38,99 €
Mehr erfahren.
- Herausgeber: John Wiley & Sons
- Kategorie: Fachliteratur
- Sprache: Englisch
The best source for cutting-edge insights into AI in healthcare operations AI in Healthcare: How Artificial Intelligence Is Changing IT Operations and Infrastructure Services collects, organizes and provides the latest, most up-to-date research on the emerging technology of artificial intelligence as it is applied to healthcare operations. Written by a world-leading technology executive specializing in healthcare IT, this book provides concrete examples and practical advice on how to deploy artificial intelligence solutions in your healthcare environment. AI in Healthcare reveals to readers how they can take advantage of connecting real-time event correlation and response automation to minimize IT disruptions in critical healthcare IT functions. This book provides in-depth coverage of all the most important and central topics in the healthcare applications of artificial intelligence, including: * Healthcare IT * AI Clinical Operations * AI Operational Infrastructure * Project Planning * Metrics, Reporting, and Service Performance * AIOps in Automation * AIOps Cloud Operations * Future of AI Written in an accessible and straightforward style, this book will be invaluable to IT managers, administrators, and engineers in healthcare settings, as well as anyone with an interest or stake in healthcare technology.
Sie lesen das E-Book in den Legimi-Apps auf:
Seitenzahl: 466
Veröffentlichungsjahr: 2020
Ähnliche
Table of Contents
Cover
Title Page
Introduction
What Does This Book Cover?
Reader Support for This Book
CHAPTER 1: Healthcare IT and the Growing Need for AI Operations
A Brief History of AI and Healthcare
Digital Transformation of Healthcare
AIOps Platform Strategy
Summary
CHAPTER 2: AI Healthcare Operations (Clinical)
Clinical Impact of AIOps
Security and Privacy
Summary
CHAPTER 3: AI Healthcare Operations (Operational Infrastructure)
Getting Started with AIOps
AIOps Platforms, Products, and Services Selection
Workflow and Event Management Design
Summary
CHAPTER 4: Project Planning for AIOps
Project Planning Requirements
Deploying AIOps
Summary
CHAPTER 5: Using AI for Metrics, Performance, and Reporting
System Performance Metrics
Information Technology Metrics
Developing Usable AIOps Metrics
Summary
CHAPTER 6: AIOps and Automation in Healthcare Operations
Automation, Workflow, Process, and Intelligence Design
Designing the Framework for Automation
Configuring and Using AIOps Automation
When Should You Use AI and ML?
Summary
CHAPTER 7: Cloud Operations and AIOps
Understanding the Cloud
Deploying to the Cloud
Managing in the Cloud
Summary
CHAPTER 8: The Future of Healthcare AI
The Dynamically Changing World of AI
The Future of AI
Artificial Intelligence and Healthcare Innovation
Future Innovation Merging Clinical and IT Operations
The Future and Beyond
AIOps, the Cloud, and Security
Summary
CHAPTER 9: The Convergence of Healthcare AI Technology
Overview of Convergence
Systems Integration
Convergence of AI, HIT, and HIE
IoT and AI
IoT Management
AIOps Management and Security
Summary
APPENDIX Sample AIOps Use Cases and Examples
Scenario 1: Failing Application
Scenario 2: No Access to EMR
Summary
Index
Copyright
Dedication
About the Author
About the Technical Editor
Acknowledgments
End User License Agreement
List of Illustrations
Chapter 1
Figure 1.1: Design, build, and run model
Figure 1.2: The modules of ITIL
Figure 1.3: A heuristics engine using AI
Figure 1.4: Convergence of platforms
Figure 1.5: Analyzing KPIs for AIOps
Chapter 2
Figure 2.1: AI, AIOps, and the cloud
Figure 2.2: Redundancy in the system
Figure 2.3: Healthcare delivery connects many people and systems.
Figure 2.4: AIOps helps solve issues with an EMR.
Figure 2.5: AIOps service performance metrics
Figure 2.6: AI converged
Figure 2.7: AI security framework
Chapter 3
Figure 3.1: Reviewing a scope of work
Figure 3.2: Viewing Operational Intelligence
Figure 3.3: Splunk Dashboard Executive View
Figure 3.4: Splunk Dashboard Detailed View
Figure 3.5: ServiceNow service management
Figure 3.6: ServiceNow Insights Explorer
Figure 3.7: Viewing the Dynatrace dashboard
Figure 3.8: Drilling into the data with Dynatrace
Figure 3.9: Service mapping dashboard
Figure 3.10: Service design with AIOps
Chapter 4
Figure 4.1: The Scrum framework's sprints
Figure 4.2: A sample project plan in Microsoft Project
Figure 4.3: A sample project RACI matrix
Figure 4.4: Communications plan
Figure 4.5: Cloud-hosted versus internal solution
Figure 4.6: ServiceNow event correlation
Figure 4.7: Event management and CMDB
Chapter 5
Figure 5.1: Building an MTTR storyboard
Figure 5.2: AIOps data collection
Figure 5.3: AI event process handling
Figure 5.4: Using Tableau
Figure 5.5: Using RapidMiner
Figure 5.6: Dynatrace performance metrics
Figure 5.7: Dynatrace custom reporting
Figure 5.8: Splunk
mstats
for CPU
Figure 5.9: Splunk
mstats
with spikes
Figure 5.10: Splunk
mcatalog
Figure 5.11: Splunk
msearch
Figure 5.12: Splunk Metrics Workspace
Figure 5.13: AI for ServiceNow ITOM
Figure 5.14: ServiceNow business workflow
Figure 5.15: Business services and ML
Figure 5.16: Multiple AIOps platforms
Figure 5.17: Translating to workflow
Figure 5.18: Clinical use of IT
Figure 5.19: Readily accessing clinical data
Figure 5.20: Real-time access to clinical data is supported by AIOps.
