Humans are complicated, and changing our behavior is hard. 1, 2 Despite all the hype about artificial intelligence and personalization, most consumer-facing behavior change tools are incredibly unsophisticated, relying on basic self-tracking and superficially-tailored feedback to change behavior. Techniques known to be highly effective, such as such as disrupting habit streaks, and linking contextual cues to behavior, 3, 4 are noticeably absent in most digital health products. Maybe this is why most of these products don’t actually work.

Using science to sell apps—but not to build them

Most consumer-facing apps are not scientifically informed. 5, 6, 7 And even when they are based on evidence, implementation of the science into the product is often poor. 8, 9 This lack of science, however, does not prevent companies from using science to sell apps. On the contrary: A recent review found that over 40% of the most popular mental health apps invoked scientific language to support their effectiveness claims yet just one of these apps linked to published literature. 10

Many of the solutions being pitched or sold to us are behavior change solutions – buy this, wear that, ingest these insights about yourself, and you will be freed from pain, sleep better and lose weight! But peel back the marketing claims, and investors and consumers alike may bristle to find out that—to mention just a couple of examples— sleep-tracking apps can make insomnia worse, 11 and the published benefits of a daily blood pressure monitoring app (whose makers just raised $12 million 12) were based on just 2% of those sampled. 13

An alarming number of companies are publicizing results using inappropriate statistical techniques. For example, conducting what is known as a completers- only analysis involves selectively analyzing only those data from people who completed the trial/experiment, and ignoring the data from people who quit. This approach makes it way more likely you will conclude your product is amazing, because the people who make it to the finish line are inherently more motivated. We want to see your intent-to-treat results, which include the people who dropped out. It is also worth pointing out that the same statistical rules apply to both big and small data. Yet amid the promise of big data, many people have grown increasingly comfortable eschewing the fundamentals. 14, 15, 16 We need to remember that methods matter too: Applying the right statistical analysis can’t overcome bad execution or study design.

The mainstream media, investors, consumers, and industry players have been sold on the idea that behavior change is one appropriately timed nudge away and that we can educate our way toward healthier living. The expectation that exists in the space is wrong: You cannot simply click ‘like’ to change your behavior. 17

So how do you change behavior?

Behavioral science can help us design for behavior change, and build technologies that not only spark change but sustain it. 18

Behavioral science is the empirical study of human behavior across the lifespan. It encompasses fields such as psychology, cognitive science, public health and economics. Behavioral science emphasizes how context, and the social and physical environments, play profound roles in behavior, beliefs, and decision- making. People are different, context matters, and things change. 19 This table lists some behavioral interventions shown to effectively address common health problems.

Links for references below correspond with number bubbles above.

20 21 22 23 24 25 26 27 28 29 30 31 32 33 34

Behavior change science can tell us what works and what doesn’t. 35 If digital health technology companies ignore behavior change science, they will fail to produce meaningful, long-lasting results. 36

Behavioral science + 21 st century technology

Behavioral science has undergone radical transformation in step with the technological revolution. No longer must we rely on humans to self-report what they are doing—now wearables and sensors can passively detect behavior throughout the day. And in instances where we must ask people questions, new technology-enabled methods can illuminate the behavioral context, such as what the person is doing while sedentary, having cravings, and feeling blue. 37

The data accrued from these newer methods are reinforcing long-held (but previously untestable) hypotheses about the non-linear nature of change. 38, 39 They also enable us to intervene at the right time. 40, 41 In other words, we can get closer to automating human support than ever before. These new data also provide strong evidence that conclusions drawn from group-level data are extremely imprecise for individuals. 42 Fortunately, we can now optimize interventions at the individual level and realize the power of behavioral phenotyping. 43

For example, let’s consider how people react differently to self-tracking. We know that some people reject it because of waning motivation – they get negative feedback and stop tracking themselves. Measuring individuals’ determination, resilience, and/or coping style can provide insight into who will benefit from consistent tracking and feedback. For those who respond poorly to consistent tracking, reframing failure as within one’s control can overcome the tendency to avoid. 44

Wellness apps: A behavior change opportunity

Historically, our focus has been on helping people once they get sick, focusing our investment dollars, product pipelines, and healthcare reimbursement strategies toward treating disease. On its face, this makes sense: There is a lot of sickness to treat. Six in 10 US adults have at least one chronic disease, and four in 10 have two or more. 45 By 2030, the number of adults with three or more is estimated to almost triple from 31 to 83 million. 46 Including lost productivity, the economic burden of chronic disease is estimated to be a staggering 20% of our gross domestic product or $3.7 trillion. 47

But, what if we helped people before they got sick? What if we could recoup some of that $3.7 trillion, and pump it back into our economy? Lifestyle behaviors drive most chronic disease incidence, morbidity, and mortality. 48 These behaviors include smoking, excessive alcohol consumption, insufficient physical activity, and poor sleep habits. 49 With the help of science, these behaviors can be changed. And while there exists a dizzying number of consumer-facing wellness apps that claim to change these behaviors, very few have published evidence to indicate that they do so in any meaningful way. 50

Shockingly, even highly valued healthcare startups do not publish peer-reviewed evidence on their apps’ effectiveness. 51 If we are to move the needle on health, this indifference toward evidence and limited use of science must come to an end. We simply cannot afford to fall victim to the illusory truth effect whereby we accept evidence to be true based on how often we hear it repeated.

