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

This is not an experiment—it is a systematic analysis using the Global Burden of Disease (GBD) 2019 database. The researchers tested a new way of measuring "effective coverage" of health services across 204 countries and territories over 30 years. Instead of just asking whether people can access a doctor (which is what simpler UHC indices measure), they created a composite index that combines:

**23 specific health service indicators** covering promotion (e.g., vaccination), prevention (e.g., screening), and treatment (e.g., cancer therapy)

**Five population-age groups**: reproductive-age women, newborns (0–28 days), children (1–4 years), adults (5–64 years), and older adults (≥65 years)

**Outcome-based measures** like mortality-to-incidence ratios for cancers (how many people die versus how many get the disease) to approximate whether the care people receive is actually *good enough* to save lives

The comparator was the existing WHO UHC Service Coverage Index (SDG indicator 3.8.1) and the World Bank's UHC index. The outcome was a single 0–100 UHC Effective Coverage Index score per country per year, plus estimates of how many "population equivalents" (people with adequate coverage) exist globally.

The entire global population of 204 countries and territories from 1990 to 2019. This is not a sample of individuals—it is a modelled estimate based on thousands of data sources including:

National health surveys (e.g., Demographic and Health Surveys, Multiple Indicator Cluster Surveys)

Administrative data (hospital records, disease registries)

Vital registration systems (birth and death certificates)

Published literature on disease incidence, mortality, and treatment coverage

The unit of analysis is the country-year, not the individual person. The GBD 2019 database includes over 1 billion data points across all causes of death and disability, but the UHC index specifically draws on 23 indicators for which country-level data were available for at least some years between 1990 and 2019.

The researchers constructed the UHC Effective Coverage Index using a multi-step process:

1. **Indicator selection**: 23 indicators were chosen based on WHO's GPW13 measurement framework. These included:

- **Reproductive and newborn**: Skilled birth attendance, antenatal care (4+ visits), facility delivery, postnatal care for mothers and newborns, and coverage of three doses of diphtheria-tetanus-pertussis (DTP3) vaccine

- **Child health**: Oral rehydration therapy for diarrhoea, care-seeking for pneumonia, measles vaccination (first dose), and DTP3 coverage

- **Infectious diseases**: Tuberculosis treatment coverage, HIV antiretroviral therapy coverage, insecticide-treated net use for malaria, and coverage of isoniazid preventive therapy for TB

- **Non-communicable diseases**: Cervical cancer screening, breast cancer screening, diabetes treatment coverage, hypertension treatment coverage, and mortality-to-incidence ratios for breast cancer, cervical cancer, colon cancer, lung cancer, and stomach cancer

- **Health service capacity and access**: Hospital beds per capita, physician density, and a composite "health access and quality" index based on amenable mortality (deaths that should not occur with good medical care)

2. **Scaling**: Outcome-based indicators (like mortality-to-incidence ratios) were transformed to a 0–100 scale using the 2.5th and 97.5th percentiles of all country-year values as anchors. For example, a country with the best mortality-to-incidence ratio for breast cancer (fewest deaths per case) would score near 100, while the worst would score near 0.

3. **Weighting**: Each indicator was weighted by its associated potential health gain, measured in disability-adjusted life-years (DALYs) for that location-year and population-age group. This means indicators that prevent more death and disability (e.g., hypertension treatment, which prevents strokes and heart attacks) get more weight than indicators for less burdensome conditions.

4. **Index construction**: The weighted average of all 23 indicators produced the final UHC Effective Coverage Index score (0–100) for each country-year.

5. **Frontier analysis**: They plotted each country's UHC score against its pooled health spending per capita (government health spending + prepaid private spending + development assistance, adjusted for purchasing power parity). The "frontier" represents the maximum UHC score achievable at a given spending level, based on the best-performing countries.

6. **Population equivalent estimation**: They calculated how many people would need to be covered to move from the current UHC score to a perfect 100, then tracked changes over time to estimate progress toward the GPW13 target of 1 billion more people benefiting from UHC by 2023.

**Study design**: This is a systematic analysis of observational data—specifically, a repeated cross-sectional ecological study using modelled estimates from the Global Burden of Disease 2019. It is not a randomized controlled trial, not an experiment, and not a meta-analysis of trials. It is best described as a descriptive epidemiological study with predictive modelling.