Chapter 6
Figure 6.1: Process workflow map
Figure 6.2: The CMDB, CIs, and service mapping
Figure 6.3: Horizontal discovery
Figure 6.4: Vertical service mapping
Figure 6.5: Service discovery process
Figure 6.6: Service mapping process
Figure 6.7: Manual discovery options
Figure 6.8: Pattern Designer
Figure 6.9: Using Operator Workspace
Figure 6.10: Viewing a critical alert
Figure 6.11: Alert of disk space issue
Figure 6.12: Viewing other dependencies
Figure 6.13: Viewing alert details and history
Figure 6.14: Alert Insight
Figure 6.15: Using Flow Designer
Figure 6.16: Creating automation
Figure 6.17: Splunk Service Analyzer
Figure 6.18: Splunk Service Analyzer tree view
Figure 6.19: Analytics and outliers
Figure 6.20: Splunk ITSI
Chapter 7
Figure 7.1: Traditional network connectivity
Figure 7.2: Cloud network connectivity
Figure 7.3: Hybrid cloud connectivity
Figure 7.4: Migrating data from a local source to the cloud
Figure 7.5: CA/Broadcom AIOps Service Analytics dashboard
Figure 7.6: ServiceNow dashboards and reports in the Cloud
Figure 7.7: Dynatrace dashboard accessed via the cloud
Chapter 8
Figure 8.1: The growth of AI and ML
Figure 8.2: Current challenges to AI/ML
Figure 8.3: Crowdsourcing big data
Figure 8.4: Remote consultation with a physician
Figure 8.5: A telehealth and AIOps blueprint
Figure 8.6: Using telehealth AI
Figure 8.7: Using AI, ML, and blockchain
Figure 8.8: Understanding SD-WAN.
Chapter 9
Figure 9.1: The convergence of AI and HIT
Figure 9.2: The use of interconnected systems
Figure 9.3: The convergence and use of HIE
Figure 9.4: The complexity of IoT in healthcare
Figure 9.5: Using IoT in the NICU
Figure 9.6: Use of an insulin pump and meter
Figure 9.7: Overview of IoT and AI usage
Figure 9.8: IoT management with AIOps
Figure 9.9: ServiceNow Flow Designer
Figure 9.10: IoT and server failure
Figure 9.11: IoT attack vectors
Figure 9.12: IoT sensors and healthcare
Appendix
Figure A.1: Review of failure map
Figure A.2: Network resiliency
Guide
Cover Page
Table of Contents
Begin Reading
Pages
i
xvii
xviii
xix
xx
xxi
xxii
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
259
260
261
262
263
264
265
266
ii
iii
v
vi
vii
ix
x
267
AI in Healthcare
How Artificial Intelligence Is Changing IT Operations and Infrastructure Services
Rob Shimonski
Introduction
Today's systems, network, data, and critical infrastructure are all dependent on becoming more efficient and “self-aware,” while also self-healing. Artificial intelligence (AI) in operational models has become the de facto standard of how services will be designed and deployed in the future. This fusion of AI and operations has spawned the newly infamous acronym AIOps. As operations becomes the foundation of integrity that critical enterprise systems rely on, it's no mystery that AIOps has become incredibly popular as of late.
Service delivery is the lifeblood of all companies looking to be and remain competitive. This is especially true in highly critical services such as those delivered by healthcare where lost time, mistakes, and red tape impact the patient experience negatively. With more push toward zero downtime and less impact to 24/7/365 systems, having the intelligence to weather this storm is one of the most exciting prospects turned reality of our time and can be fully realized with the use of AI-based systems. Using this book, you will not only learn about AI and its importance, but be given practical examples and advice on how to actually deploy it in your environment to achieve these positive outcomes. It is also within this book that years of experience not only in other verticals but in the specific nature of healthcare delivery becomes evident as you progress through the chapters.
In this publication we explore what AI and ML are and how they are used in companies today. While this book covers machine learning and artificial intelligence concepts in depth, it also covers the practicality of deploying them into your enterprise. Another critical concept that remains a constant through this book is what AI and ML concepts are and how they apply to the myriad of tools, systems, and services offered today where corporate executives and technical engineers need to understand the “how to” when they need to make a decision and deploy the correct toolsets for the best outcomes. There is too much confusion and not enough real data surrounding how to select a good tool (or tools) to get this accomplished in operational departments today. This book also focuses on the primary sector where AI and ML are creating some of the biggest impacts today: healthcare IT (HIT).
