Menu
Menu
Menu
Menu
Background: Achieving universal health coverage (UHC) involves all people receiving the health services they need, of high quality, without experiencing financial hardship. Making progress towards UHC is a policy priority for both countries and global institutions, as highlighted by the agenda of the UN Sustainable Development Goals (SDGs) and WHO’s Thirteenth General Programme of Work (GPW13). Measuring effective coverage at the health-system level is important for understanding whether health services are aligned with countries’ health profiles and are of sufficient quality to produce health gains for populations of all ages. Methods: Based on the Global Burden of Diseases, Injuries, and Risk Factors Study (GBD) 2019, we assessed UHC effective coverage for 204 countries and territories from 1990 to 2019. Drawing from a measurement framework developed through WHO’s GPW13 consultation, we mapped 23 effective coverage indicators to a matrix representing health service types (eg, promotion, prevention, and treatment) and five population-age groups spanning from reproductive and newborn to older adults (≥65 years). Effective coverage indicators were based on intervention coverage or outcome-based measures such as mortality-to-incidence ratios to approximate access to quality care; outcome-based measures were transformed to values on a scale of 0–100 based on the 2·5th and 97·5th percentile of location-year values. We constructed the UHC effective coverage index by weighting each effective coverage indicator relative to its associated potential health gains, as measured by disability-adjusted life-years for each location-year and population-age group. For three tests of validity (content, known-groups, and convergent), UHC effective coverage index performance was generally better than that of other UHC service coverage indices from WHO (ie, the current metric for SDG indicator 3.8.1 on UHC service coverage), the World Bank, and GBD 2017. We quantified frontiers of UHC effective coverage performance on the basis of pooled health spending per capita, representing UHC effective coverage index levels achieved in 2019 relative to country-level government health spending, prepaid private expenditures, and development assistance for health. To assess current trajectories towards the GPW13 UHC billion target—1 billion more people benefiting from UHC by 2023—we estimated additional population equivalents with UHC effective coverage from 2018 to 2023. Findings: Globally, performance on the UHC effective coverage index improved from 45·8 (95% uncertainty interval 44·2–47·5) in 1990 to 60·3 (58·7–61·9) in 2019, yet country-level UHC effective coverage in 2019 still spanned from 95 or higher in Japan and Iceland to lower than 25 in Somalia and the Central African Republic. Since 2010, sub-Saharan Africa showed accelerated gains on the UHC effective coverage index (at an average increase of 2·6% [1·9–3·3] per year up to 2019); by contrast, most other GBD super-regions had slowed rates of progress in 2010–2019 relative to 1990–2010. Many countries showed lagging performance on effective coverage indicators for non-communicable diseases relative to those for communicable diseases and maternal and child health, despite non-communicable diseases accounting for a greater proportion of potential health gains in 2019, suggesting that many health systems are not keeping pace with the rising non-communicable disease burden and associated population health needs. In 2019, the UHC effective coverage index was associated with pooled health spending per capita (r=0·79), although countries across the development spectrum had much lower UHC effective coverage than is potentially achievable relative to their health spending. Under maximum efficiency of translating health spending into UHC effective coverage performance, countries would need to reach $1398 pooled health spending per capita (US$ adjusted for purchasing power parity) in order to achieve 80 on the UHC effective coverage index. From 2018 to 2023, an estimated 388·9 million (358·6–421·3) more population equivalents would have UHC effective coverage, falling well short of the GPW13 target of 1 billion more people benefiting from UHC during this time. Current projections point to an estimated 3·1 billion (3·0–3·2) population equivalents still lacking UHC effective coverage in 2023, with nearly a third (968·1 million [903·5–1040·3]) residing in south Asia. Interpretation: The present study demonstrates the utility of measuring effective coverage and its role in supporting improved health outcomes for all people—the ultimate goal of UHC and its achievement. Global ambitions to accelerate progress on UHC service coverage are increasingly unlikely unless concerted action on non-communicable diseases occurs and countries can better translate health spending into improved performance. Focusing on effective coverage and accounting for the world’s evolving health needs lays the groundwork for better understanding how close—or how far—all populations are in benefiting from UHC. Funding: Bill & Melinda Gates Foundation.
