Actuarial Projection

Actuarial Projection of Health Spending, Claims and Populations Over Time

An actuarial projection in health economics carries a covered population and its costs forward year by year, so that a payer or government can see what health care will cost and whether its funding will keep pace. The answer is always conditional, stating what spending, claims or a fund balance would be if the assumed population, prices, use of services and policy turn out as specified. Official examples include the national health expenditure projections of the Office of the Actuary at the Centers for Medicare and Medicaid Services, the annual Medicare Trustees Report and the long-term health spending projections of the UK Office for Budget Responsibility (OBR). Projections also supply the baseline for budget impact analysis and for sustainability work. Pricing and reserving for a single scheme are covered under actuarial analysis.

The projection identity of people, cost per person and growth

Most health projections rest on one identity, in which spending equals the number of people in each age and sex group multiplied by cost per person in that group, summed across groups. Growth then splits into a demographic part, from changes in the size and age mix of the population, and a non-demographic part, from changes in cost per person at a given age. In its simplest form the identity is written as follows.

$$C_t = \sum_{a} N_{a,t} \, c_{a,0} \, (1 + g)^{t}$$

where $C_t$ is projected real spending in year $t$, $N_{a,t}$ is the projected population in age group $a$ in year $t$, $c_{a,0}$ is real cost per person in age group $a$ in the base year, and $g$ is the assumed annual growth in real cost per person at a given age. Dividing $C_t$ by projected GDP gives health spending as a share of GDP, the usual summary in sustainability work.

The non-demographic term carries most of the judgement. The CMS methodology models real per capita private spending after adjusting for a demographic index built from the population's composition by age, sex and time to death, so that the dependent variable is the growth that would occur if that composition stayed constant. It treats income as acting on health spending with a lag, fitted over seven years, and reports a current-period income elasticity of 0.19 and a long-term elasticity of 1.62 of private spending. Prices and quantities are projected separately.

Projecting the population with the cohort component method and life tables

The population term comes from demographic projection. The Office for National Statistics (ONS) produces UK national population projections with the cohort component method, carrying the mid-year population forward under assumptions for fertility, migration and mortality. The 2024-based principal projection assumes completed family size of 1.42 children per woman by mid-2049, annual mortality improvement of 1.1% at ages 0 to 90 by 2049, and net international migration of 230,000 a year from mid-2027. It projects the number aged 85 and over to double from 1.75 million in mid-2024 to 3.6 million by mid-2049. ONS publishes 13 variant projections alongside the principal one.

Mortality assumptions enter through the life table. ONS explains that a period life table applies the mortality rates of one year throughout a lifetime, while a cohort life table combines observed and projected rates for people born in the same year and so allows for future improvement. Period life expectancy matches cohort life expectancy only if age-specific mortality rates never change, so a projection that applies period rates to future years understates survival when mortality is improving. The life-table method of survival analysis, also called the actuarial method, which estimates survival from grouped follow-up data, is a separate topic.

How three official health projections are built

The three projections below share the identity above but differ in horizon, scope and policy basis. All three publish their assumptions with the results.

US national health expenditure projections

The CMS Office of the Actuary projects all categories of US national health expenditure ten years ahead by type of service, payer and sponsor. The projections follow a current-law framework, so they do not assume legislative change. Medicare, Medicaid and the Children's Health Insurance Program (CHIP) enter as separate actuarial projections, while private spending and smaller public programmes are modelled econometrically. The 2025 to 2034 projections have nominal health spending growing at an average of 5.4% a year against GDP growth of 4.1%, raising its share of GDP from 18.0% in 2024 to 20.6% in 2034. The methodology states that uncertainty increases with the projection horizon and at lower levels of aggregation.

The Medicare Trustees Report

The Medicare Trustees report each year on the Hospital Insurance (HI) and Supplementary Medical Insurance trust funds over a 75-year period under intermediate assumptions, with low-cost and high-cost alternatives for HI. The 2026 report projects HI trust fund depletion in the second quarter of 2033, after which income would cover 89% of scheduled benefits. It defines the trust fund ratio as reserves at the start of a year as a percentage of that year's cost, and expresses the 75-year actuarial deficit as 0.56% of taxable payroll. Medicare spending is projected to rise from 3.9% of GDP in 2025 to 7.5% in 2100, or 9.8% under an illustrative alternative that phases down legislated payment constraints. The Trustees ask readers to treat the current-law figures as the outcome if the required cost reductions can be sustained.

