AI-First

Pandemics like COVID-19 create severe strain on healthcare systems ranging from shortages of personal protective equipment, insufficient or inaccurate diagnostics tests, overburdened clinicians, and imperfect information sharing, to name a few. More importantly, any healthcare crisis such as COVID-19, or emergence of Human Immunodeficiency Virus (HIV) in the 1980s highlight the stark reality of shortcomings in our health systems. We must reimagine systems of care and healthcare systems, as healthcare crises accentuate current problems:

• Inequitable access to healthcare,

• Insufficient on-demand healthcare services,

• High costs and lack of price transparency

• Fragmented, siloed systems

• High business frictions and poor consumer experiences

• Record-keeping frozen since the 1960s and more

As we focus on these challenges, we should recognize these challenges are interdependent, giving the illusion of healthcare as complex when, in reality, healthcare is delivered through complex systems. But we can build better systems with less complexity, providing better care, and making the healthcare system work for everybody in healthcare. AI is a crucial enabler for simplicity in healthcare, building intelligent systems of care.

We think of AI as synonymous with machine learning, and by doing so, we don’t think about building complete AI systems accounting for processes, structures, experiences, and patterns of care delivery, which can be enabled by machine learning models, natural language processing and more. Framing the problem this way helps us understand and set the stage for developing AI systems, not just machine learning models, building simple, intelligent, robust systems that embody remarkable, frictionless experiences for all stakeholders in the healthcare ecosystem. That’s why we wrote this book, a primer for employing AI in healthcare without an exclusive focus on machine learning. Each chapter incrementally moves us further in understanding how we make AI the center of everything we do without focusing on AI, and this is what we mean by AI-First.

There is no magical solution or technology such as AI that fixes healthcare, just like there isn’t a single technology solving all problems in banking, retail, automotive, tech, or any industry. Repairing complex systems is not the answer; we must rethink how we build tools, experiences, and intelligent systems using data, ready for prime time that works for doctors, nurses, healthcare workers, patients, and care facilities. Today, in one process or one tool within one specialty, AI is showing up and ready for prime time identifying: cancer, eye diseases, abnormal imaging studies, the onset of Alzheimer’s, depression, and more. Think about the Internet and the shift to web pages, and then mobility as we moved to apps. Now with AI, we will embrace even more modalities, like voice, but the experience of how people machine interface must change along with the underlying systems.

Some truths to get used to:

• AI systems improve every day and continuously

• AI systems and tools may be the only way to accelerate healthcare to the underserved

• AI systems will get trusted as they explain themselves

• AI systems will endow a single doctor with the experience of millions of doctors

• AI like mobility is the way of life for children born post-2010

Each doctor’s mistakes and successes have to be learned by experience, and eventually becomes part of a standard of care and best practices. Doctors learn from other doctors, from research studies, they learn from drug and device companies pushing products, and they learn from the successes and failures with their own patients. Each doctor’s mistake has to be discovered and actualized and sometimes at the disadvantage of their patients. Some doctors routinely fool themselves by thinking a diagnostic is correct, or a treatment works when it’s contrary to evidence supported by studies or outcomes with thousands of patients. Sometimes a doctor is unaware of studies and evidence of new treatment care pathways or better diagnostic tips. But, today, doctors have direct accessibility to the experiences and best practices of thousands of cohorts; they don’t need to wait for best practices to be codified into standards of care. With AI, we can change this calculus and move at a faster scale than a single doctor or institution can do on it’s own.

The era is past when a doctor should be touting. “In my experience, this treatment works” and instead should be saying, “In my experience plus the experiences of hundreds of thousands of patients, fellow doctors, and clinical studies give me the confidence of pursuing this treatment path.” But how would a doctor have at their disposal, at their fingertips, the knowledge of thousands of clinical studies, experiences of hundreds of thousands of patient treatment pathways, and the collective experience of thousands and thousands of doctors? This requires technology; AI. As humans, clinicians are subject to cognitive and cultural biases, and we can minimize, maybe even eliminate the impacts of some such biases within AI.

AI can evolve faster, spreading best practices, cumulative knowledge, and experiences of hundreds of thousands of doctors at the doorstep of a doctor more swiftly than any previous technology. But AI needs to be embedded in our entire healthcare ecosystem spreading like wildfire, as ubiquitous as electricity so it can be used and expanded to lift the practices and skills of every medical professional, for this reality to materialize. This is what we mean when we use the term AI-First. But for AI to truly drive solutions for the most pressing problems, we have to think about how to develop holistic, intelligent systems, AI systems, not just machine learning models. We must think about the structures and processes which include patterns of diagnostics, treatments, and care delivery that can be enabled by machine learning, computer vision, natural language processing, ambient computing, and more.