Some of the key target areas that this book covers in detail include the foundations of AI and ML and what you need to know to be able to recommend and then deploy the technology. Concepts such as AI and how it fused with IT operations and specifically healthcare IT are covered in depth. There is also a deep look into clinical operations and how infrastructure services and IT operations support the clinical role and how both can interact successfully for mutual advantage, leverage AI for maximum potential, and increase successful outcomes for clinicians and their client or patients. Key target areas include but are not limited to the following:
Healthcare IT
AI clinical operations
AI operational infrastructure
Project planning
Metrics, reporting, and service performance
AIOps in automation
AIOps cloud operations
The goal of this book is to help build confidence in deploying a technology that will radically change how operations are done today. It requires understanding the versions of available AIOps platforms and systems, the vendors involved, and what infrastructure is needed. Once it is completely defined and understood from this perspective, we will explore applications of AI and ML in specific settings (or verticals) such as healthcare and how to strategize for these deployments in a cost-effective manner. The key to doing this well is to know how to use project management fundamentals to create a successful project. Project planning is the primary focus here with all of the planning done up front before the rollout to maximize the ROI for these large-scale deployments of technology. In the healthcare setting, there is an enterprise operational use of AI, and there is a clinical use, and it's important to know how they differ and how they can be integrated. This is where true innovation takes place. Although the book does cover many concepts of clinical AI and how it works in the grand scheme of IT, operational use of AI for the integrity of IT and healthcare assets, proactive and automated systems based on learned data, how to gauge service performance, and ultimately how to better deliver services (service delivery) of healthcare via IT is the primary focus and how it relates to AIOps.
Lastly, the book will cover a brief history of AI from 30 years ago until today, including where it came from, where it grew from, and ultimately where it is going (the direction it is heading). It is important to know how AI, ML, and AIOps work with cloud technologies, IoT, and other emerging technologies so that there are no missed opportunities due to lack of knowledge. It is the goal of this book to prepare you not only to understand but to be successful at a current and future rollout, implementation, and ongoing support framework for operations using AIOps in your healthcare setting.
What Does This Book Cover?
This book covers the following topics:
Chapter 1, “Healthcare IT and the Growing Need for AI Operations,” opens the book by providing a brief but thorough history on artificial intelligence (AI), machine learning (ML), and healthcare information technology (HIT). The chapter brings focus to current operations and how HIT is expanding and growing and how the digital transformation of providing healthcare requires a more focused view on technology operations, infrastructure services, providing care through technology, and how to innovatively change the digital footprint to provide reliable services for patient care. Other important topics covered are how artificial intelligence operations (AIOps) brings these different functions together and gives the users of the technology more insight into their technology investments.
Chapter 2, “AI Healthcare Operations (Clinical),” builds on what was covered in Chapter 1 by providing a different view into AI and ML and how it directly impacts clinical operations. Topics such as intelligent cloud, data analytics, informatics, convergence, and other methods to merge innovative efforts between technology support and clinical operations under the AIOps umbrella are discussed in great detail. Other topics include the need for security in the clinical technology space and why service performance is critical to providing reliable patient care on stable systems.
Chapter 3, “AI Healthcare Operations (Operational Infrastructure),” covers the strategy pillars and fundamentals required to get started with developing, strategizing, conceptualizing, and selecting products and vendors for your AIOps deployment in your organization. Topics covered include creating the project scope for vendor selection; selecting platforms, products, and services from tool vendors; and sizing the request correctly. Product vendors such as ServiceNow, Dynatrace, and Splunk are covered to help you design the correct deployment for your enterprise. Other topics include event and fault management and how these functions tie into advanced workflow and automation topics to help bridge the gap between AI and manual intervention.
Chapter 4, “Project Planning for AIOps,” builds on the concepts learned in Chapter 3 when project scope was introduced to help with vendor selection. In this chapter, you learn how to finish building the project plan and how to bring AIOps into enterprises consisting of large infrastructures. Project management concepts are covered, such as how to select a good project manager, how to build the project team (and why it’s critical to success), what a good project plan looks like, and how to build a program into your portfolio. The chapter also discusses deploying AIOps in your environment using a project plan, communicating status updates, and keeping executives informed of project milestones.
Chapter 5, “Using AI for Metrics, Performance, and Reporting,” details what you need to know post-deployment. Once you have deployed AIOps and are running it in your organization, you need to ensure your return on investment (ROI) by covering service performance metrics, KPIs, CSFs, and other important metrics that show how your investment in AIOps is creating a positive impact in both your IT environment and your clinical environment. Using AI for metrics, performance, and reporting allows you to feel confident in your AIOps platform by looking at how well it is performing and by building and viewing dashboards that help tell a story of success. Other tools helpful to building and showing metrics are covered as well as what you can pull directly from tools such as ServiceNow and Splunk.
Chapter 6, “AIOps and Automation in Healthcare Operations,” discusses how to develop advanced automation for real-world healthcare operations. By looking at tools such as ServiceNow and others, building processes, designing workflows, and other automation functions, you get a full understanding of how this helps to reduce outages, increase visibility, and increase the availability of systems. Through warning detection, incident engagement, and event handling, the framework for automation is covered in great detail to help create good-quality control in your environment. The chapter includes advanced discussions on how to create and build machine learning into process automation, how artificial intelligence self-learns, and other advanced service intelligence topics.
Chapter 7, “Cloud Operations and AIOps,” covers the movement of operations from in-house solutions to hosted solutions with product vendors or other managed service providers. Regardless of where you decide to host your platform, the details of doing so are covered in depth. Other topics that are covered are the strategy of moving an operation into the cloud from an in-house solution, what offerings and service types are available, how to do a request for proposal (RFP), when cloud should (and shouldn’t be) an option, how to manage your instance in the cloud, and what you need to know about security of the cloud within a healthcare environment.
