When companies first try to explore the potential of new technology, they tend to think within current frameworks, but often miss its deep implications. Echoing the quote attributed to Henry Ford—“If I had asked people what they wanted, they would have said faster horses”—we start using new technology to improve what we already do before realizing that we should question what we do in the first place.
With hindsight, the most significant effect of data-driven technological progress has been the transformation of the system (or game) itself. Consumers have experienced this change in many industries such as media, transportation, and retail. So far, this has yet to happen in healthcare. Our use of technology in healthcare has been superficial at best. For example, we’ve used IT systems to computerize paper processes that take filing cabinets, load their contents into computers, and call those electronic health records (EHRs).
Today, a number of powerful technologies allow us to define some of healthcare’s greatest challenges in terms of data and how we manage it. Through genomics, wearables, and digitally connected consumers, we can generate health “signal” at a scale we had never dreamt of. Through advances in machine learning and artificial intelligence (AI), we have the ability to reason about and utilize this data at an unprecedented scale in order to predict, prevent, and treat disease more effectively.
“Through advances in machine learning and artificial intelligence, we have the ability to reason about and utilize data at an unprecedented scale in order to predict, prevent, and treat disease more effectively.”
Similar to the mobile industry in the early 2000s, we’ve reached an inflection point in healthcare. Today, the healthcare “product” is so inferior relative to what is technically possible that a combination of superior products and the entry of powerful players not beholden to the status quo—Apple, Google, Amazon, and others—have the opportunity to remake the landscape.
As it has done in other industries, technology will remake the definition of healthcare itself. At the center of this transformation is data and how we harness uncertainty through machine learning. It will progress non-linearly, rapidly quickening after a painfully long buildup. And most importantly, this transformation won’t fix what is broken about our current system—it will redefine the system itself. It won’t improve things we dislike about hospitals; it will change how we even conceive of them.
So, what will a system-level transformation of healthcare look like when we see it? What rules will guide its progression? And how will AI and other advanced technologies accelerate the process? I believe history holds a number of important clues. Below, I discuss four core principles through which modern technology—and data—remade other industries and marketplaces; and why I think the same rules will soon make healthcare today as unrecognizable as dropping quarters into a pay phone or calling for a cab.
1. The Technology Isn’t The Product
Early fans of a technology treat it as if it were a product before realizing it has the power to transform industries.
When a new technology enters an industry, early applications tend to be first-order applications—basically thin wrappers around the technology. These first applications are rarely the ones that have the biggest impact or market value. One example is the advent of GPS APIs in mobile phones, which enabled the creation of location-based mobile apps. The first wave of these, “check-in apps” such as Foursquare, were briefly all the rage but barely remembered today.
Once a technology is assimilated into culture and people almost forget its presence, the real value of the technology can emerge. One of the most valuable company founded since 2008—Uber—is a geo-based application, but no one thinks of it that way. Uber would not exist were it not for your phone’s ability to track location. And even though GPS is a fundamental building block behind Uber, it is a small part of a much, much bigger application.
“Once a technology is assimilated into culture and people almost forget its presence, the real value of the technology can emerge.”
This principle applies to healthcare. The introduction of big technological building blocks into healthcare is still early, which is why we think of many of these in a first-order way.
Take genomics, a sector where many companies are developing genetics applications and initiatives, such as DNA testing for cancer risk. The most important applications driven by genetics in the years to come won’t be thought of as genetics applications. Instead, we will have much better versions of building blocks of healthcare that will produce dramatically better outcomes.
Every day around the world, millions of people take the wrong medications and suffer adverse or sometimes lethal reactions. We chalk most of this up to bad luck, yet they are anything but. Your genome and information about medications you take deeply inform how you will react. Both sets of data are readily available today.
Machine learning and AI will eventually close the loop to harness the predictive power of the data to ensure you don’t receive the wrong medication. The pharmacy of the future will reliably ensure you receive prescriptions that are appropriate for you, thanks to genetics and AI. That will happen automatically—you’ll simply expect that from your pharmacy the same way you expect your car’s collision-avoidance system to prevent a crash.
Similarly, the emergency room will soon morph into something very different. Rather than waiting for symptomatic escalation, many medical events will be detected and intercepted upstream. For example, we know that rapid weight gain in certain populations precedes heart failure. But instead of a heart attack bringing you to the ER, tomorrow’s ER should come to you before it happens. You’ll just expect that from new emergency services powered by AI, wearables, IoT, and tons of streaming data.
