Universal Health Coverage Index

Universal Health Coverage Index and How the SDG 3.8.1 Score Is Built

The Universal Health Coverage Index, formally the UHC service coverage index (SCI), is the number that WHO reports for Sustainable Development Goal indicator 3.8.1. It condenses 14 tracer measures of whether people receive the health services they need into one unitless score, where 100 is the optimal value. Health economists meet it as the service half of universal health coverage monitoring, as an output variable in cross-country efficiency studies and as a headline that has to be unpacked before it says anything useful about priorities. Because WHO revised the method in 2025, two values can only be compared when they come from the same method edition. The global value was 54 in 2000 and 71 in 2023, with most of the gain made before 2015.

Where the index sits in UHC monitoring

SDG target 3.8 has two indicators. The UN metadata for indicator 3.8.1 states that it should always be interpreted together with indicator 3.8.2, which measures financial protection through out-of-pocket spending on health, so a high service score says nothing on its own about catastrophic health expenditure or impoverishing health expenditure. WHO is the custodian agency and compiles the index from existing series produced by UN agencies and countries. It does not run new estimation to produce tracer values.

The first baseline values were released in December 2017 and updates follow every two years. Hogan and colleagues, who developed the original index, used 16 tracers across 183 countries and reported a median national value of 65 out of 100 for 2015, with a range of 22 to 86. The current series, labelled the 2025 methodology, uses 14 tracers, and since 2021 the index has been estimated every two years for all 194 WHO Member States.

The 14 tracer indicators in four service domains

The tracers were chosen because they are well established and widely reported, and the metadata describes them as indicative of service coverage rather than a complete list of the services that UHC requires. The table follows the 3.8.1 metadata.

The 2025 revision changed three tracers. Family planning and health workforce were aligned with SDG indicators 3.7.1 and 3.c.1, and diabetes treatment coverage replaced mean fasting plasma glucose, a proxy adopted provisionally in 2017 because coverage data were lacking.

Putting the tracers on a common scale

Each tracer is placed on a scale where 0 is the lowest value and 100 the optimal one. Three tracers need conversion or rescaling, and the rules are fixed in advance by the metadata rather than chosen after seeing the data.

Tobacco use is converted to non-use so that a higher number means improvement, and then stretched using a non-zero minimum of 30%.

$$T = \frac{N - 30}{100 - 30} \times 100$$

where $T$ is the tobacco tracer score and $N$ is the age-standardised prevalence of tobacco non-use in adults, in per cent.

The two density tracers are capped. The metadata explains that very high bed densities are not necessarily an efficient use of resources, so capacity above a threshold earns no extra credit.

$$H = \min\left(100, \frac{b}{18} \times 100\right)$$

where $H$ is the hospital access score and $b$ is hospital beds per 10,000 population. The workforce score is built the same way against a threshold of 6 doctors, nurses and midwives per 1,000 population, the 5th percentile among high-income countries in 2013. Any tracer that comes out at zero is set to 1 and any above 100 is set to 100 before aggregation.

Population-weighted geometric means

Within each domain the tracers are combined by a weighted geometric mean, and the four domain scores are then combined by an unweighted geometric mean. The weight for each tracer is the size of the population it refers to.

$$D_k = \left( \prod_{i \in k} x_i^{P_i} \right)^{1 / \sum_{i \in k} P_i} \qquad \text{SCI} = \left( D_1 D_2 D_3 D_4 \right)^{1/4}$$

where $D_k$ is the score for domain $k$, $x_i$ is the rescaled score of tracer $i$ in that domain, $P_i$ is the denominator population of tracer $i$ (for example adults with hypertension, or infants for immunisation) taken mainly from UN population estimates, and SCI is the overall index.

Before 2025 every tracer within a domain had equal weight. The 2025 global monitoring report found that, under equal weights, over 75% of the change in the global SCI between 2000 and 2021 came from infectious disease treatment, and antiretroviral therapy alone accounted for over 60% of it. The new weights reflect how many people each tracer concerns. A geometric mean also means that a weak domain cannot be fully offset by a strong one, since the score falls below the arithmetic mean whenever domains are uneven and is pulled towards zero by any very low component.