Chapter 8, “The Future of Healthcare AI,” brings about the next offerings in AI and ML technology to include telehealth services, the Internet of Things (IoT), and the continual merge of clinical operations and the reliance on IT systems. Other topics include Big Data, DataOps, analytics, and informatics, which are crucial in today’s health environments where the size of stored (and protected) data is growing exponentially every day. The use of AI, ML, and AIOps to manage this data and these new technologies is becoming more and more important, and vendors are looking at new ways to build their services that allow for these innovative drivers to develop into AI solutions for enterprise management and monitoring.
Chapter 9, “The Convergence of Healthcare AI Technology,” covers more advanced topics that include convergence of critical systems managed by AIOps. As more and more healthcare operations rely on IT systems, the deployment of overarching enterprise management and monitoring solutions becomes more and more apparent as well as important. In this chapter, we cover AIOps for systems integration and overall systems management; the convergence of AI, HIT, and HIE; and other systems such as IoT end points.
Lastly, the appendix, “Sample AIOps Use Cases and Examples,” shows real-world examples of problems in the healthcare environment in the form of use cases that are solved by many of the topics in this book. Understanding what this book covers can help provide guidance on how to navigate real challenges such as outages that can be healed by AI technologies to reduce impact to the system and potentially save lives in the process of doing so.
Reader Support for This Book
If you believe you've found a mistake in this book, please bring it to our attention. At John Wiley & Sons, we understand how important it is to provide our customers with accurate content, but even with our best efforts an error may occur.
To submit your possible errata, please email it to our Customer Service Team at [email protected] with the subject line “Possible Book Errata Submission.”
CHAPTER 1Healthcare IT and the Growing Need for AI Operations
Shall we play a game?
—Joshua (from WarGames)
In today's ever-changing business model of do it faster, do it better, and do it without flaws, there needs to be a balance between those who create the technology and technology having a mind of its own. As in the famous quote “Shall we play a game?” healthcare operations is anything but a game and lives hang in the balance. In today's organizations, that balance is being established as technologies such as artificial intelligence (AI) are being implemented. There needs to be a way to do more work efficiently and with greater intelligence while still ensuring that the work is performed correctly. With the boom of healthcare advancements and the need to keep up with technological change, those who rely on all of the newest technology for clinical operations need an enterprise system that ensures that the technology continues to work for us and not against us. That combination of technology and clinical advancement comes in the form of a successful merging of intelligence and strategy, using the correct tools for the job, and planning and designing a platform that works for you, not against you. This is AI operations (AIOps) in healthcare.
This chapter explores the healthcare market and how technology continually changes it, specifically within the realm of AIOps. In these pages I will discuss the growing need for technology in this space, how healthcare has been fundamentally (and forever) changed by the digital landscape, and all of the specifics revolving around AIOps. This includes how AIOps is being used to create efficiency, reduce downtime, increase time to respond to issues, improve the ability to automate efforts to reduce waste and time spent doing computation work, and ultimately create better customer experiences for all patients, clinicians, and everyone involved in the healthcare space.
In the first portion of this chapter, I will cover the basic history of artificial intelligence (AI) and machine learning (ML). Although some could say we have always been in a perpetual state of “machine learning” for as long as we have had machines and in a constant state of computational (or artificial) intelligence as long as we could compute things, there are some significant milestones in the ML and AI timeline. For one, as long as we have been playing games, there has always been a study of game theory and outcomes through games. Many military and war strategists believed in game theory, and this became even more apparent when IBM began testing ML theory with gaming to produce the first machine learning game in the 1950s when someone played checkers and the program was able to learn from the outcomes of the game, the players' choices, and so on. I think this real story was likely the predecessor to the movie WarGames decades later. Checkers, chess, backgammon, and other games were all tested to see how a machine could learn.
As more and more technology (machines) was created and advanced, the same questions and theories were applied to it. When cars were made, how could we get them to learn? What about if we made a robot? Could it learn? The same theories from a long time ago all waited until technology caught up and provided for computers, robotics, and other major technological advancements that could be fused with machines to allow them to learn. Once computers were added to cars, then cars could start to learn. Now we drive in cars that can predict a possible crash and take action. This development went way beyond the abilities of game theory, but it should be noted that the mathematical equations, usage, and logic behind it still remained the same. It was only advancing as quickly as the technology did and was expanded on.
Another major installment of ML and AI development came with the World Wide Web (WWW), the Internet, and the Internet of Things (IoT), where the interconnected nature and development of all of technology was able to fuse and share data as well as save it. The saving of large quantities of data (or big data theory) allowed for more math to be applied for machine learning capabilities. Also, the growth of large-scale search engines (like Google) continued to allow for even more ML and AI abilities due to the analytics that could be applied to “customize” an experience for every user. Augmented reality (AR), wearable technology, DNA collection, mobile technologies, social media, and so many other advancements bring us to a stage in life where ML and AI are able to be used not only in any one of these advancement areas but also across them as they too interconnect.
This is where we begin our journey into the development and fundamental layout of AI, ML, and AIOps in the healthcare world. Healthcare is the largest user of all the technologies I just mentioned and the biggest connector of intelligence usage to increase the use of treatment, medicine, and patient satisfaction. We now need to know how to keep all of these systems running and available, continue to perform their computations, and allow for the continued interconnection and learning to provide the best care possible now and in the future.