OBR long-term health spending projections

The OBR projects UK health spending over 50 years in its Fiscal risks and sustainability report, using the ONS population projections. In the July 2026 baseline, health spending rises from 8% of GDP in 2030-31 to 13% in 2075-76, real growth of 2.5% a year. Demographic factors contribute 0.4 percentage points of annual real growth, income effects 1.2 points (an income elasticity of 0.8), and other cost pressures 1 point (components rounded), attributed to the Baumol effect, a gently rising prevalence of chronic conditions and a residual that may reflect technology. Without the other cost pressures, health spending would reach around 9% of GDP.

Scenarios, sensitivity analysis and stochastic ranges

A single path hides how much the answer depends on its assumptions, so official projections publish ranges. Alternative sets, one-at-a-time tests and stochastic methods answer different questions.

Coherent alternative sets change several assumptions together. In the Medicare report the high-cost and low-cost HI assumptions raise or lower annual growth in the cost rate by 2 percentage points for 25 years, and ONS combination variants such as "old age structure" pair low fertility, high life expectancy and low migration. One-at-a-time tests, a form of deterministic sensitivity analysis, vary one assumption while holding the rest at intermediate values, as the Trustees do for the real interest rate and health cost growth. Stochastic methods set the range from statistical variation observed in the past; the Trustees use stochastic modelling of this kind, based on historical variation in the increase factors, for a short-range test of the Part B account. Scenario analysis of population health follows the same logic, and the OBR's lower and higher healthy life expectancy scenarios give 15% and 12% of GDP against the 13% baseline, which assumes half of life expectancy gains are spent in good health.

Worked example of ageing versus cost growth over ten years

All figures are illustrative. A health system covers 10 million people, with 8.0 million under 65 costing £2,000 each a year and 2.0 million aged 65 and over costing £6,000 each. GDP is £500 billion and grows by 1.5% a year in real terms.

1. Base year. Spending is 8.0 million × £2,000 + 2.0 million × £6,000 = £16 billion + £12 billion = £28 billion, a GDP share of 5.6%.

2. Demographic change alone. In year 10 the projected population is 7.8 million under 65 and 2.5 million aged 65 and over. At base-year costs, spending is 7.8 million × £2,000 + 2.5 million × £6,000 = £15.6 billion + £15 billion = £30.6 billion, a demographic factor of 30.6 / 28 = 1.0929. The population grows by 3%, but spending grows by 9.3% because the older group is costlier.

3. Non-demographic growth. In the central scenario real cost per person at a given age grows by 2.5% a year, which gives the following.

$$C_{10} = 30.6 \times 1.025^{10} = 30.6 \times 1.2801 = 39.17$$

where $C_{10}$ is real spending in year 10 in £ billion. The combined factor is 1.0929 × 1.2801 = 1.399, a real rise of 39.9%.

4. Share of GDP. Real GDP grows by a factor of 1.1605 over ten years, to £580.27 billion, so spending reaches 6.75% of GDP.

5. Scenarios. If cost per person grows at 1.5%, the same rate as GDP, spending is 30.6 × 1.1605 = 35.51 and the share is 6.12%, the rise from demographic change alone. At 3.5% a year spending is 30.6 × 1.4106 = 43.16 and the share is 7.44%.

6. Interpretation. Demographic change adds about half a point of GDP over the decade (5.6% to 6.12%), while the assumption about cost growth per person moves the result by a further 0.63 points in the central case and 1.32 points in the high case. The spread across cost-growth assumptions exceeds the whole demographic effect, as in the OBR baseline, where demography explains 0.4 of 2.5 points of growth.

Projections in budget impact and sustainability analysis

An actuarial projection supplies the baseline against which a change is costed. The ISPOR good practice report on budget impact analysis lists its key elements as the eligible population, the current and expected treatment mix, the cost of those mixes and changes in condition-related costs. It allows a cohort or patient-level model, accounting for people entering and leaving the eligible population over time, where a simple cost calculator cannot capture changes in population size, severity mix or treatment patterns.