Our primary analysis involved three main steps: first, to use intervention coverage or compute proxy measures of effective coverage for 23 indicators; second, to calculate the fraction of potential health gains associated with each effective coverage indicator based on each location’s disease burden profile; and third, to construct the overall UHC effective coverage index by weighting each effective coverage indicator relative to its health gains fraction. We then did secondary analyses, assessing UHC effective coverage performance relative to health spending and current trajectories towards the GPW13 UHC billion target. Each step is summarised below and further described in appendix 1 (pp 12–61). This analysis uses estimates from the broader GBD 2019,34, 35, 36 covering 204 countries and territories from 1990 to 2019. Details of disease-specific, injury-specific, and coverage-specific data inputs and processing, statistical synthesis approaches, and final models are available in the accompanying GBD 2019 capstone publications.34, 35, 36 This study complies with the Guidelines for Accurate and Transparent Health Estimates Reporting (GATHER) statement,37 with further information provided in the appendix 1 (pp 69–72). Development of the UHC effective coverage measurement framework and selection of effective coverage indicators was based on consultation, methods testing, and refinement via the WHO ERG on the GPW13 from 2017 to 2019;7, 33, 38, 39 the background and details of this process are provided in the appendix 1 (pp 12–28). The resulting framework (figure 1) and currently included effective coverage indicators (table 1) sought to represent the range of different health services that populations need across their lifespans while recognising present data gaps and appeals for measurement parsimony (appendix 1 pp 18–28). UHC effective coverage measurement framework Additional information about the framework development process and selection of effective coverage indicators can be found in appendix 1 (pp 12–28). ART=antiretroviral therapy. DTP3=diphtheria-tetanus-pertussis vaccine, 3 doses. IHD=ischaemic heart disease. CKD=chronic kidney disease. COPD=chronic obstructive pulmonary disease. LRI=lower respiratory infection. MCV1=measles-containing-vaccine, 1 dose. MNCH=maternal, neonatal, and child health. NCDs=non-communicable diseases. TB=tuberculosis. UHC=universal health coverage. Details of the 23 effective coverage indicators included in the UHC effective coverage index, by health service type Additional information about the framework development process and selection of effective coverage indicators can be found in appendix 1 (pp 12–28). UHC=universal health coverage. DALYs=disability-adjusted life-years. MNCH=maternal, neonatal, and child health. DTP3=diphtheria-tetanus-pertussis vaccine, 3 doses. MCV1=measles-containing-vaccine, 1 dose. LRI=lower respiratory infection. MIR=mortality-to-incidence ratio. NCDs=non-communicable diseases. ART=antiretroviral therapy. MPR=mortality-to-prevalence ratio. IHD=ischaemic heart disease. RSDR=risk-standardised death rate. CKD=chronic kidney disease. COPD=chronic obstructive pulmonary disease. As applied in this analysis, the UHC effective coverage measurement framework involves 30 unique cells from a matrix of five health service types—promotion, prevention, treatment, rehabilitation, and palliation—against five population-age groups (reproductive and newborn, children younger than 5 years, children and adolescents aged 5–19 years, adults aged 20–64 years, and older adults aged ≥65 years). Treatment is subdivided into two separate groups: first, communicable diseases and maternal, newborn, and child health; and second, non-communicable diseases. Effective coverage indicators were then mapped to these cells to represent needed health services across the life course. 23 effective coverage indicators were included in the present analysis (table 1). As recognised in previous studies,19, 20, 21, 22, 23, 24, 25, 26 data for directly measuring effective intervention coverage are rarely available across health services, locations, and over time. Subsequently, we used viable proxy measures and analytical techniques to approximate effective coverage for conditions considered amenable to health care.40, 41, 42, 43 Criteria set forth by the WHO ERG guided selection of effective coverage indicators and preferred measurement approaches (appendix 1 pp 12–28).33 Such criteria stipulated that effective coverage indicators should be currently measurable (ie, data and methods that support indicator measurement today); reflect differences in effective health services and not factors outside the immediate scope of health systems and UHC (eg, tobacco taxation and physical infrastructure such as roads and water systems); and use indicators already encompassed within the SDGs and GPW13, or draw from data systems required for monitoring of SDGs and GPW13. Several other indicator candidates were considered from 2017 to 2019 (appendix 1 pp 12–28), but inadequate data availability, access, or quality, or a combination