2. Tech Remakes The Building Blocks
Technology creates new units of value creation.
One of the system-level impacts of technology is that it often redefines the boundaries, or set of functions, for the building blocks in our world, whether those are people or tools.
Many of us grew up with phones being bulky items with a rotary dial that plugged into a wall. Today, of course, a “phone” is nothing like its predecessors—it’s a pocket-sized supercomputer. It hasn’t just become mobile, it has also subsumed other tools we once needed—cameras, notepads, calendars, stereos, Rolodexes, voice recorders, and even computers to some extent.
When functions like those jump from one building block to another, people take a while to fully utilize new capabilities. When phones first replaced everyday cameras, they simply copied basic camera functions—we didn’t yet recognize what you could do with a camera embedded in a tiny computer that was connected to the Internet. Instagram eventually connected the dots: It used the phone to make photos more beautiful and spread them across social media.
In healthcare, we treat doctors a bit like old rotary phones. They remain the core building blocks of health delivery, but haven’t been given an opportunity to redefine their functionality through technology. (It should come as no surprise why thousands of doctors today suffer from job burnout and many are leaving the profession.)
“In healthcare, we treat doctors a bit like old rotary phones. They remain the core building blocks of health delivery, but haven’t been given an opportunity to redefine their functionality through technology.”
Scores of health practices are designed around managing these scarcities—a physician’s time and her tools, such as MRI machines; and her limitations, such as how much data the human brain can process. We sacrifice another scarce resource—patients’ time—for better utilization of doctors’ care. The fact that waiting rooms account for 10–25% of healthcare real estate shows how big that sacrifice is.
How does this relate to technology? Look at a couple functions we still associate with what doctors do, such as measuring your heart rate or writing down your basic health information on clipboards when you visit a doctor for the first time. If you think of checking your heartbeat as a physician’s task, it’s extremely expensive and limited—the doctor listens for a few seconds at a time. If you wear a watch with built-in ECG features, you replace the old-school measurement with 24×7 heart monitoring year-round.
Similarly, collecting health history on forms during office visits happens infrequently and includes only coarse measures. With a digitally connected population, these questions can be answered well ahead of time without doctors or nurses involved. Your phone could easily alert you to health issues if it detects a shift in sleep patterns, for example. More importantly, making sense of all the data to preemptively detect problems is precisely the task at which machine learning excels.
Critical heart data has now become infinitely cheap. Collective knowledge on your general health is no longer bound by your physical presence in the waiting room. And the data has become dramatically easier to interpret through machine learning systems that derive insights at a scale and accuracy far beyond human capabilities.
These are just two examples of what will soon be a long list of functions doctors can offload to technology over the coming years, and in doing so, dramatically scale the powerful insights medicine has accumulated over the years.
3. Data Will Turn Healthcare Economics Inside Out
The value of data is driven by utility, not scarcity.
When we estimate the value of data, we often do it within the confines of its current accessibility and utility. But technology makes data free to replicate and distribute. When we encounter a previously constrained data set, we tend to underestimate its utility at scale once it has become freely accessible.
In the early days of the Internet, copyright owners were terrified about potential theft of their content. They assumed that before we started creating new content for the Internet, we had to figure out how to transfer all the existing “good” content already available. (Remember Encyclopedia Britannica?) But the Internet created such a massive source of demand for new content and reduced the barriers to production that new content was created de novo.
One perennial healthcare topic is how to deal with EHRs and the wealth of data and insights that they supposedly house. But it is important to realize a few things about EHRs.
Doctors once used filing cabinets where they kept a file for each patient to track their health records. EHRs are primarily a digital version of the same thing. That affords a few convenient features, such as backups, tools to help a doctor with workflows (clinical decision support systems) or help navigating the complexity of healthcare billing.
But EHRs were conceived in a world that predated the current generation of digital technologies, which I believe will completely change where the “critical mass” of health data resides in the first place. Today, for most consumers, unless they are in treatment for a disease, there is far more health information outside the health system than inside it.
“Today, for most consumers, unless they are in treatment for a disease, there is far more health information outside the health system than inside it.”