Sustainability analysis asks whether projected income covers projected spending. For a fund, the balance moves as follows.

$$R_t = R_{t-1} + I_t - C_t$$

where $R_t$ is reserves at the end of year $t$, $I_t$ is income from contributions, taxes and interest, and $C_t$ is benefit and administrative spending. The Medicare trust fund ratio, depletion date and actuarial balance are summaries of this path. A cost-effectiveness analysis asks whether a technology is good value; the projection asks whether a payer can afford its commitments, and the two answers can differ.

Limits and common misreadings

ONS states plainly that its population projections are not forecasts. Several misreadings recur.

1. Treating the central path as a prediction. A 2020 CMS review of its national health expenditure projections found that projected figures had tended to be slightly over-projected, with the earliest years somewhat more accurate.
2. Attributing growth mainly to ageing. Demography is the smallest of the three OBR growth components.
3. Reading current law as likely policy. The Trustees' illustrative alternative ends 2.3 points of GDP higher in 2100.
4. Using period mortality for future years. Period life expectancy ignores future mortality improvement and so understates the number of older people to be covered.
5. Ignoring the health of extra years. Whether longer lives are healthy lives moves the OBR projection from 13% to between 12% and 15% of GDP.

Sources

• Boards of Trustees, Federal Hospital Insurance and Federal Supplementary Medical Insurance Trust Funds. 2026 Annual Report of the Boards of Trustees of the Federal Hospital Insurance and Federal Supplementary Medical Insurance Trust Funds. Washington, DC; 2026. https://www.cms.gov/oact/tr/2026
• Centers for Medicare and Medicaid Services, Office of the Actuary. Projections of National Health Expenditures and Health Insurance Enrollment: Methodology and Model Specification. Baltimore: CMS; last updated 24 June 2026. https://www.cms.gov/Research-Statistics-Data-and-Systems/Statistics-Trends-and-Reports/NationalHealthExpendData/downloads/projectionsmethodology.PDF
• Centers for Medicare and Medicaid Services, Office of the Actuary. National Health Expenditure Projections 2025-2034: Forecast Summary. Baltimore: CMS; 2026. https://www.cms.gov/files/document/nhe-projections-forecast-summary.pdf
• Centers for Medicare and Medicaid Services, Office of the Actuary. Analysis of National Health Expenditure Projections Accuracy. Baltimore: CMS; November 2020. https://www.cms.gov/research-statistics-data-and-systems/statistics-trends-and-reports/nationalhealthexpenddata/downloads/projectionaccuracy.pdf
• Office for Budget Responsibility. Fiscal Risks and Sustainability. London: OBR; July 2026. https://obr.uk/frs/fiscal-risks-and-sustainability-july-2026/
• Office for National Statistics. National population projections: 2024-based. Newport: ONS; 28 April 2026. https://www.ons.gov.uk/peoplepopulationandcommunity/populationandmigration/populationprojections/bulletins/nationalpopulationprojections/2024based
• Office for National Statistics. National population projections quality and methods guide. Newport: ONS; updated 28 April 2026. https://www.ons.gov.uk/peoplepopulationandcommunity/populationandmigration/populationprojections/methodologies/nationalpopulationprojectionsqmi
• Office for National Statistics. National population projections, variant projections: 2024-based. Newport: ONS; 2026. https://www.ons.gov.uk/peoplepopulationandcommunity/populationandmigration/populationprojections/methodologies/nationalpopulationprojectionsvariantprojections2024based
• Office for National Statistics. Period and cohort life expectancy explained. Newport: ONS; 2023. https://www.ons.gov.uk/peoplepopulationandcommunity/birthsdeathsandmarriages/lifeexpectancies/methodologies/periodandcohortlifeexpectancyexplained
• Sullivan SD, Mauskopf JA, Augustovski F, Caro JJ, Lee KM, Minchin M, Orlewska E, Penna P, Rodriguez Barrios JM, Shau WY. Budget impact analysis: principles of good practice. Report of the ISPOR 2012 Budget Impact Analysis Good Practice II Task Force. Value in Health. 2014;17(1):5-14. https://doi.org/10.1016/j.jval.2013.08.2291