduce the number of cases to 88 per 100,000 people by 2021 and extend treatment coverage to 90% of patients (WHO, 2018b; Sonrueng, 2018). Under the leadership of the Bureau of Tuberculosis, Ministry of Health, Thailand, a great deal of research and numerous policies and activities have been implemented towards the goal of “Free TB Thailand” by 2035 as part of the global initiative by the WHO (Fig. 4.9). Accessibility to healthcare is limited in Thailand, where resources can be poor. Neglected populations in Thailand include prison inmates, taxi drivers, workers and illegal immigrants who live in crowded and poor living conditions. These people are at risk of contracting TB.  In 2017, screening was conducted in 143 prisons across Thailand. Of a total of 282,816 prisoners, 24,436 inmates were thought to have contracted TB. Of these, 3,368 inmates were confirmed to have contracted TB (Thai PBS, 2017). This initiative has precipitated a highly successful cure rate of 87–88% (The Nation, 2018a) in the past three years as many cases were

4  APPLICATION OF ARTIFICIAL INTELLIGENCE IN HEALTHCARE 

83

Fig. 4.9  TB screening during the “World TB Day 2018” campaign (Sonrueng, 2018)

Fig. 4.10  TB screening in prisons in Thailand (The Nation, 2018b)

detected early on. Hence the benefits of early detection in underserved demographics are obvious. In order to perform regular screening, the presence of many healthcare workers (HCWs) is needed, and the complexity of diagnosing the disease (Fig. 4.10) requires trained specialists and physicians for consultation and supervision. Traditionally, multiple tests such as the sputum test, CXRs and cultures would be prescribed to confirm and diagnose the stage of cases for TB by physicians in order to provide accurate interventions, as mentioned earlier. Unfortunately, many procedures are time-consuming and they may subject healthcare workers to exposure to the bacteria. In the past, the incidence of HCWs infected by TB was reported, and many of these cases resulted in multi-drug-resistant TB (von Delft et al., 2015), which was distressing for many HCWs. Several global studies showed that HCWs are at higher than average risk for TB (Baussano et al., 2011). In low- and middle-income countries, HCWs were at an annual risk of TB infection, ranging from 3.9% to 14.3%, which is higher compared to nor-

84 

J. CHANCHAICHUJIT ET AL.

mal occupational exposure (2.6–11.3%) (Joshi, Reingold, Menzies, & Pai, 2006). These findings called for action to improve the protection of physicians and healthcare workers. However, many middle-income countries, including Thailand, still failed to provide adequate TB transmission prevention initiatives in many hospitals due to limited resources. Involvement, activities and initiatives from associated governmental bodies are required to ensure the highest standards posted by the national guidelines are being met. This will help to create a safe environment for healthcare workers to work in, along with (Unahalekhaka, 2010) maintaining high morale at work. The mismatch in demand and supply contributes to an underestimation of cases and incidence in many countries including Thailand. Many assistive technologies for screening using AI have been shown to address unmet needs such as speed and accuracy, along with taking up the shortfall in healthcare professionals. Using an AI-integrated platform for evaluation of CXRs is one widely used technique due to its high effectiveness regarding both time and cost. The use of AI platforms, furthermore, helps to minimize unnecessary contact between disease-carrying persons and HCWs, and promotes the reduction of transmission. However, the reliability of AI-assisted CXR systems is still a question, and the WHO has not yet officially recommended any platform as the “gold standard” for TB detection (WHO, 2016). 4.3.2  Shortfalls of Current AI Platforms in TB Diagnosis The current platform using AI technologies specifically for CXR has been implemented in many areas. The CAD4TB project is one of the most promising developments in feasible digital technology to enhance and help non-expert readers to detect TB in resource-limited settings, (WHO, 2016). In a systemic review of CAD4TB, it was revealed that the CAD4TB programme was developed with training CXRs from a database of presumed TB patients from Zambia and South Africa. Its diagnostic performance has been evaluated in CXRs of patients from Zambia, Tanzania, England and South Africa. Consequently, its diagnostic accuracy is supported only by a small number of studies that still have methodological limitations (Pande, Cohen, Pai, & Ahmad Khan, 2016). The reviewers concluded that the system cannot offer accuracy of the same standard as expert readers without sacrificing specificity. Other platforms such as Sematic and AiQure for CXR evaluation potentially face similar limitations based on criteria in population selection and the small size of the database. The best

4  APPLICATION OF ARTIFICIAL INTELLIGENCE IN HEALTHCARE 

85

scenario is to diagnose patients using a platform derived from training CXRs of a similar demographic (e.g. country, career, social demographic). The quality and resolution of the images is another factor to consider. Typically, a CXR provided to radiologists is of high resolution in order for them to magnify any particular section of interest. In many AI platforms, raw CXR images may be processed and distorted before training and reading as part of the process. This might contribute to the problem in accuracy, and therefore using the optimum resolution should result in a more accurate and reliable diagnosis. 4.3.3  Alternative Models and Key Success Contributors for the Implementation of AI in TB As mentioned earlier, CXRs are effective and provide many benefits in TB detection, but they still have several limitations and do not provide enough information that leads to a specific intervention; further tests are required. Architectural design and links between parameters play a crucial role in achieving a successful AI algorithm. Therefore, the developer must thoroughly understand the nature of the disease and diagnosis pathway. Clear impactful outputs and outcomes will help developers to design algorithms towards a clear target. Therefore close collaboration with policymakers, governments and clinical users will be extremely helpful in refining these expected outputs and should increase the impact of AI in TB control. A more recent development in addressing unmet needs is that of DCNNs for the automated classification of pulmonary TB-associated radiographs (DAC4TB), developed by the Thai Health AI Foundation. In addition to the radiologist’s interpretation of CXRs, clinical signs and symptoms, sputum AFB, sputum culture and GeneXpert data are also employed in more comprehensive training, and the algorithm learns from data across different tests to establish a complex hidden layer. DAC4TB was designed to follow the WHO’s recommendations on the systematic screening for active TB by integrating multiple tests into the analysis (Fig. 4.11). This is a revolutionary example that uses multiple test data to increase the capability of CXR reading in TB control, and provides information that leads to intervention. In the DAC4TB process, Deep Learning allows the creation of complex hidden layers, and establishes a relationship between all the test data. After the training is complete, algorithms are able to read a new CXR and statistically correlate the results to the probability of the disease being TB by four methods of confirmation, TB, AFB,

86 

J. CHANCHAICHUJIT ET AL.

Fig. 4.11  Schematic diagram of the process and algorithm of DAC4TB to diagnose a CXR

Xpert and sputum culture, as shown in Fig. 4.11. When presented with the predicted test data of different platforms, TB, AFB, Xpert and sputum culture, a specialist should be able to use this data to make a recommendation for suitable treatment without the need to wait several months for the culture test or use of the “gold standard”. With innovative design and methodology developed in DAC4TB, the limitations of TB diagnosis by CXR are being resolved, and the information to decide upon intervention may be provided in the future. The success of the project and accuracy requires large amounts of data for training other than the CXR data. The acquisition, management and organization of large databases in all four different tests methods require strong collaboration and partnership with governments, healthcare institutes and international organizations. The Thai Health AI Foundation has been working closely with the Bureau of Tuberculosis, Ministry of Health, in order to use the database accrued over many years for Deep Learning. There has also been a pilot project set up to implement the platform in targeted health facilities to verify the accuracy of the platform compared to reading by radiologists, and to continue to gather new data. To ensure the highest accuracy, DAC4TB guarantees that high-resolution files of CXRs are used in the process, and the algorithm derived from a specific ­demographic will be used for diagnosis of the same group of population only. The segmentation and use of different algorithms must be evaluated once derived from different databases. For instance, algorithms derived from inmate data will be used on inmate populations and not on taxi drivers, or algorithms derived from Southeast Asia will be used on the people within the region and not on Europeans. Preliminary results suggest a high degree of accuracy, but more data will be examined for confirmation after the pilot project in Thailand. Researchers have a good relationship

4  APPLICATION OF ARTIFICIAL INTELLIGENCE IN HEALTHCARE 

87

with the Bureau of Tuberculosis, which means that the DAC4TB will be able to be implemented in many Thai health facilities and then presented for a system review by the WHO. With a greater amount and variety of data, Deep Learning should be achieved and testing then executed in order to perfect the desired algorithm(s).

4.4   Conclusion Chest X-ray evaluation is one of the most widely used, efficient and cost-­ effective methods of diagnosing TB. To overcome a lack of experts in low- and middle-income countries, AI was used to assist in CXR-based diagnosis. However, the algorithm presented to date fell short of the requirements for the WHO. Hence there is currently no recommendation regarding an AI-TB platform that is endorsed and recommended by the WHO for TB diagnosis. Other than limitations of accuracy, a CXR does not usually provide information that leads to diagnosis of stages, drug resistance or evidence of the bacteria. While the “gold standard” is still the TB culture test, the waiting time is too long, in that if the disease is present it may go on to cause serious complications and reach a stage where it is beyond curable. This has prompted physicians and researchers to seek new solutions that require both a unique design and improvements to AI in order to meet requirements and be specific enough for disease detection. Training and Deep Learning should contain data from other TB tests, and in this way AI machines will be able to compare and cross-analyse this data in a similar way to the processes carried out by human experts in the field. A group of physicians and researchers at the Thai Health AI Foundation has developed a new platform that could potentially revolutionize the AI platform in TB detection whilst still following the guidelines posted by the WHO. DAC4TB includes the data of multiple TB tests, CXR, AFB, GeneXpert and TB cultures of many patients and design algorithms, thus allowing the learning and analysis of relationships between these inputs. The screening of CXRs carried out by DAC4TB would still be done normally, but with a unique and complex analysis that refers to data from four different tests. The algorithm can reveal the statistical resemblance of the CXR if tested by other methods as well. With more comprehensive results, physicians and healthcare workers can make better recommendations for treatment and provide suitable interventions for patients. The success of the DAC4TB and other AI platforms for TB diagnosis depends on the design for the use of the database and the selection of the

88 

J. CHANCHAICHUJIT ET AL.

database. In the case of DAC4TB, a large amount of data is one of the key factors in achieving a new way of training, with data from four different types of tests of patients. A firm collaboration with the Thai government makes the acquisition of large amounts of data possible, and this could be implemented by executing a top-down policy. Another key success factor is that the design of the algorithm must reflect the thought processes and the requirements posted by authorized organizations such as the WHO. Lastly, the segmentation of diagnosis must be created, in that the algorithm developed from one set of data must be used to detect TB only on a similar demographic to that database in order to achieve the most reliable accuracy. The initiatives developed are still in the early stages, but they show great promise. A preliminary result has given a highly accurate degree of detection. With more data acquisition and training, the Thai Health AI Foundation hopes to present a unique and reliable method for TB detection with specifically designed algorithms for Thailand and the world. The completion of the project will hopefully reduce the cost of TB detection and increase the number of reported cases. It could eventually assist TB control to be on track with the aims of the WHO to end TB altogether. Acknowledgement  The authors would like to express their sincere gratitude to Dr Krit Pongpirul for his generous support and time in sharing his knowledge and information about the use of AI in his work in this chapter and Mr. Sinarta Wirawan, Dr. Albert Tan’s student at Curtin University, Singapore in conducting a literature review for this chapter.

References AbuHassan, K. J., Bakhori, N. M., Kusnin, N., Azmi, U. Z., Tania, M. H., Evans, B. A., …, Hossain, M. (2017). Automatic diagnosis of Tuberculosis disease based on Plasmonic ELISA and color-based image classification. Paper presented at the Engineering in Medicine and Biology Society (EMBC), 2017 39th Annual International Conference of the IEEE. Barga, R., Fontama, V., Tok, W.  H., & Cabrera-Cordon, L. (2015). Predictive analytics with Microsoft Azure machine learning. Springer. Bastawrous, A., & Armstrong, M. J. (2013). Mobile health use in low-and high-­ income countries: An overview of the peer-reviewed literature. Journal of the Royal Society of Medicine, 106(4), 130–142. Baussano, I., Nunn, P., Williams, B., Pivetta, E., Bugiani, M., & Scano, F. (2011). Tuberculosis among health care workers. Emerging Infectious Diseases, 17(3), 488.

4  APPLICATION OF ARTIFICIAL INTELLIGENCE IN HEALTHCARE 

89

CDC. (2016). Chapter 2: Transmission and pathogenesis of Tuberculosis. Retrieved from www.cdc.gov/tb/education/corecurr/pdf/chapter2.pdf. CDC. (2018). TB testing & diagnosis. Retrieved from www.cdc.gov/tb/ topic/testing. Center, U. o. M. M. (2016). Sputum Gram stain e overview. Retrieved from www. umm.edu/ency/article/. Chui, K., Alhalabi, W., Pang, S., Pablos, P., Liu, R., & Zhao, M. (2017). Disease diagnosis in smart healthcare: Innovation, technologies and applications. Sustainability, 9(12), 2309. Cohen, T., & Murray, M. (2005). Incident Tuberculosis among recent US immigrants and exogenous reinfection. Emerging Infectious Diseases, 11(5), 725. Dande, P., & Samant, P. (2018). Acquaintance to Artificial Neural Networks and use of artificial intelligence as a diagnostic tool for Tuberculosis: A review. Tuberculosis, 108, 1–9. https://doi.org/10.1016/j.tube.2017.09.006 Doshi, R., Falzon, D., Thomas, B.  V., Temesgen, Z., Sadasivan, L., Migliori, G. B., & Raviglione, M. (2017). Tuberculosis control, and the where and why of artificial intelligence. ERJ Open Research, 3(2), 00056–02017. Dowsland, K. A., & Thompson, J. M. (2000). Solving a nurse scheduling problem with knapsacks, networks and tabu search. Journal of the Operational Research Society, 51(7), 825. https://doi.org/10.1057/palgrave.jors.2600970 Er, O., Temurtas, F., & Tanrıkulu, A. Ç. (2010). Tuberculosis disease diagnosis using artificial neural networks. Journal of Medical Systems, 34(3), 299–302. Falzon, D., Timimi, H., Kurosinski, P., Migliori, G.  B., Van Gemert, W., Denkinger, C., … do Valle Bastos, L. G. (2016). Digital health for the end TB Strategy: Developing priority products and making them work. European Respiratory Journal, 48(1), 29–45. Fanning, A. (1999). Tuberculosis: 6. Extrapulmonary disease. CMAJ: Canadian Medical Association Journal  =  journal de l’Association medicale canadienne, 160(11), 1597–1603. Farrugia, S., Yee, H., & Nickolls, P. (1993). Implantable cardioverter defibrillator electrogram recognition with a multilayer perceptron. Pacing and Clinical Electrophysiology, 16(1), 228–234. George, F. L. (1992). Artificial intelligence: Structures and strategies for complex problem solving. Benjamin/Cummings. Gesing, B., Peterson, S., & Michelsen, D. (2018). Artificial intelligence in logistics. Retrieved from https://www.logistics.dhl/content/dam/dhl/global/ core/documents/pdf/glo-ai-in-logistics-white-paper.pdf. Hans, R., & Marwaha, N. (2014). Nucleic acid testing-benefits and constraints. Asian Journal of Transfusion Science, 8(1), 2. Health, N. J. Tuberculosis program. Retrieved October 1, 2018, from https:// www.nationaljewish.org/treatment-programs/directory/tb.

90 

J. CHANCHAICHUJIT ET AL.

Healthwise. (2017). Tuberculin skin test. Retrieved from https://www.cigna. com/individuals-families/health-wellness/hw/medical-tests/tuberculin-skintest-hw203560.. Jereb, J., Etkind, S. C., Joglar, O. T., Moore, M., & Taylor, Z. (2003). Tuberculosis contact investigations: Outcomes in selected areas of the United States, 1999. International Journal of Tuberculosis and Lung Disease, 7(12 Suppl. 3), S384–S390. Joshi, R., Reingold, A. L., Menzies, D., & Pai, M. (2006). Tuberculosis among health-care workers in low-and middle-income countries: A systematic review. PLoS Medicine, 3(12), e494. Kanabus, A. (2018). Information about Tuberculosis. Retrieved from https:// www.tbfacts.org/countries-tb/. Lakhani, P., & Sundaram, B. (2017). Deep learning at chest radiography: Automated classification of pulmonary Tuberculosis by using convolutional neural networks. Radiology, 284(2), 574–582. Lemley, J., Bazrafkan, S., & Corcoran, P. (2017). Deep Learning for Consumer Devices and Services: Pushing the limits for machine learning, artificial intelligence, and computer vision. IEEE Consumer Electronics Magazine, 6(2), 48–56. Lin, P. L., & Flynn, J. L. (2010). Understanding latent Tuberculosis: A moving target. Journal of Immunology, 185(1), 15–22. https://doi.org/10.4049/ jimmunol.0903856 Lindsay McKenna, A. Z., & Lessem, E. (2014). An activist’s guide to Tuberculosis drugs. New York: Treatment Action Group. Retrieved from http://www.treatmentactiongroup.org/sites/default/files/2016%20Activists%20Guide%20 to%20TB%20Drugs.1.5.pdf Little, J. V. (2004). Non-neoplastic disorders of the lower respiratory tract: Atlas of nontumor pathology. Chest, 125(3), 1176–1177. https://doi.org/10.1378/ chest.125.3.1176-a Marr, B. (2016). How machine learning, big data and AI are changing healthcare forever. Forbes Magazine. Retrieved from https://www.forbes.com/sites/bernardmarr/2016/09/23/how-machine-learning-big-data-and-ai-are-changinghealthcare-forever/#429051561a1c. MayoClinic. Retrieved August 30, 2018, from https://www.mayoclinic.org/ diseases-conditions/tuberculosis/symptoms-causes/syc-20351250. Nienhaus, A., Schablon, A., Costa, J. T., & Diel, R. (2011). Systematic review of cost and cost-effectiveness of different TB-screening strategies. BMC Health Services Research, 11, 247–247. https://doi.org/10.1186/1472-6963-11–247 Pande, T., Cohen, C., Pai, M., & Ahmad Khan, F. (2016). Computer-aided detection of pulmonary Tuberculosis on digital chest radiographs: A systematic review. The International Journal of Tuberculosis and Lung Disease, 20(9), 1226–1230.

4  APPLICATION OF ARTIFICIAL INTELLIGENCE IN HEALTHCARE 

91

Pongpirul, K., Sathitratanacheewin, S., Sunanta, P., & Kampa, K. (2018, 3–7 October 2018). Deep learning for automated classification of abnormal chest radiograph associated with Tuberculosis (DAC4TB) in the U.S.  Hospital-scale CXR database. Paper presented at the IDWeek 2018, San Francisco, USA. Qure.AI. Retrieved from http://qure.ai/qxr/. Research, M.  F. f. M.  E. a. Mycobacterium Tuberculosis complex, molecular detection, PCR, Paraffin. Retrieved from https://www.mayocliniclabs.com/ test-catalog/Clinical%C3%BEand%C3%BEInterpretive/62203. Samuel, A. L. (1959). Some studies in machine learning using the game of checkers. IBM Journal of Research and Development, 3(3), 210–229. Satheeshkumar, K., & Raj, A. N. J. (2006). Developments in computer aided diagnosis used for Tuberculosis detection using chest radiography: A survey. Journal of Engineering and Applied Sciences, 11(9), 5530–5539. Retrieved from https://pdfs.semanticscholar.org/aa1d/0fec10dda57fe549768ba51513a1 b802d39a.pdf Sathitratanacheewin, S., & Pongpirul, K. (2018). Deep Learning for Automated Classification of Tuberculosis-Related Chest X-Ray: Dataset Specificity Limits Diagnostic Performance Generalizability. Schmidhuber, J. (2015). Deep learning in neural networks: An overview. Neural Networks, 61, 85–117. Scott, L., Gous, N., Cunningham, B., Kana, B., Perovic, O., Erasmus, L., … Stevens, W. (2011). Dried culture spots for Xpert MTB/RIF external quality assessment: Results of phase 1 pilot study from South Africa. Journal of Clinical Microbiology, 49(12), 4356–4360. Shamshirband, S., Hessam, S., Javidnia, H., Amiribesheli, M., Vahdat, S., Petković, D., … Kiah, M. L. M. (2014). Tuberculosis disease diagnosis using artificial immune recognition system. International Journal of Medical Sciences, 11(5), 508. Singh, P., Kant, S., Gaur, P., Tripathi, A., & Pandey, S. (2018). Extra pulmonary Tuberculosis: An overview and review of literature. International Journal of Life-Sciences Scientific Research, 4(1), 1539–1541. Sonrueng, V. (2018). กรมควบคุมโรค รณรงค์ “ผู้ขับขี่ปอดสะอาด ปราศจากวัณโรค”. Retrieved from http://www.btripnews.net/?p=22348. Syam, N., & Sharma, A. (2018). Waiting for a sales renaissance in the fourth industrial revolution: Machine learning and artificial intelligence in sales research and practice. Industrial Marketing Management, 69, 135–146. Tania, M. H., Lwin, K., AbuHassan, K., Bakhori, N. M., Azmi, U. Z. M., Yusof, N.  A., & Hossain, M. (2017). An Automated Colourimetric Test by Computational Chromaticity Analysis: A Case Study of Tuberculosis Test. Paper presented at the International Conference on Practical Applications of Computational Biology & Bioinformatics.

92 

J. CHANCHAICHUJIT ET AL.

Tasillo, A., Salomon, J. A., Trikalinos, T. A., Horsburgh, C. R., Jr., Marks, S. M., & Linas, B.  P. (2017). Cost-effectiveness of testing and treatment for latent Tuberculosis infection in residents born outside the United States with and without medical comorbidities in a simulation model. JAMA Internal Medicine, 177(12), 1755–1764. https://doi.org/10.1001/jamainternmed.2017.3941 Thai PBS. (2017). Over 3,000 prison inmates found to have TB. Retrieved from http://englishnews.thaipbs.or.th/3000-prison-inmates-found-tb/. The Nation. (2018a). 388 suspected TB cases found among inmates in Phitsanulok prisons. Retrieved from http://www.nationmultimedia.com/detail/breakingnews/30355321. The Nation. (2018b). Inmates in Northeast screened for TB.  Retrieved from http://www.nationmultimedia.com/detail/national/30341409. Tsara, V., Serasli, E., & Christaki, P. (2009). Problems in diagnosis and treatment of Tuberculosis infection. Hippokratia, 13(1), 20–22. UN. Sustainable development goals. Retrieved from https://sustainabledevelopment.un.org/topics/sustainabledevelopmentgoals. Unahalekhaka, A. (2010). Situation analysis of nosocomial TB transmission prevention among hospitals in Thailand: Presentation Number 8-086. AJIC (American Journal of Infection Control), 38(5), E75–E76. Uplekar, M., Weil, D., Lonnroth, K., Jaramillo, E., Lienhardt, C., Dias, H. M., … Raviglione, M. (2015). WHO’s new end TB strategy. The Lancet, 385(9979), 1799–1801. https://doi.org/10.1016/S0140-6736(15)60570-0 Vikas K.  Saket, R.  K. a. P.  S. (2017). Tuberculosis in animals and humans: Evolution of diagnostics and therapy. Asian Journal of Animal and Veterinary Advances, 12(4), 177–188. von Delft, A., Dramowski, A., Khosa, C., Kotze, K., Lederer, P., Mosidi, T., … von Delft, D. (2015). Why healthcare workers are sick of TB. International Journal of Infectious Diseases, 32, 147–151. WebMD.  What is a sputum culture? Retrieved October 1, 2018, from www. webmd.com/lung/sputum-culture. WHO. (2006a). Diagnostics for Tuberculosis: Global demand and market potential, 26. Retrieved from http://www.who.int/iris/handle/10665/43543. WHO. (2006b). Tuberculosis Fact sheet N°104—Global and regional incidence. Retrieved from https://www.who.int/en/news-room/fact-sheets/detail/ tuberculosis. WHO. (2008). New laboratory diagnostic tools for Tuberculosis control. Retrieved from https://www.who.int/tdr/publications/documents/diagnostic-tool-tb.pdf. WHO. (2010). Treatment of Tuberculosis: Guidelines. Retrieved from https:// www.who.int/tb/publications/2010/9789241547833/en/. WHO. (2011a). Fluorescent light-emitting diode (LED) microscopy for diagnosis of Tuberculosis: Policy statement. Retrieved from https://www.who.int/tb/ publications/2011/led_microscopy_diagnosis_9789241501613/en/.