A Brief History of AI and Healthcare
No industry is bigger, more important, more dynamic, and exploding with change than healthcare. Some may argue that it is similar to the industrial revolution in its transformative scope. The electronic medical record (EMR) and other technological advancements are changing the way we see, deliver, and expect to receive our healthcare. One of the most interesting things about the healthcare industry is we are all our own clients, customers, and patients, which makes our industry unique in that it is something we hope we never have to use but absolutely must have in the form of benefits to cover our families and ourselves. Working in this field can be the most rewarding experience one can have, and seeing its growth and being a part of its transformation can be a once-in-a-lifetime experience.
As the healthcare field grows in every aspect, we must consider the technology used to bolster this revolutionary expansion. We call this technological field healthcare information technology (also known as healthcare IT or HIT). In this chapter, I will explain healthcare IT, some of its history, and why technology has expanded it exponentially. I will also start to talk about the need for another popular and growing technological advancement called artificial intelligence (AI). Beyond that, we will bridge the two technologies—healthcare IT and AI—and explore a third topic called healthcare operations so we can create a fusion between them all, which is known today as artificial intelligence operations (AIOps). Let's begin our journey in this chapter and this book by starting from the beginning, which is the expansion of healthcare as we know it today.
THE CORONAVIRUS AND COVID-19
When the coronavirus (COVID-19) pandemic spread around the world in 2020, it affected the world of medicine in a dramatic way. As I will cover in Chapter 8, “The Future of Healthcare AI,” there have been radical changes to the use, delivery, and expectations of healthcare service.
With guidance by the World Health Organization (WHO), Centers for Disease Control and Prevention (CDC), and local, state, and government leaders, the world population had to make radical changes to slow the spread of COVID-19. For one, all healthcare systems (hospitals, practices, etc.) needed to find ways to reconfigure to handle the surge into the emergency department (ED) and intensive care unit (ICU) areas of the system. Temporary facilities were set up, and many new ways of practicing emergency medicine were considered. New medications, new treatments, and a world of new research into finding a cure were put in place. The world actively worked to slow the spread and focus on a cure. New technologies emerged, the importance of keeping systems up and running found a renewed priority level, and the use of older technologies saw a resurgence, like telemedicine.
Healthcare IT Expansion and Growth
The radical expansion and growth of IT and healthcare IT was just the beginning. It provided the needed building blocks to get to where we are today so that we can collect data in large amounts (big data), analyze it, and make predictions and assessments to create better outcomes. Big data analysis helps our knowledge and handling of population health, the need to reduce hospital stays, what can be done inside an acute versus nonacute facility, and how to make predictions on outcomes in a geographic area. An example of a prediction in an area would be how seasonal flu may impact certain areas and why that may be. This only scratches the surface of what we can leverage big data for.
To get to this point, we needed to get all data into the computer systems. By creating the electronic medical record (EMR) and having clinical staff adding this data to these systems, the building block was in place to start expanding this practice across all health systems. This provided many benefits, one of which was leveraging the resources of the many instead of the few. The increasing cost of healthcare put a lot of stress on smaller hospitals that could no longer afford to continue to build technical systems while still expanding their operations outside of information technology. Other important factors include (but are not limited to) security, risk, and compliance where privacy became paramount by law.
Compliance, meaningful use, regulatory bodies, inspections, laws and even the dominance of the Health Insurance Portability and Accountability Act (HIPAA) passed by Congress in 1996 continued to drive more and more healthcare providers to join forces with others to share resources so they could stay afloat. One of the many benefits was the ability to leverage the administrative functions (like technology) between providers, clinicians, healthcare facilities, practices, hospitals, and even insurance. Another advantage was the chance to scale up on all of these shared resources, which allowed hospitals and practices to share operational data so that they could model best practices and standards to keep all of these technology systems operational, resilient, and well-positioned for future innovation.
BIG DATA AND ITS IMPORTANCE
Data is the fundamental building block to everything that we do in technology. Think of a simple, traditional network. A network is useless unless you have something to share. Would you spend all of this money to set up countless connections simply for the sake of having connections? Of course not. You need to share data from printers, faxes, email, text messages, and files.
Big data is the same building block for AI, ML, and AIOps. Without data, what are you collecting, and why would you need it in the first place? The key to doing mathematical and computational analysis is to find ways to solve equations, and when you apply large volumes of data and a plan for finding trends, you wind up with predictive analytics. Analytics is what you do to mine the data for whatever specific information you need. So, if you are trying to see where there are systematic flu outbreaks in an area and why the flu outbreaks are occurring there, you need to collect data, analyze it, and create a mathematical prediction based on the trends. Big data is the large-scale collection of data that would need to be mined to identify trends when you conduct your analysis.
Data Overload
The many sources of data and the need for it to be shared started to become overwhelming. IT environments continued to grow exponentially, outpacing the underlying resources needed to run them. For example, extremely large enterprise storage system platforms with terabytes of space and high-speed delivery became essential to keep up with the needs for storing and leveraging data, particularly for the data scientists who looked at, analyzed, and computed this data (big data) for many use cases.
Data overload is a potential issue. You can collect a ton of data and not know what to do with it because you lack the strategy, tools, understanding, applications, services, or staffing needed to unlock what is within it. You may also run into issues where you cannot collect, store, mine, or access in a reliable way the large data sets you collect. You may also run into issues with sharing these large data sets. For example, I have seen in the past where research teams conducting important studies could not share files with other research teams remotely because of their large size. What people may not know is that (as an example) a magnetic resonance imaging (MRI) file with large slicing capability could be terabytes in size depending on the resolution of the file in use. When sharing hundreds of these over a typical 10 megabit connection with virtual private network (VPN) encryption, you could expect issues!
