Community Health Centers vs. the World: What a Dozen Countries' Primary Care Models Actually Deliver

The scoreboard, stated plainly

Start with the uncomfortable frame. In the Commonwealth Fund's 2024 ten-country comparison, the U.S. ranks last overall and dead last on health outcomes — Americans live the shortest lives and have the most avoidable deaths — while spending far more per person than any peer. The one category where the U.S. scores near the top is care *process*: when an American gets into the room, the care delivered is good. The failure is upstream of the room.

One number explains a lot of it. Across OECD countries, about 14% of health spending goes to primary health care. In the U.S., depending on the definition you use, primary care is roughly 4–5% of spending, and it was drifting down — 4.60% in 2013 to 4.35% in 2017 on a narrow definition. Those two figures are not perfectly comparable — OECD's definition of primary health care is broader than the narrow U.S. accounting — but no reasonable harmonization closes a gap that wide. America buys the least of the thing that every high-performing system buys the most of.

Costa Rica makes the point almost rudely. It spends roughly a tenth of what the U.S. spends per person on health and its life expectancy is approaching 81 years against a U.S. figure just under 79 and, for years, falling. A PNAS analysis of the two countries found U.S. adult male mortality runs 18% higher than Costa Rica's, and 10% higher among middle-aged women — with the inequality gradient far steeper in the U.S. A Costa Rican low on the socioeconomic ladder outlives an American in the equivalent position.

What America actually built — and buried

Before the comparison turns critical, give the U.S. model its due, because on one dimension it leads the world. The Health Center Program's governance rule — a majority of the board, at least 51%, must be patients who represent the population served, under §330(k)(3)(H) — is a structural transfer of power that Brazil, Turkey, Portugal and the Netherlands simply do not have. Their systems are excellent and public; they are not patient-governed. The only close international cousins are Australia's Aboriginal Community Controlled Health Organisations and Alaska's tribally owned Nuka System — and, as we will see, those two happen to post the most striking outcome numbers in this entire comparison. That is not a coincidence worth ignoring.

The scale is real too. Health centers served nearly 34 million people in 2024, up from 32.5 million a year earlier, with about 18% uninsured and about half on Medicaid. And the model performs on the metric American payers care about most: an American Journal of Public Health study found Medicaid enrollees served by health centers had 24% lower total spending than those in other primary care settings — 27% lower inpatient, 14% lower emergency, 11% lower on drugs. This is not a charity that costs money. It is the cheapest care in the system.

So the American anomaly is not the design. It is the *financing posture*. Everywhere the outcomes are strongest, community primary care is treated as permanent public infrastructure with population-based funding. In the U.S. it is a grant-funded program for low-income people, paid largely per visit, with its core fund expiring on a date certain. No other country in this comparison asks its community primary care network to re-earn its existence on a two-year cycle.

Archetype 1: the universal team on a map

The most-studied model worldwide is a salaried multidisciplinary team assigned to a defined geography, responsible for everyone in it, with community health workers doing routine home visits. Brazil, Costa Rica, Turkey, Cuba, Iran and Thailand all run versions of it, and the evaluation literature is unusually consistent.

• **Brazil — Family Health Strategy.** Teams of a doctor, nurse, nurse assistant and four to six community health agents cover a mapped territory. Coverage grew from 66.8% of the population in 2009 to 84.7% in 2023, delivered by 236,000 community health agents across roughly 33,000 teams. In Health Affairs, a fixed-effects analysis of 1,622 municipalities found that going from 0% to 100% coverage was associated with a 6.8% reduction in amenable mortality — 11.0% in municipalities with the strongest health governance versus 4.3% in the weakest. Read that second clause twice: the same model produced more than double the mortality benefit where local governance was strong. Structure is necessary; competence at the local level is what multiplies it.
• **Costa Rica — EBAIS.** A doctor, a medical assistant, a community health worker and a records clerk are responsible for an empaneled population of about 4,000. In the *Journal of Health Economics*, nine years after a Health Area opened, the assigned population showed a 13% lower age-adjusted mortality rate, concentrated among people over 65 and in cardiovascular deaths — and the mechanism was partly visible in utilization, as more people used primary care and fewer used emergency rooms.
• **Turkey — Family Medicine Program.** Every citizen was assigned to a salaried family physician at a walk-in neighborhood health center, free at the point of care, phased in from 2005 and nationwide by 2010. An analysis published in the *Journal of Public Economics* attributed to it a 25.6% drop in infant mortality, 22.9% among children aged 1–4, and 7.7% among the elderly — and, notably, a narrowing of mortality disparities *between provinces*.
• **Iran — health houses and the behvarz.** Two local villagers per health house, trained for two years, living in the community they serve. Rural infant mortality fell from 123.7 per 1,000 live births in 1976 to 30.2 by 2000, and 23.7 by 2003, with the rural–urban gap closing as it went.
• **Cuba — consultorio and polyclinic.** A family doctor–and-nurse pair lives in the neighborhood and is responsible for as few as 900 to 1,500 people, spending mornings in clinic and afternoons on home visits and public health work. Life expectancy around 78, infant mortality about 4.5 per 1,000 — outcomes a rich country would recognize, on a poor country's budget.
• **Thailand — district health system + universal coverage.** Nearly 10,000 sub-district health centers and close to a million village health volunteers were built *before* universal coverage was switched on in 2002 — infrastructure first, insurance second. Catastrophic health spending fell from 6.0% of households in 1996 to 2% in 2015, and the infant mortality gap between richer and poorer provinces narrowed.