4  APPLICATION OF ARTIFICIAL INTELLIGENCE IN HEALTHCARE 

93

WHO. (2011b). Guidelines for intensified case-finding and isoniazid preventive therapy for people living with HIV in resource constrained settings. Retrieved from https://apps.who.int/iris/bitstream/handle/10665/44472/9789241500708_ eng.pdf?sequence=1. WHO. (2014). Global Tuberculosis Report 2014. Retrieved from http://apps. who.int/iris/bitstream/10665/137094/1/9789241564809_eng.pdf. WHO. (2015a). Digital health for the end TB strategy: An agenda for action. Retrieved from https://www.who.int/tb/publications/digitalhealth-TBagenda/en/. WHO. (2015b). Systematic screening for active Tuberculosis: An operational guide (WHO/HTM/TB/2015.16). Retrieved from https://apps.who.int/ iris/bitstream/handle/10665/181164/9789241549172_eng. pdf?sequence=1. WHO. (2016). Chest radiography in Tuberculosis detection: Summary of current WHO recommendations and guidance on programmatic approaches. Retrieved from https://www.who.int/tb/publications/chest-radiography/en/. WHO. (2017). Global Tuberculosis Report 2017. Retrieved from http://www. who.int/tb/publications/global_report/en/ WHO. (2018a). Latent Tuberculosis infection: Updated and consolidated guidelines for programmatic management. Retrieved from http://apps.who.int/ iris/bitstream/handle/10665/260233/9789241550239-eng. pdf?sequence=1. WHO. (2018b). Tuberculosis country profiles. From World Health Organisation https://www.who.int/tb/country/data/profiles/en/. Xiong, Y., Ba, X., Hou, A., Zhang, K., Chen, L., & Li, T. (2018). Automatic detection of mycobacterium Tuberculosis using artificial intelligence. Journal of Thoracic Disease, 10(3), 1936–1940. Yahiaoui, A., Er, O., & Yumusak, N. (2017). A new method of automatic recognition for Tuberculosis disease diagnosis using support vector machines. Biomedical Research, 28(9). Zink, A.  R., Grabner, W., & Nerlich, A.  G. (2005). Molecular identification of human Tuberculosis in recent and historic bone tissue samples: The role of molecular techniques for the study of historic Tuberculosis. American Journal of Physical Anthropology, 126(1), 32–47. https://doi.org/10.1002/ajpa.10409

CHAPTER 5

Optimization, Simulation and Predictive Analytics in Healthcare

Abstract  This chapter discusses the use of operations research techniques such as optimization, simulations and predictive analytics in healthcare. The chapter introduces optimization problems in healthcare, from strategic resources and capacity planning to operational and clinical issues such as resource scheduling and treatment planning. Case studies using operations research in healthcare in Singapore will be presented, followed by some insights into improved healthcare delivery. Keywords  Operations research • Optimization • Healthcare service planning • Shift capacity planning • Bed management • Inpatient flow

5.1   Overview of Operations Research in the Healthcare System Operations research (OR) is a discipline that deals with the application of advanced analytical methods to help stakeholders make better and more informed decisions. OR has strong ties with many disciplines including mathematics, business, computer science, economics, engineering and statistics. OR encompasses a wide range of methods such as mathematical optimization, stochastic modelling, simulation, scheduling, forecasting, decision trees and queueing systems (Rais & Viana, 2010). OR originated in the UK in the armed forces during World War II and ever since has been © The Author(s) 2019 J. Chanchaichujit et al., Healthcare 4.0, https://doi.org/10.1007/978-981-13-8114-0_5

95

96 

J. CHANCHAICHUJIT ET AL.

used in decision-making in various industries worldwide. Healthcare is one such industry in which decision-making processes are mainly related to supplying allocation solutions, analysing the big data of Electronic Medical Records and Electronic Health Records, activity and personnel scheduling, and healthcare service planning. As pointed out in 1952 (AJPH, 1952), OR has drawn considerable attention for its potential contribution to public healthcare. Its applications have been intensively studied for more than four decades to improve healthcare delivery, with a focus on decision support tools with operational, tactical and strategic approaches. The diverse contributions and approaches of OR have made it one of the most popular decision-making and analysis tools in healthcare (Denton, Alagoz, Holder, & Lee, 2011). OR involves systematic research that investigates how to plan, implement and evaluate activities within a system. Healthcare problems are often complicated, complex and uncertain. Among the many methodological approaches in OR, two methodologies are particularly important, deterministic optimization modelling and stochastic modelling, which are commonly used in OR applications in healthcare. Optimization modelling, which is often considered a core part of OR, comprises three major components: objective functions, decision variables and a set of constraints. Optimization methods have been applied to problems of capacity management and location selection for both healthcare services and medical suppliers. Patient scheduling, provider resource scheduling and logistics are additional large areas of research in the application of optimization methods to healthcare. Healthcare decisions inherently involve uncertainty, attributed to random patient visits and uncertain lengths of stay, the complexity of diseases, high variability of patient responses to treatments, partially observable physiological processes, errors in diagnostics tests and so on. Stochastic models such as decision trees, Markov models and simulation often offer appropriate solutions for the uncertainty of healthcare problems (Capan et  al., 2017). OR applications in healthcare have extended to, for instance, using • stochastic optimization modelling for ambulance allocation problems, • hybrid models incorporating queueing and robust optimization techniques for bed assignment and patient scheduling, • Markov models for patient admission processes and patient flow management,

5  OPTIMIZATION, SIMULATION AND PREDICTIVE ANALYTICS… 

97

• Markov decision processes based on models and Machine Learning for health screening, disease classification and prediction, and personalized medicine, • simulation models such as discrete event simulation and microsimulation for predicting patient disease progression, the trajectory of population health conditions and the trend of system costs of a population. An array of commercial specialized software can be used to solve healthcare problems using optimization modelling and simulation, such as AIMMS, Analytica, AMPL, Arena, CPLEX, Gurobi, GoldSim, LINDO API, LINGO, Maple, Mathematica, MATLAB, Minitab, OPL Studio and SIMUL8.

5.2   How to Apply Operations Research in Healthcare 5.2.1  Healthcare Service Planning Today the importance of planning in healthcare can hardly be overemphasized, as providing an adequate, accessible and sustainable healthcare service is a key concern for many countries in the world. With growing longevity and a rapidly increasing ageing population, many countries are under pressure to find extra money and resources to meet healthcare needs. Important healthcare issues using OR methods include estimation of future demand for services in order to build enough capacity, selection of hospital locations for covering a target population, and design of community care, primary care and emergency care facilities for efficient handling of patients. One powerful tool in healthcare planning is a simulation model such as discrete event simulation (P. R. Harper & Gamlin, 2003). In addition, time series regression and artificial neural network models can be used to forecast demand on healthcare services. Integer programming models are often applied to assess suitable sites for healthcare facilities and to maximize accessibility. Stochastic integer optimization models are used to solve such problems as the best base locations for a limited number of emergency vehicles, or bed capacity planning for achieving optimal resource utilization (P.  Harper, 2002; P. R. Harper & Shahani, 2001).

98 

J. CHANCHAICHUJIT ET AL.

5.2.2  Healthcare Management and Logistics Patient scheduling, resource allocation and scheduling, and healthcare logistics are the most extensively referenced management problems. In particular, nursing staff scheduling, operating theatre scheduling and bed assignment have drawn much attention from hospital managers and researchers. The Markov decision process (MDP) model has been employed to address dynamic patient schedules with different priorities according to patient acuity category to public healthcare facilities. Queueing models are often used for outpatient appointment problems and for improving the efficiency of hospital contact centres (Cayirli & Veral, 2003). In general, nursing staff scheduling consists of assigning shifts to nurses with different skill sets, while satisfying many soft constraints and personal preferences as much as possible. To address this problem, two-stage stochastic or deterministic optimization models or multiple-objective models can be developed by incorporating both nurse preferences and hospital requirements. Linear programming models or the robust optimization approach can be established for assigning operating theatres, physician scheduling and bed allocation. Simulation can also be used to analyse patient surgical scheduling, emergency department (ED) waiting time and specialist outpatient clinic appointments (Cayirli & Veral, 2003). Meng, Teow, Teo, Ooi, and Tay (2019) predicted 72-hour patient reattendance in EDs using mixed integer programming via discriminant analysis with electronic medical records. Logistics in healthcare are concerned with drug and medical material stock levels and delivery of medical items from suppliers to hospital resource centres and from hospital warehouses to pharmacy departments, and the reverse logistics of medical products. Integer programming and constrained optimization models often serve to streamline processes and standardize the medical materials and inventory levels needed in hospitals, based on their corresponding importance and service levels. Linear programming models can also be used to analyse the optimal allocation of a budget to a set of healthcare resources or a set of alternative intervention programmes based on their potential outcomes, such as quality adjusted life years and utility theory. In addition, perishable inventories in supply chain management can be applied to blood bank management policies, such as managing red blood cell supplies based on expiration dates and demand such as urgent and non-urgent.

5  OPTIMIZATION, SIMULATION AND PREDICTIVE ANALYTICS… 

99

5.2.3  Clinical Practice In addition to supporting stakeholders in improving the operations processes in healthcare, OR approaches can also be employed to assist in improved decision-making from a clinical perspective; this is termed clinical OR. OR applications in this area include clinical treatment decisions, infectious disease prevention and control, pandemic preparedness, health screening and disease diagnosis, emergency response, and organ donation and transplants (Capan et al., 2017). Optimization models, the MDP and simulation models are often used in analysis (Alagoz, Hsu, Schaefer, & Roberts, 2010; Denton, Kurt, Shah, Bryant, & Smith, 2009; Kurt, Denton, Schaefer, Shah, & Smith, 2011; Mason, Denton, Shah, & Smith, 2014). For instance, optimization-based data-mining techniques can be applied to assess normal brain activity and epilepsy episodes in patients with mental health issues, and to classify malignant and non-malignant tissues in patients with lung cancer. Chin, You, Meng, Zhou, and Sim (2018) studied schizophrenia with combinatorial regularized support vector machines and sequential region of interest selection using structural magnetic resonance imaging and predicted patients with schizophrenia with a high accuracy. Diabetes is one of the most common chronic diseases, and treatment decisions for patients with diabetes have been studied intensively using OR methods. In particular, optimal treatment strategies can be determined in terms of timing and sequencing of medications for managing blood pressure, cholesterol or HbA1c levels for patients with diabetes using the MDP. Optimal frequency for health screening for breast cancer can be examined using the MDP based on models. Policies around dynamic organ donation allocation, such as kidney transplants to waitlisted patients, can be also be created by using queueing techniques and simulation models.

5.3   Case Studies This section presents three case studies using OR applications. Bed management, nurse staffing and patient waiting time are key concerns for hospital management today. These have become problematic due to rapidly increasing healthcare demands and budget constraints of healthcare departments.

100 

J. CHANCHAICHUJIT ET AL.

5.3.1  Case Study 1. Managing Elective Admissions Using the Robust Optimization Approach Background Overcrowding in the EDs of hospitals creates long waiting times, which are associated with patient satisfaction and, more importantly, morbidity and mortality (Librero, Marn, Peir, & Munujos, 2004; Sun, Teow, Heng, Ooi, & Tay, 2012; Thompson & Yarnold, 1995). ED overcrowding is often due to the availability of hospital beds. However, beds are a critical resource in hospital operations and highly trained personnel are required to manage these beds. Day-of-Week patterns of a hospital exhibit a wide range of variations. Emergency admissions are unpredictable, while elective admissions are scheduled by the hospital. Nevertheless, often the relative variation is largest in elective admissions, and larger in discharges than emergency admissions. On days with high bed occupancy, long wait times are encountered. On days with low bed occupancy, beds are under-­ utilized. There is tightness of usage on one hand and looseness on the other. However, it is not the desired state. Elective surgeries account for the majority of elective admissions, though medical electives (non-surgical cases) do make up some of these admissions. Elective surgeries are procedures planned in advance and they can be divided into day surgery (DS), same day surgery admission (SDA) and inpatient admission (IP). DS cases do not generally take up beds, while SDA cases require beds to accommodate the patient the day after surgery. IP cases require beds one day before the surgery. In general, hospitals will admit all patients with urgent needs. Figure 5.1 shows a diagram of inpatient flow in a hospital. In Singapore, there are four levels of patient acuity category (PAC) for ED attendance, with PAC1 being the most serious and PAC4 the least. All PAC1 patients and a large portion of PAC2 patients usually need to be admitted to hospital. As such, in a tight bed situation, the trade-off is to reduce the number of beds designated for elective admissions. But a more prudent and sensible approach would be to make adjustments on a dynamic basis. What this would entail is that when emergency cases are fewer, more beds could be assigned to elective cases, and vice versa. This would lead to an optimal control policy, which would maximize bed utilization on a daily basis by controlling the number of elective admissions. This would require a more prudent scheduling of operating theatre sessions. However, a higher level of complexity in

5  OPTIMIZATION, SIMULATION AND PREDICTIVE ANALYTICS… 

EMERGENCY ARRIVALS Requests for bed (Demand)

101

Cancel operation

ELECTIVE ARRIVALS Admit to bed Same-day surgery admission (SDA) Elective surgeries

Inpatient admission (IP)

Free beds (Supply)

Day surgery (DS) DISCHARGE

People in beds/inpatients

Length of Stay (LOS)

Fig. 5.1  An illustrative diagram of inpatient flow in hospital

­ lanning ensues because of the high degree of uncertainty involved in bed p availability and its effect on admission rates. Various models for managing patient admissions have been proposed in the literature. In general, queueing theory and stochastic simulation are the main methodological tools in studies of bed allocation and bed capacity. The underlying rationale for researchers relying on these methodological tools is the uncertain nature of the hospital unit vis-à-vis the number of patients as a result of random arrivals and random lengths of stay. These studies were based on the blanket assumptions of Poisson’s patient arrival distribution and negative exponential distribution for length of stay. The admission of emergency inpatients is unscheduled and they are usually placed in a ward within hours. In contrast, the admission of elective patients is less pressing and they can be put in the ward on the day of admission or even several weeks later. The flexibility vis-à-vis elective patients allows hospitals to manage the flow in such a way as to smooth out the daily bed occupancy, a modus operandi known as elective smoothing. This ensures that on days with spikes in emergency cases, the admission rate for elective patients can be reduced. Figure 5.2 demonstrates examples of two bed allocation policies. In the figure, although the daily bed quota in both scenarios is less than the daily capacity, scenario 1 with the elective smoothing strategy is more favourable than scenario 2. Some ­hospitals in

102 

J. CHANCHAICHUJIT ET AL.

An illustration of different bed allocation scenarios 80 70

Number of beds

60 50 Scenario 1

40

Scenario 2 Bed capacity

30 20 10 0

1

2

3

4

5

6

7

Fig. 5.2  An example of different bed allocation policies

Singapore have already incorporated this mechanism into their decision support systems and it has led to improvements when elective patient flow is high. In these hospitals, the admission quotas for elective patients are obtained by solving a deterministic linear optimization problem without taking into account the variability of patient arrivals and stay durations. While this achieves smoothing in expectation, it is conceivable that efficacy would diminish when variability is high. Due to the difficulties of obtaining true probability distributions and solving stochastic optimization problems, it is common in real-world deployment of optimization technology to ignore uncertainty. A fine level of analysis would be required to obtain the distributions of patient admission and departure profiles as a function of admission quotas, which may not necessarily lead to a computationally tractable optimization problem. In recent years, robust optimization (RO) has offered an attractive alternative for addressing uncertainty in optimization modelling without having to specify exact probability distributions (Ben-Tal & Nemirovski, 1998; Bertsimas & Sim, 2004; Chen, Sim, & Sun, 2007; Delage & Ye, 2010; Goh & Sim, 2010). In many interesting cases, the approach has led to computationally tractable optimization problems. In classical robust optimization, uncertainty is represented by an uncertainty set, which is

5  OPTIMIZATION, SIMULATION AND PREDICTIVE ANALYTICS… 

103

usually a simple geometric convex set, such as a Euclidean ball intersected with the support set, the minimal convex set that contains the uncertainty. The modeller is required to address the ambiguity attitude by specifying the budget of uncertainty parameters, which relate to the size of the uncertainty set against which the modeller seeks immunity. Model Description In this study, we use the distributionally robust optimization approach for managing elective admission in hospitals, where uncertainty is characterized by a support set and a restricted ambiguous set of probability distributions or “the ambiguity set”. Briefly, the ambiguity set is the family of all possible probability distributions describing the uncertainty of both emergency and elective patient arrivals and their length of stay, defined by the information of first- and second-order moments of these underlying random variables with the patient administrative data. Similar to the uncertainty set in classical robust optimization, the proposed ambiguity set is adjustable via a so-called budget of variation parameter, which is the bound on the coefficient of variation of the uncertainty parameters. The ambiguity set is enlarged by increasing the budget of variation, which leads to greater uncertainty in patient movement. An approach to optimize the budget of variation is proposed, while ensuring that the worst-case expected maximum bed requirement over the planning horizon falls below the bed capacity of the hospital. This approach is inspired by an actual problem which has allowed us to have access to real data. We describe a non-parametric approach for characterizing the uncertainty of patient admissions and departures, using information obtained from patient movement records. The aim is to introduce a model of uncertainty without imposing an excessive burden on the information requirements, which may otherwise deter practical implementation. Instead of ignoring variability and assuming deterministic parameters taking values at their empirical averages, which is usually done in practice, we assume that the parameters are random variables with known means. However, their precise distribution is unavailable but belongs to a restricted ambiguity set. To avoid being overly conservative, we control the size of the ambiguity set by specifying the budget of variation, which is the upper bound of the coefficients of variations of all the uncertainty parameters. Using the duality theory, the proposed RO model is transformed to a

104 

J. CHANCHAICHUJIT ET AL.

second-order cone programme, which can be solved by state-of-the-art commercial solvers such as CPLEX. Numerical Experiment We studied the performance of our robust optimization models using patient-level administrative data from a public hospital in Singapore. We had access to one year of data (366 days) for the purpose of evaluating and comparing the performance of various models. Our data set consisted of daily admission and length of stay of both emergency and elective patients throughout 2008. Emergency patients, averaging about 119 daily, accounted for about 82% of daily admissions. Their mean length of stay at 3.57 days exceeded that of elective patients by about one day (see Table 5.1). The admission pattern of emergency patients over a typical week shows the highest admission rates on Monday and Tuesday and a relatively less busy period from Thursday to Sunday (see Fig. 5.3). In numerical experiments, we compared the solutions of four different strategies: Model 1. Uniform quota model: the quota is distributed evenly to maintain a weekly total quota; Model 2. Deterministic model: the quota is obtained by solving the deterministic model; Model 3. Robust model: the quota is obtained by solving the robust optimization model with different budgets of variation; and Model 4. Optimized robust model: the quota is obtained by solving the optimization problem of the robust model with respect to the budget of variation. Table 5.2 presents different configurations of the numerical study with each configuration differing in the parameters, total bed capacity, weekly bed quota and planning horizon. The evaluation period is fixed to 143 weeks (i.e. 1001 days). We reported numerical results in Tables 5.3 and 5.6 and used bold text to highlight the model with the best performance. Table 5.1  Descriptive statistics of patients Patient type Daily admissions: mean Daily admissions: std. dev. Daily admissions: coefficient of variation Length of stay: mean Length of stay: std. dev. Length of stay: coefficient of variation

Emergency patient

Elective patient

119.10 16.15 0.14 3.57 2.97 0.83

25.98 13.12 0.51 3.21 2.70 0.84

5  OPTIMIZATION, SIMULATION AND PREDICTIVE ANALYTICS… 

105

Fig. 5.3  Average emergency admissions by day of the week Table 5.2  Different scenarios in numerical study Scenarios Total beds Bed quota per week Length of planning period (day)

1

2

3

4

5

6

7

8

9

600 203 7

600 203 14

600 203 21

620 245 7

620 245 14

620 245 21

650 301 7

650 301 14

650 301 21

Table 5.3 shows the total bed shortages of different models under different configurations. Note that the optimized robust model performs relatively well against other models. In particular, it has significant performance improvements over the uniform quota and deterministic models. In Fig. 5.4, the performance of the optimized robust model was compared against the deterministic and uniform quota models with higher bed capacities. Bold text is used to emphasize the total bed shortage values that are closest to those obtained via the optimized robust model. Observe that the uniform quota model would require about 11 additional beds while the deterministic model would require about 4 additional beds to

106 

J. CHANCHAICHUJIT ET AL.

Table 5.3  Difference of bed shortage compared with Model 1 Scenario

1 2 3 4 5 6 7 8 9

Bed shortage

Difference of bed shortage compared with Model 1

Model 1

Model 2 Model 3 Model 4_0.01 Model 4_0.02 Model 4_0.05

2438 2438 2438 2969 2969 2969 2012 2012 2012

−1020 −1009 −876 −1191 −1173 −950 −777 −782 −633

−1290 −1323 −1353 −1528 −1499 −1543 −1030 −1070 −1050

−1366 −1263 −1256 −1500 −1433 −1515 −987 −1018 −1088

−1370 −1350 −1306 −1460 −1534 −1480 −1101 −1095 −1025

−1043 −347 −49 −1170 −634 −496 −904 −834 −686

Comparison of bed shortages of Model 4 with Models 1-2 with extra beds 1950 1750 1550 1350 1150 950 750

1

2

3

4

5

6

7

Model 1(+8)

Model 1(+10)

Model 1(+12)

Model 1(+14)

Model 2(+4)

Model 2(+5)

Model 2(+6)

Model 4

8

9 Model 2(+3)

Fig. 5.4  Comparisons of bed shortages of Model 4 with Models 1 and 2 with additional beds

achieve a similar performance to the optimized robust model. Although the saving in beds may be a small fraction of the total bed capacity, it is a still noteworthy improvement, since increasing the number of beds is usually not an option. Figure 5.5 presents the performance of the different models based on the fraction of the evaluation period with bed shortfalls. The robust optimization model yielded better results than the other models.