You can wear an Apple Watch, which generates continuous signal about your heart, activity, and sleep. You can sign up for genetic testing to combine data from your genome, family history, and lifestyle in order to determine your risk of cancer and heart disease, as well as how you metabolize medications. Each time your phone sees your picture it may be able to detect fatigue, weight changes, or other telltales about your health. Think about it: Walmart, the grocer for millions of Americas, likely knows a lot more about your dietary balance and alcohol consumption so can better predict your likelihood of becoming diabetic than any health system.
Each of these information sources has two important attributes. First, you own and control them. Second, they are virtually free to generate. If you’re looking for just one place where machine learning can shine in healthcare, look for the places where it taps into new torrents of high-signal information rather than the expensive and messy data in EHRs.
As we move more toward disease prevention than treatment, the decision systems that drive healthcare will not be evolutions of the EHR, but new technology stacks. Software systems will tap into vast new data sets unencumbered by legacy systems and processes. The volume of health “signal” will grow exponentially. And our ability to reason about this data through AI and machine learning will turn this raw tsunami of information into powerful new insights.
4. Data Creates Direct Links To Value Creation
Technology enables new and more efficient marketplaces for value exchange.
In a pre-digital environment, the two sides of a marketplace are subject to physical constraints—atoms instead of bits. In a digital one, pools of supply and demand can be connected virtually. Advertisers once needed to purchase billboards to help generate demand for a product, but they offered little visibility into how they actually drove purchases.
Later, in the evolution of Internet advertising, the advent of search ads changed the game. Instead of showing you a billboard (physical or virtual) and hoping it would coincide with your needs, the Internet enabled a seller to connect with you at the time when you were trying to solve a problem. This brought supply and demand closer than ever before.
Today, the way we manage health outcomes is similar to how we once purchased billboards. For example, most countries recommend that women have an annual mammogram at the age of 45. They act as if women younger than 45 don’t have a risk of breast cancer and that on their 45th anniversary, suddenly they are at risk and need to be screened.
That isn’t connected at all to actual risk—or the value to the individual and society of early detection. Our health practices assume we have no way of knowing a woman’s actual risk or of interfacing with her to adjust her screening.
This is no longer a valid assumption, given we can use machine learning to effectively stratify risks for diseases such as breast cancer through a combination of genetics (what you inherited from your parents), physical attributes (for example, breast density), and behavior (smoking, alcohol consumption). Up to now, most of these risk factors were considered sporadically and in isolation, but can now be combined into a more uniform view of risk.
One of the biggest effects of a data-driven health system is the set of linkages that connect data to insights, and insights to outcomes. When we can connect the value of a mammography to a specific woman’s risk instead of assuming she is part of an average, the business model changes; we will pay for the results of our work instead of a loose connection to targeted outcomes.
When healthcare organizations create more links from data to outcomes, they create direct markets and sidestep a massive middle layer of inefficiency, opacity, and cost. Today, pharma companies are compensated through negotiated prices with payers, but without hard links to actually curing disease. What happens when those companies get paid only for the results of their drugs? They will start to differentiate themselves around driving outcomes. A similar evolution can happen with doctors and hospitals as they shift more into a data-driven care model.
“When healthcare organizations create more links from data to outcomes, they create direct markets and sidestep a massive middle layer of inefficiency, opacity, and cost.”
As we grapple with healthcare’s monumental challenges, we should consider whether our biggest opportunities are not just ones of medical science or technology, but ones closer to the ground—rethinking how to use technology and data to redefine healthcare itself.
I believe that we are entering the most exciting phase of technology intersecting with healthcare. So far, we have seen the use of technology as a thin layer on top of healthcare. Your experience with a doctor today is almost identical to what it was 20 years ago—except that instead of looking at you, the doc is begrudgingly documenting your exchanges into EHRs.
Based on the system-level shifts technology brought to other industries, we are entering the window in time where the landscape of health will start to be redefined. This transformation will encompass sweeping changes in the pools of data we rely on; in the functional building blocks of the “work” of healthcare, such as doctors, hospitals, and emergency rooms; and in the economic undercurrents and data streams that will reset how the marketplace rewards value.
Rather than thinking of technology as a way to just make it easier to do the “job” of healthcare today—to reduce friction and costs of existing tasks—we should all think more in terms of how technology, and especially AI, will reinvent the job itself.