Capitation

Capitation as Per-Person Payment, with Risk Adjustment and Provider Incentives

Under capitation the money follows the person on a provider's list, not the services that person receives, so a practice or health plan is paid the same whether an enrolled person visits ten times a year or never. The provider gains when care costs less than the payment and loses when it costs more, which is why capitation can reward prevention and efficient care but can also reward doing too little or enrolling healthier people. Rate design depends on risk adjustment. Payment methods in general are on the provider payment page, and the per-person amount and its contract arithmetic on the capitation payment page.

How a capitation rate is built

The World Bank manual on provider payment edited by Langenbrunner, Cashin and O'Dougherty describes per capita payment as a predetermined fixed rate, paid in advance, for a defined set of services for each person enrolled with the provider for a fixed period, usually one month or one year. Because payment is not linked to inputs or to the volume of services, some financial risk moves from the purchaser to the provider. Costs above the per capita budget fall on the provider, and a provider that spends less can usually keep and reinvest the surplus.

A scheme must define who is covered, by enrolment, registration or attribution, which services the rate covers, from primary care alone to all or most care, the payment period and the base rate. With risk adjustment, the total paid to one provider is the base rate multiplied by its weighted list, as follows.

$$P = r \sum_{g} n_g w_g, \qquad w_g = \frac{\bar{c}_g}{\bar{c}}$$

where $P$ is the payment for the period, $r$ is the base rate per weighted person, $n_g$ is the number of enrolled people in risk group $g$, $w_g$ is the group's weight, $\bar{c}_g$ is the average expected cost of group $g$ and $\bar{c}$ is the average expected cost across all groups. The World Bank manual defines this weight, the risk adjustment coefficient, as the ratio of a group's average expected cost to the average for all groups, used to scale the base rate up or down. Normalising the weights so that the weighted population equals the actual population keeps the scheme budget neutral, meaning total payments equal the resources allocated.

Risk adjustment as payment for expected need

Without adjustment a provider is overpaid for people likely to cost little and underpaid for people likely to cost a lot. The World Bank manual states that much of the predictable variation in primary care cost is explained by age and sex, and that adding age and sex coefficients can significantly improve a per capita primary care payment. The Medicare Advantage model described below also uses recorded diagnoses, so that payment reflects prior health conditions as well as demographics.

Prospective risk adjustment uses last year's information, so a provider still gains from keeping costs down; concurrent risk adjustment uses the same year's diagnoses, which predicts cost better but pays more as more care is delivered and recorded, weakening that incentive. A risk score is an average prediction for similar people, not an entitlement for one person.

The incentives of restraint, underservice and selection

The World Bank manual sets out the incentives. A capitated provider wants more people to enrol, which it may pursue by improving quality, and bears the cost of every input, so it wants fewer inputs per person. It can achieve this by improving the input mix, by shifting care towards health promotion and prevention, by reducing quality or underproviding needed services, or by attracting healthier people, which the manual calls risk selection or cherry picking, unless risk adjustment compensates for predictable differences in need.

Ellis's model of provider competition shows that prospectively paid providers cream low-severity patients (over-provision) and skimp on high-severity ones (under-provision), and that dumping, the explicit avoidance of high-severity patients, comes with skimping. Selection can work quietly, through location, appointment systems or extra plan benefits, so cream skimming remains possible after adjustment.

Because an extra visit earns nothing under pure capitation, purchasers often add fees for priority services or quality payments; the WHO South-East Asia brief cites a small fee for immunisation inside primary care capitation to counter the incentive to underprovide. Such mixtures are on the blended payment page.

Capitation in practice in the NHS global sum and Medicare Advantage

Both schemes pay per person with risk weights. The NHS scheme pays general practices, and Medicare Advantage pays private health plans.

NHS England and the GP global sum

English general practices on the General Medical Services contract receive a global sum that works as weighted capitation. The General Medical Services Statement of Financial Entitlements Directions 2026 describe it as a contribution towards the cost of delivering the contracted core services and minor surgery, including staff costs, revised quarterly and paid monthly. Each practice's registered list is adjusted by the Global Sum Allocation Formula to give its weighted population, which is multiplied by GBP 130.07 for 2026/27, plus a London adjustment.