**Data sources**: The GBD 2019 study synthesizes data from over 200,000 sources including surveys, censuses, vital registration, hospital records, and published studies. For the UHC index, the researchers used a subset of these sources that contained information on the 23 selected indicators. Missing data were imputed using statistical models that borrow strength from similar countries and time periods.

**Statistical approach**:

Uncertainty intervals (95% UIs) were generated using 1,000 draws from the posterior distribution of each modelled estimate

Trends were assessed using annualized rates of change (percent per year)

Correlation between UHC index and health spending was assessed using Pearson's r

Frontier analysis used quantile regression to estimate the 90th percentile of UHC performance at each spending level

**What this design can prove**:

It can describe global and national trends in health service coverage over 30 years

It can identify which types of services (e.g., NCD care vs. maternal health) are lagging

It can show associations between health spending and coverage outcomes

It can generate hypotheses about why some countries outperform others at similar spending levels

**What this design cannot prove**:

It cannot prove causation. The association between health spending and UHC coverage does not mean spending more money *causes* better coverage—reverse causation (better coverage attracts more funding) and confounding (countries with stronger institutions have both higher spending and better coverage) are possible

It cannot tell you whether a specific intervention (e.g., a new vaccine program) works at the individual level

It cannot account for within-country inequality. A country could score 80 on the index while the richest 20% get excellent care and the poorest 20% get none

The ecological fallacy applies: country-level averages do not describe any individual person's experience

**Major methodological weaknesses**:

**Data quality varies enormously**: High-income countries have complete vital registration and survey data; low-income countries often rely on sparse surveys with large uncertainty intervals. The model imputes missing data, but imputation introduces error

**Outcome-based indicators are imperfect**: Mortality-to-incidence ratios reflect both healthcare quality and underlying disease biology, genetics, and lifestyle factors. A country with a low mortality-to-incidence ratio for breast cancer might have excellent treatment, or it might simply have a population with less aggressive tumour types

**Weighting by DALYs is controversial**: This gives more weight to conditions that cause more death and disability, which makes sense for population health but means the index is dominated by a few high-burden conditions (e.g., cardiovascular disease, cancer) and may not reflect the full breadth of health services

**No measure of financial protection**: UHC includes both service coverage and financial hardship, but this index only measures service coverage. A country could score high while still bankrupting its citizens with medical bills

**The frontier analysis assumes efficiency is achievable**: The "maximum" UHC score at a given spending level is based on the best-performing countries, but those countries may have unique historical, cultural, or political advantages that cannot be replicated

**Global UHC Effective Coverage Index improved** from 45.8 (95% UI 44.2–47.5) in 1990 to 60.3 (58.7–61.9) in 2019—an increase of 14.5 points over 29 years, or roughly 0.5 points per year

**Massive country variation in 2019**: Japan and Iceland scored ≥95; Somalia and the Central African Republic scored <25. The gap between best and worst performers was >70 points

**Sub-Saharan Africa accelerated after 2010**: Annualized increase of 2.6% (1.9–3.3) per year from 2010 to 2019, compared to 1.1% (0.8–1.4) from 1990 to 2010

**Most other regions slowed down**: In 2010–2019, annualized increases were lower than in 1990–2010 for high-income countries (0.3% vs. 0.6%), Latin America (0.6% vs. 1.2%), and Southeast Asia (0.8% vs. 1.5%)

**Non-communicable disease (NCD) coverage lags badly**: In 2019, NCD effective coverage indicators (e.g., hypertension treatment, cancer mortality-to-incidence ratios) were consistently 10–30 points lower than communicable disease and maternal/child health indicators, despite NCDs accounting for a larger share of potential health gains

**Health spending correlates with coverage**: The correlation between pooled health spending per capita and UHC index was r = 0.79 in 2019—strong, but far from perfect. Many countries (e.g., the USA) spend far more than their UHC score would predict

**Spending needed for high coverage**: To achieve a UHC index score of 80, countries would need approximately $1,398 per capita in pooled health spending (adjusted for purchasing power parity), assuming maximum efficiency

**GPW13 target will be missed**: From 2018 to 2023, an estimated 388.9 million (358.6–421.3) additional population equivalents will gain UHC effective coverage—far short of the 1 billion target

