Neal Khosla, founder of Curai, writes about some of the biggest challenges for digital health companies.
For a decade, digital health has been the supposed savior of the health care system, driving health care into a data-first, low-cost industry worthy of the 21st-century.
For that reason, investors have poured over $30 billion into digital health since 2011.
But almost a decade in, what material change can we point to in health care costs or the experience of the average patient? Are there companies that qualify as major disruptors? To me, the answer is no. And I call this “the Digital Health Conundrum.”
First, let’s consider some of the digital health companies that might qualify as big wins:
Teladoc, a publicly-listed telemedicine company? Well, adoption of virtual doctor visits and utilization among users continues to be low at less than 10 percent. Despite all the hype and fast growth, Teladoc remains an infrequent way to treat urgent care conditions. And this is fine, but hardly transformative. Ask any insurance executive whether telehealth has materially impacted cost or patient access (spoiler alert: they’ll say no and they might even say it has increased costs.)
What other digital health companies could qualify? Livongo? It’s got a $2 billion market cap, down almost 50 percent from its July I.P.O. Health Catalyst? $1.2 billion in market cap and down since IPO. Phreesia? It just crossed the $1 billion market cap threshold. These companies pale in comparison to the massive wins elsewhere in software (Stripe, Uber, Airbnb, Square, and others) and the jury is still out on all of them. They have incrementally improved the efficiency of existing resources, but in my view they have not dramatically improved or scaled healthcare expertise.
Livongo has perhaps been digital health’s biggest success, but it feels like we should see more change in the industry given the investment.
If health care is such a large market ready to be disrupted by technology (17 percent of our GDP spend), where are the winners? Why are the few wins we can point to incremental rather than transformative?
Let’s face it: digital health has been one of the most disappointing investment areas of the last two decades. Venture investors have yet to formulate a strategy that works in health care. The approaches from technology entrepreneurs have been underwhelming.
Most either fall into the camp of “tech for tech’s sake” (i.e. the first generation of digital health efforts like Google Health) or over-indexing on the suggestions of so-called “health care people.”
My favorite examples of the latter are companies working on AI scribing. Yes, charting is a huge problem for doctors. Removing it would be a win. However, this isn’t transforming healthcare, it’s putting lipstick on a pig in my opinion. Insider approaches lead to incremental solutions; insiders don’t point to true innovation, they point to what they know. Incremental solutions don’t lead to venture scale change.
A quick aside on Medicare Advantage
This post addresses the payor/provider world, not pharmaceutical companies or biotech. I also ignore the pockets of value-based care, or Medicare Advantage (MA). While there is innovation in MA due to improved incentives, it is not how the bulk of the U.S. health system operates. The aligned incentives could create big wins. This opportunity is policy driven so I would argue it fits into the framework of this post.
So why isn’t digital health working?
The tech approach and the venture model have been marginal successes in digital health. The lack of original thinking on how to approach health care has surprised me. Venture investors, despite their financial incentives to find home run startups, want to back predictable go-to-market strategies. Few people want to take risks in their approach and instead have banged their heads against painstaking sales cycles and organizations with structural resistance to change.
Venture investors shy away from strategies with a 20 percent chance of being transformative and an 80 percent chance of being a zero. Why? Investors don’t have a healthy relationship with risk and want to see predictable progress. This is why investors love software-as-a-service or SaaS business — they’re understood and repeatable.
The common refrain to tech people entering healthcare is to “learn about the financial incentives.” This is a red herring. Learning about the financial incentives of any one player ignores how that player fits into the rest of the system. Even inside of an individual organization there are numerous players with their own incentives. The complex structure of the health care system and its organizations makes teasing out go-to-market opportunities almost impossible.
Here’s my take on what the real challenges in working with existing healthcare enterprises are:
1. Provider systems
Provider systems, including doctor’s offices and hospitals, have two primary challenges in adopting digital health products: misaligned financial incentives and a change-averse culture.
The fee-for-service model is viewed as the primary blocker for change, but it’s not the only misaligned financial incentive. Provider systems do a lot of revenue with minuscule margins. The median operating margin for health systems in 2018 was 1.7 percent!
