Industries / Health / Healthcare / Direct primary care
Records and software
With no insurance claims to file, these practices run on membership and billing software, keep their own records, and stay largely out of sight of the claims data that researchers and regulators count with.
4
Milestones achieved, from 2013 to 2025.
1
To watch for: announced, not yet confirmed outside.
2
Open deliverables. Nobody has shipped them.
The road
AchievedWatch forOpen
Achieved
2013
Software made for memberships
On 10 October 2013 the Atlas MD clinic in Wichita, Kansas, released Atlas.md, record and practice-management software its doctors had designed around memberships instead of insurance billing, and called it "direct care's first EMR and Practice Management software"; the Wichita Business Journal credits the family doctor Josh Umbehr with launching it. In the American Academy of Family Physicians' survey of 2025 practice, Atlas.md was the record system respondents used most (26%), taking the lead from Elation Health (53% in 2022, 26% in 2024); 52% said their system collected both diagnosis (ICD-10) and procedure (CPT) codes, and 66% said their practice had adopted tools that use AI, 49% for clinical work.
Credit
Josh Umbehr and the Atlas MD clinic; American Academy of Family Physicians (survey)
2022
Billing software without claims
Hint Health's HintOS runs enrollment, membership management and recurring billing for practices that take a monthly fee and file no claims. In June 2022 Fierce Healthcare reported a $45 million investment led by Banneker Partners and Frist Cressey Ventures, bringing Hint's total to $60 million, and Hint's own figure of more than 2,400 providers using it, at an average fee of $75 a month. Hint publishes an API for partners and practices.
Credit
Hint Health (chief executive Zak Holdsworth)
2022
Invisible to claims data
Alison Huffstetler and colleagues at Virginia Commonwealth University counted Virginia's active physicians as those with at least one claim in the state's all-payer claims database, which collects insurers' claims: 20,976 in 2019, of whom 5,899 gave primary care. They noted that "uninsured care and direct primary care are not captured... because of lacking claims with any payer." A 2011 review by K. A. Fairman and F. R. Curtiss had made the same point for drugs: claims datasets lack generic drug discount programs, free samples and other ways people get medicines outside insurance. Federal rules bar a doctor who has opted out of Medicare from submitting Medicare claims, and the DPC Alliance's 2026 survey report says working outside insurance billing limits how visible these practices are to research that relies on claims.
Lessona practice that files no claims drops out of workforce and spending counts built from claims.
pmc.ncbi.nlm.nih.govpmc.ncbi.nlm.nih.govlaw.cornell.edudpcnews.com
Credit
Alison N. Huffstetler and colleagues (Annals of Family Medicine)
2025
One state's register
Washington's 2007 law makes each direct practice register and file an annual statement, and its Office of the Insurance Commissioner turns the filings into a yearly report, the only government count of these practices found. For 2025 it counted 99 registered practices, up from 81 (23 new, 5 left the market), serving 43,841 patients, or 0.54% of the state's residents. The average of practices' monthly fees rose 20.3%, from $154.29 to $185.65, with fees from $40 to $2,000 a month and 52 of the 99 practices charging $61 to $120. Eighty-nine percent of practices had a single clinician, 19 were also in an insurer's network, and the office received no complaints.
Lessona short reporting duty in one state's law produces figures no national source has.
Credit
Washington State Office of the Insurance Commissioner
Watch for
AI inside a virtual membership
HealthTap launched Direct Virtual Primary Care for Organizations, a monthly-fee virtual primary care plan sold to employers and other groups, with built-in AI tools for patient intake and decision support. It tied the launch to the 2025 tax law that lets people in direct primary care keep a health savings account, and said 96% of Christian Healthcare Ministries members who began care with a HealthTap doctor returned to the same doctor.
Achieved whenan outside party publishes how often the intake and decision-support tools are right, or audits the continuity figure.
Claimed by
HealthTap, August 2025
Open
01
Quality reports without claims
An open tool that takes a standard FHIR export (the common web format for health records) from a direct primary care records system and computes a few standard electronic quality measures, such as CMS165 (blood pressure control) and CMS122 (diabetes with a high HbA1c), then prints a one-page report a practice could publish. Tested only on synthetic patients.