Progress is summarised by an annualised rate of change, which the report also uses to project the index to 2030.

$$r = \frac{\ln\left(\text{SCI}_{t_2} / \text{SCI}_{t_1}\right)}{t_2 - t_1}$$

where $r$ is the annualised rate of change and $\text{SCI}_{t_1}$ and $\text{SCI}_{t_2}$ are index values in years $t_1$ and $t_2$, with $t_1 < t_2$. On this measure the global index improved by 1.5% a year between 2000 and 2015 and by 0.5% a year between 2015 and 2023.

Worked example for a hypothetical country

All numbers in this example are illustrative. A country reports domain scores of 80 for RMNCH and 72 for infectious diseases, and the other two domains are computed from their tracers.

1. Rescale the tracers. Tobacco use of 20% gives non-use of 80, and since 80 minus 30 is 50 the tobacco score is 50 / 70 × 100 = 71.4. With 6 hospital beds per 10,000 population the hospital access score is 6 / 18 × 100 = 33.3, and with 2 health workers per 1,000 the workforce score is 2 / 6 × 100 = 33.3. The IHR capacity score of 60 is already on the scale.

2. NCD domain. Hypertension treatment coverage is 40 with a denominator of 2.0 million adults, diabetes treatment coverage is 50 with 0.8 million and tobacco non-use is 71.4 with 10.0 million adults, a total weight of 12.8 million. The logarithms are 3.6889, 3.9120 and 4.2683, and multiplying each by its weight gives 7.3778, 3.1296 and 42.683, which sum to 53.190. Dividing by 12.8 gives 4.1555, and exp(4.1555) = 63.8. With equal weights the domain would score the cube root of 40 × 50 × 71.4 = 142,800, which is 52.3, so the large tobacco denominator lifts the NCD score by 11.5 points.

3. Capacity domain. All three capacity tracers use the total population as denominator, so their weights are equal and $D$ is the cube root of 33.3 × 33.3 × 60 = 66,533, which is 40.5.

4. Overall index. The product of the four domains is 80 × 72 × 63.8 × 40.5 = 14,883,264, and its fourth root is 62.1. The arithmetic mean would be (80 + 72 + 63.8 + 40.5) / 4 = 64.1, so the imbalance costs the country 2 points.

5. Where gains count most. Raising capacity by 10 points to 50.5 gives 80 × 72 × 63.8 × 50.5 = 18,558,144 and an index of 65.6, a rise of 3.5. Raising RMNCH by 10 points to 90 instead gives 90 × 72 × 63.8 × 40.5 = 16,743,672 and an index of 64.0, a rise of 1.9.

6. Rate of change. If the same country scored 55.0 five years earlier, ln(62.1 / 55.0) is 0.1214, and dividing by 5 years gives 0.024, or 2.4% a year.

The example shows that the index rewards balanced progress and that the same 10-point gain is worth almost twice as much in the weakest domain. It says nothing about which gain is cheaper or produces more health, which is a separate question for economic evaluation.

Reading the score with care

The 2025 report states that, as an index score, the SCI cannot be directly interpreted as the percentage of the population covered by a set of services. Its estimate that 4.6 billion people lacked full coverage in 2023 comes from a separate regression step that converts average coverage into the share of people receiving most needed services.

Many tracers measure contact coverage, whether someone received a service, rather than effective coverage, which the report defines as the proportion of people in need who receive services of sufficient quality to obtain potential health gains. Some remain proxies, and the report notes that this can bias a domain towards a particular indicator, such as tobacco in the NCD domain. It also explains that the global SCI looked smooth through the COVID-19 pandemic because several tracers do not directly measure service delivery or react slowly to shocks. Questions of quality of care and unmet need therefore need evidence beyond the index.