FS Productions | Tetra images | Getty Images
Imagine doing $1 billion in revenue and taking home $10 million in profit. Some startup comes along and wants to change your operational processes. Why would you ever take this risk unless you’re 100% certain it will work? These organizations have optimized the crap out of their processes to achieve that 1 percent margin. A $250,000 contract materially damages that profit. The tolerance for trying things isn’t there, nor should it be. Health systems often live month-to-month. Consider this list of 19 recent hospital shutdowns. Now try articulating why they should adopt your solution. The impact of your solution represents a catch 22: if it’s too small they have bigger fish to fry and if it’s too large it creates too much risk.
Health systems also have change-averse cultures with complex internal relationships. There are discrepancies between the desires of primary care physicians and those employed in specialty care. There are discrepancies between the priorities of the health informaticians and the providers. Administrators prioritize different things than both providers and informatics folks. Buy-in and adoption require all these parties.
Someone from a large academic health system once told me, “we’ll kill the project due to a tie in the faculty senate if it’s one hundred for and one against.” Physicians have been trained since medical school to avoid malpractice at all costs. What does this all result in? Stagnancy. This change-averse culture pervades health systems as it should. In what world is this the kind of environment where venture scale change can happen? If you’re restricting your product to administrative processes your chances are better, but affecting care is challenging.
Insurers are more open to change, but are difficult customers due to their inability to change provider group behavior, lack of direct connection to the patient, and, again, misaligned financial incentives.
One of the fundamental challenges of working with payers is their inability to affect provider behavior. Except for pockets of value-based care, anything that needs changes in provider behavior to save costs or improve patient experience/access requires a separate value proposition for the providers. Providers have little incentive to adopt anything new. This kills a number of impactful businesses.
Take the case of the start-up Call9. Call9 had an amazing premise: give patients in skilled nursing facilities (SNF’s) access to telemedicine so doctors could prevent patients from unnecessary trips to the ED. The value proposition here is insane! SNF patients are an expensive patient population. Nurses are on hand to help assess the patient and communicate with the doctor. The patient’s medical history is known by the SNF. Saving a trip to the emergency room can be tens of thousands of dollars and the telemedicine unit costs are a few hundred dollars. In June, Call9 shut down. Why? Among other things, it was difficult to get SNF’s to participate despite payers benefitting from the cost savings.
Timothy Peck lived in this conference room for 3 months
Even when companies have gotten distribution through payers they have struggled with customer utilization. Teladoc markets its 8 percent utilization rate as 4 times the industry standard. Imagine Slack marketing that 8 percent of their customer’s employees used the product!
Folks at one successful company that sold through Self-Insured Employers told me they did bespoke marketing campaigns for each new customer. The adoption challenge becomes a cascading risk with a long enterprise sales cycle AND a consumer marketing challenge. Reaching the consumers is difficult: insurance databases with customer information are out of date, have incorrect addresses and contact information, and issues with duplication of patients and missing records. Good luck!
Last, medical loss ratio restrictions, among other things, creates negative incentives for insurance companies to reduce costs. If you’re interested in digging in, this post by Sidney Primas explains why insurance companies benefit when the cost of care increases. Here’s the salient part on the “80/20 rule”:
[I]nsurers can [by law] at most keep 20% of the money they collect from consumer premiums for profit and administrative costs (e.g. salaries). So, if insurers want to increase profits, they need to increase the total amount of premiums collected. However, they cannot just increase premiums since insurers need to maintain competitive premium rates as compared to other insurance companies. The solution is allowing the cost of care to increase.
What is missing from healthcare that has driven tech-enabled growth in other industries?
In healthcare, go-lives, implementation periods, and pilots grind adoption to a halt. Contrast this to the learnings from the SaaS world that the best software companies start as self-serve. Self-serve software never happens in healthcare. Why not? What’s missing to enable self-serve or other forms of rapid progress in healthcare? Three things come to mind: a lack of early adopters, a lack of platforms for low-cost experimentation, and the challenges of integration into the system.
