Prelude

Investors have deployed over $100b into digital health since 2010. But not a single generational company has come of it. Almost every industry has birthed a $100 billion company, but digital health.

Over the next several weeks, I’m going to unpack why. This week I’ll start by breaking down the challenges with tech-enabled primary care.

Primary care (your family doctor/GP)

Many of the world’s smartest companies are trying to own primary care – Amazon, CVS, and Walmart, to name a few. The idea is that once you own primary care you can control downstream health expenditures.

At the same time, many individuals are wanting to improve primary care. I’ve had at least 10 friends tell me they want to start a company in the space. So this article is also for them, and anyone else exploring primary care as a business.

Today, I’ll summarise the challenges with the four models of primary care ‘innovation’:

1. Virtual & hybrid fee-for-service care, which was not economical, leading to

2. Membership-based direct primary care, which was too expensive for customers, leading to

3. Reimbursed value-based care, which was not scalable, leading to

4. AI-powered primary care, which is trying to optimize the wrong thing

In future articles, I’ll provide tactical learnings and business model breakdowns of the companies featured in this article and hint at what can be done differently.

Part I: The four stages of primary care

1: Virtual and hybrid fee-for-service care (Teladoc, One Medical, Amwell, Carbon Health)

Teladoc is off 90% since Feb 2021. One Medical was down 85% before its acquisition. Babylon was down 95% before delisting. Amwell is off 90%. These companies have struggled for several reasons.

Diseconomies of scale at the gross profit line

Virtual-first fee-for-service care has struggled to expand gross margins.

This is because these companies compete on either (a) access, (b) quality, or (c) price, which are all inimical to unit economics.

When competing on access, you have to operate with a surplus of doctors to meet spikes in patient demand, which worsens gross margins.

When competing on quality, the key was to let doctors spend more time with patients. Moving from 30 patients per day to 15 patients per day helped with customer acquisition, retention, and engagement, but ate into economics.

When competing on price, you charge less per member, but clinician salaries are fixed, which definitionally compresses gross margins.

Inverse operating leverage at the EBITDA line

Virtual-first care companies haven’t improved margins at the operating profit line. In fact, they haven’t even improved incremental margins — in many cases we’ve seen inverse operating leverage. This means that as revenue has increased, costs have increased not just in absolute terms but also as a percentage of revenue. Crazy!

This is because, unlike your old-fashioned primary care clinics, tech-enabled clinics have two entirely new line items, which are zero-sum and commoditized.

First, sales and marketing, which is zero-sum because many digital health companies fight over the same customers. While paid marketing might work initially, it has diminishing returns in the margin as there are only so many early adopters.

The second new line item was R&D. However, all the digital health players spent money on the same tools, and therefore R&D spend generated limited alpha – it was also commoditized.

Primary care is loss-leading in the US

75% of health networks operate their primary care business at a loss. Large players use primary care as a loss leader for other services: CVS for prescription drugs, Walmart for consumer retail, and Kaiser for surgical services.

It’s hard for companies like One Medical to compete with incumbents that can cross-subsidize primary care with more expensive services.

Revenue is fixed by the reimbursement model

Since revenue is mostly fixed by the reimbursement model, tech-enabled primary care became a game of operations optimization rather than quality improvement. Some companies like One Medical were able to negotiate favourable fee-for-service contracts, but others like Carbon struggled.

These challenges have resulted in a ‘flight to quality’ and a ‘flight to gross margin’, which ushered in the era of ‘direct primary care’, where patients pay a monthly membership fee to access ‘unlimited care’.

2: Membership-based primary care (Forward, Parsley, MDVIP)

To solve for ‘quality’ whilst maintaining margins, companies like Forward and Parsley emerged, which charged $100-$200 per month for unlimited care. Consults would go for over 30 minutes, clinicians would have patient panels of 600 instead of 3,000, and services like labs were in-house.

But at $200 per month, it’s hard to acquire and retain patients — they expect a top-notch service that increases costs.

Acquisition is hard and retention harder

It’s hard to find a critical mass of patients willing to pay $2,400 per year for primary care. CACs blew up to over $2,000 at times making it challenging to scale.

Whilst high CACs can be justified by strong retention, the problem is that retention was the opposite. If patients didn’t receive $100-$200 of value each month, there was the threat of churn.

Margins don’t improve much since costs go up

To acquire customers and reduce churn, these companies had to provide an exceptional service. That meant introducing high-cost items like a larger care team, 24/7 chat, longer consults, and consumables such as genetic tests, wearables, and unlimited blood tests.

There’s an even greater scarcity of high-quality doctors

Costs blew up even further due to the scarcity of high-quality doctors. High-quality doctors were already minting money in private practice. It was challenging and expensive to lure them away. Provider shortages made it challenging to maintain quality with scale.

Incentive misalignment

These companies had to battle incentivized to have patients visit less to reduce costs versus the incentive to have patients visit more to improve retention. It was never about “what is best for patients” and rather it was about “what is best for retention and economics”. After all, if you keep someone healthy, they should hardly need to visit at all. Healthy patients ended up cross-subsidizing unhealthy patients and hypochondriacs.

This resulted in a game of optimizing perceived value to drive patient retention. Services were staggered over a 6-month initial period to keep members long enough to pay off the CAC.

Reimbursement is challenging

The ‘fair value’ of a per-member-per-month fee is closer to $50 than $150. Payers, therefore, struggled to reimburse this care model. Without payer reimbursement, it was hard to scale since patients could get similar services elsewhere, for free. That said, companies like Parsley did a good job of accessing employer reimbursement by proving a 2x ROI via a 77% reduction in need for specialists and a 65% reduction in need for chronic disease medications.