107

5  OPTIMIZATION, SIMULATION AND PREDICTIVE ANALYTICS… 

Percentage of days with bed shortfalls over evaluation period (%)

18 16 14 12 10 8 6

1

2 Model 1

3 Model 2

4

5 Model 3

6

7

Model 4_0.01

8

9

Model 4_0.02

Fig. 5.5  Fraction of days with bed shortfalls over the evaluation period

5.3.2  Case Study 2. Shift Capacity Planning for Nursing Staff Using Mixed Integer Programming B  ackground ED management is one of the most challenging fields in healthcare today. ED utilization has grown rapidly, resulting in overcrowding, consult or admission delays, and increased staff burnout (Boyle, Beniuk, Higginson, & Atkinson, 2012; Burke, De Causmaecker, Berghe, & Van Landeghem, 2004; Bursch, Beezy, & Shaw, 1993; Cheang, Li, Lim, & Rodrigues, 2003). Emergency nursing takes care of patients of all ages with perceived or actual physical or emotional changes in health that are undiagnosed or that require further intervention. A report by the World Health Organization states that managing and deploying personnel resources efficiently will be key challenges for the healthcare industry in the coming decade due to the increasing demand for care and healthcare providers. In particular, nursing staff are critical to efficient healthcare services. Studies demonstrate that unattractive schedules, poor practice environments and high workloads are key factors which lead to discontent and a high nursing turnover. On the other hand, proper personnel policies or rules were shown to have a positive impact on the nurses’ working conditions, which in turn are closely related to the quality of care (Hu, Chen, Chiu, Shen, & Chang, 2010; Trinkoff et al., 2011). These observations should motivate hospitals to adopt policies that increasingly accommodate the preferences

108 

J. CHANCHAICHUJIT ET AL.

and requests of nursing staff while ensuring suitably qualified staff are on duty at the right time. As appropriate staffing and shift scheduling of nursing staff play a role in delivering quality patient care, it is necessary to develop better decision support systems for hospital decision-makers to manage nursing staff. Nursing staff management is basically a sequential planning and control process, consisting of nurse staffing, shift planning and allocation. In general, staffing is strategic and long-term manpower planning. Shift scheduling and assignment satisfy both the minimum coverage requirements and time-related rules and practices for the nurses and the hospital. Nurse allocation determines the individual nurse’s schedule and the timetable for all nurses. In past decades, a great deal of effort has been made to investigate nurse scheduling using various methods from different perspectives (Cheang et al., 2003; Dowsland & Thompson, 2000; Jaumard, Semet, & Vovor, 1998; Meng, Teow, Ooi, Soh, & Tay, 2016). Patients may flow into different locations in the ED depending on their needs. Consequently, the workload at the ED can change dramatically depending on the time of day and the day of week. To spread out the workload over a day, some researchers studied the ideal patient-to-nurse ratio in the general ward ­setting. However, few studied the situation in an ED setting. In particular, there has been an accelerating demand for emergency care for the ageing population. It has been observed that the ability of staff to cope with patient loads not only affects the health outcomes and satisfaction of patients, but also the morale and wellbeing of the staff. Hence, it is crucial to develop an appropriate nurse staffing model to address the main concerns of healthcare providers. Model Description In this study, we aim to balance the overall nurse workloads over the period of a day looking at current staffing capacity (Meng et al., 2016). In particular, we would like to develop a mathematical optimization approach to determine optimal staffing rules for shift scheduling and capacity planning in an ED. The proposed model also incorporates shift working hours, shift times and total number of shifts as decision variables for planning. Regarding this, an immediate issue is to estimate patient loads (i.e. demands) in various areas in an ED over different time intervals in a day. In the literature, patient arrivals are usually modelled according to the Poisson process, and patient loads in different areas in the ED are estimated using a queueing theory with expected service rates (i.e. patient

5  OPTIMIZATION, SIMULATION AND PREDICTIVE ANALYTICS… 

109

treatment and care). However, this approach is often criticized due to the strong assumptions around arrivals and the high variability of patient stay in the ED by nature. In this study, a different way of assessing patient loads is used by employing large-scale patient data. Briefly, an individual patient’s movements within various areas in the ED are studied, starting at triage through to leaving the ED. At a given time of the day, a weighted sum of accumulated nursing touch points is used on average for key areas in ED demand during this time period. We investigate the nurse workload over time periods in the day by using the demand and the total number of nursing staff on duty during this period. The proposed model is a deterministic mixed integer programming model in which the objective is to minimize the mean deviation of the workloads over different time intervals of a day. The model provides optimal decision rules concerning the total number of shifts to be planned, shift working hours and the number of nursing staff for each shift. ED managers can use these suggested results in decision-­ making for nurse rostering. The hospital used in this study is an acute care hospital in Singapore, consisting of about 1200 beds. The key functional areas in the ED are triage, consult, resuscitation, trolley observation room, fever observation room and decontamination. In general, when patients arrive at the ED, nurses will triage them. Doctors will assess them in consultation rooms. After consultation, patients may proceed to observation rooms, or leave the ED. The workload intensity may vary among these areas. For instance, the workload in the decontamination area is generally much heavier than in the observation room. To evaluate the workload of the entire ED used in this study, the demand for nursing staff in different areas is adjusted with weighting, which basically reflects the differences in workload intensity between the areas. The values of these weightings are suggested by the ED. The analysis is based on the adjusted demand at various times in the day under consideration. Nursing staff with different skill sets are deployed at different locations and work in shifts. It is assumed that all the nursing staff are fully skilled. Time intervals under consideration are referred to in half-hourly time periods starting at 00:00; there are a total of 48 half-hour intervals in the day under analysis. The aim is to balance the overall workload of nursing staff during these time intervals through designing optimal nursing staff deployment policies. A patient is said to make a touch point at some area when they arrive at that location. Changes of touch points are captured when staff update patient locations in the ED information system. The touch points are used to estimate the demand. Patients

110 

J. CHANCHAICHUJIT ET AL.

could have multiple touch points for a specific location such as ­consultation areas. In this case, we only use the first touch point to represent the demand at that location. The demand at each time interval is then calculated by the weighted sum of the total touch points of the key locations in the ED (see Fig. 5.6). The ED operates three major shifts for nursing staff, that is, morning shift (AM shift), afternoon shift (PM shift) and night shift, each with shift times from 07:00 to 15:30, 13:00 to 21:30 and 21:00 to 07:30, respectively. In addition to the major shifts, there are three minor shifts, with shift times from 09:00 to 17:30, 15:00 to 23:30pm and 16:00 to 24:00. These three minor shifts operate between the major shifts. In practice, the numbers of nursing staff in minor shifts are relatively small. They are assigned to various locations in the ED as extra support staff to meet high patient loads. The present complete staff deployment distribution over the 48 half-hour time intervals is taken as the baseline schedule. This together with the estimated demand over time intervals facilitates the assessment of the ratio of the number of nursing staff to total touch points (or demand) at each half hour, which is referred to as a measure of the workload in analysis. By analysing patient administrative data and current nursing staff assignment, current workload can be seen to vary considerably over the 200

Overall daily touch points in emergency department

180 160 140 120 100 80 60 40

00

:0 01: 00:3 00 0 2: -1: am 00 30 3: -2:3 am 00 0 4: -3: am 00 30 5: -4:3 am 00 0 6: -5:3 am 00 0 7: -6:3 am 00 0 8: -7:3 am 00 0 9: -8: am 10 00- 30a :0 9:3 m 11 0-1 0am :0 0: 12 0-1 30a :0 1: m 0- 30 1: 12:3 am 00 0 2: -1:3 pm 00 0 3: -2:3 pm 00 0 4: -3: pm 00 30 5: -4:3 pm 00 0 6: -5: pm 00 30 7: -6:3 pm 00 0 8: -7:3 pm 00 0 9: -8: pm 10 00- 30p :0 9:3 m 11 0-10 0pm :0 :3 0- 0p 11 m :3 0p m

20

Fig. 5.6  Half-hourly demand at emergency department by time of day

5  OPTIMIZATION, SIMULATION AND PREDICTIVE ANALYTICS… 

111

day. In particular, the ratio of nursing staff to touch points is quite high in the middle of the night and during the early morning (03:00 to 06:00) and very low between 10:00 and 14:30. A shift-capacity planning model should minimize the mean deviation of overall workloads over 48 half-­ hour time intervals. The following issues would be particularly important: (i) How many shifts (both major and minor shifts) should be instituted? (ii) What would be the right shift times? and (iii) How many nursing staff should be allocated to each shift? The model should assist hospital decision-­makers in their capacity planning. Following this, a mixed integer programming model could be developed to determine the optimal numbers of nursing staff required in the six shifts, so as to minimize the mean deviation of the overall workload over the different time intervals in the day. Numerical Experiment A review was conducted of a total of 41,231 patients using three-months’ patient administrative data revealing approximately 448.2 daily patient attendances on average. Our analysis was based on half-hourly intervals in a 24-hour day. Patient changes to touch points in ED were captured when staff updated patient locations in the ED information system. To estimate the average touch points in each time interval, we examined patient flow pathways by identifying sequences of patients’ touch points from triage to exit. At each time interval, we calculated the average total touch points at the key areas in the ED. By assigning different weights to areas according to suggestions from clinicians, we derived the demand over the 48 half-­ hour time intervals in a day. In the numerical study, the current nursing staff capacity is 60, with a mean deviation of the workload ratios being 0.194. We considered factors concerning shift-capacity planning from a practical perspective, such as capacity requirements of manpower supply in each time interval or in each shift, shift times and total number of minor shifts. As shown in Table 5.4, we considered nine different scenarios and evaluated various shift-capacity planning rules under these different considerations. ED managers could assess the feasibility of capacity planning solutions according to respective constraints, along with other possible considerations in practice. Table 5.5 shows the optimal solutions of capacity planning rules associated with the shifts under different scenarios from the mixed integer programming model. Figure 5.7 presents mean deviations of workloads and the improvements in mean deviation compared with the current shift planning. Overall, the solution derived in scenario 9 appears to be the most promising rule

112 

J. CHANCHAICHUJIT ET AL.

Table 5.4  Different scenarios in numerical study Scenario

Maximum number of minor shifts

Minimum number of nurses over time intervals

Starting time of minor shifts

1 2 3 4 5 6 7 8 9

No requirement 3 2 1 2 2 2 2 3

No requirement No requirement No requirement No requirement No requirement No requirement No requirement No requirement 13

No requirement No requirement No requirement No requirement One shift starts at 10:00 One shift starts at 10:30 One shift starts at 11:00 One shift starts at 11:30 One shift starts at 10:30

for shift-capacity planning. In particular, the shift times ­concerning minor shifts seem ideal and the reduction rate in the mean deviation of the workload is considerably high, at 46%, compared with other reduction rates under consideration. Decision-makers can use the information as a reference in their shift-capacity planning. 5.3.3  Case Study 3. Analysis of Patient Waiting Time Governed by a Generic Maximum Waiting Time Policy Background Patients often need to wait before being attended to, due to various reasons such as limited capacity, variable demand and inefficient operations management. Long waiting times have been recognized by hospital managers as a challenge in health services management. In particular, studies have shown that waiting times in the ED accounted for more than 50% of total patient turnaround time (Boyle et al., 2012; Hwang & Concato, 2004). Waiting times in hospitals are associated with patient satisfaction and, more importantly, morbidity and mortality. In Singapore, waiting times at various stages of hospital consultations are closely monitored by the Ministry of Health of Singapore. For example, the daily median waiting time for admission to wards in each public hospital in Singapore is an important quality indicator monitored by the ministry. Hospital managers are keen to reduce waiting times by introducing interventions at various points. A variety of methods have been used to predict waiting times, such as quantile regression and discrete event simulation (Mowen,

AM shift PM shift Night shift Minor shifts

Scenario 13 13 13 5 7 9 0 0 0 0 0 0

13 2 2 2 3 2 1 1 6 2

Scenario 2

13 13

Scenario 1

13 12 9 0 0 0 0 0 0 0

13 13

Scenario 3

13 13 0 0 0 0 0 0 0 0

21 13

Scenario 4

13 3 11 0 0 0 0 0 0 0

20 13

Scenario 5

12 0 0 0 0 0 0 0

2

13

20 13

Scenario 6

Table 5.5  Number of nurses per shift by the model under different scenarios

13 1 12 0 0 0 0 0 0 0

21 13

Scenario 7

13 1 12 0 0 0 0 0 0 0

21 13

Scenario 8

13 10 8 5 0 0 0 0 0 0

14 10

Scenario 9

5  OPTIMIZATION, SIMULATION AND PREDICTIVE ANALYTICS… 

113

114 

J. CHANCHAICHUJIT ET AL.

Fig. 5.7  Mean deviations of workload and their reductions compared with the current mean deviation

Licata, & McPhail, 1993; Siciliani & Hurst, 2005; Sinreich & Marmor, 2005; Sun et al., 2012). However, most methods do not take into consideration how the waiting time will be affected when interventions are introduced. Without knowing the impact of different interventions on the waiting time, we cannot tell which interventions are effective. Queue management can be found implicitly or explicitly in hospitals at many points. For instance, in EDs, patients are often triaged and prioritized before they are seen by the physicians. Nevertheless, some lower-­ priority patients may have extremely long waiting times if priority rules are followed strictly. In practice, these patients are often seen within a defined period. For instance, in Denmark, elective patients are entitled to treatment at a private hospital locally or at a hospital abroad if the local public healthcare system is unable to provide treatment within the stated maximum waiting time guarantee. We believe that this kind of maximum waiting time policy could be applied to or integrated with other healthcare programmes as well, such as chronic care.

5  OPTIMIZATION, SIMULATION AND PREDICTIVE ANALYTICS… 

115

Model Description In this study, we considered a generic maximum waiting time policy to ensure that patients experiencing a wait longer than a given time period would be processed within a target time period. Figure 5.8 illustrates the original waiting time distribution and the maximum waiting time policy. Of particular interest was the interaction between the new mean waiting time (or the variance) and time points of the target time period and their values, gathered using the patient administrative data. Patient waiting behaviour was governed by the policy, using a piecewise distribution function. For patients who experienced a wait beyond the given time point were considered under two processing strategies, (i) processing using a constant rate and (ii) processing using an increasing rate, within the target time period, resulting in two different waiting distributions, one piecewise discontinuous distribution and one piecewise continuous distribution. Figure 5.9 demonstrates the transformed waiting distribution of two different cases under consideration. As opposed to the usual negative exponential distribution assumptions in the literature, the original waiting time (i.e. prior to the intervention) was assumed to follow general phase-type (PH) distributions.

Fig. 5.8  An illustration of original waiting distribution and maximum waiting time policy

0

Density

J. CHANCHAICHUJIT ET AL.

Density

116 

b-c

b Waiting time

Case I. Discontinuous piecewise distribution

0

b-c

b

Waiting time

Case II. Continuous piecewise distribution

Fig. 5.9  An illustration of transformed waiting distribution

PH distributions describe probability distributions of the time to absorption in a finite state Markov chain, with one or more transient states and one absorbing state (Asmussen & Nerman, 1996; Fackrell, 2009; Faddy & McClean, 1999; Olsson, 1996; Vincent & Paul, 2012). The distributions can approximate continuous probability distributions with arbitrary precision. From this can be derived closed form expressions concerning new mean waiting times, new variance and time points of the threshold period. We estimate the parameters in PH distribution by fitting PH distributions to patient waiting time data using an expectation maximization algorithm, which is then used to assess the threshold period, new mean waiting time and its variance, according to the developed analytical formulae. Numerical Experiment In the study, PAC3 (patient acuity category) patient data from an ED in Singapore from November 2011 to February 2012 was used. There are four levels of patient acuity, with PAC1 being the most serious and PAC4 the least. The main reason we chose PAC3 patient data is that this group of patients accounted for a large proportion of attendance, compared with other categories. Waiting time under consideration refers to the time from the end of nurse triage to the start of consultation by a doctor. Basic statistics of PAC3 waiting time data are shown in Table 5.6. To estimate the parameters in PH distributions, the expectation maximization (EM) algorithm was adopted along with the software package EMpht to execute the EM algorithm. To compare different fitting results,

5  OPTIMIZATION, SIMULATION AND PREDICTIVE ANALYTICS… 

117

1 phase to 14 phases in PH fitting were examined. Four popular measures of “goodness of fit” were used for model selection in the study, that is, log-likelihood (logLik), R2, Akaike Information Criterion and Kullback-­ Leibler divergence. By analysing the goodness of fit with different choices of phases in fitting, the preferred phase of the PH distribution that had an optimal fit was phase 12. With different choices of the threshold such as the 95th and 75th percentile, mean and median of the data, the new mean and variance with the obtained PH approximation were estimated as shown in Table 5.7. Table 5.6  Descriptive statistics of patient waiting time data 1 3 5 7 9 11 13 15 17 19 21 23 25

Characteristics Sample size Minimum 25th percentile Median Mean 75th percentile 95th percentile Maximum Standard deviation Variance Coefficient of variation Skewness

2 4 6 8 10 12 14 16 18 20 22 24 26

Value 27689 0 21.1 (minutes) 40.1 (minutes) 50.6 (minutes) 69.6 (minutes) 128.3 (minutes) 321.8 (minutes) 39.1 (minutes) 1528.7 0.772 1.454

Table 5.7  Mean and standard deviation of new waiting time under various scenarios Threshold

128 128 128 70 70 70 40 40 40

Length of interval

50 40 35 30 20 10 20 15 10

Discontinuous model

Continuous model

Mean

Standard deviation

Mean

Standard deviation

47.7 47.6 47.6 38.2 39 40.1 262 27.3 28.5

32.6 32.4 32.5 18.8 19.5 20.6 9.1 9.3 10.2

47.3 47.6 47.7 38.9 39.7 40.5 27.7 28.5 29.3

32.0 32.5 32.7 19.4 20.2 21.0 9.6 9.8 10.2

118 

J. CHANCHAICHUJIT ET AL.

5.4   Conclusion Healthcare decision-making problems are becoming more complicated and complex today. OR techniques can help address a diverse range of challenges inherent in decision-making in healthcare. For example, there may be uncertainty in clinical, operational and administrative performance measures, along with increasingly complex healthcare processes. OR methodologies and techniques afford the opportunity to translate the increasing amount of patient data together with uncertain healthcare problems into improved evidence-based decision-making and outcomes. As the healthcare industry moves towards a system organized around what patients and populations need, proactively engaging in OR-based solutions has the possible potential to provide information for decision-­makers, such as population health screening and treatment plans, patient flow in healthcare facilities, personalized medicine and chronic disease prevention. OR methods can guide healthcare stakeholders to make better-­ informed decisions by capturing the objectives, decisions and constraints associated with healthcare challenges using the appropriate set of mathematical, statistical and computational methods for target-oriented models.

References AJPH. (1952). Operations research and public health. American Journal of Public Health and the Nations Health, 42(10), 1306–1307. Alagoz, O., Hus, H., Schaefer, A. J., & Roberts, M. S. (2010). Markov decision process: A tool for sequential decision making under uncertainty. Medical Decision Making, 30(4), 474–483. Asmussen, S., & Nerman, O. (1996). Fitting phase-type distributions via the EM algorithm. Scandinavian Journal of Statistics, 23(4), 419–441. Ben-Tal, A., & Nemirovski, A. (1998). Robust convex optimization. Mathematics of Operations Research, 23(4), 769–805. https://doi.org/10.1287/ moor.23.4.769 Bertsimas, D., & Sim, M. (2004). The price of robustness. Operations Research, 52(1), 35–53. https://doi.org/10.1287/opre.1030.0065 Boyle, A., Beniuk, K., Higginson, I., & Atkinson, P. (2012). Emergency department crowding: Time for interventions and policy evaluations. Emergency Medicine International, 2012(2012), 838610–838610. https://doi.org/10. 1155/2012/838610 Burke, E., De Causmaecker, P., Berghe, G., & Van Landeghem, H. (2004). The state of the art of nurse rostering. Journal of Scheduling, 7(6), 441–499. https://doi.org/10.1023/B:JOSH.0000046076.75950.0b

5  OPTIMIZATION, SIMULATION AND PREDICTIVE ANALYTICS… 

119

Bursch, B., Beezy, J., & Shaw, R. (1993). Emergency department satisfaction: What matters most? Annals of Emergency Medicine, 22(3), 586–591. https:// doi.org/10.1016/S0196-0644(05)81947-X Capan, M., Khojandi, A., Denton, B. T., Williams, K. D., Ayer, T., Chhatwal, J., … Zaric, G. (2017). From data to improved decisions: Operations research in healthcare delivery. Medical Decision Making, 37(8), 849–859. Cayirli, T., & Veral, E. (2003). Outpatient scheduling in health care: A review of literature. Production and Operations Management, 12(4), 519–549. https:// doi.org/10.1111/j.1937-5956.2003.tb00218.x Cheang, B., Li, H., Lim, A., & Rodrigues, B. (2003). Nurse rostering problems– –A bibliographic survey. European Journal of Operational Research, 151(3), 447–460. https://doi.org/10.1016/S0377-2217(03)00021-3 Chen, X., Sim, M., & Sun, P. (2007). A robust optimization perspective on stochastic programming. Operations Research, 55(6), 1058–1071, 1191, 1193. https://doi.org/10.1287/opre.1070.0441 Chin, R., You, A.  X., Meng, F., Zhou, J., & Sim, K. (2018). Recognition of schizophrenia with combinatorial regularized support vector machine and sequential region of interest selection using structural magnetic resonance imaging. Scientific Reports, 8, 13858. Delage, E., & Ye, Y. (2010). Distributionally robust optimization under moment uncertainty with application to data-driven problems. Operations Research, 58(3), 595–612, 768, 772. https://doi.org/10.1287/opre.1090.0741 Denton, B. T., Alagoz, O., Holder, A., & Lee, E. K. (2011). Medical decision making: Open research challenges. IIE Transactions on Healthcare Systems Engineering, 1(3), 161–167. https://doi.org/10.1080/19488300.2011.619157 Denton, B., Murat, K., Nilay, S., Bryant, S., & Smith, S. (2009). Optimizing the start time of statin therapy for patients with diabetes. Medical Decision Making, 29(3), 351–367. Dowsland, K. A., & Thompson, J. M. (2000). Solving a nurse scheduling problem with knapsacks, networks and tabu search. Journal of the Operational Research Society, 51(7), 825. https://doi.org/10.1057/palgrave.jors.2600970 Fackrell, M. (2009). Modelling healthcare systems with phase-type distributions. Health Care Management Science, 12(1), 11–26. https://doi.org/10.1007/ s10729-008-9070-y Faddy, M., & McClean, S. (1999). Analysing data on lengths of stay of hospital patients using phase-type distributions. Applied Stochastic Models in Business and Industry, 15(4), 311–317. Goh, J., & Sim, M. (2010). Distributionally robust optimization and its tractable approximations. Operations Research, 58(4), 902–917, 1029, 1032. https:// doi.org/10.1287/opre.1090.0795 Harper, P. (2002). A framework for operational modelling of hospital resources. Health Care Management Science, 5(3), 165–173. https://doi.org/10.102 3/A:1019767900627

120 

J. CHANCHAICHUJIT ET AL.

Harper, P. R., & Gamlin, H. M. (2003). Reduced outpatient waiting times with improved appointment scheduling: A simulation modelling approach. Quantitative Approaches in Management, 25(2), 207–222. https://doi. org/10.1007/s00291-003-0122-x Harper, P. R., & Shahani, A. K. (2001). Modelling for the planning and management of bed capacities in hospitals. Journal of the Operational Research Society, 53(1), 11. https://doi.org/10.1057/palgrave.jors.2601278 Hu, Y. C., Chen, J. C., Chiu, H. T., Shen, H. C., & Chang, W. Y. (2010). Nurses’ perception of nursing workforce and its impact on the managerial outcomes in emergency departments. Journal of Clinical Nursing, 19(11–12), 1645–1653. https://doi.org/10.1111/j.1365-2702.2009.02999.x Hwang, U., & Concato, J. (2004). Care in the emergency department: How crowded is overcrowded? Academic Emergency Medicine, 11(10), 1097–1101. https://doi.org/10.1197/j.aem.2004.07.004 Jaumard, B., Semet, F., & Vovor, T. (1998). A generalized linear programming model for nurse scheduling. European Journal of Operational Research, 107(1), 1–18. https://doi.org/10.1016/S0377-2217(97)00330-5 Kurt, M., Denton, B. T., Schaefer, A. J., Shah, N. D., & Smith, S. A. (2011). The structure of optimal statin initiation policies for patients with type 2 diabetes. IIE Transactions on Healthcare Systems Engineering, 1(1), 49–65. https://doi. org/10.1080/19488300.2010.550180 Librero, J., Marn, M., Peir, S., & Munujos, A. V. (2004). Exploring the impact of complications on length of stay in major surgery diagnosis-related groups. International Journal for Quality in Health Care, 16(1), 51–57. https://doi. org/10.1093/intqhc/mzh008 Mason, J. E., Denton, B. T., Shah, N. D., & Smith, S. A. (2014). Optimizing the simultaneous management of blood pressure and cholesterol for type 2 diabetes patients. European Journal of Operational Research, 233(3), 727–738. https://doi.org/10.1016/j.ejor.2013.09.018 Meng, F., Teow, K. L., Ooi, C. K., Soh, C. K. K., & Tay, S. Y. (2016). Shift capacity planning for nursing staff in emergency department using mixed integer programming. Pacific Journal of Optimization, 12(3), 635–648. Meng, F., Teow, K. L., Teo, K. W. S., Ooi, C. K., & Tay, S. Y. (2019). Predicting 72-hour reattendance in emergency departments using mixed integer programming via discriminant analysis with electronic medical records. Journal of Industrial and Management Optimization, 15(2), 947–962. Mowen, J.  C., Licata, J.  W., & McPhail, J. (1993). Waiting in the emergency room: How to improve patient satisfaction. Journal of Health Care Marketing, 13(2), 26–33. Olsson, M. (1996). Estimation of phase-type distributions from censored data. Scandinavian Journal of Statistics, 23(4), 443–460.