1. There are few small organizations that act as early adopters of new products. This has been exacerbated by consolidation in healthcare. Physician practices, where they do exist, are not set up to adopt new software tools the way that small businesses are. On the insurance side, it is hard to even imagine the concept of small early adopters: insurance is a business of scale. New plans like Oscar Health and Clover Health may be the closest thing, but they are still figuring out the blocking and tackling of insurance as they scale.
Oscar Health leaders Josh Kushner and Mario Schlosser.
2. Continuity of care and plugging into the existing healthcare system is near impossible. There have been few platforms for building healthcare “apps” in the way we use API’s like AWS, Plaid for fintech, or Stripe for payments. This is starting to change with the rise of companies trying to build a platform layer for digital health applications like Eligible and TruePill. The biggest problem, of course, is data interoperability, which hasn’t yet been solved. This would let standalone digital health apps to exchange patient data with health systems and vice versa. Integration into these systems requires a long development cycle because of shoddy endpoints and risk aversion around sharing data.
3. Distribution channels prevent building customer-centric solutions. Enterprise players have incentives independent of what end users want (whether those end users are patients or physicians). The result is clunky user experiences that don’t get used. Hence the low utilization rates. Instead of feature checklists to enable enterprise sales (ahem, electronic health record companies…), healthcare needs products built with the customer obsession we see elsewhere in tech.
What are my takeaways? Everything is messed up in healthcare and digital health hasn’t been an exception. Change isn’t coming with linear approaches. We need radically new approaches or black swan events. We need people who think differently. Isn’t this what tech does best?
How do we change healthcare?
For over three years, I’ve spent the bulk of my time thinking about how to bring large scale change to healthcare. What I’ve come to believe is we need systems-level thinking about how to catalyze change, not singular focus on individual players and their needs. The complex interplay between healthcare players has caused this approach to fail. We call it the healthcare system yet our approaches try to solve for the individual pieces.
Creative solutions to cracking this problem exist. Perhaps the most successful go-to-market approach has been through Self-Insured Employers (SIEs). This approach was pioneered by companies like Grand Rounds and Livongo that noticed incentive alignment in SIEs. While this was great for a time, what we’ve seen is targeting the SIE as a customer has toughened as the market has become saturated.
The largest employers have heard the pitch “we’re going to use data and technology to increase the efficiency of care” hundreds of times. CFO’s and heads of HR are novices at analyzing healthcare costs and their internal incentives are to avoid paying for extraneous benefits that could get them fired. The burden of proof that you’re going to save costs has become high. They’re asking their Third-Party Administrators to handle these deals for them — and then you’re back to dealing with large insurance co’s. This is a market I’m less bullish on than most.
Full stack approaches have also lessened the burden of the complex interplay of incentives. While it’s early, I believe these can be some of the most disruptive companies in healthcare. Digital health companies like Omada Health, One Medical and others have had some success removing the challenges of provider adoption of tools by hiring their own providers and selling directly to self-funded employers and insurers. One of the challenges for these companies remains navigating an enterprise sales cycle and a consumer marketing challenge at the same time. More importantly, because of the focus on the enterprise, these businesses often lead with sales and marketing instead of product.
Digital health writ large has also struggled with finding the balance between virtualization and high touch. One set of companies have focused on delivering efficient services to low cost, low complexity patients who don’t need more care or change the underlying cost in the system (think of direct to consumer health companies like Hims, Roman, and others). Another set has focused on high cost, high complexity patients (i.e. multiple chronic conditions needing a lot of treatment and management) but run more or less completely manual services.
Consider the class of new primary care and medical management companies: ChenMed, Iora Health and Landmark Health. These are some of the most interesting companies in health care, but ironically they’ve been successful by being more intensive and manual in their care workflows. Instead of progressing towards less manual service, we’re progressing towards more! I will give this whole class of companies credit for developing innovative care models that improve quality of care and costs, but it’s a step forward in one direction and backward in another.