One honest caveat before drawing conclusions: none of these are randomized trials. They are difference-in-differences, fixed-effects and synthetic-control designs on national administrative data — strong quasi-experimental evidence, and consistent across very different countries, but association rather than proof. The rebuttal to the skeptic is not any single study; it is that six countries with different politics, incomes and cultures ran roughly the same play and got the same direction of result.

And the model does not solve everything. Ethiopia's Health Extension Program, which trained tens of thousands of health extension workers from 2003, significantly raised full childhood vaccination and bednet use in program villages but did not reduce the incidence or duration of childhood diarrhea and cough, with limited effect on most prenatal and postnatal care. India's Ayushman Arogya Mandir network — about 1.74 lakh centers, with 84.9 crore hypertension screenings and 74.2 crore diabetes screenings recorded by late 2024 — is a staggering volume of contact whose downstream mortality effect is still being established. Reach is not the same as outcome.

Archetype 2: community-governed — the FQHC's real cousins

This is the archetype American health centers actually belong to, and it is the smallest club in the world. Four members are worth knowing.

• **Alaska — the Nuka System of Care.** Southcentral Foundation, owned by Alaska Native people who are called "customer-owners" rather than patients, rebuilt its system around relationship and same-day access. The Baldrige Award profile documents a 50% decrease in emergency room and urgent care visits, a 53% drop in hospital admissions, a 65% decrease in specialty care, diabetes measures above the 90th percentile, and staff turnover falling from 37% in 2008 to 17% in 2011. Treat those as organization-reported figures from a specific era — but note that a peer-reviewed *Annals of Family Medicine* analysis independently found emergency care use fell during and after the redesign after rising beforehand. The workforce number deserves as much attention as the clinical ones: cutting turnover by more than half is the single hardest thing any FQHC leader is trying to do right now.
• **Australia — ACCHOs.** Aboriginal Community Controlled Health Organisations are governed by the communities they serve. A *Medical Journal of Australia* review found they achieve comprehensive primary health care consistent with the medical-home model, cover more than 60% of the Aboriginal population outside major metros, and show superior performance to mainstream general practice on best-practice care indicators, while reducing unintentional racism and barriers to access. The measured advantage is in process and access rather than mortality — but it is the same population, and the difference is who governs.
• **Canada (Ontario) — Community Health Centres.** Ontario runs four primary care models side by side, which makes it a natural experiment. The ICES comparison found CHCs — salaried, team-based, serving the sickest and most disadvantaged panels — had emergency department visits at 0.79 of expected, versus 1.25 at blended-capitation Family Health Networks and 1.06 at Family Health Organizations. The hardest panel produced the lowest emergency use. That is the FQHC value proposition, measured in a system that could measure it cleanly.
• **New Zealand — Māori and Pacific providers.** Kaupapa Māori and Pacific providers, plus the Very Low Cost Access scheme capping adult co-payments for practices whose enrolled population is majority high-need, were built for exactly this purpose. The honest finding is mixed: being Māori or Pacific, or living in material deprivation, remained associated with poorer outcomes across all models of care, and researchers argue funding still leaves financial barriers in place. Community governance plus underfunding still yields inequity — which is precisely the warning for the U.S.

Archetype 3: universal GP lists — and the inverse care law

The third model is the one Americans usually picture when they think "other countries": everyone registered with a general practice that acts as gatekeeper and coordinator. The Netherlands is the cleanest example — every resident registered with a GP, with out-of-hours care consolidated into GP cooperatives now often co-located with hospital emergency departments, so the cooperative absorbs the large majority of after-hours demand instead of the ED. The Netherlands is also one of the three top overall performers in the Commonwealth Fund's 2024 ranking.

Portugal shows what happens when you add performance pay to that structure. Its Family Health Units come in a Model A (salaried) and a Model B, which carries a pay-for-performance system whose incentives can reach roughly two-thirds of base salary, with teams accepting explicit accountability for access and outcomes. Model B units post the highest indicator values of any unit type. It is the closest international analog to the direction U.S. value-based care is heading — and worth reading next to our value-based care hub.

But universal registration does not, by itself, produce equity — and this is the most important cautionary tale for anyone who thinks the U.S. just needs coverage. Scotland's "GPs at the Deep End" group works in the 100 most deprived practices in the country. Their central finding, in a British Journal of General Practice scoping review, is that general practice funding per patient in Scotland's most deprived areas is roughly equal to funding in the least deprived areas, despite far greater need, and the GMS contract insufficiently accounts for complexity and workload. Of twenty interventions identified, only two were rolled out nationally, and their future is uncertain. Universal coverage with need-blind funding reproduces Julian Tudor Hart's inverse care law inside a national health service. Coverage is necessary. Weighted funding is what makes coverage fair.