Our journey of describing AI-First starts with a chapter describing AI and ends with the section telling how to make it real in healthcare at scale:

Chapter 1 - Myths and Realities of AI

In order to understand what AI-First means, we must first understand what AI is and what it is not. We must describe the myths and realities of AI, understanding the art of what’s possible. Most tales have some threads of truth, but we describe them as myths because they are either false or misleading. AI and machine learning are used synonymously and interchangeably, and this can be good and bad. Machine learning is critical to the success of building AI systems, but AI systems can be a lot more than a collection of machine learning models. Computers getting better at tasks previously only done by humans does not mean machines are getting smarter and smarter, moving towards human intelligence. It does mean we have the tools to build intelligent systems.

Chapter 2 - Human-Centered AI

The conversation about machines surpassing human intelligence is both nuanced and philosophical more than science-based. The dark and dystopian views of AI taking over the world or replacing doctors causes us to lose sight of what we can do today rather than a future we can envision but don’t have the technology to make real. Our real threat is not super-intelligent machines or systems but dumb systems. Dumb systems promote black boxes creating friction, fragmentation, and lack of interoperability. Today, the evidence is overwhelming that superior results are obtained with the pairing of humans and machines. Human-centered AI is our opportunity to usher in a new era for healthcare where access is improved, and the chance for people to live healthy lives is improved.

Chapter 3 - Monitoring + AI = RX for Personal Health

The opportunity for people to have a more significant role in their healthcare has never been higher with the proliferation of personal health gadgets, intelligent medical devices, smart wearables with sensors that monitor people’s vital signs. Infuse AI and combine with sensory abundant ambient spaces, and a prescription emerges for improving personal health. Invisible computing is emerging as we see AI in our everyday lives like the Apple Watch, looking at your heartbeat to check for an irregular rhythm that may be atrial fibrillation (afib, a risk factor for stroke). This will expand to more ordinary things like your toothbrush taking saliva samples and detecting changes that might alert people if they are at risk of metabolic disease. The arrival of smart and ambient spaces in our homes and places of work creates a future where non-invasive technology married with AI becomes a tool, a prescription for keeping people healthy.

Chapter 4 - Digitization and AI

The delivery of care should be transparent to all constituents in the healthcare ecosystem and access to services coordinated with all parties in real-time. Back office systems like claims and prior authorizations should operate like other industries in near real-time or real-time. Real-time health care must be the norm, not the exception where outcomes are immediate, and prior authorizations should behave the same as credit card authorizations, operating in seconds for the vast number of transactions. Digital healthcare starts and continues with humans, people. This requires digital platforms, the kind we see with Internet and cloud era born companies. Adoption of AI and accompanying technologies can make this a reality. Digitization requires understanding and adopting AI, not just machine learning models.

Chapter 5 - An Uncomfortable Truth

An uncomfortable truth must be addressed in today’s healthcare, the enormous amount of waste. Suboptimal consumer outcomes from clinician visits should and can be lessened with AI. Evolving to a human-machine based healthcare system, where the agency is solely in the hands of people, not “Dr. Algorithms” or “AI Doctor” will dramatically improve patient outcomes. AI and technology must be ever-present but invisible to reduce errors and waste. There is overwhelming evidence of the value AI brings in reducing waste in health care, and this chapter examines how leveraging AI can facilitate the solutions which improve efficiency and reduce waste.

Chapter 6 - Emerging Applications of Healthcare

Applications for healthcare sit in the front office often visible to consumers or patients; they live in our pockets via smart mobile devices, we wear them, and they live in the back office of payers, insurers, and healthcare service providers. AI is upending all of these applications types, and new application types are emerging, some situational applications with short lives. All of these application types must be embraced with an eye toward making healthcare operate in real-time, enabling points of care to start with the patient when necessary, and enabling healthcare to be ubiquitous and on-demand. This chapter explores these new emerging application types.

Chapter 7 - Delivering AI at Scale

Making the promises of AI come alive will quickly be an order of magnitude more complicated than our shift to mobility for a number of reasons. It will start just as it did with mobile by recognizing we must embrace a new modality and new type of application. Every organization’s journey will be different, and this chapter provides a prescriptive approach to get starting whether the organization is big or massive.

Summary

The adoption and value of AI implementations are shaped less by the intrinsic attributes of AI technologies and innovations and more by the economics investing in it, so it’s not what AI can do for me, but more what specific benefits will be achieved from the investment in AI. AI-First is not about the investment in AI but more about why it must be seen as a general-purpose technology that becomes core to every technological adoption pursued. AI integration into healthcare is harder than it looks, especially when we seek the rebuilding of care systems, not just using systems making predictions with machine learning models.

This is a time to reimagine how we do healthcare, and it is the start of a transformation of the healthcare system with AI at its core.