5  OPTIMIZATION, SIMULATION AND PREDICTIVE ANALYTICS… 

121

Rais, A., & Viana, A. (2010). Operations research in healthcare: A survey. International Transaction in Operational Research, 18(1), 1–31. Siciliani, L., & Hurst, J. (2005). Tackling excessive waiting times for elective surgery: A comparative analysis of policies in 12 OECD countries. Health Policy, 72(2), 201–215. https://doi.org/10.1016/j.healthpol.2004.07.003 Sinreich, D., & Marmor, Y. (2005). Ways to reduce patient turnaround time and improve service quality in emergency departments. Journal of Health Organization and Management, 19(2), 88–105. https://doi.org/10.1108/ 14777260510600022 Sun, Y., Teow, K. L., Heng, B. H., Ooi, C. K., & Tay, S. Y. (2012). Real-time prediction of waiting time in the emergency department, using quantile regression. Annals of Emergency Medicine, 60(3), 299–308. https://doi. org/10.1016/j.annemergmed.2012.03.011 Thompson, D. A., & Yarnold, P. R. (1995). Relating patient satisfaction to waiting time perceptions and expectations: The disconfirmation paradigm. Academic Emergency Medicine, 2(12), 1057–1062. https://doi.org/10.1111/j.1553-2712. 1995.tb03150.x Trinkoff, M. A., Johantgen, L. M., Storr, P. C., Gurses, P. A., Liang, P. Y., & Han, P. K. (2011). Nurses’ work schedule characteristics, nurse staffing, and patient mortality. Nursing Research, 60(1), 1–8. https://doi.org/10.1097/ NNR.0b013e3181fff15d Vincent, A. K., & Paul, R. H. (2012). Modelling emergency medical services with phase-type distributions. Health Systems, 1(1), 58. https://doi.org/10.1057/ hs.2012.1

CHAPTER 6

Innovative Health Technologies and Start-­ Ups Process in Healthcare Industry

Abstract  This chapter discusses an innovative health technology and development process in the healthcare industry, challenges and trends in the healthcare system, and health technologies for the future. The chapter concludes by recommending methodologies for developing health technology to start-ups and entrepreneurs. Keywords  Health technology • Digital healthcare • Personalized healthcare • Precision medicine • Nano medicine • Emerging healthcare model

6.1   Introduction to the Health Technology The advancement of technology has allowed improvements in the quality of life of people in many different ways. Many solutions to unmet needs have been introduced from the technologies and innovations developed by entrepreneurs and visionaries. The adoption of many innovations has also transformed lifestyles and changed the way people live. This can happen when a new technology is introduced. For example, the introduction of cars and planes has reimagined transportation, and the introduction of the internet and cell phones has changed the speed with which and the way in which people communicate and connect. These changes and this disruption, referred to as “technology waves”, have been observed during © The Author(s) 2019 J. Chanchaichujit et al., Healthcare 4.0, https://doi.org/10.1007/978-981-13-8114-0_6

123

124 

J. CHANCHAICHUJIT ET AL.

past decades. In the modern world, technology and innovation are t­ ypically seen as a crucial and essential factor in the economic prosperity of countries or in any economic system (Hargroves & Smith, 2006). Since the 1800s, there have been more advances in technology and in different technology waves including iron, water power, steam power, railroad, electricity, petrochemicals and aviation. Currently we are living through the fifth and sixth waves of technology, which include digital technology, biotechnology and information technology. Many of these technologies could potentially give rise to innovative technologies that could address many problems in the world. As the global population increases, health problems will increase and become more complex. As healthcare is a basic need, many healthcare challenges have been listed as a top priority to be addressed and solved. Healthy living is a most desirable state, and good healthcare should be accessible to everyone. However, with limited resources in finance, manpower and supplies, the healthcare industry is now facing numerous issues that result in poor health in many undeveloped countries. With new waves of technology and innovations, healthcare is one of the sectors that will benefit greatly and can be transformed by the new and innovative health technologies in the near future. As explained by the WHO, “health technology” is the application of organized knowledge and skills in the form of devices, medicines, vaccines, procedures and systems developed to solve health problems and improve quality of life. These include the pharmaceuticals, devices, procedures and organizational systems in healthcare (Department of Essential Health Technologies, 2011) that can be used and applied in solving the challenges in the current healthcare situation. Many healthcare problems are mostly due to the rise in world population, ageing, poor resource management and new emerging diseases. Health technology was then developed to promote efficiency in treatment and address unmet needs, thereby improving the healthcare system. Since healthcare is a human basic need, it should be able to be accessed when needed. However, immediate and personalized healthcare also comes with a high cost, and it is mainly available to the privileged. For instance, a costly mobile medical unit on the President of the USA’s private plane, Air Force One, is equipped with world-class medical facilities to ensure the safety of the President, with doctors flying on every flight (Business Insider). Similarly, a mobile medical unit is provided for Simon Cowell (net worth £325m), producer of American Idol, The X Factor and several music bands, at a cost of around $250,000 in order for him to stay healthy while on the road. On the other

6  INNOVATIVE HEALTH TECHNOLOGIES AND START-UPS PROCESS… 

125

side of the spectrum, many developing countries do not have basic health facilities or enough healthcare professionals to provide good healthcare services. For example, patients in a small village in Botswana need to travel very long distances to the capital city of Gaborone for diagnosis and consultation with a specialist, or just to replenish their medicine. The disparity in access to healthcare could not be greater, and that is the reason why many health technologies should be developed to promote accessibility at a lower cost or at no charge. Many technologies in the digital healthcare or telemedicine area could provide an inexpensive solution, and could increase accessibility and reduce the arduous commute that causes trauma, inconvenience and additional expenditure for sick patients. Other than the patients and health professionals, governments and other stakeholders in healthcare could also benefit from innovative solutions or new health technologies in terms of reduction in expenditure, promoting better health and increasing efficiency in healthcare management. For example, wearable health monitoring systems could serve in promoting preventative healthcare, and encourage people to maintain their health and reduce the risks affecting quality of life. Biosensors and rapid diagnosis could help warn the public and minimize infection during a pandemic. In this chapter, many types of promising technologies, along with their applications and their benefits, will be considered along with the commercialization process. The unavoidable transformation of the health industry which is leading to a new value chain and Healthcare 4.0 will also be discussed.

6.2   Challenges and Trends in the Healthcare Systems When it comes to the healthcare industry, every service, decision, management aspect and product is heavily regulated and required to comply with many legal and industrial standards as the safety of lives is at stake. Regulators, professional organizations and governments have tightened regulatory pathways to ensure the safety of patients and the people who provide the services. At the same time, these stringent requirements have placed various business operations under great pressure and many have struggled to adapt. For example, a decline in drug discovery was partly due to the tightened regulatory pathways in illustrating safety and efficacy profiles (Scannell, Blanckley, Boldon, & Warrington, 2012). In addition,

126 

J. CHANCHAICHUJIT ET AL.

the rise in population and an increase in new diseases are also factors ­adding to the challenges to the healthcare system. Health technology and innovation are seen as potential solutions that will help increase efficiency and address many other problems. It is therefore important for innovators and entrepreneurs to understand the current challenges and problems in the global healthcare system so that they may address unmet needs with the correct solutions and technology. Some of the main factors and trends that may have given rise to problems in the healthcare system are summarized and investigated further in this section. These include the rise in healthcare expenditure, life expectancy increases, the rise of technology and higher expectations, and the shortage of healthcare professionals. 6.2.1  Rising Healthcare Expenditure According to OECD data (2018), healthcare spending has risen at rates greater than GDP in most OECD countries. Many had average health spending exceeding 9% of their GDP in 2009, an increase of around 9% from the previous year (Sorenson, Drummond, & Bhuiyan Khan, 2013). US health expenditure alone is expected to continue increasing by an average of 5.8% annually through to at least 2024 (Morgan, 2015). Since most OECD countries have universal health coverage systems which promote equitable access to necessary health services, having relatively high and growing health expenditure is putting many governments under pressure. There are many factors believed to drive healthcare spending. Other than the increase in population, complexity of disease and illness has led to more advanced research and development in the areas of better treatments and technologies. Greater availability of magnetic resonance imaging, computerized tomography, coronary artery bypass grafts, angioplasty, cardiac and neonatal intensive care units, positron emission tomography, and radiation Oncology facilities is associated with greater per capita use and higher spending on these services (Bodenheimer, 2005). The use of improved and advanced healthcare technologies generally increases healthcare costs rather than reducing healthcare expenditures (Songul Cinaroglu, 2018). However, improved outcomes and quality of life are the magnets drawing people to pay for these technologies. Innovative technologies and new treatments are typically diffused and adopted faster in high-spending healthcare systems. In this case, the optimization for usage and strict health technology assessment should be performed before the approval of many technologies by reimbursement schemes. Many high-spending

6  INNOVATIVE HEALTH TECHNOLOGIES AND START-UPS PROCESS… 

127

countries, including the USA, that face rises in healthcare expenditure have started to embrace a tightened policy to drive down healthcare costs (Lorenzoni, Belloni, & Sassi, 2014), spend more efficiently, and reduce costly and unnecessary treatments. Although quality of care generally improves along with cost, governments and the funders are looking to balance the quality of treatment with spending. Investing in cost-effective health promotion interventions and innovation is one important way to improve value for money and reduce health inequities (OECD, 2018). 6.2.2  Life Expectancy Increase According to the WHO, most people can expect to live longer than 60 years, and the population of people above age 60 could reach two billion by 2050, compared to 900 million in 2015 (WHO, 2018). The majority of older people will live in low- and middle-income countries. Factors such as income, lifestyle behaviours, education and geographical influences have been investigated for a correlation with life expectancy (Mathers, Stevens, Boerma, White, & Tobias, 2015; Chetty et al., 2016). In high-­ income countries such as the USA, higher income was associated with greater longevity, and the differences in life expectancy were correlated with health behaviours and local area characteristics. Currently, life expectancy at birth among the OECD countries is 80.6 years on average. Japan and Spain lead a group of 25 OECD countries with life expectancies over 80 years (OECD, 2018). An increase in life expectancy can be attributed to technology (Lichtenberg, 2017), advanced healthcare facilities and health knowledge that allows healthcare professionals to provide suitable and effective treatments. At the same time, people can monitor and look after their health easily and efficiently. This helps to reduce mortality rates as seen in diseases such as circulatory diseases. Since people have been living in a healthier manner, there has been a fall in deaths, with 50% fewer deaths from ischaemic heart disease, on average, since 1990. Cancer mortality rates have also fallen, though less markedly, by 18% since 1990 (OECD, 2018). Since people are living longer, they are likely to be faced with chronic diseases and many health problems. Good physical and mental health for healthy living in older people is important, but at the same time incurs expenditure and involves high maintenance. Innovation and technologies

128 

J. CHANCHAICHUJIT ET AL.

focusing specifically on this group are being encouraged by many countries in keeping their older people healthy and active. 6.2.3   Rise of Technology and High Expectations Nowadays, the internet, social media and mobile phones have improved communications and allowed us to search for and have access to information. This encourages people to become more health-conscious and educated through the transfer of knowledge, information sharing and communication. Healthcare systems are now placed under close scrutiny by society, and patients prefer to be involved in decision-making. Numerous healthcare bodies such as regulatory, health technology assessment, academic and healthcare providers have started to incorporate patient inputs into their decision-making processes. This signifies a shift in healthcare system from “disease-centred” to “patient-centred” (du Plessis et al., 2017). Many healthcare professionals view patient-centred care to be an important aspect of high-quality care (Snyder et al., 2011). This set-up allows them to customize the treatment, preventative measures and other care to fit the needs of the individual. In many private settings, personalized treatment or patient-centred care has been offered because the care provider can spend more time with the patient and utilize technology to design custom care for those high-spending patients. However, in a resource-­ limited setting, the time per patient and technology utilization can be limited, which leads to difficulty in offering personalized care for everyone. Implementation of this premium care solution for the public is limited by the high cost involved. Health technology can contribute greatly to personalized healthcare. For example, health informatics has the potential to facilitate and offer a mechanism for patients to provide their clinicians with critical information, and to share information with family, friends and other patients. This information may enable patients to exert greater involvement and take more control over their own care (Snyder et al., 2011). With the development of new technologies towards personalization of healthcare, people can expect better quality of care. At the same time, the pressure on healthcare providers increases, and they must adapt to the change and start the transition of healthcare to a demand-driven model. For example, the Cleveland Clinic in Ohio has a clear mission to improve patient experience, and it has a board-level Chief Experience Officer lead-

6  INNOVATIVE HEALTH TECHNOLOGIES AND START-UPS PROCESS… 

129

ing the Office of Patient Experience with a mission to place the patient at the centre of everything they do. We can expect hospitals to adapt and align with this trend as patients begin to exercise their rights in selecting the options for their healthcare such as the facility, protocols or procedure for their care, and they will demand access and transparency of data and processes. As a consequence, healthcare organizations will need to focus on how quality outcomes can be published in a meaningful way for patients (INC., 2014). Patient safety is the major focus of patient advocacy groups and healthcare leaders. They will enforce higher care standards and instigate deeper investigations into such things as medication errors, hospital-­ acquired infections or wrong site surgery. 6.2.4  Shortage of Healthcare Personnel Medical graduate enrolments have been falling in many countries. According to the WHO, there were 2.4 million too few physicians, nurses and midwives to provide essential health interventions in 2007. The situation became more serious as the WHO reported the same shortage of 7.2 million in 2013 with a projected significant increased shortage of 12.9 million healthcare workers by 2035 (Global Health Workforce Alliance and World Health Organization, 2013). This trend and these projections have called for global attention and a demand to revoke the trend. The potential shortage of doctors and other medical staff may well increase the cost of healthcare or result in a reduction in service quality. Many developed countries have the potential to attract doctors, nurses and healthcare staff from other countries through high-quality training, education and better career growth and compensation. The opportunities available to health workers in seeking employment abroad have led to a complex migration pattern, characterized by a flow of health professionals from low- to high-income countries (Aluttis, Bishaw, & Frank, 2014). This trend of migration may solve a problem in one country but poses more risks to countries that already have a serious shortage of doctors, nurses and healthcare workers. Many have raised questions about the consequences for health systems worldwide, including questions about sustainability, justice and global social accountability. One of the major factors in the healthcare worker shortage is the increase in health service demands due to the population increase and ageing society. Many countries, including the US, have failed to produce doctors, nurses and professionals to match growth, as can be observed in

130 

J. CHANCHAICHUJIT ET AL.

Fig. 6.1  Graduates of allopathic medical schools in the USA, 1980–2005 (Salsberg & Grover, 2006)

the  near-zero growth in US medical school graduates (see Fig.  6.1) (Salsberg & Grover, 2006). Many professional organizations are joining together at the Third Global Forum on Human Resources for Health to address both shortages and the uneven distribution of healthcare workers. Recommendations on actions to address workforce shortages in the era of universal health coverage include: 1. Increased political and technical leadership in countries to support long-term human resource development efforts; 2. Collection of reliable data and strengthening of human resources for health databases; 3. Maximizing the role of mid-level and community health workers to make frontline health services more accessible and acceptable; 4. Retention of health workers in countries where the deficits are most acute and greater balancing of the distribution of health workers geographically; and 5. Provision of mechanisms for the voices, rights and responsibilities of health workers in the development and implementation of policies and strategies towards universal health coverage.

6  INNOVATIVE HEALTH TECHNOLOGIES AND START-UPS PROCESS… 

131

As there is a shortage of healthcare workers and professionals, health technologies could play a greater role in assisting with and reducing the workload of healthcare providers, allowing them to serve and work more efficiently. The integration of multiple technologies in life science, medical devices and digital health is the key to a solution that drives efficiency in the workplace and reduces the physical hours of the workers, as well as leading to greater personalization. Another angle of technology development is to focus on preventative healthcare, which can reduce the number of patients seeking treatment as well as unnecessary visits to hospitals. At the same time, the promotion of education using health technology can be another solution allowing medical schools to produce more graduates and professionals to meet the demands of society.

6.3   Health Technology for the Future As the WHO describes it, the purpose of health technology is to solve health problems and improve the quality of people’s lives. There are many technologies that have begun to be adopted and that have started to make an impact upon the healthcare system. Digital healthcare and personalized healthcare are the two large fields of health technologies that will be discussed in this section. These two sectors of health technology have been imagined and visualized in many fictional films and books. As these technologies are being developed, glimpses of these thought-to-be fictional worlds are becoming a reality. For example, the advancement of digital diagnosis and genomic testing, together with AI, could assist many doctors in providing rapid and accurate diagnosis in a short amount of time. Medical professionals can be consulted via telemedicine and customized drugs or biologics obtained for treatment: these innovations are becoming the norm and may well be our future. In addition to this, the use of controllable nanorobots to fight cancer cells and effect internal cures is being trialled and implemented. These rapidly changing health technologies are being discussed in more detail, along with opportunities and challenges for real-life applications and the commercialization process. 6.3.1   Digital Healthcare  igital Technology and Healthcare D For some time now, industries and businesses have been moving from traditional processes towards digitization. Digital technology is being

132 

J. CHANCHAICHUJIT ET AL.

integrated into many aspects of our lives, shifting values and changing the way we move from traditional physical processes towards digital processes. The first wave of digital technology was the internet, which gave individuals access to a variety of information. This was followed by the mobile internet, smartphones and tablet devices respectively. The era of the “Internet of Things (IoT)” was idealized in order to get many devices plugged in and connected to the internet and ready for future commands and data acquisition. However, with the limitations of current communication technology, the full deployment of IoT may not be complete until the arrival of 5G technology. Within the IoT world, there will be a massive increase in the number of internet-enabled devices. Tiny embedded sensors and computers in equipment, machinery and devices will link to the Cloud to generate new value for industry and society. We will start to see robotics and AI being used and incorporated more into our daily lives. For example, we already have applications that utilize drones to inspect agricultural areas using AI, and from the images and data acquired decisions are made on suitable plants to be planted. AI may also be used in screening and assisting in making recommendations for the suitable treatment of complex diseases such as TB (see Chap. 4). The healthcare industry has also realized the power of digital health technologies to address many existing problems in healthcare. As mentioned earlier, the benefits of adopting these emerging technologies for Healthcare 4.0 should help empower patients and push the drive to provide them with better care. Many of the benefits in adopting digital healthcare such as IoT, blockchain, AI, big data and mobile applications have been mentioned previously in this book (Chap. 1). In 2017, the digital health industry was already worth $25 billion globally, with the potential to cut healthcare costs by an estimated $7 billion a year in the US alone (Duggal, Brindle, & Bagenal, 2018). Nowadays the digitization of healthcare is unavoidable, and the amount of medical knowledge continues to grow, as does the number of new platforms and devices with digital capacity. All stakeholders, especially health providers, must be prepared to foster digital transformation and embrace the new possible disruptive ­technologies in healthcare (Meskó, Drobni, Bényei, Gergely, & Győrffy, 2017) in order to align with future trends in the healthcare industry.  lassification of Digital Health Technology C The term digital healthcare, occasionally referred to as digital health or e-health, has come about since the convergence of science and technology

6  INNOVATIVE HEALTH TECHNOLOGIES AND START-UPS PROCESS… 

133

Direct enduser technologies

- Health portals - Internet sites/app - Wearable consumer technologies

Direct-use gatekeeper technologies

- Metered-dose inhalers - Prescription Drugs

Indirect-user gatekeeper technologies

- Angiography - CT/MRI - Dialysis - Defibrillators

Fig. 6.2  Classification of technologies (Weiss et al., 2018)

- E-Heath/m-Health - Genetics (testing and diagnostics)

in a dynamic digital era. This era has resulted in the development of innovative digital health devices, services and treatments (Bhavnani, Narula, & Sengupta, 2016) allowing digital data acquisition, online access to information, communication and digital analyses. The term is used extensively to refer to all digital technologies that promote and solve problems in the healthcare services. There have been several initiatives to classify these technologies. The focus of the technology is on the user in the current move to a patient-centric era. Daniel Weiss and his team have implemented the approach used by Cotterman and Kumar (Cotterman & Kumar, 1989) to categorize digital health technologies based on end-user control. According to their work (Weiss et al., 2018), the technologies were classified into three different categories: technologies accessed and used directly by the end user (type 1, or direct end-user technologies), technologies used by the end user but accessed through someone other than the end user (type 2, or direct-use gatekeeper technologies), and technologies accessed and used by someone other than the end user (type 3, or indirect-use gatekeeper technologies), as shown in Fig. 6.2. According to Weiss, the direct end-user technologies (type 1) refer to technologies accessed and used directly by the end user. With this technology, patients have direct access, control and use of wearable devices, health information websites, e-health records and communication portals. The direct-use gatekeeper technology (type 2) refers to intermediate technologies that are used by the patient or end user but also accessed through

134 

J. CHANCHAICHUJIT ET AL.

someone other than the end user such as relevant physicians, care providers and other stakeholders. These technologies monitor, screen and acquire data to support healthy living of patients and provide alerts for treatments if needed, sharing several similar aspects with the type 1 technologies. However, the access of patients is more limited by the gatekeeper, and the end user typically has less control than the previous type. Lastly, indirect-use gatekeeper technologies (type 3) refer to the technologies that will be used by physicians or other stakeholders, and the end users do not have control, or have very little access. These technologies are typically the interventions, therapies or diagnoses that are critical to patient health and should be mainly used by certified professionals (gatekeepers). For the new technologies to be introduced or integrated into the system, the developer and the technology provider must clearly understand the patient flow, the indication for use and the current standard treatments in each disease. The plan for implementation and the control for each tier of users mentioned in the category above should be considered to visualize the fit of each product and technology into the system. Understanding the nature of users and care providers as well as their interactions is critical. Care providers must comply with clinical guidelines which are regulated heavily. Guidelines must be considered very carefully before being adopted or implemented in any new technology to be used with patients. Many professional care providers’ main concern is the outcomes for their patients. If the developer or technology providers can be clear about the value of the new technology, there is a higher chance for successful adoption.  he Application of Digital Health Technology T As mentioned earlier, digital health technology is broad, ranging from mobile health apps, or m-health, to decisional support systems that use algorithms derived through mining clinical datasets, through to biometric sensors, such as continuous glucose monitoring, consultations via video link (“telemedicine”) and electronic personal health records (Meskó et al., 2017). The implementation of these technologies has been a key driver in the shift in healthcare, and has helped in bringing down the separation between healthcare providers and patients through digital communications. Transparency in information has been enhanced. This is partly benefited by the fact that more tests have become digitally available, and doctors can make evidence-based decisions and share them with patients more easily. With more data, clinical studies and treatment options will

6  INNOVATIVE HEALTH TECHNOLOGIES AND START-UPS PROCESS… 

135

become available for many diseases, and there are more reasons to involve and collaborate with patients to plan for their care and treatment. With this type of engagement, patients are likely to be more cooperative, and comply with procedures and treatment plans. With digital technology, health information and consultations are more easily accessible through web information or telemedicine. This allows patients and many health-­ conscious people to look after themselves better and, in turn, helps reduce the number of unnecessary visits to hospital and reduces workloads. In the USA, the application of digital health technology has been shown to provide cost savings compared to the traditional healthcare system (Ekman, 2018), as well as producing better outcomes for many diseases such as diabetes (Kaufman & Khurana, 2016) and Tuberculosis (Ngwatu et al., 2018). As we have started to see more benefits in digital health technology, many countries are restructuring healthcare and engaging in digital transformation. The UK is among the nations that have made a clear commitment to digital transformation (Breen, Xie, & Cherrett, 2016). The National Health Service (NHS) has started to evaluate the current model of service provision and is trying to shift away from a sickness service into prevention and well-being promotion, which is believed to help reduce significant amounts of expenditure on expensive procedures and equipment (Powell, Newhouse, Boylan, & Williams, 2016). This could be an aspect that health technology can support and bring to fruition with appropriate designs in the healthcare system as well as policies and support from governments, especially in the legal and regulatory areas. With correct execution, digital transformation in healthcare is an opportunity to increase service proficiency and quality of care at a lower cost (Herrmann et al., 2018). However, successful transformation requires substantial initial investment, resources and a high level of commitment. The engagement with users and stakeholders can be an iterative process and often requires education. Ensuring user acceptance, ownership and a culture of data use for decision-making takes time and effort to build human resource capacity (Konduri et al., 2018). Digital transformation will result in the need for a new business model for organizations. Since many health organizations typically do not have large digital R&D units, they can benefit from collaborations with start-up companies to keep investment low and off the balance sheet. At the same time, the regulatory knowledge of established corporations or organizations might help these start-ups to

136 

J. CHANCHAICHUJIT ET AL.

kick off digital disruption in the healthcare sector (Herrmann et al., 2018) and create successful outcomes for all parties. 6.3.2  Personalized Healthcare In recent years, personalization and customization have been introduced into many industries. The revolution of digital technology and data sciences allowed business to offer tailored solution and products, driving more sales and business efficiency. Personalization has been a trend implemented in healthcare for more than 10 years. It has been introduced and recognized under different terms such as “personalised medicine”, “personalised health care (healthcare)”, “P4 Medicine—Predictive, Personalised, Preventive/Preventive, and Participatory Medicine”, “individualised medicine”, “precision medicine” and “systems (bio)medicine” (Cesuroglu et al., 2016). The personalization of care is typically extremely expensive and therefore only available to the privileged classes. However, with advancements in digital technology, the personalization of healthcare is becoming more and more affordable and much more manageable. When we look closely at organizations that offer healthcare personalization, there are four essential components in its implementation and execution: (1) customer relationship management, (2) big data and IT uses, (3) training in managerial and data response competencies and (4) patient engagement (Minvielle, Paccaud, Peytremann-Bridevaux, & Waelli, 2017). Many digital health technologies were designed to enhance one of these capacities. The most important aspect of a personalized healthcare system is the insight gained from data that is accessible. The flow of the data has to be designed carefully from acquisition, to data processing, to analyses, through to visualization and recommendation. Treatments and care for complex diseases are among the first candidates in personalized healthcare where a great impact can be made. Cancer, genetic diseases and blood transfusions have been shown to benefit from personalization. In addition, the emphasis on preventative and precision medicine is gaining more attention, and these are key drivers of personalized medicine. We have started to see more and more wearable IoT devices that provide ­customized lifestyle, dietary and exercise plans through their websites or applications, as people become more health-conscious. In this section, several key personalized approaches and important technologies contributing to healthcare personalization are highlighted with examples.

