The Researcher Who Measured DPC's Growth Raised the Hardest Question About It. Her Data Answers Half of It.

Jane Zhu led the team that counted how fast direct primary care has grown. In a Healio report published September 1, she also raised the hardest objection to it.

“A key question is whether they ultimately increase overall primary care capacity or simply reallocate care toward patients who are able to access these models,” Zhu told Healio.

Every serious argument about this model eventually lands on that sentence. Her own paper gets partway to an answer, then stops just short of the number the argument actually turns on.

The Article Healio Ran

Healio’s September 1 piece put DPC’s defenders and its critics in the same nine-minute read, which is rarer than it sounds in trade coverage of this model. Zhu, an associate professor of medicine at the Oregon Health & Science University School of Medicine, was lead author on the Health Affairs analysis that found concierge and DPC practice sites grew 83.1 percent between 2018 and 2023, with participating clinicians up 78.4 percent. Healio’s piece separately put the national count at roughly 3,163 DPC practices, citing DPC Frontier Mapper.

Around her, Healio quoted Ann Greiner of the Primary Care Collaborative, Josh Umbehr of Atlas.md, Philip Eskew of DPC Frontier, Carolyn Engelhard of the University of Virginia School of Medicine and internal medicine physician Norman Dy of Johns Hopkins Community Physicians.

The reporting is even-handed. AAFP’s fee ranges run $50 to $100 a month for adults and $20 to $49 for children, with family plans starting around $100. AAFP calls the adult figure a general range rather than a set price. Umbehr’s dispensing economics get a full airing: four cents a pill for generic Lexapro in his office against $12 a pill for the name brand, because practices buy at wholesale and pharmacy benefit managers sit outside that transaction. A 2025 Health Compiler report cited in the piece puts employer-sponsored memberships at 58 percent of all DPC memberships in 2024.

Then the piece turns.

The Subtraction Everyone Runs

Healio put a typical DPC panel at about 413 patients against about 1,700 in a typical family medicine practice, attributing both numbers to AAFP. Anyone can do the arithmetic from there. A physician who moves from one to the other appears to leave roughly 1,287 patients behind her.

Dy said it without decoration. If the census of traditional primary care drops, “you’ll need more providers, and we don’t even have enough providers now.”

Zhu attached a condition to her version. If more physicians move to DPC “without a corresponding expansion of the primary care workforce,” this could limit access for patients who choose a traditional fee-for-service model. That condition is carrying real weight in her sentence, and it’s the part that tends to fall off when the argument travels.

What the Data Actually Tracks

The Health Affairs team identified practices through a national online directory, then linked those sites to NPI records and billing data, following individual clinicians year by year rather than taking two snapshots. Zhu is careful about the limits: the analysis “wasn’t a census and likely underestimates concierge and DPC practices nationally.”

It does record where the entering clinicians came from, and that answer has been sitting in Exhibit 4 the whole time. Of the physicians who entered concierge or DPC practices after 2018, 841 came from a health system or integrated delivery network, 726 from an independent practice, and 530 from another corporate-owned or affiliated practice. Only 133 left a government, federally qualified health center, or community health center setting. Another 655 had no prior employer the data could see, and 40 percent of those were within three to nine years of finishing medical school.

Read that against the subtraction and it cuts both ways. Most entering physicians did come out of active clinical settings, which is the premise the capacity worry needs. But for those 655 the data records no prior employer at all, so there’s no panel in it to subtract.

So the origin question has an answer. The one underneath it doesn’t.

The subtraction assumes a DPC physician was carrying something near 1,700 patients the year before she converted, and would have kept carrying them for the rest of her career. Exhibit 4 records the setting she left. It doesn’t record the panel she carried inside it, and a physician can walk out of a health system having already cut to a half-day schedule. Burnout-driven attrition in primary care is well documented as a general phenomenon. The study is descriptive and carries no comparison group, so nothing in it shows what these specific clinicians would have done had DPC never existed.

Without those two numbers, 1,287 is a ceiling on the loss rather than an estimate of it. The real figure sits somewhere below, and no one currently knows how far.

Greiner Moved the Frame

Ann Greiner offered a different read on why any of this is happening at all.

The rise of these models may reflect “a failure on the part of policy,” she told Healio, pointing to what she described as inadequate support for primary care, “and so people are finding these arrangements outside of the insurance system.” She went further: “If we were adequately supporting primary care, and we didn’t have all the administrative burden that is on primary care, perhaps these models wouldn’t be in existence.”

That reframing changes what the capacity worry implies. If DPC is a symptom of underinvestment, slowing DPC leaves the shortage exactly where it was and removes one of the few exits physicians have found on their own.

The Counterargument

The strongest version of the case against DPC doesn’t need the panel arithmetic at all, and knocking the arithmetic down doesn’t clear the model.

Set aside where DPC physicians came from. Affordability stands on its own. Dy called an $80 to $150 monthly fee “considerable for many patients” and put it in terms nobody can argue with: “That’s the same cost as utilities or paying for gas to get to work.” Greiner asked whether the model is “really only for folks who have more disposable income and can afford to pay this monthly subscription cost.” A model priced past the people with the least access is a distribution problem whether or not a single physician gets subtracted from the fee-for-service pool.

Engelhard’s objection is narrower and harder to answer. DPC physicians are limited in “what they can offer if medical problems get chronic and expensive,” she said. “They have to be able to refer to specialists for cancer care and other kinds of specialty treatment. Some DPC PCPs have those relationships, but specialty doctors will not do ‘contract’ or ‘membership’ medicine like DPC, so patients need to know that up front.”

Both of those survive intact. The capacity claim is the one still open, and the distinction matters because the three get bundled together and presented as a single settled verdict on the model.

Zhu asked for evaluation, not restriction. Taking her seriously means running the study that would finish what Exhibit 4 started: count how many patients these physicians were actually carrying before they converted, and what share of them were on their way out of clinical medicine regardless. Until somebody funds that, the loudest objection to direct primary care rests on a panel size nobody has gone back and counted.