The formula, set out in Annex B of the Directions, adjusts for the age and sex structure of the list, additional needs related to morbidity and mortality, list turnover, patients in registered nursing and residential homes, and unavoidable costs through a staff market forces factor and a rurality index. Against a reference index of 1 for a male aged 5 to 14, a male aged 85 or over attracts 6.27 times the resources, new registrants in their first year carry an uplift factor of 1.46, and each nursing or residential home patient a factor of 1.43. The Quality and Outcomes Framework is paid separately under the same Directions.

US managed care and Medicare Advantage

In US managed care, Robinson, applying agency theory, judges fee-for-service and capitation both flawed ways of motivating physicians and describes blends such as capitation with fee-for-service carve-outs. The largest capitated programme is Medicare Advantage. The Medicare Payment Advisory Commission (MedPAC) payment basics note of November 2025 states that Medicare pays private plans, including health maintenance organisations, a capitated rate for enrolled beneficiaries, who made up 55% of eligible beneficiaries in 2025, with payments of USD 494 billion in 2024. Each plan's monthly base payment comes from its bid relative to a benchmark price, set by county for local plans, and is adjusted for each enrolee using the CMS hierarchical condition category risk model, which uses age, prior health conditions and Medicaid, disability and institutional status.

MedPAC's March 2026 report shows the cost of imperfect adjustment. It defines favourable selection as enrolment of beneficiaries whose risk scores overpredict their spending, and coding intensity as the tendency for more diagnoses to be recorded for the same people in Medicare Advantage than in traditional Medicare. It projects that in 2026 Medicare will spend 14%, about USD 76 billion, more for these enrolees than in traditional Medicare, with favourable selection accounting for about USD 57 billion. After the statutory minimum coding adjustment of 5.9%, risk scores are projected to remain about 4% higher than in traditional Medicare, worth USD 22 billion. The CMS Innovation Center also pays a set amount per patient for a set period and set of services in several models, which it notes is sometimes called capitation.

What the evidence shows

Direct evidence is thin. A Cochrane review by Gosden and colleagues found four eligible studies of primary care physicians. Compared with capitation, fee-for-service produced more visits, specialist visits and diagnostic and curative services, fewer hospital referrals and repeat prescriptions, and higher compliance with a recommended number of visits. The authors judged the generalisability unknown.

Worked example of one practice with two need groups (illustrative)

All numbers are illustrative. A practice has 2,000 enrolled people, 1,500 with low need, each expected to cost GBP 100 a year, and 500 with high need, each expected to cost GBP 300. Expected cost is 1,500 × 100 + 500 × 300 = GBP 300,000, an average of GBP 150 per person.

1. Unadjusted capitation at GBP 150. The practice receives 2,000 × 150 = GBP 300,000 and breaks even in expectation, but it gains GBP 50 on each low-need person and loses GBP 150 on each high-need person. If 100 high-need people leave and 100 low-need people join, revenue is unchanged and cost falls by 100 × (300 − 100) = GBP 20,000, which is the selection incentive.

2. Risk-adjusted capitation. The high-need weight is 300 / 150 = 2 and the low-need weight is 100 / 150, or two thirds. The weighted population is 1,500 × 2/3 + 500 × 2 = 2,000, equal to the list size, so the scheme stays budget neutral. Payment is GBP 100 for a low-need person and GBP 300 for a high-need person, and the same swap of 100 people now lowers revenue and cost by GBP 20,000 each, removing the gain.

3. Selection the formula cannot see. If the practice attracts 100 high-need people whose true expected cost is GBP 250, because their condition is well controlled in ways the risk model does not record, it gains 100 × (300 − 250) = GBP 5,000. This is favourable selection in MedPAC's sense.

4. Coding. If a qualifying diagnosis is recorded for 60 low-need people whose need has not changed, the weighted population becomes 1,440 × 2/3 + 560 × 2 = 2,080, and revenue rises by 80 × 150 = GBP 12,000 with no change in cost. Scaling all risk scores by 2,000 / 2,080, a cut of about 3.8%, would offset this practice's gain, but it would also cut payment to practices that coded no more intensively.