Unorganized data is another potential issue. When you work with data, you need to tag it correctly so that it can be accessed via field searches in databases. If you do not tag it correctly, when you run scans to find keywords or other criteria, you may not find what you are looking for, or worse, you may miss large portions of data, which can skew or create kurtosis of your results.
Unusable data is a related concern, meaning the data is missing entirely (orphaned), misunderstood (outliers), or corrupted and therefore unusable.
In any of these cases, you can have a data grooming exercise conducted to make sure that a sample set of the data shows that these problems may or may not exist and be fixed if possible. This is generally part of an organization's master data management (MDM) program. Regardless of what state your data is in, you need to recognize that it could be problematic and cause issues in your deployments, use of AI and ML, and other future endeavors of innovation. One of the most important things to consider when conducting any mathematical or computational exercise is that your data is usable and trustworthy.
CLEANING UP YOUR DATA WITH MDM
Data is the fundamental building block to everything that we do in technology. The best way to handle your data is to start fresh and make sure that your organization utilizes a master data management (MDM) program. An MDM is more of a governance committee made up of both technology and business leaders that puts rules into effect on how all company data will be tagged, listed, used, controlled, stored, and named (conventions). This includes identifying who has access, who owns what data, who maintains the data, and who ensures the lifelong accuracy and accountability of all data shared within the company. Without a program of this kind, you will run into issues later when your data grows so huge that it will take large project-like efforts and create massive disruption to address these and other issues. If you plan on working with data (which is the building block of AI and ML and thus AIOps), make sure that you know what the state of your data is before you begin innovative projects that require the use of big data.
Digital Transformation of Healthcare
The digital transformation of healthcare takes two paths: the clinical and the operational. Although I will cover them both in this chapter, the focus of the book really only goes deeply into one of these two paths. Although they are intertwined, they are in fact different when it comes to their stated purpose and scope. Make no mistake, they are both connected, and one may argue that you can't have one without the other. Case in point, if you do not have a stable environment in which to work (AIOps), how can you have a platform to conduct informatics to prevent illness or disease? On the contrary, why would you have a large AIOps platform if you weren't doing this type of the work in the first place? You can also argue that without the development and growth of AI and ML over the years, the fusion of this technology into newer tools like event and fault management platforms, service management tools, and help desk systems would be pointless when it comes to building on AI and its benefits. As you can see, they are both interdependent.
To further explain these two paths, path 1 is the clinical informatics space where AI and ML are used to do informatics work. Big data and its dissection are the future of clinical healthcare. As we collect more and more data, successfully analyzing it allows us to predict behaviors, solve clinical problems, create cures, or create new treatment plans. Looking at this data can tell you historically how many times the flu came through an area and at what impact rate so you can make sure that you market the flu shot effectively or logistically ensure that you have sufficient vaccines in that area. The use of this data is important to innovation and research for moving healthcare forward into the future.
NOTE For the clinical side of AI, ML, informatics, and using data to create positive outcomes, there is a natural fear that the storage and usage of this data would somehow create a privacy concern. However, all healthcare systems, providers, and users of this data are mandated to only use copies of the data that have been de-identified for use. This means all identifying data that can link any medical data to a patient by name or any other form of ID is removed completely before any data is used to study. This allows for privacy to remain intact.
Path 2 is the AIOps path where AI and ML are the focus of using enterprise platform systems that keep all of the data up and running alongside the EMR and every other clinical system, application, program, and system in use in the health provider environment.
For example, consider a patient needing to go in for an annual checkup, complete with routine blood work and urine sample. The doctor reviews the results of blood and urine tests to give the patient an assessment of the current state of their health. The doctor may also look at the patient's chart to see what they are predisposed to, their family history, their age, and many other factors before giving advice, a clinical path to follow, medications, or a referral to a specialist for further work. The patient's chart is reviewed to see what their last checkup or blood labs showed, and trends and patterns are examined so that nothing is missed. Various systems support the bloodwork lab, the urinalysis lab, the pharmacies, the specialists, the patient's historic files, the patient's current general condition, the ability to recognize patterns and predispositions, and the ability to project potential areas of concern, and AI and ML play a role in synchronizing and coordinating all of this data.
One of the reasons why understanding this is so important is that it sets the basis for keeping all of these operations up, running, stable, and operational. All of this technology must be highly available, recoverable from outage or disaster, and manageable. This can be done through traditional methods, but it can also be done with artificial intelligence. There are pros and cons to both, and we will explore these considerations throughout the book. Remember, the reason we want to design and set up an enterprise system like AIOps is so you, your family, your community, and your world can get the quality healthcare you expect and, in some cases, demand. If any of these systems are down, not operational, or corrupted, you can get delayed care, no care, or, in some rare cases, the wrong care, which can bring about life-altering experiences.
WARNING When dealing with a pandemic like COVID-19, the stability of infrastructure, the integrity of systems, and the reliability of key applications are beyond mission critical. They are always a priority, but with the need to keep healthcare workers focused on providing clinical care during such a dire crisis, the use of AIOps can be instrumental if deployed correctly. You need to strategize, configure, plan, deploy, and use AIOps to achieve that goal.