6  INNOVATIVE HEALTH TECHNOLOGIES AND START-UPS PROCESS… 

137

 ext Generation Sequencing Enabling Precision Medicine N The concept of using genetic information to help personalize patient care and treatment has been around for many years. DNA and genetic information are unique and specific to an individual; important data can be utilized to design customized care and treatment to improve the outcome of the patient. Using Next Generation Sequencing (NGS), an entire human genome can be sequenced within a single day, which is much faster than Sanger sequencing, first-generation sequencing, which could take years to deliver the same results (Behjati & Tarpey, 2013). This technology has opened doors to evidence-based medicine which promotes effective decision-­making in healthcare and is associated with improved patient outcomes. NGS is powerful and is allowing the transformation of the diagnosing of gene-related diseases (Tarailo-Graovac, Wasserman, & Van Karnebeek, 2017). In addition, it has been used in blood typing to identify all the antigens at the surface of red blood cells for truly personalized blood transfusion (Johnsen, 2015) and to prevent alloimmunization, which can be deadly. With NGS, the promise of today is that a complete genome can be sequenced in a few days for less than $1000 per genome. Even though we are not there yet, the implications and the impact of NGS in understanding the biological processes of diseases like cancer and in personalizing patient care are unprecedented (Kamps et al., 2017). Although NGS has proven to be advanced and far superior to conventional Sanger sequencing, it has not yet been incorporated widely in personalized healthcare for all diseases, with the exception of cancer. As cancer is a genetic disease caused by mutations, NGS will have a significant impact on the detection, management and treatment of the disease (Meldrum, Doyle, & Tothill, 2011). For example, sequencing can help to identify mutation detection in inherited cancer syndromes based on DNA-­ sequencing, detection of spliceogenic variants based on RNA-sequencing, DNA-sequencing to identify risk modifiers and applications for pre-­ implantation genetic diagnosis, cancer somatic mutation analysis, pharmacogenetics and liquid biopsies (Kamps et  al., 2017). For NGS to be efficiently translated to clinical or hospital settings, there are several requirements, including the infrastructure to handle large and complex data and facilities for sample collection and testing. The data derived from NGS is typically complex and there is a need for acquiring and analysing complex data. This requires genomic experts capable of regulating and managing genome-wide data and making high-quality, accurate data interpretation, in close collaboration with clinicians, to interpret and produce

138 

J. CHANCHAICHUJIT ET AL.

meaningful results. While NGS sequencing technology, and computational, and clinical resources continue to be improved, the preparation and readiness of healthcare providers has not been moving at the same speed (Tarailo-Graovac et al., 2017). Cancer treatment has been transformed by the ability to read the whole genome in a short period of time at a relatively lower cost, together with advancement in the development of small molecule inhibitors and antibodies against specific gene targets (Meldrum et  al., 2011). The use of these therapeutic agents could be enhanced more effectively in combination with companion diagnosis such as NGS. This would help identify the suspected genes, allowing the clinician to make appropriate decisions and to choose the right agents for a more positive prognosis. This approach is known as precision medicine, and seeks to use genomic data to help provide the right treatment to the right patient at the right time. NGS will continue to provide more insights and unlock potential to personalize treatment in complex genetic diseases. It will be interesting to see how healthcare providers will adapt to this technology and the changes in other disease diagnosis and treatment. It is hoped that other disciplines will follow Oncology’s example and embrace NGS change. Precision Medicine As mentioned earlier, the rise of genetic sequencing and analysis has unlocked the potential for physicians to customize treatment in complex diseases in order to achieve the best outcome. This approach has been recognized as “precision medicine”, where healthcare delivery relies mainly and heavily on data analytics and information. The field is largely inclusive of all data available, of which genetic data is a part. Many consider it as a branch of “personalised medicine” or personalized healthcare, but the difference is that personalized medicine still recognizes the preferences, beliefs, attitudes, knowledge and social context of patients more than does precision medicine (Ginsburg & Phillips, 2018). Therefore, precision medicine is important in treating diseases that are complex and unpredictable. Cancer treatment is a good example and candidate for precision medicine because it is a very complex disease with a wide range of treatments available. However, drugs or chemotherapy may have failed to improve the outcome of patients where treatment was not specific enough to the subtype. Cancer treatment is costly, and bypassing ineffective treatments would reduce cost and patient trauma. Using as much data as is

6  INNOVATIVE HEALTH TECHNOLOGIES AND START-UPS PROCESS… 

139

Fig. 6.3  Old and new healthcare paradigm with digital health technology. Source: Deloitte Healthcare Solutions (Taylor, 2015)

available would help provide more accurate prognoses as to outcomes and may increase the chance of successful treatment. An example of the implementation of precision medicine is illustrated by Garralda and his team (Garralda et  al., 2014). Their approach (see Fig. 6.3) was an integration between exome sequencing and Avatar mouse models to select drug candidates for the treatment. To start the process, patients had an exome characterization of tumours and normal tissue and bioinformatic analysis to determine the most biologically-relevant somatic mutations, while an Avatar mouse model from the same patient was created. The drug candidates initially suggested by genomic analysis were later tested in the Avatar mouse to narrow down the best and most effective drug candidates for patients. A retrospective study reported that this approach revealed evidence of a correlation between the drug response of candidate treatments in Avatar models and the clinical response, and helped to select empirical treatments in some patients with no actionable mutations. However, the application of precision medicine in a mainstream healthcare system is still yet to be discussed and needs to be approved by the regulators. Despite this, there is a high potential for better outcomes in many diseases. We should see more research and ­development in this area as well as more innovative treatment approaches in the near future (Fig. 6.4). Many countries including the USA have been considering utilizing precision medicine in everyday practice. The launch of the United States

140 

J. CHANCHAICHUJIT ET AL.

Fig. 6.4  A personalized treatment approach tailored by the integration of exome sequencing and Avatar mouse models (Garralda et al., 2014)

Precision Medicine Initiative is due to the belief that this method of practice will help to reduce costs. At the same time, other technology such as the adoption of EMRs, the growth of data science and AI (see Chap. 4) and other genome-based technology platforms have become available. However, the success of the implementation of precision medicine also relies on components other than the technology. Depending on the design of the treatment scheme, a lot of data, such as genomic data, clinical information and patient preferences could be incorporated into the clinical decision-making process, but the acquisition of this data results in costs to the patient and insurer. Without conclusive proof of the efficiency and need for data acquisition, the insurer or payers might hesitate and not allow the reimbursement of costs in their policy schedules. Similarly to treatments and drugs, the funder wants to see proof of improved outcomes such as tumour shrinkage response rates in cancer treatment or a lessening of side effects (Morash, Mitchell, Beltran, Elemento, & Pathak, 2018) (Fig. 6.5). In order for precision medicine to be adopted, it requires input and collaboration between all the stakeholders: physicians, patients, payers (insurers) and regulators. Successful treatment that promotes improved outcomes benefits all stakeholders, and therefore the outcome evaluation

6  INNOVATIVE HEALTH TECHNOLOGIES AND START-UPS PROCESS… 

141

Fig. 6.5  Outline of precision medicine in Oncology (Morash et al., 2018)

of each patient should not only be updated on their treatment history but also be communicated to healthcare policymakers. Once enough evidence is amassed to show the clear benefit of a certain treatment, changes in healthcare policy should follow, resulting in new clinical guidelines which will affect the physicians’ consultation in designing care, the types of treatments that health insurance policies cover, and the cost of treatment to the patient (Morash et al., 2018). Wearable Technology Over the last decade, the design and development of wearable biosensor systems for health monitoring has gained increasing attention from the industry and been welcomed by health-conscious people who can afford such systems. These technologies include wearable devices with embedded sensors and analytic algorithms which allow monitoring and analysis to provide personalized behaviour guides to improve the quality of health of the user (Schüll, 2016). There have been many investments and developments through start-up companies, as well as large electronics companies, that see opportunities in this area to transform the future of healthcare by enabling proactive personal health management and monitoring of a patient’s health condition. In previous years, there have been major developments in physiological sensors, transmission modules and processing capabilities, and this sort of progress facilitates low-cost wearable unobtrusive solutions for continuous health status and monitoring being devel-

142 

J. CHANCHAICHUJIT ET AL.

oped and commercialized (Pantelopoulos & Bourbakis, 2010). This field is largely driven by the knowledge that preventative care reduces the prevalence of disease and helps people live longer, healthier lives. It has provided significant cost reductions and increased the cost-effectiveness of patient care (costing between $50,000 and $100,000 per Quality Adjusted Life Year: Cohen, Neumann, & Weinstein, 2008; Neumann & Cohen, 2009). Many healthcare systems and providers are embracing the idea, and hence this type of technology is immensely desirable. Wearable and biosensor systems have transformed traditional views of diagnosis, and created a unique opportunity that focuses on identifying problems much earlier, due to real-time monitoring. In order to fully integrate the systems and move towards preventative and personalized healthcare, current wearable systems are now being utilized in mobile electronic devices. These can be embedded in the user’s outfit or used as part of their everyday clothing or as an accessory (such as watches, socks, gloves, headbands, glasses, pads or tattoos). They can be made operational and accessed without or with very little hindrance to user activity (Lukowicz, Kirstein, & Tröster, 2004). Although the future of wearable technologies in the real healthcare industry is still being discussed, the technologies have been applied to teaching facilities and institutes to foster and promote teaching and training in decision-making for the trainees. For example, wearable technology such as “Google Glasses” was used to explore different scenarios in cardiovascular practice where practitioners could learn from the video streaming of data transmitted through a smartphone, tablet or computer. This arrangement allows trainees to be exposed to many cases, and to share their knowledge and experience. However, privacy and confidentiality of patients must be maintained (Vallurupalli, Paydak, Agarwal, Agrawal, & Assad-Kottner, 2013). 6.3.3   Nanorobotics and Nanomedicine Nanotechnology is a science that focuses on small particles and has been an important driver of change in many industries. Together with biomedical sciences, the field of nanomedicine derived from both sciences focuses on the development and creation of materials and devices designed to interact with the body at sub-cellular scales with a high degree of specificity. This technology could potentially offer specific targeted therapeutic drugs or targeted sub-cellular treatment that would present locally limited adverse effects. Nanomedicine is believed to be important in many life-­

6  INNOVATIVE HEALTH TECHNOLOGIES AND START-UPS PROCESS… 

143

threatening diseases such as cancer, cardiovascular problems, lung, blood and neurological diseases, diabetes, inflammatory/infectious diseases, and Parkinson’s or Alzheimer’s disease in which the current gold standard treatment is still not sufficient (Saha, 2009). Fiction and film have portrayed nanomachines that can be introduced into our bodies to fix certain areas or cure diseases. In the near future, we may see some of these devices emerge from the fictional into the real. Along with the rise of nanorobotics, there is an emerging field of nanotechnology which focuses on developing and designing functional atomic, molecular cellular scale devices that can be controlled to achieve specific purposes. These nanorobots will be miniature and small enough to travel within the human body through blood vessels. They will be equipped with sensors to detect disease and assist in navigating towards the target areas where diagnosis and/or treatment can be performed at the site diseases of the targeted molecules (Manjunath & Kishore, 2014). For example, nanorobots such as respirocytes, microbivores and clottocytes are being designed to act as artificial substitutes for blood. The respirocytes are designed to mimic all the important functions of red blood cells and are also used in the treatment of anaemia, heart attack and lung diseases where the clottocyte mimics the natural process of haemostasis and the microbivore follows the process of phagocytosis to destroy blood-borne pathogens. With all this potential, many have projected that the nanorobot might be available for mainstream use by 2030. Currently, many researchers have been making great progress in discovering methods of motorizing and controlling the movement of nanorobots, for example with the assistance of magnetic fields (Tierney et al., 2011) or the use of natural materials such as sperm to deliver drugs to target cervical cancer tumours (Xu et al., 2018). These designs and visions are maturing at an extremely rapid pace, which means they will be materializing into real-world applications soon. Thus, healthcare may soon be executed in a completely different manner to that of today.

6.4   Development of Health Technology and Healthcare Start-Ups As the healthcare system is facing many challenges, innovative technologies are being developed and introduced into the system to address these problems (as mentioned in section 6.2). In parallel, the new technologies

144 

J. CHANCHAICHUJIT ET AL.

have also caused disruption in healthcare by shaping new forms of treatment, diagnoses and analyses. Healthcare is being unavoidably transformed; therefore health technologists, entrepreneurs and researchers must be aware of the innovative disruptions, technology landscapes and regulatory requirements, and they must have a good business plan and strategy in mind. In this section, we will discuss and introduce several components that are important in development innovation solutions and technology in the healthcare industry. 6.4.1  Smart Healthcare Industry: Future Value Chain in Healthcare Smart healthcare is the beneficiary of digital technologies adopted into the healthcare system. Digital transformation is slowly disrupting the traditional views of the healthcare process by shifting the focus from physical value chains towards digital value. The technologies highlighted in this book including IoT, big data, Machine Learning or AI, and health technologies such as wearable devices and innovative diagnosis, are playing their part in completing the picture of digital transformation. Smart healthcare is seen as the integration and flow of health data that allows care providers to diagnose and treat patients more effectively. This practice of information and data-sharing between patients and health service providers is believed to offer improved treatment to patients, and enhances the quality of their lives. A cost-effective and sustainable healthcare information system relies on the ability to collect, process and transform healthcare data into information, knowledge and action (Demirkan, 2013). Digital and health technologies will become important pieces in the “jigsaw” and help integrate fragmented physical value chains, allowing a complete flow of health data. As we can see nowadays, large physical facilities or devices are now being replaced by more efficient mobile and compact devices with IoT capability and cloud-connection. The integration and high connectivity allows borderless collaboration, and creates a potential global digital value chain in the healthcare system. In digital transformation and digitally connected value chains in several fields, we have started to see that customization and personalization is critical. For example, dental implantation has always been faced with the challenge of an inaccurate transfer where a physical impression is taken and then transferred to a gypsum cast, which might suffer shrinkage and distortion of the impression materials or unsta-

6  INNOVATIVE HEALTH TECHNOLOGIES AND START-UPS PROCESS… 

145

ble repositioning of the analogue during the laboratory process (Christensen, 2008). Digital technologies, along with material 3D printing technologies, have transformed the patient journey, and the stakeholders can now work together globally and digitally. This means that a patient can have a digital scan in a dentist’s office in Argentina, and the data can be uploaded to the cloud to be viewed by a surgical expert in the USA and a manufacturer in India simultaneously. All stakeholders will be able to collaborate and work on the design of the products (customized 3D-printed implants and a drill guide) together before they are created and sent back for use in surgery in the dental office in Argentina. A similar trend has been found in orthopaedic implants, in which patients could benefit from higher precision and quality of care. Many leading companies have realized this disruptive change and have already been developing a platform that integrates relevant physical value chains in proven significant fields. As connectivity increases, patient engagement and healthcare can start outside hospitals. The integration between healthy living, prevention, diagnosis, treatment and homecare has been adopted into the development strategies of large companies such as Philips. This type of smart healthcare should promote inclusive and personalized care, allowing people to receive treatment when they need it, as well as reducing waste in the future. Technology and Innovation Development Processes 6.4.2   in Health Technology As society desires more advancement and better technology, the path towards technology development and commercialization still remains difficult, time-consuming and full of challenges. A typical development process normally starts with fundamental research, and later evolves through product development, proof-of-concept, regulatory approval, clinical acceptance, reimbursement and marketing, as shown in Fig. 6.6. The largest and most common hurdle in developing technology and innovation is known as “the valley of death”. For many years, people believed that the most important and the only “valley” that the innovator or entrepreneur needed to cross was that of achieving regulatory approval. As the health technology sector has grown, getting clinical approval and acceptance, or someone to financially back your technology, has proven to be challenging and much harder than previously thought. The uniqueness of health technology in development is that there are two major mile-

146 

J. CHANCHAICHUJIT ET AL.

Fig. 6.6  Physical value chains are disrupted and transformed by digital technologies (digital value chain: https://slideplayer.com/slide/10496307/)

stones or two “valleys of death” (see Fig. 6.6) that every entrepreneur has to traverse. This means that in addition to regulatory approval, a reimbursement plan should be in place at an early stage. The success of disruptive health innovations will rely on good regulatory and reimbursement strategies in order to enter the target countries. To create effective plans, the developer must have a thorough understanding of the classification, legal requirements and indications for use of their products. For example, developing a medical device product or a drug-device product (see Fig. 6.7) can be different from developing a pharmaceutical product (see Fig. 6.8). When we talk about health technologies, the majority of the technologies and innovations are mostly considered to be medical devices, and will follow the path shown in Fig. 6.7. As illustrated, the expertise required to drive the technology development ranges from R&D, engineering and

6  INNOVATIVE HEALTH TECHNOLOGIES AND START-UPS PROCESS… 

Fundamental research

Fundamental research

Product development

Valley 1

Regulatory approval

Regulatory approval

Clinical acceptance

Reimbursement

147

Marketing

Implementation

Valley 2

--> time

Fig. 6.7  A typical health technology (medical device and pharmaceutical products) development pathway (https://www.ttopstart.com/news/the-occurrenceof-a-second-valley-of-death-during-medical-device-development; https://www. slideshare.net/AmberHolHoreman/design-for-reimbursement-in-medicaldevice-development-50344101)

clinical research through to insurance and reimbursement insight, as well as production and marketing. The components of the skills needed must align with the growth and maturity of the technology in the development roadmap, and the start-up model seems to be one of the best fits for developing disruptive technologies. When we focus on the two “valleys of death”, the resources to overcome the two valleys require large increases in capital investment to gain clinical acceptance by generating reliable data and evidence for the public. For regulatory approval, a high level of safety for humans must be developed in both medical devices and pharmaceutical products (see Fig. 6.8). Typically, the regulator would require safety and efficacy evidence from three phases of clinical trials. Due to the complexity of the biological system in our body, the trial sizes of groups for testing pharmaceutical products are usually greater than those for medical devices. It may be advisable to discuss strategy and planning with the appropriate regulator if there is such an option. This may help avoid unnecessary costs when designing a clinical trial. After the clinical evidence is submitted and reviewed by a regulator, such as the USFDA, the developer will need to plan for the technology being adopted by physicians or its use included in guidelines. At the same time, engagement with the payers, insurer or policymakers cannot be neglected. This level of engagement needs good strategy and support from clinical field influencers, along with guidance from experienced people. Typically, in the industry, up to five years of safety profiles are required

148 

J. CHANCHAICHUJIT ET AL.

Fig. 6.8  Development of a medical device from fundamental research to marketing with the two “valleys of death” (https://www.ttopstart.com/news/theoccurrence-of-a-second-valley-of-death-during-medical-device-development; https://www.slideshare.net/AmberHolHoreman/design-for-reimbursement-inmedical-device-development-50344101)

and these need to be demonstrated to the community before they may want to make decisions to adopt a new technology as the new “gold standard”. However, the process can be longer or shorter depending on the safety risks to the users, as well as other factors such as the number of alternative technologies, current clinical unmet needs and the urgency of the need for treatment of the particular disease. As a developer, incorporation of these factors in the development would help create a clear roadmap to follow, and allow preparation for any changes in the ecosystem.

6  INNOVATIVE HEALTH TECHNOLOGIES AND START-UPS PROCESS… 

149

6.4.3  Regulatory and Legal Requirements The healthcare industry is one of the most heavily regulated, as people’s lives and safety are at stake. Laws and regulations are generally created to ensure safety and reduce risks from unethical acts by providers and business operators. However, strict and tight regulations have been shown to slow down innovation and technology development. There must be balance on both sides, as regulators and medical professional associations try to factor into their decisions that the use of technology can help improve outcomes. Regulators from different regions may take different approaches. For example, European regulators have a tendency to focus mainly on safety, performance and technology, while the USFDA has factored health benefits into their considerations. In health technology, the US market is one of the largest and most attractive single markets, with many physicians being eager early adopters. However, US regulators are known to be among the most stringent in their approval of medical devices and pharmaceutical industry products. This has prompted the USFDA to provide more approaches to promote innovation, such as De Novo submission. This trend is now being seen in other countries and regions. Once technology has passed the R&D phases, if it is proven feasible in terms of technology, patentability and marketing, the next step that developers will take is to obtain clinical approval which encompasses such areas as complying with safety standards, providing necessary clinical evidence and post-marketing surveillance plans. There are many factors which should be considered and seriously planned for. The first thing that a developer should do with regard to regulatory strategy is to understand the competitive landscape, select a market to enter and decide on which regulatory pathway to follow. A typical classification or regulatory pathway for the same product in one country or region could be very different to that of another. It should be clear from the beginning whether the product will be classified as a drug or a medical device, which agency is responsible for it and what sort of timeline is needed for the process. The USA is one of the countries with numerous technologies being developed, and they have high a technology adoption rate. The USFDA has been trying to keep abreast of such technologies, particularly as the arrival of the digital healthcare era has increased the use of software, IoT, AI and (micro) electronics. The FDA is one of the fastest regulators to introduce guidelines for these new digital technologies. Figure 6.9 illustrates the evolution of the USFDA towards the new era of Healthcare 4.0.

150 

J. CHANCHAICHUJIT ET AL.

Fig. 6.9  Development pathway of pharmaceutical products (https://steveblank. com/2013/08/19/reinventing-life-science-startups-evidence-based-entrepreneurship/)

Traditionally, the FDA mainly focuses on the safety and efficacy of technologies for human use and application. Quality assurance must be of a high standard. This means that products must be produced, stored and distributed by certified manufacturers and operators. The validation or evidence to confirm compliance with all the standards during all the processes must be demonstrated. With the IoT and the digital era, the FDA must also concern itself with cybersecurity, safety and security in software, data handling and connectivity. Figure 6.10 shows an example of the relevant standards and compliance needed to commercialize and meet regulatory requirements for an electronic device with software, sometimes known as an electromedical device. In addition to safety evidence, following the biocompatibility and performance tests of each product, any product considered as an electronic medical device needs to comply with electronic safety regulations, such as IEC 60601-1. Figure  6.10 provides an overview of the connection between all the standards in three different categories: quality management (ISO 13485: manufacturing, ISO 14971: risk management), software lifecycle (IEC 62304: life cycle management) and electronic safety (IEC 60601 series). Several regulatory bodies following the risk management concept would like to see the risk profile of a product reduced according to all three pillars. A concept of design control which includes regulatory inputs when designing a product to ensure regulatory compliance could be implemented to break down potential risks to be addressed by the product’s designer or technology developer to achieve the level of compliance required by the regulator. Other than regulatory requirements, professional policy and legal requirements must also maintain currency and be updated to keep pace

6  INNOVATIVE HEALTH TECHNOLOGIES AND START-UPS PROCESS… 

151

A shift in how the FDA thinks about regulating medical devices

Evolving

Traditional

Traditional considerations meet technology

Safety

Is a medical device safe for use in humans? Does it cause adverse events? Are its risks tolerable in relation to its benefits?

Efficacy

Is a device effective for its given purpose? What is the magnitude of the effect?

Quality

After approval, a device must be kept safe and effective through adherence to quality manufacturing standards established by FDA

Security

Once a medical device is networked with other devices or the internet, is it still safe, or is it vulnerable to potentially serious problems?