Digital health won’t advance without behavioral science

Digital health companies should have behavioral scientists embedded in their product development teams from the very beginning. Companies should also include behavioral science in their medical affairs departments, for both evidence generation and strategic leadership. The most forward-thinking companies will have chief behavioral science officers. 52

Within product development, behavioral design should operate in concert with user- experience design. 53 Behavioral design involves translating science into products and services. In addition, behavioral science should be positioned to work alongside analytics and data science. Establishing behavioral-data science architecture is necessary for many reasons, including planning experiments and interpreting users’ engagement data. Digital health companies need to go beyond a focus on the number of monthly or daily active users and ask themselves, what effective engagement looks like. 54, 55

Four questions for evaluating a product’s behavioral claims

• What evidence supports these claims?

• How was this evidence generated?

• How is behavioral science informing product/service design?

• Who is responsible for pilot tests/experiments?

Remember: Behavior change is hard. Science helps.

In my 10 years as a medical executive in health tech start-ups, I noticed that medical/science people do not necessarily appreciate the difference between the different types of engineers or designers. To be honest, when I came to Silicon Valley, I considered all people who do anything with computers to simply be “IT people.” Likewise, many people from the tech industry don’t necessarily know the difference between the different types of scientists or the different types of medical doctors—or even the difference between a medical doctor (MD) and a research scientist (PhD).

Good Science Doesn’t Instantly Translate into a Viable Health Product 

Tech investors are increasingly interested in the $4-trillion US healthcare industry. Many are new to this sector. Although they use lawyers to do their legal diligence, they often use their own associates to do their “medical diligence.” Understanding the difference between medical training and scientist training helps to explain why the passion of a scientist entrepreneur may not necessarily translate into a viable health product or why the associate who did your medical diligence missed all the red flags that blocked your investment from widespread market adoption. 

Scientific validity is not the same as clinical validity, yet most of the literature provides scientific validity. In addition, not all MDs are the same. Some specialties get more training in clinical trials, while others develop more expertise in digital imaging, informatics, or device/test development. A dermatologist MD may not be the best person to handle medical diligence for a home laboratory testing start-up.

Medical Review Isn’t the Same as Scientific Review 

The confusion about medical and scientific training can impact efficiency in health tech startups. Once, after working at a particular startup for a couple years, I got frustrated at the commercial team for not getting medical review before releasing new collateral, despite multiple requests. When we met to talk about it, they explained that they did have medical review because a PhD scientist had approved the collateral. I said, “It’s good to get a scientific review, but it’s not the same as a medical review.” One of my teammates asked, “What’s the difference?” When I explained it, many of them seemed genuinely surprised. The purpose of this blog is to shorten the learning cycle on that epiphany.

(In a future blog, I’ll try to explain the different kinds of engineers to the healthcare industry. And I’ll tell them not to call software engineers “IT people.” I will also explain the difference between scientific and clinical validity, and what is required to bridge the gap between interesting science and a viable medical product.)

Table 1: The training and expertise of medical doctors versus research scientists.

Not All MDs Are the Same

The next table addresses the high-level differences between the different kinds of MDs. This classic joke sums up the stereotypes quite nicely:

“A medical student, an internist, a radiologist, a surgeon and a pathologist go duck hunting. They barely find their duck blind before the first duck flies over. The medical student is the first to raise his shotgun, but unable to tell if the duck is really a duck, he does not shoot. The internist aims his shotgun, but can not tell for sure what subspecies of duck it is and wants to order more tests. He does not shoot. He asks the radiologist to take a picture of the duck. The surgeon quickly raises his shotgun, aims, and without pause shoots. The duck falls to the ground. The surgeon turns to the pathologist. “Go figure out if that’s a duck or not.” Funny, right? Even other doctors don’t really know what pathologists do. Pretty sure my family doesn’t. Now I can refer them to this blog.

Table 2: Different types of medical doctors

Not All Lab Directors Are Created Equal

Some health tech companies choose to operate their own clinical laboratories, if their product includes a clinical test. The decision to operate one’s own lab rather than outsource the testing to an established clinical testing laboratory has important pros and cons that will be unpacked in a future blog. But one of the requirements of a clinical laboratory is that it is overseen by appropriately credentialed (and in some states, licensed) clinical laboratory professionals. So it is relevant to this blog to know who those professionals might be. 

Most types of clinical testing must be overseen and interpreted by a MD. However, a few types of clinical tests can be overseen by credentialed PhDs who have done a post-doctoral clinical laboratory fellowship. These include germ-line genetic testing, microbiology, chemistry, and toxicology. For many tech people, the difference between MD pathologists (who are trained to run all types of clinical labs) and subspecialty-focused clinical PhDs is not obvious. Using the example of genetics, the following table outlines the difference between the MDs and the PhDs.