Other companies going full stack (Virta Health, Omada, and so on) are also heavily manual. In my experience, the industry’s dirty little secret is that “tech enabled service” means “service where we built some software that hardly makes a difference.” They still have a person doing most of the talking directly to the patient. Almost every company in the space is guilty of this. And this is another form of rampant incrementalism. People haven’t charted a path to dramatically changing the underlying cost structure of care delivery. One of the key questions is how we make the service of care delivery a good that we can mass manufacture. This is a key idea to unlocking an equitable and efficient future economy: the “goodification of services.” We need to dramatically scale healthcare expertise, especially when it comes to complex patients who need more access. We need near-zero marginal cost medicine.
There’s a lesson here though: you can drive change in healthcare if you’re willing to think about how the parts fit together. There’s a fundamental misunderstanding when people say health care is slow. Yes, the timelines are slow due to healthcare’s local nature and natural rate of change. The reality is when disruptions come to the healthcare system, they come fast. My questions: Can we find examples of that change? Can we understand their underlying causes? How can we recreate this change going forward?
1. Policy changes
Ahh, everybody run for your life! I’ve just screamed Voldemort in the middle of Sand Hill Road.
Now back to reality. In health care, policy changes can drive massive change and disruptive companies have benefited from this. Some good examples of this include the electronic medical record companies, Epic Systems and Cerner. Cerner’s market cap is over $20 billion and Epic, which has stayed private, is thought to have about half Cerner’s revenue. Both companies have been around since 1979 but didn’t become behemoths until the last 10 years or so. This timing reflects the growth of the electronic medical record industry, which has erupted over the last decade. Just look at this table, which shows adoption of the technology in hospitals in a 6-year span. Massive change. Venture scale companies. Fast.
Why did this happen? In 2008, the Obama administration enacted the HITECH Act that gave financial incentives (read: lump sums) to provider groups that adopted electronic medical records and also required adoption on certain timelines. Many companies have turned healthcare policy changes into large economic opportunities. I’ll also point to Oscar Health, which sprung up around Obamacare, and many of the Medicare Advantage startups.
A couple enrolls in Obamacare with an insurance agent in Atlanta.
2. Existential threats
If you don’t want to wait around for the federal government to figure things out, I’ve got good news for you: there’s another way to change the system. I like to call it existential threats. The other way people have been able to build venture scale businesses is by scaring the health care establishment. As discussed, the incentive for the system as a whole is to resist change, let the costs of care go up, and continue to profit from that over time. Thin margin corporations with large revenue streams resist change until there’s a black swan that represents, in practice or perception, an existential threat to their business. Turns out they start moving fast when they see this. What are some good examples of this? My favorite is urgent care.
We’ve seen about 9,000 urgent care centers in the US since the early 1990’s (source). Guess how many Starbucks we’ve seen in a similar time period? About 14,000. Brick and mortar healthcare growing at about 65 to 70 percent of consumer coffee. Take a moment to appreciate that. It’s bananas. Why did this happen?
It started as incentive alignment: consumers loved the convenience and cost of urgent care and so did insurers. What kicked it into hyper-growth was when health systems realized that these little mom and pop centers were draining their profit-making emergency rooms. Today, every major health system in the US has its own urgent care or is associated with external urgent care. When these players get scared, they move.
USCareWays Urgent Care
Source: Phoenix Sky Harbor Airport
Examples of fear prompting movement haven’t always been successful products. The best example was IBM Watson. Watson put ads on the New York City subway that spooked people into thinking cancer diagnostics and treatment were changed forever. Add in the Jeopardy narrative and it was a perfect storm. Watson Health had collaborations with the best oncology centers in the US: Memorial Sloan Kettering, MD Anderson, etc. MD Anderson spent $62 million on the collaboration. Of course, this all came crashing down when people realized Watson was vaporware and the marketing hype was just that — hype. In the end, I believe that Watson was a selfish, unethical, and incompetent foray increasing long-term skepticism towards technology in doctors and health systems.