Missing values are filled by a simple algorithm. Gaps between observed years are linearly interpolated, the last observed value is carried forward to the current year, and a country with no data on a tracer receives a regional median, or a regression or subregional estimate for two tracers. The metadata asks that the extent of this imputation be reported with the index. Hogan and colleagues found that, on average, countries had primary data since 2010 for 72% of the indicators in the baseline set.

Revisions affect levels as well as trends. The 2025 report states that the revised SCI keeps the same name but gives higher levels and a slower improving trend than the original method, so a value from an older report should not be set against a value from the current series. Trend analysis of the index is only meaningful within one method edition.

National averages can hide large gaps. The metadata treats equity as central and suggests computing the index by province or by urban and rural residence. For household wealth, the most complete data are for RMNCH tracers. In the baseline study, coverage among the poorest fifth of households was between 1% and 66% lower than in the national population across 52 countries. Measures such as the concentration index apply to the tracers where microdata allow, and health inequality analysis adds the distributional view that a single score lacks.

Uses in health economic analysis

The index is a common output variable in studies of how efficiently health systems turn spending into coverage. Jordi and colleagues used data envelopment analysis for 172 countries with 2015 data, taking service coverage and financial risk protection as outputs, and reported a mean bias-corrected technical efficiency score of 85.7%, with 68.9% for low-income and 95.5% for high-income countries. Similar questions are studied with stochastic frontier analysis and summarised as an efficiency index. Such results inherit the features of the output measure. The index is bounded at 100, capped for beds and workforce, partly imputed and redefined between editions, so an efficiency score estimated on the pre-2025 series may change when the current series is used.

Decomposition shows which tracers drive change. Under the 2025 method, the infectious diseases domain accounted for 52% of the change in the global SCI between 2000 and 2023, and insecticide-treated nets, basic sanitation and lower tobacco use together accounted for 78%. Raising the score quickly therefore favours tracers with large denominators and low values, which need not be the services offering the most health per unit of spending.

The index places no value on health gains and contains no costs, so it cannot rank interventions or show whether an expansion is worth its opportunity cost. Its role in priority setting is diagnostic, showing where coverage is weakest, after which economic evaluation compares the options. A related measure, the UHC effective coverage index from the Global Burden of Disease Study 2019, combines 23 effective coverage indicators weighted by potential health gains measured in disability-adjusted life years. Its authors reported that it generally performed better than the WHO service coverage index of that time on their validity tests, and they set it against pooled health spending per capita (correlation 0.79). The two indices answer different questions and are not interchangeable.

Sources

• GBD 2019 Universal Health Coverage Collaborators. Measuring universal health coverage based on an index of effective coverage of health services in 204 countries and territories, 1990-2019: a systematic analysis for the Global Burden of Disease Study 2019. The Lancet. 2020;396(10258):1250-1284. https://doi.org/10.1016/S0140-6736(20)30750-9
• Hogan DR, Stevens GA, Hosseinpoor AR, Boerma T. Monitoring universal health coverage within the Sustainable Development Goals: development and baseline data for an index of essential health services. The Lancet Global Health. 2018;6(2):e152-e168. https://doi.org/10.1016/S2214-109X(17)30472-2
• Jordi E, Pley C, Jowett M, Abou Jaoude GJ, Haghparast-Bidgoli H. Assessing the efficiency of countries in making progress towards universal health coverage: a data envelopment analysis of 172 countries. BMJ Global Health. 2020;5(10):e002992. https://doi.org/10.1136/bmjgh-2020-002992
• United Nations Statistics Division. SDG indicator metadata: indicator 3.8.1, coverage of essential health services. New York: United Nations; last updated 12 December 2025. https://unstats.un.org/sdgs/metadata/files/Metadata-03-08-01.pdf
• World Health Organization, International Bank for Reconstruction and Development / The World Bank. Tracking universal health coverage: 2025 global monitoring report. Geneva: WHO and World Bank; 2025. ISBN 978-92-4-011780-8. https://www.who.int/teams/health-financing-and-economics/global-monitoring-report/2025