The briefCredit · unclaimed02
An open count of practices
A public, dated list of US direct primary care practices built only from public sources: the DPC Frontier mapper, other open directories, Washington's register of direct practices, CMS's monthly list of clinicians who have opted out of Medicare, and NPPES, the national register of provider numbers (NPIs). Match each practice to its clinicians' NPIs, check every listing twice a year for a live website, a listed phone number, a current state business registration or a closure notice, and release dated snapshots, so that openings and closures can be counted for the first time.
The briefCredit · unclaimed
Open 01 · the brief
Quality reports without claims
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
An open tool that takes a standard FHIR export (the common web format for health records) from a direct primary care records system and computes a few standard electronic quality measures, such as CMS165 (blood pressure control) and CMS122 (diabetes with a high HbA1c), then prints a one-page report a practice could publish. Tested only on synthetic patients.
Why it is still open
The first national study of these practices found that quality data were lacking, and in 2026 the DPC Alliance's survey report said robust data remain scarce because practices work outside the insurance billing that research relies on. Opted-out doctors file no Medicare claims, so claims-based quality programs do not see them. What is published is patient-reported: Hint Health's 2026 benchmark scored 12 clinics with the Person-Centered Primary Care Measure, a patient survey, not with clinical measures, and only 23% of the data requests employers make of their practices concern quality measures, by the AAFP's 2026 survey. The parts exist: CMS's Cypress tool, the official test of whether software calculates these measures correctly, and the open fqm-execution library, which calculates them from FHIR data. A GitHub search for "direct primary care" found practice websites, directories and one small records system, but no quality tool.
jabfm.orgdpcnews.comget.hint.comaafp.orghealthit.govgithub.com
Built with what exists
Synthea synthetic patients; fqm-execution; Cypress test patients and expected results; a HAPI FHIR server; the test accounts or APIs that practice software vendors offer, such as Hint's.
First step
Run fqm-execution on Cypress's test patients for CMS165 and compare its results with Cypress's expected results.
Done looks like
For four measures, results match Cypress's expected results on every Cypress test patient; the same pipeline runs on 1,000 synthetic patients loaded into at least two practice records systems, with a published list of the data each system's export lacked.
Open 02 · the brief
An open count of practices
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
A public, dated list of US direct primary care practices built only from public sources: the DPC Frontier mapper, other open directories, Washington's register of direct practices, CMS's monthly list of clinicians who have opted out of Medicare, and NPPES, the national register of provider numbers (NPIs). Match each practice to its clinicians' NPIs, check every listing twice a year for a live website, a listed phone number, a current state business registration or a closure notice, and release dated snapshots, so that openings and closures can be counted for the first time.
Why it is still open
Counts disagree: about 1,600 by the Direct Primary Care Coalition (as cited in 2024), 2,125 on the mapper in October 2023, about 3,163 in September 2026 and 3,286 in November 2026, and more than 3,600 by Hint Health's estimate for early 2025. Zhu and colleagues (Health Affairs, 2025) built the largest study sample from the FindMyDirectDoctor directory, NPPES and proprietary IQVIA data, did not separate direct primary care from concierge practices, and counted growth, not exits. Washington counts exits only for its own register (5 practices in 2025), and the only broader claim found, a 2021 post reporting a "0% closure rate" for membership practices in the pandemic, rests on one consulting firm's cohort. CMS's September 2026 opt-out file lists 59,071 clinicians, 2,981 of them in family practice and 1,812 in internal medicine (this session's count of the download); no published work found links it to a practice list.
pmc.ncbi.nlm.nih.govhealio.comdpcnews.compmc.ncbi.nlm.nih.govinsurance.wa.govdpcnews.comdata.cms.gov
Built with what exists
The sources above; Internet Archive snapshots; state business registries; Python record linkage on names and addresses; a browser agent such as Browser Use (see the Note "Deleting 80% to start over") to visit listings; an AI model asked fixed yes-or-no questions about each practice website, such as whether it bills insurance (the site's Note "A thousand questions, one pass" covers Gemini Decision 1, built for such questions), with a person checking every listing marked closed or unclear.
First step
Take the 99 practices in Washington's 2025 appendix, match them to NPPES and the opt-out file, and see whether the five exits can be named from public records.
Done looks like
At least three snapshots; at least 90% of listings matched to an NPI or marked unmatched with a reason; a hand-checked random sample of 200 practices with the classifier's accuracy reported; counts of openings and closures between snapshots, with survival by year of opening.