Five design features that travel

Strip away the flags and the same five features recur wherever community primary care produces measurable outcome gains. Each has a direct FQHC translation, and the U.S. already has three of them.

• **Empanelment to a defined population, not a visit stream.** Costa Rica's ~4,000 per team, Cuba's 900–1,500 per consultorio, Turkey's assigned physician. Health centers know their patients but are paid per visit under PPS, so the accounting unit is the encounter, not the person. Every value-based contract an FQHC signs is a step toward the international norm — which is why our per-state salary and revenue data and the ECM/care-management path matter more than they look.
• **Community health workers at real density, doing routine home visits.** Brazil fields 236,000; Iran two behvarz per village; Thailand nearly a million volunteers. The U.S. equivalent exists and is proven — the Penn IMPaCT model returns $2.47 to Medicaid per $1 invested, and globally the investment case put the return at about $10 per $1 with up to 3 million deaths averted annually — but U.S. density and financing are nowhere near Brazil's. California's own CHW/P/R Medi-Cal benefit is the test case.
• **Funding that follows the population, weighted for need.** This is the one the Deep End data proves by its absence, and where the U.S. is most exposed: base funding on a grant cycle with a hard expiration date is the opposite of population funding. The countries with the best mortality trends fund primary care as infrastructure.
• **Local governance with real authority.** Brazil's own data shows the same model producing 11.0% versus 4.3% amenable-mortality gains depending on governance strength. ACCHOs and Nuka post the strongest results in this comparison. The U.S. mandates patient-majority boards by statute — a genuine world-leading design that deserves to be defended as a clinical asset, not treated as paperwork.
• **A front door that is genuinely first contact.** Dutch GP cooperatives absorb after-hours demand instead of the ED; Ontario CHCs hit 0.79 of expected ED use; Nuka cut ER use in half with same-day access. Emergency-department displacement is the most reliably reproduced outcome in the entire global literature — and it is the one an individual health center can move on its own, this year, with access redesign.

Where U.S. health centers genuinely lead

A fair comparison runs both directions. There are things American health centers do that most of these systems do not, and they should be named rather than apologized for.

• **Patient-majority governance in statute.** Not a consultative committee — control of the board, legally required, at ~1,400 organizations. Nowhere else does this at this scale.
• **Enabling services as a funded expectation.** Interpretation, transportation, eligibility assistance, case management and outreach are inside the model rather than bolted on by a charity. Many universal systems assume these needs away and then wonder why utilization is uneven.
• **A sliding fee scale tied to income, with care regardless of ability to pay.** In systems where care is free at the point of use this is unnecessary; in a country without universal coverage it is the mechanism that keeps 18% uninsured patients in care.
• **Language-concordant, community-hired workforce.** The bilingual staffing depth of a California health center is not standard practice in most national systems, which typically centralize interpretation rather than hire from the community.
• **340B and integrated services.** Pharmacy margin funding uncompensated care, plus dental, behavioral health and vision under one roof, is a degree of integration many national primary care systems have never achieved.

Which yields the actual verdict of this comparison: the United States is a design leader and a financing laggard. The model is not what is failing. The way the country pays for it, and the fact that it is offered to the poor instead of built as the spine of the system, is what produces the last-place scoreboard.

What to do with this on Monday

International comparison is only useful if it changes something local. Three moves are available now, without waiting for Congress or a treaty.

• **For leaders — argue the Ontario number, not the moral case.** When a county, plan or legislator asks what a health center buys, the strongest available evidence is comparative: 0.79 of expected emergency visits on the hardest panel in Ontario, 24% lower total Medicaid cost in the U.S., a halving of ER use at Nuka. Put those three side by side in the board deck and the budget testimony. Our Board Brief and Risk Radar are built for exactly that conversation.
• **For operations — copy the access redesign before the financing model.** Same-day access, empanelment discipline and an after-hours front door are the mechanisms behind the ED numbers everywhere they appear, and none of them require a new payment system. This is the one part of the international playbook a single center can run alone.
• **For the workforce — the global evidence is a career argument.** Every high-performing system in this comparison is built on community health workers, care managers and team-based roles rather than on more specialists. That is the direction of travel for hiring, and it is what our Career Roadmap, Academy and live job board are organized around. Turnover is the metric to watch: Nuka's fall from 37% to 17% did as much for its outcomes as any clinical intervention.

The bottom line is not that another country has a system worth importing wholesale — none of them would survive the transplant. It is narrower and more useful than that. The countries beating the United States on avoidable death are not doing anything Americans have failed to invent. They are funding it as permanent infrastructure, offering it to everyone, and weighting the money toward need. The U.S. built the world's most radical version of the idea, proved it costs less, and then attached an expiration date to it.