The Cost Equation for New Primary Care Models in Existing Frameworks (Part I)

Kevin O'Leary
Oct 10, 2017 · 14 min read

TL;DR: traditional primary care costs $25 PMPM, new models cost somewhere around $50 — $75 PMPM. For insurers or employers to be interested they probably have to at least breakeven or demonstrate cost savings, so they need to demonstrate $25 — $50 PMPM of cost savings to break even, which is a large challenge. Part I of this goes through the cost structure differences; Part II outlines the challenges in demonstrating cost savings & some thoughts on moving forward.

Innovative Primary Care Models In the News

New primary care models have been in the news recently — from Forward raising $30+ million, Google spinning out a Medicaid focused startup called Cityblock, to a number of companies shutting down (Zoom, Harken, Qliance), to Sherpaa recapitalizing and bootstrapping, to Iora Health and Oak Street Health growing like weeds in the Medicare Advantage space, to Oscar Insurance opening their own primary care clinic.

In some form or another, these primary care centric models seem like they have to be core to the future of health care — one where we reduce costs by paying for prevention instead of treating sickness. Having a well functioning preventive primary care system is the most logical way to achieve this from my perspective. These models are at the forefront of doing this, and anecdotally have great results, particuarly in the sickest / most complex populations.

However, these new primary care models face a challenge in that they’re stuck being a new model within the constraints of an old system. This presents a number of challenges to the adoption of these primary care models, particularly financially. We confronted a number of these challenges head on at Harken Health and weren’t able to overcome them, but here are some lessons I took away from the experience.

Needing To Interact With The Traditional Healthcare System

Most innovative primary care models I’ve seen attempt to integrate with the traditional healthcare system for one of two primary reasons:

  1. Primary care isn’t an island and needs to be integrated into the overall healthcare system.
    This is the most logical reason to work with insurers / employers — primary care can’t meet all of the needs of everyone and needs to refer people to other specialists. In order to do that, primary care models either formally partner with insurers and employers to integrate the primary care offering into a product (this is what Iora did with Harken Health, for example), and/or informally build a network of friendly physicians that they refer these patients to (this is what concierge / direct primary care models like Forward do). Coordinating this downstream specialist care, as we’ll see later, is actually one of the biggest benefits these primary care models can offer to the broader healthcare system. It’s why insurers and employers are so interested in these models.
  2. They’re startups that are attracted to the growth potential. (more members!)
    Building new primary care models is a particularly capital intensive endeavor. A lot of these companies have attracted significant amounts of venture capital funding to build out the early version of the model (i.e. Forward raising $30 million). The problem is that adoption of direct-to-consumer primary care models is low at the moment. There isn’t a large population of people out there looking to pay $25 — $150 per month on top of health insurance to pay for primary care. So going after this market results in really slow growth. This is problematic when a company is venture capital backed and needs to demonstrate expontenial growth in order to provide a return on the investment. That exponential growth seems achievable by partnering with an insurer or employer who can sell your primary care product to their members.

Harken provided a good example of both these reasons — the entire concept was designed around the first thought… that by integrating primary care into the broader system you could improve outcomes and lower costs. Harken also ‘raised’ $65 million from its investor, UHC, and needed to grow rapidly in order to justify the investment. There certainly is an appeal to the growth potential of these models, as this example demonstrates below:

If Harken is like other Iora practices, there are typically in the neighborhood of 3,000 patients per practice (this would suggest there are roughly 30,000 patients in year one for Harken). Iora has proven they can scale quite rapidly and have demonstrated successful practices in every part of the country… Generally these sorts of models can, at least, double in reach every year. Like all exponentials, the numbers are relatively unimpressive in the early years. In a market like Chicago, that could mean a half million members in 5 years.
Dave Chase describing Harken Health’s Primary Care Model

Now, 20,000+ members in year 1 (which Harken achieved) and 500,000 members in one metro area in 5 years would start to change the game. All of a sudden these care models could start moving from the periphery to the mainstream. This rapid growth potential is what gets VC money excited about the space. It is not nearly as exciting to grow slowly to 30,000 members in 10 years for a VC fund that has a time frame of 5–7 years to generate a return on investment.

While that growth potential is really enticing —enticing enough to send a lot of these primary care companies down the path of partnering with insurers and the existing system — the economics of these new care delivery models present a real challenge to such growth currently.