Fig. 6.10  The regulatory path and approach of the USFDA to regulate medical devices (https://www.slideshare.net/Healthegy/breakout-session-cybersecurityin-medical-devices)

with the changes in technology. For instance, telemedicine platforms have been developed to increase accessibility to healthcare and reduce costs. However, legal liability and insurance policies have been slowed, thus hindering the technology from offering all its benefits. Healthcare systems that change fast enough can enjoy the benefits that these technologies can bring to the system. The USA is among the first of the countries that have a high rate of implementation due to many positive factors. The use of telemedicine requires FDA approval as well as professional organizations to advocate and form new laws regarding this new technology. After indications for use are set, the FDA and the national medical council can set up new guidelines, regulations and requirements around the terms. In the USA, a special licence is required in many states for physicians who want to perform telemedicine, or who wish to issue telemedicine licences. The recognition of telemedicine as “medical treatment” is increasing, and this will allow patients to receive coverage from funders. More laws will be implemented to cater to and address changes as more disruptive technologies are introduced. For example, privacy laws and regulations have forced service providers to obtain patient consent before using or releasing any pertinent information. These situations have taught us that when introducing or developing a new disruptive

152 

J. CHANCHAICHUJIT ET AL.

IEC 60601-1 Security of Electromedical Devices

ISO 13485 QMS

Requires

ISO 14971 Risks Management

Affects Affects

Requires

Design and maintenance of software in MD

IEC 60601-1-2 EMC of Electromedical Devices

Affects

IEC 62304 Software Lifecycle

Fig. 6.11  Regulatory standards for an electromedical device with software (https://blog.cm-dm.com/post/2013/04/12/MD-and-IVD-standards%3AIEC-60601-1-and-IEC-61010-1%2C-versus-IEC-62304-Part-2)

technology, new pathways must be forged and legal issues dealt with for successful commercialization. Start-Ups and Entrepreneurship in Health Technology 6.4.4    ealth Technology and Innovation Landscape H As mentioned earlier, innovation in health technology is known to be time-consuming, challenging and highly regulated. The development of traditional pharmaceutical and medical device paths is becoming more difficult and comes with many challenges. The rise of digital technology and innovation has been proven to add value to the health industry, and the start of digital transformation is already occurring. Many entrepreneurs see unique opportunities in digital technologies, and this is leading to a growing trend in start-ups and investment in digital innovation for healthcare. In 2017, 17% of venture capital was invested in digital health technology companies by 11% of companies (see Fig.  6.12), ranging from technology such as EHR, Web destinations and communication to the healthcare and medical community, telehealth, genomics and insurance payments. A Venture scanner monitored health tech start-ups in 2017,

6  INNOVATIVE HEALTH TECHNOLOGIES AND START-UPS PROCESS… 

153

Innovation quadrant for health technology ESTABLISHED EHR/EMR

HEAVYWEIGHTS

Clinical Adminstration Population Health Healthcare Robotics

Average Age

Healthcare Marketing Patient Engagement Communications Healthcare Communities teleHealth Gamification PIONEERS Health Apps

IoT Health

Remote Monitoring Medical Big Data Doctor Network Nutrition Healthcare Search

Medical Device

Insurance/Payments

Health Destination Sites IoT Fitness

Genomics DISRUPTORS

Average Funding

Fig. 6.12  Health technology innovation quadrant when considering average funding and average age of the technology (https://www.slideshare.net/ NathanPacer/venture-scanner-health-tech-report-q3-2017)

and suggests a division of health tech using two parameters, average funding and average ages, as shown in Fig. 6.11. There are several disruptive technologies being developed, such as genomics, medical devices, insurance, payments, health destinations and IoT. These are relatively new technologies with large investments required but with the promise of making a great impact or potentially changing certain aspects of healthcare. However, smaller capital intensives defined as “pioneers” are seen on a larger scale. These include popular healthcare communities, telehealth, IoT health and medical big data. The search for effective healthcare is very popular with entrepreneurs, but it can result in “red ocean space” (cut-throat competition) which requires different strategies and game plans (Fig. 6.13). Roland Berger projected that the digital health market will grow at an average of 21% per annum from 2015 to 2020, with a market size of 206 billion US dollars. It is definitely a sector that the investor is interested in, but it requires a unique knowledge of development plans and commercialization. Mobile health will continue to grow rapidly and is a key growth

154 

J. CHANCHAICHUJIT ET AL.

Investing into the sector is occurring across a variety of categories The Digital Medical Devices category accounts for 17% of venture funding and 11% of total companies in the health technology sector

VC Funding

Count

300 250 200 150 100

Company Count

50 t s s s s a h e h g n es ies ng ics lth rk ps te lth es ent tio Dat arc car en alt MR ice rin e /E ev to Si ea two Ap evic nit keti bot hea a tn e u r i n H lth gem H e h Fi ym tr o ig S T /Pa inis l B are Hea nga ion HR al D on atio tele r N ealt al D mm Ma e R n of o E ic t I ce m a c M r o ct le H edic Co are hca atio d ote estin n Ad edic alth IoT nt E pula o e a c e c i t r al r l ifi D ob D tie Po M He su l M ca lth a l M em Pa In linic M gita alth Hea He am ita R alth e i G ig C H D He D

-

Di

gi

tal

M

ed i

ca lD ev G ices en om ic s

Total Funding ($B)

Venture Investing in Health Technology 10 9 8 7 6 5 4 3 2 1 0

Fig. 6.13  The venture capital investment profile in the heath technology sector (Pacer, 2017)

driver in this sector. In addition to the capital required, the speed to complete the milestones and building traction for capital investments still remain challenging for the entrepreneur. Venture capital investors typically expect a large return, and therefore the technology or innovation idea should be something that will have a large impact. This means that the idea must address a real unmet need or a major problem in the sector. Discussion with potential customers, users and even acquisitors can be helpful in screening the idea or developing the plan, and it can allow the developer to understand and formulate the value of their technology. Without a real potential need, an entrepreneur may wish to change the plan or terminate the project rather than running the risk of developing a technology that does not have a clear value proposition. Since start-ups comprise smaller teams with innovative minds, they are a good vehicle for inventing and coming up with new technologies quickly, compared to large multinational corporations. However, these corporations still want to innovate and come up with new technology to maintain

6  INNOVATIVE HEALTH TECHNOLOGIES AND START-UPS PROCESS… 

155

their competitive advantages. In addition to VC, the corporations that are already in the industry are also potential investors or may be partners with start-ups as well. If the entrepreneur is able to align needs and maximize commonly shared interests, the multinationals could add great value to developments, especially in the marketing phase where a large salesforce and numerous distribution channels will be needed. These partnerships can come in different forms, depending on the strategy of the start-ups and the portfolio of the technologies, to create a win-win situation. Or they could potentially be an acquisitor of the company, and place these new technologies in their pipeline. Either way, there is still a lot of interest and potential in the health technology sector. If the entrepreneur or start­up knows how to navigate and orchestrate all the components correctly, they have a high chance of being successful in this area.

6.5   Conclusion There is a need for disruptive and innovative technologies to address many unmet needs to improve the outcome of patients and the health of people in general. At the same time, the traditional healthcare system is facing challenges and is struggling to maintain a high standard, and healthcare expenditure is a burden in many countries. Longer life expectancy and higher expectations of healthcare treatment are driving providers to seek improved technologies and solutions. The use of digital technology in offering personalized care is one of the main solutions in the new era of healthcare, and it will digitally transform all areas of healthcare including preventative care, genomics and precision medicine. There are multiple challenges involved in commercializing these technologies successfully for users, including regulatory approval and reimbursement adoption. Due to the intensity of capital investment and know-how, start-up models are among the most suitable routes for developing many of these technologies. Digital health technology will still remain an area of interest for the investor and service provider. The entrepreneur and the start-up must align with and orchestrate stakeholders and partners to bring about the right skills and create a good strategy. The acquisition of technology is common in health technology and with medical devices, and therefore the entrepreneur should incorporate an exit strategy into their plan and start to initiate potential partners to achieve that goal. With a clear plan and de-risk mechanism, it will be feasible for investors to invest in developing the technology and the company.

156 

J. CHANCHAICHUJIT ET AL.

References Aluttis, C., Bishaw, T., & Frank, M.  W. (2014). The workforce for health in a globalized context—Global shortages and international migration. Global Health Action, 7(1), 23611. Behjati, S., & Tarpey, P. S. (2013). What is next generation sequencing? Archives of Disease in Childhood. Education and Practice Edition, 98(6), 236–238. Bhavnani, S. P., Narula, J., & Sengupta, P. P. (2016). Mobile technology and the digitization of healthcare. European Heart Journal, 37(18), 1428–1438. Bodenheimer, T. (2005). High and rising health care costs. Part 2: Technologic innovation. Annals of Internal Medicine, 142(11), 932–937. Breen, L., Xie, Y., & Cherrett, T. (2016). Where are you? A preliminary examination of the track and trace mechanisms in place to facilitate effective closed-loop medical equipment retrieval in the National Health Service (NHS) (UK). Retrieved from https://bradscholars.brad.ac.uk/handle/10454/8664. Cesuroglu, T., et al. (2016). Other side of the coin for personalised medicine and healthcare: Content analysis of ‘personalised’ practices in the literature. BMJ Open, 6(7), e010243. Chetty, R., Stepner, M., Abraham, S., Lin, S., Scuderi, B., Turner, N., … Cutler, D. (2016). The association between income and life expectancy in the United States, 2001–2014. JAMA, 315(16), 1750–1766. Christensen, G. J. (2008). Will digital impressions eliminate the current problems with conventional impressions? The Journal of the American Dental Association, 139(6), 761–763. Cohen, J. T., Neumann, P. J., & Weinstein, M. C. (2008). Does preventive care save money? Health economics and the presidential candidates. New England Journal of Medicine, 358(7), 661–663. Cotterman, W. W., & Kumar, K. (1989). User cube: A taxonomy of end users. Communications of the ACM, 32, 1313–1320. Demirkan, H. (2013). A smart healthcare systems framework. IT Professional, 15(5), 38–45. Department of Essential Health Technologies, W. H. O. (2011). Health technology assessment of medical devices. du Plessis, D., Sake, J. K., Halling, K., Morgan, J., Georgieva, A., & Bertelsen, N. (2017). Patient centricity and pharmaceutical companies: Is it feasible? Therapeutic Innovation & Regulatory Science, 51(4), 460–467. Duggal, R., Brindle, I., & Bagenal, J. (2018). Digital healthcare: Regulating the revolution. BMJ, 360, k6. Ekman, B. (2018). Cost Analysis of a Digital Health Care Model in Sweden. PharmacoEconomics—Open, 2(3), 347–354. Garralda, E., Paz, K., López-Casas, P.  P., Jones, S., Katz, A., Kann, L.  M., … Hidalgo, M. (2014). Integrated next-generation sequencing and avatar mouse models for personalized cancer treatment. Clinical Cancer Research, 20(9), 2476.

6  INNOVATIVE HEALTH TECHNOLOGIES AND START-UPS PROCESS… 

157

Ginsburg, G.  S., & Phillips, K.  A. (2018). Precision medicine: From science to value. Health Affairs (Project Hope), 37(5), 694–701. Global Health Workforce Alliance and World Health Organization. (2013). A universal truth: No health without a workforce. Third Global Forum on Human Resources for Health Report, p. 104. Hargroves, K., & Smith, M. (2006). The natural advantage of nations: Business opportunities, innovation and governance in the 21st century. Routledge. Herrmann, M., Boehme, P., Mondritzki, T., Ehlers, J. P., Kavadias, S., & Truebe, H. (2018). Digital transformation and disruption of the health care sector: Internet-based observational study. Journal of Medical Internet Research, 20(3), e104. INC., C. G. (2014). Healthcare Challenges and Trends. Johnsen, J.  M. (2015). Using red blood cell genomics in transfusion medicine. ASH Education Program Book, 2015(1), 168–176. Kamps, R., Brandão, R. D., van den Bosch, B. J., Paulussen, A. D. C., Xanthoulea, S., Blok, M. J., & Romano, A. (2017). Next-generation sequencing in oncology: Genetic diagnosis, risk prediction and cancer classification. International Journal of Molecular Sciences, 18(2), 308. Kaufman, N., & Khurana, I. (2016). Using digital health technology to prevent and treat diabetes. Diabetes Technology & Therapeutics, 18(Suppl. 1), S56–S68. Konduri, N., Aboagye-Nyame, F., Mabirizi, D., Hoppenworth, K., Kibria, M. G., Doumbia, S., … Mazibuko, G. (2018). Digital health technologies to support access to medicines and pharmaceutical services in the achievement of sustainable development goals. Digital Health, 4, 2055207618771407. Lichtenberg, F. R. (2017). The impact of biomedical innovation on longevity and health. Nordic Journal of Health Economics, 5(1), 45–57. Lorenzoni, L., Belloni, A., & Sassi, F. (2014). Health-care expenditure and health policy in the USA versus other high-spending OECD countries. The Lancet, 384(9937), 83–92. Lukowicz, P., Kirstein, T., & Tröster, G. (2004). Wearable systems for health care applications. Methods of Information in Medicine, 43(3), 232–238. Manjunath, A., & Kishore, V. (2014). The promising future in medicine: Nanorobots. Biomedical Science and Engineering, 2(2), 42–47. Mathers, C. D., Stevens, G. A., Boerma, T., White, R. A., & Tobias, M. I. (2015). Causes of international increases in older age life expectancy. The Lancet, 385(9967), 540–548. Meldrum, C., Doyle, M. A., & Tothill, R. W. (2011). Next-generation sequencing for cancer diagnostics: A practical perspective. The Clinical Biochemist Reviews, 32(4), 177–195. Meskó, B., Drobni, Z., Bényei, É., Gergely, B., & Győrffy, Z. (2017). Digital health is a cultural transformation of traditional healthcare. Mhealth, 3, 38.

158 

J. CHANCHAICHUJIT ET AL.

Minvielle, É., Paccaud, F., Peytremann-Bridevaux, I., & Waelli, M. (2017). Personalized Medicine: A doorway to an effective health care delivery system? Journal de gestion et d’économie médicales, 35(1), 3–5. Morash, M., Mitchell, H., Beltran, H., Elemento, O., & Pathak, J. (2018). The role of next-generation sequencing in precision medicine: A review of outcomes in oncology. Journal of Personalized Medicine, 8(3), 30. Morgan, L. (2015). US healthcare annual spending estimated to rise by 5.8% on average through 2024. American Health & Drug Benefits, 8(5), 272–272. Neumann, P., & Cohen, J. (2009). Cost savings and cost-effectiveness of clinical preventive care. Ngwatu, B.  K., Nsengiyumva, N.  P., Oxlade, O., Mappin-Kasirer, B., Nguyen, N. L., Jaramillo, E., … Schwartzman, K. (2018). The impact of digital health technologies on tuberculosis treatment: A systematic review. European Respiratory Journal, 51(1), 1701596. OECD. (2018). Health spending (indicator). https://data.oecd.org/healthres/ health-spending.htm Pacer, N. (2017, August). Health technology: Q3 startup market trends and insights. Venture Scanner. www.venturescanner.com Pantelopoulos, A., & Bourbakis, N. G. (2010). A survey on wearable sensor-based systems for health monitoring and prognosis. IEEE Transactions on Systems, Man, and Cybernetics, Part C (Applications and Reviews), 40(1), 1–12. Powell, J., Newhouse, N., Boylan, A.-M., & Williams, V. (2016). Digital health citizens and the future of the NHS. Digital Health, 2, 2055207616672033. Saha, M. R. (2009). Nanomedicine: Promising tiny machine for the healthcare in future-a review. Oman Medical Journal, 32(4), 177–195. Salsberg, E., & Grover, A. (2006). Physician workforce shortages: Implications and issues for Academic Health Centers and Policymakers. Academic Medicine, 81(9), 782–787. Scannell, J. W., Blanckley, A., Boldon, H., & Warrington, B. (2012). Diagnosing the decline in pharmaceutical R&D efficiency. Nature Reviews Drug Discovery, 11, 191–200. Schüll, N. D. (2016). Data for life: Wearable technology and the design of self-­ care. BioSocieties, 11(3), 317–333. Snyder, C. F., Wu, A. W., Miller, R. S., Jensen, R. E., Bantug, E. T., & Wolff, A. C. (2011). The role of informatics in promoting patient-centered care. Cancer Journal, 17(4), 211–218. Songul Cinaroglu, O.  B. (2018). The relationship between medical innovation and health expenditure before and after health reform. Health Policy and Technology, 7(4), 379–387. Sorenson, C., Drummond, M., & Bhuiyan Khan, B. (2013). Medical technology as a key driver of rising health expenditure: Disentangling the relationship. ClinicoEconomics and Outcomes Research: CEOR, 5, 223–234.

6  INNOVATIVE HEALTH TECHNOLOGIES AND START-UPS PROCESS… 

159

Tarailo-Graovac, M., Wasserman, W.  W., & Van Karnebeek, C.  D. M. (2017). Impact of next-generation sequencing on diagnosis and management of neurometabolic disorders: Current advances and future perspectives. Expert Review of Molecular Diagnostics, 17(4), 307–309. Taylor, K. (2015). Connected health: How digital technology is transforming health and social care. Retrieved from https://www2.deloitte.com/content/ dam/Deloitte/uk/Documents/life-sciences-health-care/deloitte-uk-connected-health.pdf. Tierney, H. L., Murphy, C. J., Jewell, A. D., Baber, A. E., Iski, E. V., Khodaverdian, H.  Y., … Sykes, E.  C. H. (2011). Experimental demonstration of a single-­ molecule electric motor. Nature Nanotechnology, 6, 625–629. Vallurupalli, S., Paydak, H., Agarwal, S.  K., Agrawal, M., & Assad-Kottner, C. (2013). Wearable technology to improve education and patient outcomes in a cardiology fellowship program—A feasibility study. Health and Technology, 3(4), 267–270. Weiss, D., Rydland, H. T., Øversveen, E., Jensen, M. R., Solhaug, S., & Krokstad, S. (2018). Innovative technologies and social inequalities in health: A scoping review of the literature. PLoS ONE, 13(4), e0195447. WHO. (2018). Ageing and health. Retrieved November 14, 2018, from https:// www.who.int/news-room/fact-sheets/detail/ageing-and-health. Xu, H., Medina-Sánchez, M., Magdanz, V., Schwarz, L., Hebenstreit, F., & Schmidt, O. G. (2018). Sperm-hybrid micromotor for targeted drug delivery. ACS Nano, 12(1), 327–337.

CHAPTER 7

Transforming and Managing Healthcare Projects

Abstract  In this chapter, we will explain how the process transformation methodology can be used to redesign healthcare processes in order to meet changing customer needs, based on input from customers. This methodology has been in used for many years in Singapore, and some of the successful implementations include the National Library Board, National Healthcare Group and the Port of Singapore Authority. The intention is to achieve drastic improvements in process efficiency, along with a possible reduction in operating costs. In addition to the methodology, we would like to introduce Agile project management for implementing such projects since the requirements from customers and users are bound to change over time. Agile project management will be able to cope with changes in project scope, especially for IT systems. Keywords  Transformation healthcare • Agile methodology • Waterfall methodology • Healthcare project • Assumption surfacing • Migration plan

© The Author(s) 2019 J. Chanchaichujit et al., Healthcare 4.0, https://doi.org/10.1007/978-981-13-8114-0_7

161

162 

J. CHANCHAICHUJIT ET AL.

7.1   Process Transformation The following list shows the key steps in process transformation methodology (Tan, 2013): 1. Assumption surfacing; 2. Challenging assumptions; 3. Idea generation; 4. Idea grouping; 5. Idea evaluation; 6. Ideas integration and digitization; 7. Process validation and implementation; and 8. Project management and change management. To illustrate using this methodology, a specialist private clinic (SPC) example is used (these are still found in some underdeveloped countries). In a typical SPC, a patient must go to different counters in the clinic for registration for different services. Even though the patient has a prior appointment with the specialist, they may need to get an X-ray first. So, they must proceed to the X-ray room, register and wait their turn. This takes at least 30 minutes. After the X-ray, the patient keeps the result and waits to see the doctor. This can take another 30 minutes. Upon seeing the doctor, the consultation must be paid for and a prescription obtained and a new appointment made if necessary. The patient then takes the doctor’s prescription to the pharmacist. A patient may typically spend half a day in a clinic, as illustrated in Fig. 7.1. We will now apply the process transformation methodology to improve SPC process flow. Step 1: Assumption Surfacing 7.1.1   Assumption surfacing involves the identification of stakeholders and the surfacing of assumptions. With respect to any process transformation, the identification of stakeholders and the surfacing of assumptions are of central importance. Process transformation is concerned with changing old ways of thinking, and to achieve these, assumptions will need to be surfaced and challenged. At first glance, this may seem quite simple; one can use the customer’s needs as a benchmark when deciding which changes to make. However, what has not been considered is who is the customer and are the assumptions being made about their wants and needs reasonable.

7  TRANSFORMING AND MANAGING HEALTHCARE PROJECTS 

163

Fig. 7.1  Current process flow for a specialist private clinic

The perceived customer is not necessarily the same as the real customer. It is likely that there will be many stakeholders involved and the most important stakeholder is not necessarily the ultimate customer. In addition to the above, assumptions tend to persist long after they are no longer true, and they may also be “invisible”, unless a deliberate attempt is made to identify and examine them. Outdated assumptions are a major obstacle to making radical changes in the way we do things. We recommend that simple templates as listed below be used in surfacing assumptions (Table 7.1). These templates are meant as a guide in surfacing assumptions and need not be used in the strictest manner. Experienced consultants can surface assumptions without being restricted to them. It is useful at this stage to include team members who are not involved in this process for assumption surfacing. These members will be able to question fundamental assumptions that could have been taken for granted by others. These fundamental assumptions can make a quantum improvement to performance if they can be challenged successfully. You will notice that some of the templates overlap where there are the same assumptions, and this is acceptable as long as they have been listed in any of the templates. Any assumption that can be perceived as an obstacle is included and, if challenged successfully, can achieve the process transformation objectives.

164 

J. CHANCHAICHUJIT ET AL.

Table 7.1  Templates for assumption surfacing WHAT Template  •  Getting cured is important to our patients  •  Collecting payment is important to us  •  Specialized equipment is required for lab tests WHO Template  •  A doctor is required to perform diagnosis/prescribe treatment  •  A patient must be ill before visiting SPC  •  A patient must move from one test station to another WHEN Template  •  A patient must be diagnosed before they can be treated  •  A patient must undergo an initial consultation before taking tests  •  A patient must pay for consultation/treatment before leaving SPC WHY template  •  A patient must undergo tests in order to determine their condition  •  A patient’s condition must be known in order to give them treatment  •  An initial consultation is required to determine what tests are appropriate HOW Template  •  A patient can only be diagnosed before the undertaking of tests  •  A patient can only be tested by taking a blood sample or doing a scan and so on

7.1.2  Step 2: Challenging Assumptions After identifying potential customers through stakeholder analysis, the next step is to question or challenge those assumptions that have been made by the stakeholder group. In some cases, it is not necessary to find out what the customer needs and instead team members may make some assumptions about what they think is important to their customers. We propose that the team members surface as many assumptions as possible in order to generate more ideas at a later stage. After surfacing all the assumptions, the next step is to challenge them by either questioning the relevance of the assumption in today’s context or contradicting it with the aim of a radical improvement in performance. This can be illustrated using the specialist private clinic example (Table 7.2). Step 3: Idea Generation 7.1.3   For each challenge posed, one attempts to generate as many ideas as possible to overcome that challenge. Challenges specify what is to be achieved while ideas specify how they are to be achieved. Brainstorming is one method to keep ideas flowing fast and freely. It is a lateral thinking process

7  TRANSFORMING AND MANAGING HEALTHCARE PROJECTS 

165

Table 7.2  Assumptions and challenges regarding a private clinic WHAT Template Getting cured is important to our patients Collecting payment is important to us

Specialized equipment is required for lab tests WHO Template A doctor is required to perform diagnosis prescribed treatment A specialist technician is required to perform certain tests/scans (e.g. X-rays) A patient must be ill before visiting SPC A patient must move from one test station to another

 • Staying well is important to our patients  • Getting cured quickly and with the least cost, risk and hassle is important to our patients  • Being remunerated is important to us: In money? In other terms? From elsewhere?  • Continuing to exist and to be of service is important to us  • Some tests can be performed with commonly available equipment  • Some diagnoses and treatment prescriptions can be done by other staff doctors not within SPC  • Certain tests/scans can be performed by ordinary nursing staff/staff not within SPC  • A patient can visit SPC on occasions other than when he is ill  • A patient can perform all his tests in one location  • A patient’s movement through all stations should be minimized  • Tests can be brought to the patient  • A patient can be conveyed from one test station to another without effort

WHERE Template A patient can only be diagnosed at the SPC

 • A patient can be diagnosed at home/a location close to home/conveniently accessible to them A patient can only be treated at the  • A patient can be treated at home/a location SPC close to home/conveniently accessible to them A patient must collect medication in  • Medication can be delivered to patient/ person collected from a convenient location A doctor must be at the SPC to  • A doctor can diagnose/treat patient from perform diagnosis/treatment another location within/outside SPC   • A doctor can diagnose/treat patient at a different time than during patient’s visit to SPC WHEN Template A patient must be diagnosed before  • A patient can be diagnosed and treated at the they can be treated same time   • A patient can be treated without diagnosis A patient must undergo initial  • A patient can take tests without initial consultation before tests consultation (continued)

166 

J. CHANCHAICHUJIT ET AL.

Table 7.2 (continued) A patient must pay for consultation treatment before leaving SPC A patient can only fix the next appointment after paying for treatment and consultation WHY Template Patient records must be retrieved the day before to ensure availability at SPC on appointment day A patient must wait at the clinic so that he is available when his name is called (for consulting treatment) so that SPC resources are efficiently utilized A patient must collect and carry their test results with them to ensure availability at main consultation HOW Template A patient can only be diagnosed by going through tests A patient can only be tested by taking a blood sample or doing a scan and so on

 • A patient can make payment after leaving SPC   • A patient can make payments in instalments  • A patient can fix the next appointment while paying for treatment consultation   • A patient can fix the next appointment from home (after leaving SPC)  • Patient records should be available anytime/ all the time  • A patient can wait in more pleasant surroundings or do other things while waiting to be called, without adversely affecting efficiency  • Test results should be available at the main consultation without patient having to bring them along  • A patient can be diagnosed without going through tests/scans  • A patient can be tested remotely   • A patient can test themselves

and is designed to help one break out of one’s thinking patterns and find new ways of looking at things. It works by focusing on a challenge, and then coming up with many creative solutions to it. Ideas should deliberately be as broad and innovative as possible. Group brainstorming can be very effective as it uses the experience and creativity of all members of the group. When individual members reach their limit on an idea, other members’ creativity and experience can take the idea to the next stage. Therefore, group brainstorming tends to develop ideas in more depth as compared to individual brainstorming. During brainstorming sessions, there should be no criticism of ideas. You are trying to open up possibilities and remove outdated assumptions that limit the process performance. Judgements and analysis at this stage will stunt idea generation. Valuable but unusual ideas may appear stupid at first glance. Thus, care must be taken to ensure nobody will crush these

7  TRANSFORMING AND MANAGING HEALTHCARE PROJECTS 

167

new ideas and leave group members feeling embarrassed. Ideas should only be evaluated once the brainstorming session has finished, but do not spend too much time embellishing ideas or trying to build complete solutions. Some ideas for process transformation consideration include: 1. Relocate work to/from patients/SPC • Can SPC do more work on behalf of patients? • Can SPC get the patients to do some of their work? 2. Minimize and simplify interface between patients and SPC • Can interfaces with patients be reduced by automating some of the standard enquiries and simple transactions using automated voice response systems and information kiosks? 3. Task compression and integration for the process • Can manual intervention be reduced so that there will be fewer delays and errors? 4. Concurrent processing • Can activities be done concurrently instead of sequentially? 5. Standardization of process • Can components of services be standardized so that they are capable of being combined to provide products or services with a customer focus? • Can processes be made less complex by limiting the number of alternative inputs? 6. Deferred decision-making • Can products and services be tailored to suit the customer’s needs? 7. Outsource non-value-added activities • Can non-value-added activities be outsourced to third-party providers? Using the specialist private clinic example, the following ideas can be illustrated (Table 7.3).