of these factors, impeded their inclusion in the current analysis. Four effective coverage indicators were measures of intervention coverage and 19 were mortality-based measures to proxy access to quality of care (table 1; appendix 1 pp 30–32). For the mortality-based measures, we primarily used mortality-to-incidence ratios (MIRs) and mortality-to-prevalence ratios (MPRs) for chronic or longer-term conditions (eg, diabetes or asthma). Without better data on effective coverage, such mortality-based measures are viewed as suitable proxies,33, 44, 45, 46 providing good signals on what access to quality care should, at minimum, avert or protect against even if the onset of disease cannot be wholly prevented. The main exception was ischaemic heart disease, for which GBD input data coverage and quality on non-fatal outcomes were less robust than data on causes of death and related risks; subsequently, we used risk-standardised death rates instead of MIRs or MPRs to proxy effective coverage. As a statistical approach used in previous GBD analyses41, 43 and further described in the appendix 1 (pp 31–32), risk standardisation aims to better isolate variations in mortality associated with health-care access and quality from differences in underlying risk exposures mainly related to factors outside the health system. Effective coverage indicators for intervention coverage were kept on their natural scale (0–100%), whereas the 19 other effective coverage indicators were transformed to values on a 0–100 scale (appendix pp 31–33). Across locations and from 1990 to 2019, 0 was set by values at the 97·5th percentile or higher (ie, “worst” levels of MIRs) and 100 by the 2·5th percentile or lower (ie, “best” levels of MIRs). As outlined by previous work,14, 15, 16, 17 population-level measures of effective coverage should represent the fraction of total health gains a health system could potentially provide, given currently available interventions, that a health system actually delivers. This construct is thus grounded in the principle of comparability—all health systems ought to maximise potential health gains for their populations—but also requires accounting for local health needs and epidemiological profiles. For instance, if a country currently experiences a high burden of diabetes and a comparatively lower burden of HIV, at least equal or even higher priority in expanding services for diabetes should occur relative to HIV in order to further support health gains. To construct the UHC effective coverage index, we weighted each effective coverage indicator relative to their health gain weights, a metric approximating the population health gains potentially deliverable by health systems for each location-year. More detail is provided in the appendix 1 (pp 32–35), but in brief, calculations were based on three inputs for each effective coverage indicator and corresponding population-age group: estimates on the 0–100 scale, targeted disease burden, and effectiveness categories of associated interventions or services (table 1). For effectiveness, incremental values were assumed by category (ie, 90% effectiveness for category 1, 70% for category 2, 50% for category 3, and so on), as informed by studies published in the Cochrane Database of Systematic Reviews, the Tufts Cost-Effectiveness Analysis Registry and Global Health Cost-Effectiveness Analysis Registry, and Disease Control Priorities, third edition (DCP3); sensitivity analyses on shifting each effective coverage indicator by one category (ie, moving each category 2 indicator up to category 1 and then down to category 3) showed high correlations with current assignments (appendix 1 p 35). As shown in figure 2, UHC effective coverage index estimates based on health gain weighting and an unweighted average across effective coverage indicators were positively associated (r=0·95); however, effects differed across countries. Comparing the UHC effective coverage index in 2019 with health gains weighting to the unweighted index (unweighted average of effective coverage indicators) in 2019 Locations are colour-coded by GBD super-region, and are abbreviated according to their ISO3 codes. ISO3 codes and corresponding location names are listed in appendix 1 (pp 64–68). UHC=universal health coverage. GBD=Global Burden of Diseases, Injuries, and Risk Factors Study. Since no gold-standard measures of UHC service coverage currently exist, we used three types of validity testing to compare UHC effective coverage index performance to previously published multi-country indices of UHC service coverage: the WHO UHC service coverage index for 2017;21 UHC service coverage index from GBD 2017;26 and service coverage index values from the World Bank.24 Further details of these analyses are provided in the appendix 1 (pp 38–52), with results summarised in table 2. Results for content, known-groups, and construct validity across multi-country health service indices for UHC service coverage measurement Content validity was evaluated on the basis of the percentage of 30 matrix cells of health service types against population-age