The issue, as we found out with Harken, and it seems like others that have raised VC have found as well, is that the model presents some really challenging economics for insurers and employers in a typical commercial population. It works in really sick and costly populations anecdotally (i.e. Medicare Advantage) where some savings can be shown, but in the broader general population, it costs too much. Why?

Traditional Primary Care Is Cheap Relative To The Rest of Healthcare…

Spending on primary care today is typically less than, or approaching 10% of overall medical spend for a population of individuals. As an example, the charts below illustrate primary care spending in the state of Oregon on a per member per month (PMPM) basis for different insurers in the state. If you’re not familiar with insurance business models, numbers are almost always articulated on a per member per month basis — so the rest of this article will use that approach.

Data from Oregon on what insurers pay for primary care PMPM in the commercial and Medicare Advantage markets. Accessed here: Carrier 3 is an interesting outlier in terms of spending on primary care.

Using the state of Oregon as an example (they published some cool data on primary care spend at the link above that I believe is pretty representative of what’s happening elsewhere in the country), most insurance companies spent between $8.46 and $30.09 PMPM on primary care spend on a commercial population, anywhere from 3% — 10% of total medical spend. In a Medicare Advantage population that is sicker and costs more than a commercial population, insurers spend slightly more on primary care — ranging from $22.43 to $54.45 PMPM (although its slightly less as a percentage of total spend). But note that’s in the same general ballpark of 3% — 7% of total medical spend.

Setting a Cost Benchmark for New Models

Using Oregon’s data as a proxy for the rest of the country, if an insurer is looking at it’s existing primary care spend to set a benchmark for what a new model should cost, it is going to be somewhere in the ballpark of $25.00. Note that Medicare Advantage models spend slightly more, so their benchmark might closer to the range of $30.00 — $40.00. That is going to be what an insurer uses as a baseline in determining how much they’re willing to pay for a new primary care model.

It is worth noting that these benchmarks are likely the wrong place to start in evaluating whether these new primary care models are increasing or decreasing overall costs. Since primary care is undervalued in the traditional system, looking at data from the traditional system to price a new model is going to undervalue the new model. That is true, and there are a number of efforts underway to change how primary care is paid for. But, for now, the conversations with employers and insurers are going to revolve around whether their costs go up or down, which means the existing data is the benchmark. So $25.00 PMPM for all primary care spend for the population is the bogey these models need to measure themselves against.

…But New Primary Care Models Are (& Probably Should Be) More Expensive

Delivering really good primary care costs more than what currently exists. That shouldn’t come as a surprise given the state of primary care (and healthcare more generally) in America. How much more though? Lets go through what an innovative primary care model might cost…

Headcount is expensive in a primary care clinic

The biggest driver of costs in operating a primary care clinic is paying for headcount. Iora Health provides a good example of what headcount might look like for a primary care staffing model — it’s been published that their typical primary care clinic would have 2–3 physicians, a nurse, 6–9 health coaches (3 health coaches per physician), a behavioral health specialist, and a clinic manager. The typical member panel size for a physician with this staff would be 1,500, so each clinic at full capacity would have 3,000–4,500 members, dependent on whether it’s two or three physicians. The chart below shows what the financials would look like to staff such a model:

Buildup of headcount costs for a sample primary care practice for a startup like Iora Health

The chart above outlines headcount costs for a sample primary care practice. Note that headcount alone is already 40% higher than what an insurer traditionally pays on a PMPM basis. Headcount costs alone in this model are $37.92 PMPM, already 40% higher than what an insurer would typically expect primary care to cost. This is the biggest cost driver for these new models. It’s natural to begin to wonder whether headcount can be reduced from these levels as a way to reduce cost. The answer to that might be yes, although it’s problematic for a few reasons:

  1. The biggest cost driver here is the assumption that the PCP is seeing 1,500 members. Given the PCPs annual cost of ~$250k, they alone cost $13.89 PMPM when seeing only 1,500 members. It isn’t hard to come up with ideas as to how a PCP might see more than 1,500 members. However, this assumption is already higher than the ratio for most direct primary care practices, which have ratios of ~600–800 members per physician. Once you’re at 2,000–3,000 members per physician, it quickly starts to look like the old fee for service model, and the benefits of these models for physicians (additional time with each person, less burnout, etc) are lost. Enabling PCPs to email, text, perform video visits, can help make these ratios larger without sacrificing the model.
  2. You could also reduce costs by reducing headcount of the other staff around the PCP, for example health coaches. These would probably be the first cost buckets you’d look at when trying to reduce costs. However, that seems a bit short sighted — those less costly roles supporting a physician are core to the long term solution. The more labor you can have performing primary care-like services that cost less than a PCP, the better. They are the key to these kinds of models. While it’s hard to quantitatively measure the value of these non-clinical resources currently, cutting those roles to cut costs just makes these models more like the existing system. It’s a losing battle.

Other Costs — Rent, Insurance, SG&A

In addition to headcount, there are a number of other costs that come with operating a primary care practice — rent, medical malpractice insurance, etc. These costs all add up, as the chart below demonstrates. This chart walks through my best guess of all the costs of operating a primary care clinic.

A fully staffed and supported primary care clinic would cost approximately ~$2.25 million per year to operate based on my analysis. At 3,000 members per clinic at this level, it would cost $62.69 on a per member per month cost basis (compared to ~$25.00 PMPM for the traditional model).

The total cost of operating a primary care clinic for the year based on my analysis would be ~$2.25 million. This would be for two PCPs seeing 3,000 members in total. The primary costs accounted for here include:

  • Rent: Assumed $40 per sq ft @ 6,000 ft. This is likely a generous number — clinics may be able to get away with less than 6,000 sq ft, although 5,000–6,000 feels like a good number. Rent will obviously depend on the specific market.
  • EMR / Tech: This is a monthly cost per physician based off some cloud-EMR pricing models. This number could rise quickly for a company like Forward that is choosing to build it’s own tech platform in house… I think you’d allocate that tech cost across each practice equally, although I guess finance folks could debate doing that. Regardless, it’s a big cost that will need to be accounted for somewhere on a P&L
  • Insurance: Providers need medical malpractice insurance; this is a ballpark number
  • Clinic Build / Upkeep / General: This is a catchall of other costs that a clinic will probably incur. There will be some cost associated with building / creating the new space and then upkeep of that space, supplies & equipment, and then costs related to sales & marketing in the local area

While the individual buckets could be higher or lower for each of these categories, I think ballpark it seems like a pretty reasonable estimate. This build results in a monthly PMPM cost of $62.69 — so lets say it’s somewhere in the range of $50 — $75 PMPM to operate a primary care clinic like this.

Fixed Costs vs Variable Costs

Before moving on from a cost build, it’s worth noting that the new primary care model is shifting costs to a fixed cost structure from a variable cost structure in the traditional model. This switch can create some real headaches for companies. Unfortunately, this shift is a foreign concept for insurers and employers that are used to dealing with a 100% variable cost structure for their medical spend (i.e. everything is expressed in PMPM terms that increases and decreases linearly based on the number of members).

The practical challenge with this is that when primary care startups are building/negotiating partnerships with insurers and employers, a financial model needs to be build with cost assumptions. People generally assume things will work out well and maximum capacity will be achieved, which reduces the fixed cost on a PMPM basis. What is often overlooked is that these primary care centers aren’t at maximum capacity for a long time. This results in a primary care PMPM spend that is much higher than was modeled in the negotiation. Here’s a chart to illustrate the impact of lower membership on the PMPM numbers.

Since the clinics are essentially a fixed cost (it would be hard to hire / fire staff based on membership levels… it take a long time to recruit a good primary care provider), the monthly spend on primary care balloons at low membership levels.

This presents a real challenge for these models if demand ramps up slowly over time, yet clinics need to staff up to be operational. This is particularly challenging since it’s really hard to hire good PCPs / staff, so it is not as though a company can just hire and fire on a whim based on demand.

The result is that low membership levels create real pressure on partnerships between primary care companies and insurance companies or employers that are used to dealing with medical spend on a PMPM basis. The PMPM cost jumps from $62.69 to $376.17, which is almost the entire spend that an insurer is used to paying for primary care.

Everyone references the cost of the model when it is at capacity, but when the clinic opens up and is relatively empty for the first few months (or years) that PMPM skyrockets and can sink a business. Something to keep an eye on.