168 

J. CHANCHAICHUJIT ET AL.

Table 7.3  Ideas generated for a private clinic WHAT Template Getting cured is important to our patients

Collecting payment is important to us

Specialized equipment is required for lab tests

 • Staying well is important to our patients  • Getting cured quickly and with the least cost, risk and hassle is important to our patients  • Being remunerated is important to us— in money? In other terms?

 • Some tests can be performed with commonly available equipment

 • Establish preventive care programme—new line of business  • Better patient education and counselling regarding treatment options

 • What are alternative sources of funding to supplement health insurance and patient’s own funds?  • Continued provision of high standard of affordable health care  • Better categorization of patients; ensure each category’s needs are adequately met  • Diagnostic test-kits (to allow self-administered tests)— patients can perform the test at home at their convenient time

WHO Template A doctor is required to perform diagnosis/prescribe treatment

 • Some diagnoses and  • Phone or digital (preliminary) treatment consultations with patient’s prescriptions can be practitioner (GP) may eliminate done by other staff/ unnecessary referrals; reduce doctors not within load on SPC of conditions not SPC requiring specialist care; ensure that appropriate specialization is approached A specialist  • Certain tests/scans  • Allow patient’s GP to collect technician is required can be performed by samples/perform simple tests to perform certain ordinary nursing on SPC’s behalf tests/scans (e.g. staff/staff not within X-rays) SPC A patient must be ill  • A patient can visit  • Preventative care education— before visiting SPC SPC on occasions tailored to specific high-risk other than when groups they are ill (continued)

7  TRANSFORMING AND MANAGING HEALTHCARE PROJECTS 

169

Table 7.3 (continued) WHERE Template A patient can only be  • A patient can be diagnosed at the SPC diagnosed at home/a location close to home/ conveniently accessible to them A patient can only be  • A patient can be treated at the SPC treated at home/a location close to home/conveniently accessible to them

A patient must collect medication in person

WHEN Template A patient must be diagnosed before he can be seated A patient must undergo initial consultation before taking tests

 • Medication can be delivered to patient collected from a convenient location

 • A patient can be diagnosed and treated at the same time  • A patient can take test without initial consultation

 • Outsource basic testing to other polyclinics  • Establish a “Service Centre” to conduct tests in a more conveniently located place— major train station?  • Service Centre can also perform certain types of treatments—for some disciplines only  • Allow patient’s GP to treat them in consultation with SPC specialist (for follow-up treatment only)  • Mobile laboratory to dispense medications at patient’s home  • Establish electronic prescription arrangements with major pharmacies—patient only has to go to a pharmacy branch  • Service Centre can also dispense medications  • Allow patient’s GP to perform “presumptive treatment” under guidance of SPC specialist  • Patient’s GP to prescribe tests under consultation with SPC specialist—patient will go for tests on first SPC visit (without need for initial consultation)  • Artificial intelligence system available online to replace initial consultation (some conditions only)—to recommend what tests to perform or whether to see specialist first; allow for booking of tests online (continued)

170 

J. CHANCHAICHUJIT ET AL.

Table 7.3 (continued) A patient must pay for consultation/ treatment before leaving SPC

WHY Template Patient records must be retrieved the day before to ensure availability at SPC on appointment day A patient must wait at the clinic so that they are available when their name is called (for consulting/ treatment) so that SPC resources are efficiently utilized A patient must collect and bring their test results with them to ensure availability at main consultation HOW Template A patient can only be tested by taking blood sample or doing a scan and so on

 • A patient can make payment after leaving SPC  • A patient can make payments in instalments

 • Patient will be billed later— option for direct charges to credit cards or retirement funds  • Establish electronic fund transfer  • Patient can claim from their health insurance

 • Patient records  • “Electronic medical records” should be available (EMR) system anytime/all the time

 • A patient can wait in  • Establish an alert system to more pleasant call or message patient 15 min surroundings or do before appointment—allow other things while patient to go to cafeteria, book waiting to be called, shop, education centre or without adversely business centre while waiting effecting efficiency for main consultation/ treatment  • Test results should  • All test results scanned in be available at main immediately after results are consultation without available—no more test results patient having to on paper—eventually to be bring them along integrated into EMR.

 • A patient can be tested remotely   • A patient can test themselves

 • Remote diagnostic equipment   • Test-kits for self-administered testing

7.1.4   Step 4: Idea Grouping The next step after idea generation is to organize the list into meaningful groups of related ideas, which will eventually become options. Since there will be many ideas generated during brainstorming, it is better to consolidate them into options that are built around a central theme. This will allow an individual project or option to be implemented independent of other options. The following questions are helpful in grouping the ideas together into options. These questions are intended to guide you in grouping options.

7  TRANSFORMING AND MANAGING HEALTHCARE PROJECTS 

171

• What ideas are similar or seem to be based on the same theme? For example, some ideas may be built on outsourcing. • In what way is each option different? If there is any overlapping of ideas between options, it may be better to regroup them under different criteria. • Which ideas do not seem to fit anywhere? There will always be some ideas that do not fit into any of the options and it is fine to leave them aside. • Are there any options that are related to each other? This is not recommended, as the main objective is to allow any option to be implemented individually. The following options are proposed for the specialist private clinic. Option 1: Tie-ups with GPs/polyclinics • Phone/electronic consultation; • Collect samples/perform simple (preliminary) tests on behalf of SPC; • Perform presumptive treatment in consultation with SPC specialist; • Perform follow-up treatment in consultation with SPC specialist; • Recommend tests in consultation with SPC specialist (patient can go directly for tests on first visit, avoiding initial consultation); and • Prescribe/renew medications on advice of SPC specialist. Option 2: Rearrange facilities to minimize patient movement • General laboratory combining urine/blood testing and non-invasive cardio testing; • One general laboratory on each floor (only special equipment labs to be centralized); • Locate frequently co-occurring tests close together; and • Locate laboratories close to clinics which most frequently order them. Option 3: Self-registration kiosks • Patient registers himself/herself on arrival; and • Prints routing card with shortest/fastest route.

172 

J. CHANCHAICHUJIT ET AL.

Option 4: SPC On Wheels—home diagnosis and/or treatment • Equipped with basic test facilities and electronic link to SPC (transmission of test results); • Remote diagnostic equipment for direct, remote examinations by doctor back at SPC; • Perform treatment in (remote) consultation with doctor at SPC; • Specially trained nurses to carry out treatment under (remote) instruction; and • Dispense medications remotely. Options 5: Travelator system • For quick, effortless movement along major SPC corridors; and • Arrange clinics, labs and facilities in a circular arrangement. Option 6: Patient Service Centre(s) • Conveniently located and more accessible than main SPC; • Provide basic testing capabilities (may be expanded later); • Consultation/treatment for some types of disciplines; • Prescribing/dispensing medications; and • Dispensing diagnostic test-kits. Option 7: Remote diagnosis/recommend treatment • Need remote diagnostic equipment (if diagnosis is synchronous); and • Need electronic test result (if diagnosis is asynchronous). Option 8: Electronic Medical Records • Complete online availability of all patients’ records; and • Replaces all paper records pertaining to patients. Option 9: Electronic Test Results • Scanned-in test results available online immediately; and • Eventually integrated into EMR.

7  TRANSFORMING AND MANAGING HEALTHCARE PROJECTS 

173

Option 10: More payment options • Direct charge to credit cards or retirement funds; and • Instalment payments or claim from health insurance. Option 11: Reorganize post-consultation process • Combine billing for tests, consultations, treatments and medications; and • Single person to handle payment receipt and appointment booking. Option 12: Reorganize pharmacy • Electronic prescription sent to pharmacy immediately after main consultation; • Pharmacy consolidates medications by clinics; • Delivers medications to clinics (ready for patient to collect after treatment/post-consultation); and • Pharmacy prepares medications for patients to pick up after post-­ consultation (without waiting). Option 13: Patient free movement • Patients allowed to visit bookshop, cafeteria, education/business centre while waiting for main consultation; and • Alert system to call or message patients 15 minutes before consultation. Option 14: Patient kiosks • Located at common areas in SPC; • Provide electronic map for direction; and • Provide healthcare information/advice. Option 15: Health Education Centre • Preventative care programme for high-risk groups; and • General education on healthcare options, costs, and so on.

174 

J. CHANCHAICHUJIT ET AL.

7.1.5  Step 5: Idea Evaluation This step involves selecting one or more options based on their impact upon customers and the organization. This selection process is not aimed to reject or eliminate alternatives, but to establish priorities instead. All the options could be impactful but they must be given priority because of the limited amount of resources available. The objective of this step is to allow the team involved to obtain a consensus on which options (or ideas) must be given priority. Before the evaluation can start, everyone participating in the evaluation and selection process must share a clear vision of the ideas to be evaluated: objectives, implications, feasibilities, and so on. The selection criteria must be established for evaluation and be built on three dimensions: customer benefits, organization benefits and aversion factors. Criteria for customer benefits include improving quality and effectiveness, increased speed, responsiveness and reduced wait times, and enhancing convenience, accessibility and comfort. The criteria for organization benefits include operational savings, increased productivity and efficiency, increased capacity, strategic advantage, and enhanced image and reputation. Aversion factors relate to the negative impact of an option that reduces its feasibility. Criteria for aversion factors include the cost of implementing the option, the difficulty of implementing the option and the risk of implementing the option. Once the criteria have been finalized for evaluating each option, weight can be added to each criterion from 0 to 1, based on importance. The coordinator (person leading the evaluation process) may then ask the team members to score each option against each attribute for customer and organization benefits based on a score of −3 to 3 (−3 denotes negative contribution, 0 denotes no effect and 3 denotes significant contribution). In the case of the aversion factor, the scoring is based on a score of −3 to 0 since all the criteria refer to negative impact on performance. Table 7.4 gives an outline of how to conduct the evaluation. The team members must understand that this method is not a scientific tool. Rather, it is based on individual opinion and its strength lies in its simplicity. Even though the criteria are clear, the options defined are not easy for comparison, and in this case more deliberation about each option does not necessarily produce better results. For each rating of each attribute, every participant must question the impact each option has against the rest. Once every participant has classified all the ideas against every criterion, the coordinator will then create a final table recording the total score of

7  TRANSFORMING AND MANAGING HEALTHCARE PROJECTS 

175

Table 7.4  Format to evaluate each option

Others

Risk

Difficulty

Implementation

Aversion Factors

Others

Image

Strategic Advantage

Capacity

Productivity/ Efficiency

Organizational Benefits Operational Savings

Others

Convenience/ Accessibility

Speed/ Responsiveness

Options

Quality/ Effectiveness

Customer Benefits

each option. Using the final score, the score is averaged and multiplied by each criterion’s weight (if any), and the result is added into each column to compute the weighted score of each option. After these steps are taken, quantitative evaluation is available to measure options against each criterion that is deemed important. Therefore, we have passed from individual and subjective perceptions to a unified conclusion. Following this, the values obtained in the tables may be examined. Results based on scoring by the whole team make it easier for each member to accept the options selected later. It is not necessary that only options with high scores be selected. Options that collectively satisfy all benefit criteria to an acceptable degree are much preferred to achieve a win-win situation. Another consideration is to ensure that the selected options do not interfere with each other negatively, and that they are easy to integrate. For example, it is advised to avoid selecting one option that involves outsourcing a task and another option that involves building capabilities for the same task. One should select options that reinforce each other or which are able to achieve economies of scale. Table 7.5 illustrates the evaluation of options for a specialist private clinic.

176 

J. CHANCHAICHUJIT ET AL.

Table 7.5  Scoring of each option for a private clinic

Others

Risk

Difficulty

Aversion

Implementation

Others

Image

Strategic Advantage

Capacity

Productivity/Efficiency

Organizational Benefits

Operational Savings

Others

Convenience/Accessibility

Options

Speed/Responsiveness

Quality/Effectiveness

Customer Benefits

GPs/Polyclinics Rearrange SOC Self-reg. kiosks travelators Doctors at office EMR ETR ES on web Reorg. post-consult JIT pharmacy Roaming patients Patient test labs SOC on wheels Roving pharmacy Pharmacy tie-ups

7.1.6   Step 6: Ideas Integration and Digitization The next step is to integrate the selected options into a new process. This involves developing a “sketch” of how selected options will be integrated and how they will work together by trial and error. This may require additional ideas or enhancements, and information technology can be included at this point to improve overall design. Examples of information technology used in this step include the following: • Automation—Can each activity be automated to remove human intervention? • Accessibility—Can information be captured at the source?

7  TRANSFORMING AND MANAGING HEALTHCARE PROJECTS 

177

• Multi-tasking—Can the sequence of processes be changed and performed concurrently using IT? • Tracking—Can IT be used to monitor real time? • Analysis—Can IT be used to analyse information to improve decision-making? • Integration—Can IT be used to integrate tasks and data from multiple countries? • Dis-intermediary—Can IT be used to eliminate intermediaries from the process? The new process for a specialist private clinic is mapped out as a result of integrating the selected options, as shown in Fig. 7.2. In the new process, patients can now perform simple testing and diagnostics at home before going to an SPC. At the SPC, the patient can register themselves at the kiosk and proceed to the test centre with the smallest queue. Once the tests are completed, they are free to shop or eat at a nearby cafe. Fifteen minutes before the actual consultation, they can be alerted via their smartphone. After consultation, they can either receive treatment or go directly

Fig. 7.2  New process flow for specialist private clinic

178 

J. CHANCHAICHUJIT ET AL.

to the pharmacy to collect their medication. They end their visit by paying all their bills at once and they can make their next appointment by checking their Google calendar. The new process must be validated to ensure that it is feasible and to solicit buy-in from the stakeholders. 7.1.7   Step 7: Process Validation and Implementation The last step is to examine the major changes implied by the new design and identify the major obstacles that potentially stand in the way of their realization or acceptability. There will be a need to identify problems that might be introduced by the new process design, or any erosion of existing benefits from the current process by the new process design. These problems could be resolved by: • Verifying major assumptions; • Seeking necessary in-principle approvals; • Studying impact of external dependencies; and • Seeking assurances of acceptability and cooperation from stakeholders. As a result, one may need to modify the new process in view of feasibility considerations or to reinstate benefits from the current system that might be removed from it. It is also important to establish the financial, manpower and skills training needed in order to implement the new process. Since the budget for the healthcare industry is limited, one must decide the priority of implementation by outlining various considerations. The priorities are then translated into a preliminary implementation schedule. The implementation plan can include: • Management structure for implementation; • Financial cost and benefit analysis; • Considerations for prioritizing IT applications; • Implementation options; and • Preliminary implementation schedule.  eview Business and Policy Developments R Business and government policy developments will also need to be reviewed and this would include reviewing changes in healthcare policies,

7  TRANSFORMING AND MANAGING HEALTHCARE PROJECTS 

179

laws or regulations that are likely in the near future; new initiatives that might affect the new process, changes in location for business units, new buildings or infrastructure developments; and changes in management. The review would include any impact upon the scope, resources and schedule of the new processes during implementation.  uantify Projected Benefits Q For the purposes of deriving the projected benefits, several workshops are proposed. These would be held with the process teams in order to identify the benefits derived from the transformation project to meet patient outcomes. Based on the schedule drawn up in the overall migration path, as well as the costs derived, it is possible to draw up a year-by-year expenditure plan. From the completion time-table for each of the systems, it is possible to derive a year-by-year projection of the benefit stream. Together, these two components form the complete investment-flow analysis, from which one can derive the break-even time-frame as well as the present value of the investment and benefit stream.  ropose New Organization Structure P The implementation plan should clearly outline the management structure necessary to support the new process. The expected number of staff needs to be identified, and their relationship to the project sponsors or a steering committee outlined. This can be done by setting up a project plan as per the project management standards. An example of the new staffing requirements to support a new process is shown in Table 7.6. J ustify IT Investments Applications development costs can be estimated by soliciting input from various system integrators or other vendors. The descriptions of the functionalities and sizing of the applications that have been identified can serve as a basis for such initial discussions. Comparison with other similar-sized projects can also serve as a gauge or benchmark upon which to base the estimations. These application development costs should be considered separately from the hardware costs, which will be covered in the costs of infrastructure. These alleviate the need to amortize the costs of the infrastructure across each application. The costs of the infrastructure, including the servers, PCs and network, are usually much easier to estimate. These costs can easily be obtained from vendors.

180 

J. CHANCHAICHUJIT ET AL.

Table 7.6  Staffing requirements to support new process Department

Accomplishment Transaction Value Estimated Number Daily time volume time number of of staff taken per (annual) (mins) staff staff (hr)

Department A Supplier registration Department A Supplier edition Department A Customer registration Department A Customer editing Finance Setting up electronic payment gateway Department A Generate and send pin mailer Department A Service preapproval Department A Service edit/ resubmit Finance Batch payment to finance Team Preliminary check Team Onsite audit

2250

30

1.41

2

2.8

1626 10,000

30 20

1.02 4.17

1 4

4.1 4.2

843

20

0.35

1

1.4

12,250

10

2.55

3.00

3.4

120

0.50

1

2.0

18,428

15

5.76

6.00

3.8

13,036

5

1.36

1.00

5.4

655

15

0.20

1

0.8

360

12

0.09

0

0.0

3240

24

1.62

2

3.2

200

I dentify Skills Gaps Typically, skills required in various departments are IT literacy skills, analytical skills, customer management and handling, and so on. In parallel with the IT applications development and migration path, the relevant IT literacy and technology management skills have to be developed within the organization. A time-table for the internal development or external recruitment of such skills needs to be created. Such a time-table is usually worked out in conjunction with the managers as well as the human resource section of the department in question. Plans should also include training, recruiting or purchasing the services of the required project managers who will be implementing the project if necessary.  evelop Migration Plan D A workshop can be carried out with the steering committee of the project or with the staff planning section or sub-committee. During this

7  TRANSFORMING AND MANAGING HEALTHCARE PROJECTS 

181

workshop, each IT application or process needs to be assessed on a variety of criteria. These criteria can be proposed by the facilitators but should be verified by the business managers. It is important to identify and agree on the criteria before assessment begins. Once a consensus of the priority of the applications has been reached, facilitators, taking into consideration business and policy developments, will need to derive a migration path for the existing systems. An example of a migration path for a new IT system is shown in Table 7.7. 7.1.8  Step 8: Project Management and Change Management Why Is Project Management Methodology Important in the Healthcare Industry? Trying to manage a project without project management is like trying to play a football game without a game plan. (K. Tate, past board member of the Project Management Institute)

Project management methodology is a structured approach to the initiation, planning, execution, monitoring and closure of projects. It provides standardized management information and discipline to a project team to ensure “consistency and predictable project outcome”. A project team should not be surprised if a project control signal turns from “green to red”, warning that a project may be out of control. For a large healthcare project such as the implementation of a new system, or an upgrade, such as EMRs, Hospital Information System, RIS/PACS Systems (Radiology Information System and Picture Archiving and Communication System), there is a need for a proper project management methodology to govern the project. This allows guidance on processes, tools and techniques to help achieve the goal and empower health professionals to have better control of their projects. Project Management can be viewed as a map to help you when driving in unknown territory. Without a map, you may easily get lost and fail to reach your destination. Waterfall Methodology Versus Agile Methodology There are two common methodologies used to manage projects, namely the Waterfall and Agile methodologies. The Waterfall methodology (sometimes called Predictive or Plan-driven) has been around for decades and has become a traditional methodology for project manage-

User requirement documentation by Dept A User requirement study and deign documentation (by vendor) Development and testing by vendor UAT

UAT fixes and user testing Training ops details (including IT helpdesk) System live

2

6 7

1.2.2 Send out remainder 1.2.3 Data capture

Transition Plan Data conversion plan Account Send out form Send out remainder Data capture Old system account import into new system 1.2 Product pre-approval 1.2.1 Send out form

II I 1.1 1.1.1 1.1.2 1.1.3 1.1.4

8

5

4

3

Activity IT Plan New system upgrade Hardware PO Hardware delivery Re-conf iguration Testing

SI no I 1 1.1 1.2 1.3 1.4

Oct 05

Nov 05

Table 7.7  Migration plan for new IT system Dec 06

Jan 06

Feb 06

Mar 06

Apr 06

(continued)

May 06

182  J. CHANCHAICHUJIT ET AL.

Communication Plan Corp comm. to come up with the communication plan To inform user of the new enhancement to the current system (e.g.email, advertisement)

Public launch for customers and users Internal comm. (change management) to all depts.

Operation Plan

Policy changers

Organization chart

III

4

IV

1

2

5

2

1

User self-learning/ chargeable classroom training

5

1.2.4 Old system account import into new system Management approval 2 for training and temp staff To prepare training for 3 customers 4 Mail out training kit

Table 7.7 (continued)

7  TRANSFORMING AND MANAGING HEALTHCARE PROJECTS 

183

184 

J. CHANCHAICHUJIT ET AL.

ment. On the other hand, the Agile methodology is newer and in the last 15 years has proven to be practical for software application development. Many people are exploring Agile as an alternative methodology. These methodologies are examined below. Waterfall Methodology This is a traditional project management methodology which emphasizes the logic of the sequential process. All of the project planning is carried out at the beginning, and is then followed by the execution of the plan, together with monitoring and control. All project requirements (e.g. scope, time, cost, resources, quality, risk, etc.) are gathered, analysed, documented and approved before the start of the project. If there is a change during the project, for example a physician may request a mass customization of clinical reports, a project manager is called in. They then analyse the impact of the change and may submit it to a project committee board for approval. This is only needed if that change has a significant impact upon the project’s cost and timeline. Some changes may necessitate the reworking of the project and some may actually derail the project. There are many well-known project management frameworks that use the Waterfall methodology, such as the PMBOK Guide (Project Management Body of Knowledge) from the Project Management Institute (PMI) (2017) widely used in the USA, and Prince II (Project IN Controlled Environments) used extensively by the UK government (Fig. 7.3). Agile Methodology This is a methodology that is commonly used for software development projects. Agile methodology assumes that project requirements cannot all

Fig. 7.3  Waterfall methodology

7  TRANSFORMING AND MANAGING HEALTHCARE PROJECTS 

185

be completed at the beginning due to many factors. These include that a customer/user may not fully engage and/or understand during the requirements-gathering phase. As time goes by, when a product starts to take shape, the user can learn and realize what could have been done better. Working on this assumption, Agile uses iterative and incremental approach building and refining of products/applications with frequent reviews and delivery of the work in small pieces at a time. With this approach, a customer/user will be involved throughout the project, ­providing feedback and suggestions for changes that may be required during reviews. The iterative approach is based on a cyclical process of refining a product or process. During a product/application development, more than one iteration of the product/application development cycle may be in progress at the same time. This is known as an incremental build approach. There is no designated project manager in Agile but there is a team facilitator. A classic example of Agile methodology is Scrum. Scrum is an iterative and incremental framework for managing application/product development projects. Scrum uses time-boxed (the goal of timeboxing is to define and limit the amount of time dedicated to an activity) and phases called “sprints”. Sprints are basic units of time development generally less than 30 days long. Sprint is a regular, repeatable work cycle during which work is completed and made ready for customer/user reviews and evaluation on a daily basis and at the end of sprint demonstrations. Planning is done at the start of each sprint together with duration, list of deliverables and prioritizing. If a sprint cannot be completed, that work will be reprioritized and the information is kept for future sprint planning (Fig. 7.4). Table 7.8 presents a high-level comparison of Waterfall and Agile methodologies. Sprint

Sprint

Sprint

Finish

Start Sprint

Fig. 7.4  Agile methodology

Sprint

186 

J. CHANCHAICHUJIT ET AL.

Table 7.8  Agile versus Waterfall methodology Project factor

Waterfall methodology

Agile methodology

Requirements

Can be defined, described or fixed Focus on timeline Focus on cost Identified upfront

Highly uncertainty or dynamic

Schedule Cost Quality

Communication

Customer/user involvement

Solely relies on project manager to manage and provide updates to project team As per project milestone

Focus on time-boxes vs sprint Focus on customer/user needs  • Quality can be increased/optimized for learning during project  • Focus on quality and functionality at each sprint—improve patient safety  • More frequent—earned out daily and at the end of sprint—project team and customer user can review the deliverable and provide feedback  • High customer/user involvement during each sprint   • Reducing risk of misunderstanding  • Can be identified at the start of each sprint cycle  • Can be identified and managed in a small group at a time as per each sprint cycle

Risk

Identified upfront

Stakeholder

Identified upfront and relies on project manager to manage from project start to finish Restricted changes using  • Welcome changes triple constraints (scope, time, cost) and quality to measure impact Overall solution delivered  • Customer user can receive subset of at the end of project the overall solution during project Benefit can be realized  • Can be benefited throughout project only at the end of project which delivers immediate value to business Identified upfront  • Can be defined during project   • Using small-scale experimentation   • Prioritization by business value ensures the most valuable features are built first   • Reducing risk of complete failure Based on fixed schedule  • Based on its capacity to start work  • Focus on subject  • Focus on subject matter experts, matter experts team work and interpersonal skills  •  Focus on   • Focus on cross-functional teams and collaboration within increased collaboration the project team

Change request

Deliverables Business value

Work items

Workflow Project team—Skill set

(continued)

7  TRANSFORMING AND MANAGING HEALTHCARE PROJECTS 

187

Table 7.8 (continued) Project factor

Waterfall methodology

Agile methodology

Project manager

 • Using team facilitator for facilitating collaboration and coaching

Documentation

Dedicated project manager for managing project and project team Identified upfront

Contract term

Suitable for fixed fee

 • Can be identified at the start of each sprint cycle  • Suitable for time and material   • May not permit partial deliverables if project is terminated

In summary, the Waterfall methodology is suitable for projects that have been implemented previously either by experienced vendors or internal resources. The project outcome must be well defined, with repeatable steps for requirement gathering, planning, building, testing and implementing. Waterfall is also suitable for projects that have fixed deadlines and costs. Agile methodology works well when dealing with an unknown project outcome (research, innovation or start-up), a hypothesized ­solution, or where there is high uncertainty requiring a lot of changes. Agile can be good for a project that requires fast results which will support the organization. This is because Agile can deliver a subset of the overall solutions during the project, providing immediate value to the business.  hen to Choose the Waterfall or Agile Methodology for a Healthcare W Project Since the Waterfall and Agile methodologies have their own strengths and weaknesses, they can co-exist in a healthcare project where business value can be realized and project risks can be reduced. An example of this would be the implementation of big data analytics within an organization for better prevention, treatment and management. Traditional Waterfall methodology can be used for the implementation of a data warehouse to store clinical data. At the same time, predictive analytics tools can be built using the flexibility of the Agile methodology to help refine AI analytics algorithms and reports for predicting chronic diseases such as diabetes, high blood pressure and high cholesterol. In short, when it comes to building a platform or healthcare IT infrastructure using experienced vendors, Waterfall is well suited. Most vendors have their own methodologies and most of them are still based on the Waterfall model. Agile can be useful in the case of building something

188 

J. CHANCHAICHUJIT ET AL.

specific or something new, such as an application for a clinical trial, new research or innovation, predictive analytics or diagnostic tools. To work with Agile methodology, a high-level plan is required along with the promotion of an “Agile mind set” within an organization, including customers, to make it truly work. For example, the management team may not fully understand Agile methodology and continue to manage the project in a conservative style. They will not allow their project manager to choose their team members and resources, which contradicts Agile principles. The Agile method cannot help address resource capacity issues, since productivity may or may not increase depending on the Agile process and management. In addition, Agile cannot create extra time, and one option is to consider hiring more staff to solve the issue. Lastly, continuous improvement in the Agile process is required along with continual refining for consistency.