Table 3: Different types of laboratory directors allowed by CLIA ‘88 regulations.

The bottom line: Know your experts and what each is best suited for. Tapping the precise expertise you need will get your health product faster into the hands of the people who need it.

In a recent Business Insider article, John Ioannidis, Racquel Bracken, and other health tech experts lamented that the Silicon Valley tech ethos of “move fast and break things” is not being counterbalanced by the healthcare principle of “do no harm. (Why everybody gets duped by hot health and science startups, June 2019). Ironically, tech’s desire to move fast is slowing its ability to achieve widespread market adoption and profitability for health-related products. The goal of this blog is to help both industries better understand the importance of their cultural differences in order to get high-impact products to consumers more quickly.

 The healthcare perspective is underrepresented in health tech. I spent the past decade serving as the Chief Medical Officer of several health tech startups. With each one, I was initially energized by the potential for responsible disruption. The engineering, design, and product talent—and the money—was unlike anything available to physicians in a traditional healthcare setting. It truly felt like an embarrassment of riches. With each company, I was the only medical professional on the executive team, and usually the only MD in the company. With each one, I knew that we could have been more successful more quickly if the company had had an earlier understanding of the expectations of the healthcare industry and sensitivity to healthcare culture. I observed this same phenomenon in many of the health tech startups evolving into and out of existence around me.

Tech Needs to Get Curious about Medical Culture

 When I left traditional practice to embed myself in Silicon Valley, I recognized quickly that my tech teammates spoke differently than I did. They solved problems differently than I did. They worked differently than I did. If I was going to effectively partner with them to disrupt health care, I needed to learn about tech culture. I read every Silicon Valley holy book I could find—The Lean Startup, The Hard Thing About Hard Things, Measure What Matters, The Four, In the Plex, Hatching Twitter, The Upstarts, The Everything Store, From Good to Great, etc. I learned about OKRs, MVPs, 20% time, flywheels, doom loops, BHAGs, pivots, dogfooding, flat org charts, and scrums. I read about the history of the tech industry and the stories of how the FAANGs became FAANGs. (In a future blog, we will unpack all these terms. For now, just realize that in order to respect a different culture, it’s important to learn about why that culture is the way it is.) Too often, my tech teammates did not reciprocate my willingness to understand their culture. This limited the company’s ability to design appropriate products, close deals, and compile realistic financial projections. It cost real time and real money.

 Granted, the tech industry has successfully disrupted almost every industry in the last 20 years. Tech entrepreneurs are somewhat justified in feeling like their way is the right way and everyone else should get out of the way. But the reality is that the healthcare industry is different. One of the key differences is that it is not a free market. Full stop. 

In addition, there are multiple, very powerful stakeholders in health care—insurers, regulators, physicians, professional societies, pharma, and patient advocacy groups. Consumers often don’t feel like they should have to pay for legitimate health products out of pocket, so simply delighting consumers and getting them to vote with their feet is not enough to achieve widespread adoption, let alone topple the incumbent system. Unlike other industries, a direct-to-consumer product can’t circumvent the incumbent stakeholders. That’s because consumers will inevitably take the results of their health products to their healthcare providers. If the healthcare provider says, “What is this? This is junk. You’ve been swindled,” it doesn’t make for a positive consumer experience. The flywheel cannot gain momentum. You simply have to understand the values and social mores of the healthcare stakeholders to win.

Being Awesome is Not Enough

 So, how do you do that? With data, credibility, and transparency. It is not sufficient to think your product is awesome and convince consumers of this. You have to prove that your product is awesome. And not just awesome, but medically beneficial and economically worthy enough to justify the high cost of implementation. Generating this proof needs to be part of your go-to-market strategy and financial strategy from inception. What’s more, there is a well-established formula for how to do this. That means there is no excuse for confusion or delay, if investors and founders do their homework.

An obvious way to ensure success is to balance your leadership team with the right tech and medical professionals. If you go too far in the medical direction, you are essentially a traditional device company. If you go too far in the tech direction, you will waste too much time and energy learning the basics of market access and product-market fit. In general, I recommend hiring most of your engineering, product, and design teams from tech and most of your commercial and medical affairs teams from health care. Medical affairs can help make sure your product solves a real problem in health care and design the proof studies. Experienced sales, business development, and marketing professionals not only have the right contacts, but they also know how to deliver the right messages through the right channels at the right time. A future blog will elaborate on how to do this.

Culture Clash

This table highlights some of the cultural differences I’ve observed between healthcare and tech. 

More details on the Nuremberg Code referenced above.

Finding the Right Balance

We need to find the right balance of healthcare and tech cultures to get impactful products to patients quickly and responsibly, and to successfully usher in the much needed disruption of the incumbent healthcare system. The longer we delay prioritizing this balance, the longer we all have to live with the current broken system, and the longer we have to watch billions of dollars wasted on health tech startups  that are doomed to fail because the founders do not understand healthcare culture—both are pretty unbearable!