Ultimately, reimbursement matters in healthcare. Whilst there will always be a small minority willing to pay out-of-pocket, the majority will take free healthcare over an equivalent paid service. To solve for both quality and reimbursement, the move was to value-based models, where, in simple terms, companies would make money if they kept patients healthy, and patients would get the service for free via payers, be that self-funded employers, Medicare Advantage, or Medicaid.

3: In-person value-based care (Chenmed, Oak Street, VillageMD)

Oak Street, ChenMed, and VillageMD pioneered in-person value-based primary care. The value-based payment model rewarded companies for keeping patients healthy. To simplify, the idea was that a company gets paid $1,000 for a patient, and then they bear all the costs associated with that patient’s healthcare. If they keep the patient healthy, out of hospital, and off expensive medications, they pocket the savings. But if the opposite happens, they bear the cost of care.

This dynamic means that value-based care companies see cohort margin expansion over time. This means that patients who cost $1000 in 2018, might cost $950 in 2019, $900 in 2020, and so on. The theory is that costs decrease and margins expand over time as the company gets better at caring for its patients. Companies use this analysis to justify poor (and often negative) margins in the present.

The kicker is that cohort margin expansion doesn’t continue forever — it asymptotes, as is the case with all competitive service-led markets. The reason for this is two-fold.

First, as soon as margins get too large, competition or the government forces margins back down. For Medicare populations, the government reduces value-based payments if providers are extracting excessive rents. For commercial populations, competition puts downward pressure on margins, and there’s no monopolistic scale advantage to offset this.

Second, the scarcity of high-quality doctors means that it’s harder to reduce costs as you take on more patients. Taking on more patients can increase your average cost of care rather than lowering it. In many respects, healthcare favors boutique, old-school private practices.

Ultimately, value-based care became a game of optimizing the traditional model of care and gaming the billing/coding system. New care models were unable to drive step-function improvements. Expertise remained a scarce resource. It proved challenging to simultaneously solve the trilemma of cost, quality, and access without material care model innovations. That’s what led to AI-powered primary care which aims to provide always-on, high-quality care at zero marginal cost.

4: AI-powered primary care (Curai, Babylon, K Health)

AI, to date, hasn’t met Silicon Valley’s hype. A large reason for this is that companies have been too obsessed with AI diagnosis, which misses the forest for the trees for several reasons.

First, the process of diagnosis takes a minority of a doctor’s time. The majority of it is in history taking, testing, empathizing, explaining treatment plans, and managing chronic disease.

Second, most consults don’t involve patients who are diagnostic odysseys. Most of the time, the doctor knows the correct course of action. And the cases which are diagnostic odysseys are the very cases in which a statistical AI system is least likely to work.

Third, AI tools are already becoming commoditized in care delivery. This is for two reasons. First, the marginal benefit is minimal for the AI startup that is 85% correct versus the incumbent at 80% accuracy because AI will be a copilot (rather than autopilot) for the foreseeable future. Second, foundation models will likely sit horizontally across healthcare players. These factors mean that distribution will matter more than accuracy. As tends to be the case in healthcare, incumbents win.

Where AI is interesting is where it can materially drive care delivery innovations versus just being used for diagnosis — Twin Health and Forta are interesting here, albeit not without their own challenges.

Part II: Summing up why nothing has worked

1: Incremental innovation is commoditized

All the ‘innovations’ to date have been incremental. Incremental companies end up playing fiercely competitive optimization games. This is because every company can incrementally improve. Everyone is doing AI differential diagnosis. Everyone is doing AI scribing. Everyone is doing AI care plans. AI isn’t the right way to productize healthcare (I’ll share a better way in coming months).

There is little defensibility in incrementalism. Incremental innovation is a commodity. It favors incumbents with distribution. It favors superior process optimizers. It favors economies of scale.

2: The cost to build and distribute is too high

In the past 3 years, over 1,000 digital health companies have wasted millions of dollars building the same features. Once they’ve built those features, GTM is notoriously challenging.

In other industries, platform functionality brought the cost to build and distribute down to zero, which increased the rate of innovation. The iPhone did this for mobile apps. Amazon and Shopify did this for ecommerce websites. We haven’t fully seen this yet in healthcare.

I think of the cost to build and distribute as similar to a ‘cost of capital’.

In capital markets, growth is related to the ‘cost of capital’. As the cost of capital approaches zero, innovation and risk-taking increase. For innovation to increase in healthcare, the cost to build and distribute healthcare products has to approach zero (I’ll share how this can be done in coming months).

3: Structurally, the system is not set up for innovation

Median operating margins in the US aren’t conducive to risk-taking and innovation, and neither are VCs’ risk appetites.

There’s not much risk you can take on a 1% margin.

At the same time, many VCs incentivize incremental innovation. Venture investors de-risk their investments by backing proven strategies that go something like this:

1. Tell me what population you are targeting

2. Explain who is going to pay for it

3. Show me which large health system you worked at previously

VCs aim to deploy fast money for “safe” founders, as if resumes are an investment thesis.

Despite evidence that the power law works, investors often prefer predictable outcomes incentivizing proven business models. When was the last time a generational company took that approach? Incrementalism rarely works.

Conclusion and What’s Next

It’s hard to change healthcare. But the hard is often what makes things great. If something was easy, someone would have done it already. Difficulty creates defensibility. The kingdom easy to conquer is hard to keep; the kingdom hard to conquer is easy to keep.

I’m keeping my eyes out for outsiders working alongside insiders thinking about healthcare from first principles. Outsiders bring a fresh perspective and haven’t internalized the dogma of how things are done. Outsiders unapologetically place consumers first. But insiders realize that you need to take into account the realities of reimbursement and systematic incentives. You can break road rules, but you can't break laws of gravity.

In the coming months, I’ll break down the business models of individual companies and start suggesting solutions.

Subscribe to my newsletter to keep up to date!