Another great example is that of Intuitive Surgical. Intuitive has a $60 billion market cap and makes robots to improve surgical outcomes. The narrative is amazing. Robots are coming to replace your surgeon’s job with their steady motors and literal surgical precision. Intuitive has built a behemoth selling its Da Vinci robot into health systems. Here’s the catch: turns out, they might not be so good. Intuitive’s robots appear to worsen outcomes. They might have other serious issues, too. And here’s the problem: nobody cares. If a patient goes to UCSF and says, “Hey, I’m considering getting this surgery at Stanford instead, they have these fancy new robots, do you?” and UCSF tries to cite a bunch of data saying their surgeons are better, it’s over. The patient is going to Stanford. The narrative is captivating and the threat it represents to surgical profit centers means adoption from health systems is necessary. In my view, the shared trait between Intuitive and Watson? Incredible marketing machines based on existentially threatening stories, despite underwhelming technology.
Where are some of the existential threats we’re seeing in healthcare? One big area is direct to consumer health companies like Hims, Nurx, and Roman. Hims went from zero to $100 million in revenue in their first year of operating selling erectile dysfunction and hair loss pills. That kind of growth doesn’t exist elsewhere in healthcare. While it remains to be seen where these companies go, it’s clear they’re creating a lot of buzz, attention, and fear. Incumbents are unhappy about them, arguing everything from “they are pill mills” to “patient safety” to “it’s a toy.” Ever heard that one before?
In many ways, healthcare isn’t different than any other industry in its ability to resist digitalization. I’ve been asking deeper questions: What does the evolution of this kind of D2C model look like? How do we broaden the scope of these approaches to address serious and complex healthcare issues? What can we borrow from these models to spark large scale adoption and patient empowerment?
3. Inflation and incrementalism
The third way people change the system is by introducing inflationary or incremental things. Inflationary things work great in healthcare. One of the challenges of reducing cost is you’re taking away somebody’s revenue. When you’re inflationary, everyone makes money so people will adopt.
Incrementalism reigns in healthcare. If you want to sell into large healthcare players, you’ve got to spoon-feed them. They can’t understand grand technological visions so take a simple, well-known problem and solve that in a way they can understand. It’s hard to isolate simple yet material problems, but this is the best approach for getting a foot in the door. Unlike other areas, people have struggled to expand out from that initial beachhead. Patient Ping is an example of such a company. They’ve found an isolated, well-understood problem, and grown by solving it in a simple way.
The other way to simplify your story is to accumulate enough evidence that your solution works. Virta Health did this with published studies. This is hard and takes time, but it’s one way to paint something that’s not incremental as approachable to large players. These companies have long timelines and 7 to 8 year return cycles from venture investors don’t allow time for running a 2 year randomized controlled trial. The best middle ground might be Omada Health’s approach. They commercialized a tech-enabled version of the pre-existing research behind the Diabetes Prevention Plan (DPP).
I’m tired of the same old conversations about long sales cycles and promising companies with good products dying. I want people to think systematically about how to drive change into a system that needs it. Investors need to start taking business model risk — if you agree the underlying business model of healthcare is broken, why are you investing in things that fit into it instead of having a plausible chance to change the model? Founders should be shooting for bigger, more impactful companies. You need vision and understanding of why your startup can make a material dent in a problem that threatens to bankrupt our society.
Payors and providers should establish sandboxes, creating environments where small companies can test rapidly and prove their solutions work. Understand many of these will fail. Focus on developing products not testing perfect ones. Use your pockets of capitated care to enable powerful solutions that don’t fit into your traditional business model. Those pockets don’t have to be large: 1 percent of your patient population. Make sure these projects retain velocity by supporting them with the right internal champions. Allow companies to test and iterate fast in a responsible way.
Policymakers need to really understand the second order impacts of the incentive systems they set up and to talk more to people in the valley about how technological change could be enabled by policy changes. Things as varied as reimbursement changes for digital health tools, Stark reform, data interoperability, deployment of AI technologies, and support for consumer first healthcare approaches could all unlock new business models. Digital health policy might be one of the most effective levers we have to spark change.
This quest to catalyze change in healthcare has dominated my life for the last 3-and-a half years and it’s clear to me we’re on the precipice of big things. It’s time to stop thinking incrementally and being locked on what is. Let’s focus on what could be. There’s positive change coming to U.S. healthcare and we need to make it happen.
Neal Khosla is the founder and CEO of two health start-ups, Curai and Totemic.