**3.1 billion people still lacking in 2023**: Projected 3.1 billion (3.0–3.2) population equivalents will lack UHC effective coverage in 2023, with 968 million (903.5–1,040.3) in South Asia alone

The global UHC index improved by 14.5 points over 29 years—roughly equivalent to moving from a "D" grade to a "D+" on a 100-point scale. To put this in perspective:

The difference between the best-performing country (Japan, ~95) and the worst (Somalia, <25) is about 70 points—a chasm that represents life-or-death differences in access to basic care like childhood vaccines, skilled birth attendance, and hypertension treatment

The annualized improvement of 0.5 points per year globally means that at current rates, it would take 80 years to go from the current 60.3 to a perfect 100

The gap between NCD and communicable disease coverage (10–30 points) means that a person with high blood pressure in a low-income country is roughly 20–40% less likely to receive effective treatment than a child with pneumonia—even though hypertension causes far more death and disability globally

The $1,398 per capita spending threshold for a UHC score of 80 is roughly equivalent to what Costa Rica spends (which achieves a score of ~80), while the USA spends over $10,000 per capita and achieves a similar score—suggesting massive inefficiency in some systems

**What the authors acknowledge**:

Data gaps in low-income countries require statistical imputation, which introduces uncertainty

The 23 indicators do not cover all health services (e.g., mental health, palliative care, rehabilitation are excluded)

Outcome-based indicators (mortality-to-incidence ratios) reflect factors beyond healthcare quality

The index does not measure financial protection or catastrophic health spending

Within-country inequality is not captured—national averages hide disparities by income, geography, ethnicity, and gender

The frontier analysis assumes that all countries could achieve the same efficiency as the best performers, which may not be realistic

**What a critical reader would note**:

**No individual-level data**: This is entirely ecological. You cannot conclude anything about individual patient experiences

**Weighting decisions are subjective**: Choosing to weight by DALYs means the index is dominated by conditions that kill or disable the most people. This is reasonable for population health but means rare diseases and conditions with low mortality (e.g., mental health, chronic pain) are essentially invisible

**The "effective coverage" concept is still evolving**: There is no gold standard for measuring whether care is actually "effective" at the population level. The authors' approach is sophisticated but unvalidated against real-world outcomes like mortality rates

**Funding source**: Funded by the Bill & Melinda Gates Foundation, which has a strong interest in showing that UHC is achievable and that spending more money works. While the GBD study is generally considered rigorous, the framing of results (e.g., "countries can do better with the same money") aligns with the foundation's advocacy goals

**No preregistration**: This is not a clinical trial; there is no preregistered analysis plan. The researchers could have made many analytical choices (which indicators, how to weight, which frontier method) that affect the results

**The 1 billion target is arbitrary**: The GPW13 target of 1 billion more people benefiting from UHC by 2023 was set before this analysis. The finding that it will be missed is not surprising—it is a political target, not a scientific prediction

**Important caveat**: This paper is about national health systems, not individual experiments. However, the concept of "effective coverage" can be adapted for personal health tracking. Here is how:

**What to test**:

Whether your own "effective coverage" of preventive services (vaccinations, cancer screenings, blood pressure checks) matches the recommended schedule for your age and sex

Whether you are receiving "effective" treatment for any chronic conditions you have—meaning treatment that actually lowers your risk of death or disability, not just treatment that makes you feel like you are doing something

**Minimum meaningful duration**:

For preventive services: 1 year is enough to assess whether you are up-to-date on annual screenings (e.g., blood pressure, cholesterol, cervical cancer screening)

For chronic disease management: 3–6 months to see if treatment is actually controlling your condition (e.g., blood pressure below 130/80, HbA1c below 7% for diabetes)

For lifestyle changes: 6–12 months to see if changes in diet, exercise, or sleep affect your need for healthcare services

**What to measure**:

**Coverage rate**: What percentage of recommended preventive services have you received in the past year? (e.g., flu shot, blood pressure check, dental cleaning, cancer screening if age-appropriate)

**Control metrics**: For any chronic condition, measure the actual outcome (e.g., blood pressure, blood sugar, cholesterol) and compare to guideline targets

**Healthcare utilization**: Track doctor visits, hospitalizations, emergency department visits, and prescription fills. A high "effective coverage" system should keep you out of the hospital

**Out-of-pocket costs**: Track what you actually pay for healthcare. UHC includes financial protection—