The Science of Healthcare Innovation
So, with all of this technology, what are we really striving to do? Much like Maslow's hierarchy of needs, we want to reach enlightenment and integrity. Once we have learned how to survive and take care of the fundamentals such as stability and integrity, we ultimately want to innovate. Innovation is the hallmark of any civilization that has moved to the highest stages of self-enlightenment.
With healthcare, the goal will be to have healthcare IT systems that run 24/7/365 with a five nines (99.999%) uptime and immediate resilience to any disaster whether technological, weather-based, military, or biological. We also want to reduce cost, waste, our environmental footprint, and our need to react to system issues. Lastly, we want to create efficiency, innovation, and the ability to rely on the systems we deploy and use. That's a lot to ask for, right? Well, this is the underlying goal that has spurred the entire market called AIOps. AIOps provides the ability to do everything I just said and more. With AIOps, your key systems will remain stable, and intelligent decisions will be made through machine learning to reduce impact related to downtime. Downtime impacts your ability to innovate.
To have innovation, we need our experts focused on their jobs. They should not be waiting for systems to recover from outages. They should not be working on how to connect users to systems as they recover. Our clinical experts shouldn't be using pen and paper during an outage and then afterward painstakingly entering that information to the EMR without a single data error. All of this is a waste of time and energy that creates missed opportunities for greatness. It also increases the potential for error.
Can we remove all of this waste and create innovation that we can all benefit from? The emergence of a technology platform that delivers AIOps promises to fulfill this role, among other things. Does it come with its own issues? Yes, and I will break them down and how to overcome them throughout this book. Regardless, the biggest goal of AIOps today is to create the ability to not have to manage failing systems and give that time back to those who would use it to create magic like innovative new healthcare solutions.
NOTE There are many roles an IT expert can play in healthcare IT such as analyst, data scientist, technical professional, systems engineer, risk manager, and so on. Everyone in IT should be enjoined in AIOps, and all parties should be stakeholders to any project where AIOps is deployed and used in an enterprise. Others who are to be involved would be any clinicians, providers, or other staff who will use the systems. Including their voices helps ensure that systems will be developed and implemented reliably in your enterprise.
Artificial Intelligence in Healthcare
Before we can really delve into the platforms that deliver AIOps, other supporting platforms, offerings, designs, strategies, and use cases, you should have a firm understanding of the fundamental concepts of artificial intelligence, machine learning, operations research (OR), and other technologies that I will refer to throughout this book.
First, artificial intelligence is exactly how it sounds. You have a computer system (or form of technology) that is able to learn, grow smarter, or be self-aware. It has “intelligence” between its programming language and expected behavior. This is a far cry from the Cyberdyne Systems' Skynet in the Terminator movies. Will we one day have the type of adaptable, self-aware, artificial intelligence technology system like we see in science fiction? We don't know. However, for here and now, the answer is no, we have nothing like that. Today's AI is not Skynet. AI today is still in its infancy, and while it is mature in the sense that we have adapted it into many of our technology systems, the term artificial intelligence is more of a play on words than a truly conscious technology system.
In simple terms, current AI can be thought of as computer technology such as programming languages, applications, and systems that emphasize fabricated, simulated intelligence through patterns and trends. It is not conscious, but it is programmed to be intelligent. I think this is an important distinction to make right up front to make sure that you are working from the beginning with the correct definition and expectations.
As we develop AI systems, we can think of them as behaving as if they have consciousness. In doing so, it can help to further divide AI into hard AI and soft AI (also known as strong AI and weak AI). Soft AI is simple. It is usually purpose-built and performs a single or simple function. This can be thought of as a program that you operate such as a game that tracks and learns your skills, patterns, and game-playing behaviors. Hard AI is something more complex with deeper programming, abilities, functions, and architecture. A great example of hard AI would be Google Brain. Google Brain is a hard-AI, deep-learning, AI-based research program formed in the early 2010s by Google that incorporates the concepts of machine learning with the large-scale computing resources of an enterprise as large as Google. Hard AI such as this can be more successful at simulating intelligence because of the complexity and power scaled behind it.
Either way, whether hard or soft, AI is still something that needs to be developed, worked on, built, deployed, and managed, and once everything starts to really work well, it needs to be managed some more. This is an important point to keep in mind as you embark on your journey into the world of AI in healthcare. AI requires real time and effort. So, now that we have covered AI, how does machine learning fit in?
Machine learning is the science behind the AI machine. ML can be thought of as the mathematical algorithms, statistics, underlying models, and data that powers AI. ML can be seen as a subset or component of AI where AI is what is used to make decisions and ML is what gives AI the ability to learn new things to make decisions about. ML has many components and relates to many topics such as data mining, analytics, informatics, and AI. An in-depth discussion of these topics is outside the scope of this book, and each of them can be adapted into a book of its own. I will, however, point out how healthcare AIOps and ML cross-connect and that is that there is a need for ML to be the underlying technology in AIOps for the technology to work correctly.
NOTE There is another concept to note, which is operations research (OR). OR is a “decision-tree” concept of AI where ML is the underlying science to create data for decision making, AI is the system that allows the decisions to be made in an automated way, and OR is considered the overlay to contrast with AI in that it is the decisions that are made with or without the automation. The “intelligence” part comes from the automation of actions, which we will cover in depth when we talk about the enterprise systems that allow AIOps to become a reality.