7.2   Challenges Faced in Managing Healthcare IT Projects When managing a healthcare IT project as part of a transformation, there are many unique challenges that one needs to plan for and address early on. 7.2.1   Patient Safety This is a number one priority and every healthcare organization strives to improve their patients’ safety. Quality and patient safety usually go hand-­ in-­hand. As such, extra attention should be paid to the “quality” of everything, particularly data integrity, as patients’ lives depend on the accuracy of diagnostics and data. 7.2.2   IT Integration Difficulties There can be problems with integrating healthcare equipment from legacy machines to a new pharmacy system of robotic dispensing. Integration tasks should not be underestimated; they are time-consuming and need to be well organized between vendor resources (where two systems need to communicate) and internal resources. Healthcare organizations use communication protocols such as DICOM and HL7 for integration.

7  TRANSFORMING AND MANAGING HEALTHCARE PROJECTS 

189

7.2.3  Lack of Commitment Another difficulty can be found in the lack of interaction from “business” users such as physicians and nurses due to the nature of their work. Time commitments should not be underestimated; healthcare staff are often dealing with or preparing for a crisis. This contingency should be factored into the project schedule and enough time allowed for staff involvement in a project. Process transformation has become the method of choice in starting again with business processes. However, far too often, this was applied to employees and middle managers only, neglecting senior management who sometimes felt it was not necessary for them to change along with IT. Credibility for a new system project can be lost in such instances as CEOs not knowing how to use email. Management may spend money on purchasing technology, but they must also finance the training for staff to utilize the IT. 7.2.4  Resources Constraints In the early years most IT healthcare projects failed due to the use of general resources for healthcare instead of implementing specific resources (Ooi & Tan, 2016). For example, to run a healthcare project you would require people and resources that understand medical terminology. They would be capable of communicating and interpreting requirements. To address this issue HIMSS (Healthcare Information and Management Systems Society) have created specific certifications for healthcare information and management systems. Changes in Regulations 7.2.5   It is important to be aware of rapid changes in government regulations and compliance requirements to cater for health IT adoption and integration. Users must keep up to date with policy to avoid any reworking or penalties for non-compliance. Fear of Job Loss 7.2.6   When companies take on a project that will incur massive costs, senior management must have hard evidence that there will be sizeable savings somewhere. One of the easiest ways to provide a cost benefit to such a

190 

J. CHANCHAICHUJIT ET AL.

project is to cut the workforce. However, many cases have shown that this only reduces the available talent pool in an organization, as well as creating unrest among remaining staff. Downsizing can have a profound effect on the lives and productivity of individuals, whether they leave the organization or survive the downsizing. IT tends to flatten the hierarchical management structure of the past, which can cause resistance by some employees. Research has shown that stressful responses in the workplace to downsizing or dramatic changes are major factors behind illness, substance abuse and low productivity (Khubchandani & Price, 2017). 7.2.7   Change of Organizational Structure and Culture People react strongly when the prospect of change intrudes on their familiar working and living patterns. Most people believe that the necessary cooperation to achieve success will not occur unless people are assured that they will not be working themselves out of a job. Why then do companies expect their project to bring about rapid changes? Technology is changing so swiftly that we have come to expect people and organizational structures to do the same. IT project teams that promise quick results are only setting themselves up for disappointment. If process transformation is approached with inadequate skills in the employee base with very little training, few support programmes and little input during the design stage by employees, change at a moderate pace is extremely difficult.

7.3   Change Management Is an Important Part of Healthcare Transformation Projects The concept of “change management” is useful for people who are coming to grips with change management problems for the first time during transformation, and for more experienced people who wish to reflect upon their experience in a structured way. In thinking about what is meant by “change management”, the most obvious definition is “the task of managing change”. This is not necessarily unambiguous. Managing change is itself a term that has at least two meanings. One meaning of “managing change” refers to the making of changes in a planned and managed or systematic fashion. The aim is to more effectively implement new methods and systems in an ongoing organization. The changes to be managed lie within and are controlled by the organization. Perhaps the most familiar instance of this kind of change is the change

7  TRANSFORMING AND MANAGING HEALTHCARE PROJECTS 

191

or version control aspect of information system development projects. However, these internal changes might have been triggered by events originating outside the organization, in what is usually termed “the environment”. Hence the second meaning of managing change, namely the response to changes over which the organization exercises little or no control (e.g. legislation, social and political upheaval, the actions of competitors, and shifting economic tides and currents). The preceding section points out that the challenges during transformation, especially the problems of change, have both a content and process dimension. It is one thing, for instance, to introduce a new claims processing system into a functionally organized health insurance company. It is quite another to introduce a similar system into a health insurance company that is organized along product lines and market segments. The systems use different “languages” and have different “values and cultures”. At a detailed level, the problems differ; however, the overall processes of change and change management remain almost the same. It is this fundamental similarity in the change processes across organizations, industries and structures that makes change management a task, a process and an area of professional practice. Managing the kinds of changes encountered by and instituted within organizations requires a set of comprehensive skills. 7.3.1   Political Skills Organizations are first and foremost social systems. Organizations can be complex and political. Change agents should be aware of these elements and attempt to understand them whilst remaining objective with regard to judgement. 7.3.2   Analytical Skills Clear, rational and well-argued analytical skills can often be used to justify the efforts involved in transformation. There are two particular sets of skills which are very important: (1) workflow operations or systems analysis and (2) financial analysis. Change agents must learn to take apart and reassemble operations and systems in novel ways, and then determine the financial and political impact. Change agents must be able to start with some financial measure or indicator or goal, and make their way quickly to those operations and systems that, if reconfigured a certain way, would have the desired financial impact.

192 

J. CHANCHAICHUJIT ET AL.

7.3.3  People Skills The human element must be the primary consideration in an organization. There is great variety in the race, culture and gender of individuals who find themselves working for one organization, dimensions along which people vary. The skills most needed in this area are those that typically fall under the heading of communication or interpersonal skills. To be effective, a change agent must be able to listen actively, restate, reflect, clarify without interrogating, draw out the speaker, lead or channel a discussion, and plant ideas and develop them. This is especially so in competitive cultures where most people tend to keep information to themselves. Frame of reference is also significant. The change agent must attempt to see all points of view and to reconcile and resolve any conflicts that may arise. 7.3.4   Leadership Skills Leadership is the act of creating organizational vision, alignment and deployment. Once the organization is clear and committed to its core vision, a specific vision of the transformation required should be created, a purpose developed. The project team must be focused on its goal. The vision must then be communicated throughout the entire organization, and commitment to the transformation effort and ensuing changes encouraged. The responsibility of creating the vision in most cases falls upon senior management, but it can be revised after the more “micro” and complex work of project teams has taken place. Therefore, the skills most needed in this case are those of communicating the process vision to everyone effectively and guiding team members in the right direction to reach the vision.

7.4   Key Success Factors for Implementing Large-­ Scale IT Systems The challenges of implementing a Healthcare 4.0 project may be those of uncertain requirements, new products, new technologies, market uncertainty and government regulations. Breaking down the larger project into smaller projects allows greater control, and an appropriate methodology can be applied to each one of them. Managing the people, processes and tools should effectively contribute to successful outcomes (Ooi & Tan, 2016). It is essential to secure suitably skilled project staff and to provide for their development

7  TRANSFORMING AND MANAGING HEALTHCARE PROJECTS 

193

as the project unfolds. Additional staff should be sought where necessary to avoid placing unrealistic burdens on existing staff and thereby putting the project at risk. Project management is the responsibility of the project manager. In general, up to 90% of the project manager’s time is spent dealing with people. Interpersonal and communication skills are therefore essential, along with team guidance, mentoring and informing stakeholders of progress. If the proposed organization does not have a qualified project manager, or needs additional assistance, a professional project manager may be hired. It is most important to keep project information circulating to and amongst the project team on a constant basis. By keeping the customer/user involved and informed, problems may be dealt with in a timely manner and the project kept on track. Input, review and feedback should ensure that expectations are met and a satisfactory outcome achieved (Ooi & Tan, 2016). Executive Management Support is required throughout the project. Strong support from executive management is a good indicator of the likelihood of the project’s success (Ooi & Tan, 2016). If the top levels of management should lose interest in the project, then they are unlikely to secure funding or additional resources for the project. In the rapidly changing world of healthcare it is always good practice to keep checking the alignment of your projects with current business strategies. Such alignment should ensure that the project outcomes will still add significant general value to the organization and/or contribute to its economic value or profits. Acknowledgement  The authors would like to express their sincere gratitude to Ms Malai Williams. She contributed to  this chapter  regarding the development idea of a healthcare project and change management by using Waterfall and Agile methodology. Her expertise in IT project management, especially in healthcare, has helped us to share key success factors with the reader on transforming and managing healthcare projects.

References Khubchandani, J., & Price, J. (2017). Association of job insecurity with health risk factors and poorer health in American workers. The Publication for Health Promotion and Disease Prevention, 42(2), 242–251. https://doi.org/10.1007/ s10900-016-0249-8 Ooi, L.  C., and Tan, P. (2016). Our IT journey: One patient-one record. In Singapore’s healthcare system (pp.  337–350). World Scientific. https://doi. org/10.1142/9648

194 

J. CHANCHAICHUJIT ET AL.

PMI. (2017). PMBOK guide, Sixth edition. Retrieved from https://www.pmi. org/pmbok-guide-standards/foundational/pmbok. Tan, A. (2013). Supply chain process reengineering. Pearson Malaysia. ISBN 978-967-349-290-9.

CHAPTER 8

Conclusion

Abstract  This chapter concludes the book, with summaries and recommendations for implementing Healthcare 4.0. Keywords  Industry 4.0 • Healthcare 4.0 • IoT • Big data • Blockchain • Operations Research The healthcare sector is facing a new wave of digitalization in driving Healthcare 4.0. Digitalization is no longer just a technical aspect, for example saving imaging and laboratory data in electronic medical records. Instead, it is now shaping the management and design of complete healthcare processes and affecting healthcare providers’ business models. This development is driven by the fact that in a wide variety of areas, enormous amounts of data are now available, while at the same time storage and computing costs are decreasing drastically and Artificial Intelligence methods are advancing. If healthcare services are to incorporate Industry 4.0 core principles, they require proper guidelines or a framework within which to incorporate these. Based on the examples given in previous chapters, a set of emerging technologies is recommended for implementation into the healthcare sector. Table 8.1 demonstrates how these emerging technologies are fulfilling the core principles of Industry 4.0.

© The Author(s) 2019 J. Chanchaichujit et al., Healthcare 4.0, https://doi.org/10.1007/978-981-13-8114-0_8

195

196 

J. CHANCHAICHUJIT ET AL.

Table 8.1  Emerging technologies to support Industry 4.0 Core principles of Industry 4.0

Wearable devices

Interoperability Decentralization Virtualization Modularity Service orientation Real time capabilities

Yes Yes Yes

Internet of Things/big data analytics

Blockchain Yes Yes

Yes Yes

Yes Yes Yes

Yes

Artificial Intelligence

Yes

Yes Yes

The benefits of adopting these emerging technologies for Healthcare 4.0 include: • Wearable devices—Empowering patients to perform self-­management of medical needs, and provide channels for more interactive communication with healthcare professionals. • IoT and big data analytics—Maximizing healthcare resources, increasing the preventative and predictive components of care with the expectation of keeping individuals as healthy as possible and less dependent on curative care. • Blockchain technology—Providing real-time capturing of patient clinical records. • Artificial Intelligence—Providing more accurate predictive models of a patient’s condition.

8.1   Implementing Healthcare 4.0 The  challenges in implementing a Healthcare 4.0 project  are including uncertain requirements, new products, new technologies, market uncertainty and government regulations. Breaking down the larger project into smaller projects allows greater control, and an appropriate methodology can be applied to each one of them. Managing the people, processes and tools should effectively contribute to successful outcomes. It is essential to secure suitably skilled project staff and to provide for their development as the project unfolds. Additional staff should be sought where necessary to avoid placing unrealistic burdens on existing staff and thereby putting the project at risk. Project management is the responsibility of the

8 CONCLUSION 

197

project manager. In general, up to 90% of the project manager’s time is spent dealing with people. Interpersonal and communication skills are therefore essential, along with team guidance, mentoring and informing ­stakeholders of progress. If the proposed organization does not have a qualified project manager, or if additional assistance is needed, a professional project manager may be hired. It is most important to keep project information circulating amongst the project team on a constant basis. By keeping the customer/user involved and informed, problems may be dealt with in a timely manner and the project kept on track. Input, review and feedback should ensure that expectations are met and a satisfactory outcome achieved. Executive Management Support is required throughout the project. Strong support from executive management is a good indicator of the likelihood of the project’s success. If the top levels of management should lose interest in the project, then they are unlikely to secure funding or additional resources for the project. In the rapidly changing world of healthcare, it is always good practice to keep checking the alignment of your projects with current business strategies. Such alignment should ensure that the project outcomes will still add significant general value to the organization and/or contribute to its economic value or profit. However, one of the biggest obstacles to early adoption of Healthcare 4.0 would be the lack of coherence in digital enablement across different stakeholders. For instance, many hospitals may not have the necessary financial muscle to implement a complete digital transformation. The industry may not transform overnight but it can certainly start in areas such as nursing or procurement. This level of sophistication in the healthcare industry not only requires a high level of technological transformation but a shift in mindsets. At present, fragmentation in the sector does not allow a seamless experience for patients, who often find themselves having to negotiate multiple avenues and parties along with the struggle of storing their medical records. The complexities involved in the industry would need to be detangled first and allow more information-sharing within the industry. It is evident that current successes in IT transformation have been brought about by overcoming previous failures. Implementing healthcare innovations into large systems has always been a complex endeavour, and there are no guarantees of success. However, it is realistic to assume that future generations of healthcare professionals will progress with technology and continue to innovate to make Healthcare 4.0 a reality.

Index

A AFB, 85–87 Agile, vii, 181–188 AI, see Artificial Intelligence AIDS, 68 Algorithm, 38, 53, 64–66, 75, 78, 80–82, 85–88, 116, 134, 141, 187 Analytical skill, 180, 191 Artificial Intelligence (AI), v, vi, 2, 9, 12, 13, 63–88, 131, 132, 140, 144, 149, 187, 195, 196 Avatar mouse model, 139, 140 B Bed capacity, 97, 101, 103–106 Bed shortfall, 106, 107 Big data, v, vi, 2, 12, 13, 17–33, 73, 75, 96, 132, 136, 144, 153, 187, 196 Biosensor, 76, 82, 125, 141, 142 Bitcoin, 37, 43, 53

Blockchain, v, vi, 2, 9, 10, 12, 13, 25, 37–61, 73, 132, 196 Bureau of Tuberculosis, 82, 86, 87 C Cancer, 25, 68, 99, 127, 131, 136–138, 140, 143 Capacity planning, vi, 97, 107–114 Chest X-rays (CXR), 48, 71, 73, 76, 78–81, 83–87 CXR, see Chest X-rays Chronologically-based, 39 Clinical, vi, 9, 11, 12, 18, 19, 22, 26, 28, 33, 68, 71, 76, 79, 80, 82, 85, 99, 118, 134, 137–141, 145, 147–149, 184, 187, 188, 196 Cloud computing, 21, 32 Consensus protocol, 39 Constraint, 96, 98, 99, 111, 118, 189 CPLEX, 97, 104 Cryptographically, 38 Cyber-physical systems (CPS), 3–4, 7, 8

© The Author(s) 2019 J. Chanchaichujit et al., Healthcare 4.0, https://doi.org/10.1007/978-981-13-8114-0

199

200 

INDEX

D DAC4TB, 85–87 Day-of-Week, 100, 108 Decentralization, 2, 5, 7–8, 38, 45 Deep Learning, 64, 66, 78–80, 85–87 Diabetes, 28–33, 51, 68, 99, 135, 143, 187 Digital currency, 37 Digital healthcare, 32, 125, 131–136, 149 Disease-centred, 128 Distributed ledger, 3, 38 DNA-sequencing, 137 E Elective, 100–107, 114 Electronic Health Records (EHRs), vi, 9, 18–20, 25, 31, 44–45, 96, 152 Emergency, 23, 50, 97, 99–101, 103–105, 107, 108, 110 Entrepreneurs, vi, 123, 126, 144–146, 152–155 Ethereum, 37, 41, 50 Exome sequencing, 139, 140 F Fear of job loss, 189–190 G Genetic, 30, 71, 73, 78, 82, 136–138 GeneXpert, 85, 87 Genomic, 9, 26, 28, 131, 137–140, 152, 153, 155 GPOs, see Group Purchasing Organizations Group Purchasing Organizations (GPOs), 18

H Healthcare 1.0, 9, 10 Healthcare 2.0, 9 Healthcare 3.0, 9 Healthcare 4.0, v, vii, 1–13, 125, 132, 149, 192, 195–197 Healthcare delivery, vi, 8, 19, 96, 138 Healthcare expenditure, 126–127, 155 Healthcare industry, v, vi, 2, 9, 10, 13, 20, 25, 41–42, 57, 61, 107, 118, 126–127, 178, 181, 197 Healthcare process, v, vii, 118, 144, 195 Healthcare technology, 126 Health technology, v, vi, 13, 123–155 HIV, 68 I Industry 4.0, v, vi, 2–9, 12, 13, 21, 25, 195, 196 Innovation, 2, 9, 21, 82–87, 123, 124, 126, 127, 131, 144, 149, 152–155, 187, 188, 197 Inpatient, 49, 100, 101 Inpatient flows, 100 Internet of Things (IoT), v, vi, 2, 4–7, 9, 12, 13, 17–33, 45, 49, 50, 52, 55, 56, 132, 136, 144, 149, 150, 153, 196 Interoperability, 2, 5–7, 10, 23, 45, 52, 55, 61 Intervention, 30, 31, 49, 65, 69–73, 82, 83, 85–87, 98, 107, 112, 114, 115, 127, 129, 134, 167, 176 IoT, see Internet of Things IrDA, 23

 INDEX 

L Laboratory Information Management system (LIMS), 5 Lack of commitment, 189 Leadership skill, 192 Legality, 43 Life expectancy, 126–128, 155 LIMS, see Laboratory Information Management system Litecoin, 37 M Machine Learning, 27, 64–66, 75, 78, 79, 97, 144 Mean deviations, 109, 111, 112, 114 Migration plan, 180–182 Ministry of Health, 31, 82, 86, 112 Mixed integer program, 98, 107–114 Model description, 115–116 Modularity, 2, 5, 6 M. tuberculosis bacteria (MTB), 67, 68 N Nanomedicine, 142–143 Nanorobotics, 142–143 Neuro Learning, 64–66 Numerical, 116–118 O Oncology, 126, 138, 141 Operations research, vi, 95–97 Optimization, v, vi, 2, 13, 43, 95–118, 126 Organisation structure, 2, 179–180 P Patient-centred, 11, 128 Patient outcome, v, 2, 11, 137, 179

201

Patient-provider relationships (PPRs), 48, 50, 51 Peer-to-peer, 23, 37, 38 People skill, 192 Personalized healthcare, 124, 128, 131, 136–142 Personalized treatment, 31, 128, 140 Pharmaceutical industry, vi, 54–56, 149 Political skill, 191 Precision medicine, 136, 138–141, 155 Predictive analytic, v, vi, 27, 95–118, 187, 188 Privacy, 18, 24, 32, 39, 40, 47, 53, 61, 142, 151 Process transformation, vii, 162–190 Proof-of-work, 39 R Radiologist, 76, 85, 86 Real-time, 2, 4–7, 12, 18, 22, 23, 25, 31, 48, 50, 55, 57, 64, 177, 196 Regulatory, 18, 41, 125, 128, 135, 144–147, 149–152, 155 Remote control, 22, 23 Ripple, 37 RNA-sequencing, 137 S Same day surgery admission (SDA), 100 Scalability, 7, 8, 43 Scheduling, vi, 95, 96, 98, 100, 108 Security, 6, 23, 24, 32, 39, 41–43, 49, 54, 57, 150 Sensor, 3, 4, 7, 22, 26, 32, 48, 49, 55, 56, 132, 134, 141, 143 Simulation, vi, 6, 64, 95–118

202 

INDEX

Singapore, vi, 11, 29, 33, 100, 102, 104, 109, 112, 116 Singaporean, 29, 30 Smart contract, 40–43, 45–48, 50, 51, 55–58, 60 Smart factory, 4–5 Specialist private clinic (SPC), 162–164, 167, 171–173, 175, 177 Sputum, 68, 76, 82, 83, 85, 86 Start-up, v, vi, 13, 80, 123–155, 187 Stochastic, 95–98, 101, 102 Strategic resource, vi Supply chain, v, vi, 2, 7, 22–23, 54–60, 98 Symptoms, 18, 44, 51, 71, 76, 82, 85 T TB, see Tuberculosis Technology wave, 123, 124 Telemedicine, 125, 131, 134, 135, 151 Thailand, vi, 66, 82–88 Traceability, 54–57 Treatment, vi, 11, 19, 23, 26, 27, 29, 31, 41, 44, 45, 48, 49, 51,

67–69, 82, 86, 87, 96, 99, 109, 114, 118, 124, 126–128, 131–145, 148, 155, 171–173, 177, 187 Tuberculosis (TB), vi, 67, 73–75, 80–82, 85–88, 135 V Value chain, 9, 125, 144–145 Virtualization, 2, 5–7 W Waiting distribution, 115, 116 Waiting time, 71, 73, 87, 98–100, 112–118 Waterfall, 181–188 Wearable, 9, 12, 22, 23, 25, 26, 31, 32, 49, 125, 133, 136, 141–142, 196 WHO, see World Health Organization World Health Organization (WHO), 66–69, 71, 73–76, 79, 82, 84, 87, 88, 107, 124, 127, 129, 131