Adjustment removes the selection reward only for differences the formula measures, and a formula built on recorded diagnoses also rewards recording.

Capitation in economic evaluation and budget impact

Treatment depends on the perspective. From a health system perspective the payment is a transfer between purchaser and provider; the NICE technology appraisal manual (PMG36) states that costs should relate to NHS and personal social services resources, valued at prices relevant to them. A model should therefore value the care delivered, not add the capitation payment as a further cost. In a budget impact analysis for the purchaser, the capitation payment is the expenditure, and a new treatment used within a capitated service changes purchaser spending only when the rate is reset.

Boundaries with neighbouring payment concepts

Capitation pays per enrolled person per period; a global budget pays a fixed sum to an organisation whatever the size of its list; salary pays per period of a clinician's time; fee-for-service pays per item; and a bundled payment pays per episode of care. An accountable care organisation is a provider group that may be paid by capitation, not a payment method.

Common misreadings of capitation

Most errors come from reading the rate as more than an average price for a pooled list. Five recur.

1. A capitation rate is not a budget for one patient. A high-cost person is not entitled to less care because the rate is low.
2. Capitation does not guarantee prevention. A provider gains only if it pays for the avoided care and keeps the person long enough.
3. Lower use is not proof of efficiency. Reduced activity can mean better care or unmet need.
4. Risk adjustment does not end selection. It removes the reward only for what the model observes, and MedPAC's 2026 estimate of the payment effect of favourable selection is more than twice its estimate for coding.
5. Headline rates are not comparable without matching scope. A GP global sum and a Medicare Advantage rate cover very different packages.

Sources

• Langenbrunner JC, Cashin C, O'Dougherty S, eds. Designing and Implementing Health Care Provider Payment Systems: How-To Manuals. Washington, DC: World Bank; 2009. https://openknowledge.worldbank.org/handle/10986/13806
• Department of Health and Social Care. The General Medical Services Statement of Financial Entitlements Directions 2026. Signed 30 March 2026. https://assets.publishing.service.gov.uk/media/69cbe5032d120d9d5ec0f352/general-medical-services-statement-of-financial-entitlements-directions-2026.pdf
• Medicare Payment Advisory Commission. Medicare Advantage program payment system. Payment Basics, revised November 2025. https://www.medpac.gov/wp-content/uploads/2026/10/MedPAC_Payment_Basics_26_MA_FINAL_SEC.pdf
• Medicare Payment Advisory Commission. The Medicare Advantage program: status report. Report to the Congress: Medicare Payment Policy, March 2026, chapter 12. https://www.medpac.gov/wp-content/uploads/2026/03/Mar26_Ch12_MedPAC_Report_To_Congress_SEC.pdf
• Centers for Medicare & Medicaid Services, Innovation Center. Capitation and pre-payment (key concepts). Reviewed 10 September 2024. https://www.cms.gov/priorities/innovation/key-concepts/capitation-and-pre-payment
• Ellis RP. Creaming, skimping and dumping: provider competition on the intensive and extensive margins. Journal of Health Economics. 1998;17(5):537-555. https://doi.org/10.1016/S0167-6296(97)00042-8
• Robinson JC. Theory and practice in the design of physician payment incentives. Milbank Quarterly. 2001;79(2):149-177. https://doi.org/10.1111/1468-0009.00202
• Gosden T, Forland F, Kristiansen IS, Sutton M, Leese B, Giuffrida A, Sergison M, Pedersen L. Capitation, salary, fee-for-service and mixed systems of payment: effects on the behaviour of primary care physicians. Cochrane Database of Systematic Reviews. 2000;(3):CD002215. https://doi.org/10.1002/14651858.CD002215
• World Health Organization Regional Office for South-East Asia. Provider payment methods and strategic purchasing for UHC. UHC technical brief. New Delhi: WHO SEARO; 2017. https://iris.who.int/handle/10665/258894
• National Institute for Health and Care Excellence. NICE technology appraisal and highly specialised technologies guidance: the manual (PMG36), chapter 4, economic evaluation. Published 31 January 2022, last updated 13 October 2026. https://www.nice.org.uk/process/pmg36/chapter/economic-evaluation-2