Now that you have a good understanding of AI, let's fuse this concept to healthcare. Healthcare is an ever-changing, evolving, dynamic field with many components. I like to divide healthcare into two separate realms when discussing AI: clinical and operational. These concepts were introduced earlier in the “Digital Transformation of Healthcare” section, and we will explore them further here.
Clinical AI is the use of AI, ML, and other intelligence components for clinical outcomes. As an example, if I want to trend what areas in a state are highest for the flu each year so I know where to stockpile medicine, I can data mine that information and make a clinical decision based on the results. This is the part of AI where ML allows us to use data mathematically to make decisions. Within the clinical realm, this use of AI and ML is still in its infancy, but major efforts are under way to develop innovative new technologies, tools, programs, systems, and workflows to more efficiently and effectively use clinical AI. Since this book focuses on the “operational” artificial intelligence realm, I will not delve too deeply into the clinical side except to make comparisons or show how clinical and operational AI at times will cross paths. However, it is important to know that AI in healthcare is very much a budding field where the positive clinical outcomes we seek and the innovation that is required to realize them is only just emerging.
AI in healthcare in the operational realm, or AIOps, is where most of our discussion in this book will focus. Operations can be defined as functions that keep the day-to-day business moving forward and ongoing. You can think of installing a server as a project where the management, maintenance, and patching of that server is the operational component needed to ensure the integrity of service. Not just healthcare, but all verticals, channels, fields, and roles require an operational component. Figure 1.1 shows the build-out of design, build, and run, which is a model that is commonly used to describe how we normally run an operation.
Figure 1.1: Design, build, and run model
The design, build, and run model is used when we want to do something like deploy a service (for example, a medical service like a new EMR system). First we design the service, then we build it, and then we run it. The design phase is where we put together a strategy, create a project plan, procure funds, put a team together, and so on. The build phase is considered to be an engineering component to the workflow. We make something that needs to be managed or a service that needs to be maintained. At the end of the build process, the service is tested and, assuming it functions properly, is then signed off as production ready. Next, we run it. This is the operational component to the workflow. Now, it enters into monitoring systems so that it can be safeguarded from failure.
This is normal operational behavior for designing, building, and running anything, and in the healthcare space it is no different. However, there is one major difference to note that is a priority. The biggest difference that makes healthcare IT, healthcare operations, and other functions of healthcare service delivery more important than most other fields is that human lives are on the line. It is critical to keep in mind that there is a real “life or death” aspect to the things we do in healthcare. Therefore, when we monitor systems for use in a healthcare setting, having them operational and ready for use is an absolute top priority. This is where AIOps meets healthcare operations and helps to not only bolster that priority but act on it as designed.
USING ITIL IN HEALTHCARE OPERATIONS AND AIOps
Although this book is specifically about healthcare AIOps, being aware of how operations are generally built for the AIOps platform will help when you design and deploy your operations. One of the most widely used models that establishes IT best practices and guidance that you can use to deploy tools correctly is the information technology infrastructure library (ITIL). ITIL is a model used to help create a sound and reliable IT organization and is an important framework for modeling how to deploy reliable IT services. For example, there are five major sections to this model, which include service strategy, service design, service transition, service operations, and continuous process improvement. Each component of ITIL digs deeply into how a service is created all the way from inception through transmission and into production. It also creates an iterative function with continuous service review so that things can be improved especially as organizations and their services change. You can see an example of this in Figure 1.2.
Figure 1.2: The modules of ITIL
The important takeaway here is that, as I noted with the design, build, and run model, services and systems need to be created well and managed in an ongoing and proper way so that they provide the required services. In healthcare operations, this becomes the fundamental factor of importance. When deploying AIOps, frameworks such as ITIL can help to ensure that once you get your services designed and deployed correctly, AIOps can deliver real value as an overlay and not just add confusion to poorly managed services that need some attention and focus.
Keep in mind that you can't build a good house on a weak foundation. Make sure that you consider the design suggestions listed in this section prior to your AIOps deployment to ensure that the house you build will be stronger and last longer.
Healthcare IT Operations
So, how does all of this fuse together to connect technology, healthcare, operations, and artificial intelligence? When you have a platform that can predict and respond, you have reached enlightenment. Today's operational systems usually lack interconnectivity and the ability to leverage the benefits of convergence. If we want to successfully navigate the roads of healthcare IT, we must build and pave them correctly. Only through proper planning and implementation will we deliver AI into healthcare IT operations.
The use of IT in healthcare is already deeply rooted and will continue to grow. Working in healthcare with paper files has been replaced with the EMR system and its interconnected world of computer technology that provides a clinical facility or doctor's practice more insight into how to provide care. Keeping that functionality up and running at all times can be complex and present an array of challenges. For example, if interconnectivity between systems is not operable, you may be missing components of the healthcare service you are trying to deliver. Let's look at some real-world scenarios where this might happen. If you want to get the results of bloodwork, you need the connection to the lab that processes that information, possibly with the proper credentials for security purposes. In another instance, if you want to prescribe medications, you need connections to the patient's preferred pharmacy (or the proper department if you're providing them within a hospital or other medical facility). In prescribing these medications, if your system has artificial intelligence that can check the patient's current medications and flag potential contraindications between them and your new prescription, that would be very helpful. To make such clinical decisions successfully and to the greatest benefit possible, you need the full, accurate data and functionality of all parts of the system.