Looking at Pricing for Existing PCP Companies — Harken, DPC, Concierge Medicine Pricing Models

Let’s look at some of the leading companies working on new primary care models to see if it can help inform what costs might be for these primary care models. Harken provides a good data point here as our Co-Founder & CEO publicly stated we were spending over 2x the amount of traditional insurers on primary care for a commercial population.

“We are investing over two times what normal insurers would pay for primary care to provide what we are providing within the robust set of services within the Harken Health Centers.” Harken’s CEO said in a media interview here.

This would imply that instead of spending $25.00 PMPM, Harken was spending at least $50.00 PMPM, same as the general ballpark I backed into above. Funny how that works out!

Beyond Harken, looking at the market for direct primary care or concierge medicine gives some sense of the cost structure as well. The median direct primary care monthly fee is $75 (link). At the high end there are models like Forward operating in the concierge space that charge ~$150 per month. Note I’m going to go ahead an ignore the crazy Silicon Valley type models where people are paying $40,000 for access to a primary care doc (See this NYTimes article if interested in that model. Yay capitalism!).

The low end of the spectrum is an option like Sherpaa that is online only and costs closer to ~$25 per month for the cheapest plan option (this is a really cool model that I think has the potential to avoid a lot of these cost issues since you don’t have the same physical infrastructure costs, but that’s for another day). Sherpaa prices increase to $100 PMPM for individuals to access the full suite of services they provide.

Given that revenue structure for DPC / Concierge models, and the math above, it doesn’t seem unreasonable to think that a primary care practice will cost somewhere on the order of $50 — $75 PMPM even at maximum capacity. If not at maximum capacity, that PMPM can skyrocket quickly.

Comparing Costs of New Primary Care Models vs. Traditional Models

The end result is that new primary care models tend to cost somewhere around twice as much as traditional primary care spend — they increase the cost from $25.00 to somewhere around the number I calculated of $62.69 PMPM.

Note that I’m making an assumption here that when a business adopts a new primary care model, it erases all traditional primary care spend. This assumption makes this analysis easier here, although may not in fact happen in the real world. That is, some of the people buying into a new model like this will continue to see their old primary care physician. This will bring costs even higher — the new primary care model still costs $62.69 PMPM, but the old primary care spend isn’t reduced from $25.00 to $0.00. This happened to Harken because other non-Harken PCPs remained in-network, so members could sign up for Harken and still use their old PCP if they chose. Another thing to keep an eye on.

Assuming away that issue, that increase in primary care spend from $25.00 to $62.69 is already a large increase when looking at a typical commerical population. Keep in mind total medical spend is only around $350 PMPM for a normal commercial population (the Oregon data above suggests it is ~$325 PMPM for that state, I am rounding up to $350 here). That means these primary care models are increasing primary care spend from 7% of overall spend to 18% of overall spend. That is a huge jump, particularly when considering there’s added risk that costs could run higher than 18% if the centers aren’t filled to capacity from day 1— PMPM costs could increase rapidly as demonstrated above.

This is why adopting these models presents a real financial challenge for the traditional healthcare industry. Methods to bear that increase in cost need to be developed to close the gap between what primary care should cost if done well and the current model.

So… How to Solve This Gap? Demonstrating Cost Savings

Based on the above, it’s pretty clear that these models increase the cost of primary care by 2x or more. The question then is how to offset those costs when implementing a model like this— no insurer or employer is going to want to increase the overall cost of care for it’s population.

The way these primary care models achieve that is by demonstrating downstream cost savings caused by better primary care. By generating downstream cost savings, these primary care models can remain cost neutral, or even potentially generate cost savings for insurers, via different financial arrangements (i.e. ACOs, shared savings / risk agreements, etc).

How specifically do these companies go about trying to demonstrate that they can reduce medical spend by $25+ PMPM? More on that here in Part II.

Kevin O'Leary

Written by

Tinkering in health tech. Previously ran population health and data analytics for Harken Health. Sign up for my newsletter @

Welcome to a place where words matter. On Medium, smart voices and original ideas take center stage - with no ads in sight. Watch
Follow all the topics you care about, and we’ll deliver the best stories for you to your homepage and inbox. Explore
Get unlimited access to the best stories on Medium — and support writers while you’re at it. Just $5/month. Upgrade