groups covered by each index. Known-groups validity was evaluated on the basis of the percentage of 16 country pairs correctly ranked based on country A’s UHC or health-system performance being recognised as better than country B’s performance; details are found in appendix 1 (pp 45–47). Convergent validity was evaluated on the basis of how much index performance could explain variation in HALE after controlling for levels of sociodemographic development (as measured by SDI). UHC=Universal health coverage. HALE=healthy life expectancy. SDI=Socio-demographic Index. GBD=Global Burden of Diseases, Injuries, and Risk Factors Study. SDGs=UN Sustainable Development Goals. For content validity, we computed the percentage of 30 cells (ie, combinations of health services and population-age groups) from the UHC effective coverage framework that were represented by indicators for each index. For known-groups validity, we assessed how well each index could discriminate between 16 country-pairs for which previous studies show “country A” as having better performance or progress on UHC service coverage than a similar “country B”.11, 23, 47, 48, 49, 50, 51, 52, 53, 54, 55 These pairs were selected a priori, and for each index we calculated the fraction of pairs correctly ordered on the basis of mean estimates and accounting for uncertainty where available. For convergent validity, we quantified how much variation in healthy life expectancy could be explained by each index after removing the average relationship between each index and overall sociodemographic development (as measured by Socio-demographic Index [SDI]). In general, the UHC effective coverage index based on health gain weights showed stronger performance across these three validity measures than previous UHC service coverage measures and the unweighted UHC effective coverage index (table 2; appendix 1 pp 38–52). To better understand potential drivers of UHC effective coverage, we used stochastic frontier metaregression to quantify UHC effective coverage frontiers—estimated maximum levels of UHC effective coverage index achieved given any amount of health spending per capita—and compared country-level UHC effective coverage performance relative to these frontiers. The magnitude of these gaps between the frontier and UHC effective coverage index values provides insights into potential inefficiencies, as well as measurement error, in translating health spending into improved UHC effective coverage at the population level. Further analytical details are in the appendix 1 (pp 53–59). Since UHC aims to minimise financial hardship associated with receiving essential health services, we focused on assessing the relationship between pooled health spending per capita (ie, government spending, prepaid private health spending, and development assistance for health)56 and UHC effective coverage performance. Alternative analyses, wherein out-of-pocket spending was included (ie, total health expenditure) and then development assistance for health was excluded (ie, pooled domestic health expenditures), were also done but are not reported here (appendix 2 pp 6–7). Spurred by the GPW13 UHC billion target,6 which calls for 1 billion more people benefiting from UHC by 2023, various approaches have been considered for translating performance metrics into the number of people covered by health services.20, 21, 57, 58 For this analysis, we used a similar approach currently recommended by WHO:58 we applied index estimates as fractional metrics and multiplied these values by populations to approximate population equivalents with UHC effective coverage. To assess UHC effective coverage trajectories and their contributions towards meeting the UHC 1 billion target, we first projected country-level UHC effective coverage index estimates through to 2023. These projections were based on stochastic frontier metaregression modelled relationships between UHC effective coverage index and total health spending per capita; a related method has been used previously by GBD26, 59 and is described further in the appendix 1 (pp 60–61). Taking UHC effective coverage index as a fraction, we multiplied these values by country-level GBD-based population forecasts through to 2023.60 Last, we aggregated these estimates globally and by GBD super-region, and calculated additional population equivalents with UHC effective coverage from 2018 (the GPW13 baseline) to 2023. GBD aims to propagate sources of uncertainty through its estimation process,34, 35, 36 resulting in 1000 draws from the posterior distribution for each measure by location, age, sex, and year. We incorporated uncertainty quantified for each effective coverage indicator and associated disease burden based on GBD 2019 estimates, and did scaling, index construction, and UHC effective coverage index projections at the draw-level to reflect uncertainty. We report 95% uncertainty intervals (95% UIs) based on the ordinal 25th and 975th draws for each measure. The funder of the study had no role in study design, data collection, data analysis, data interpretation, or writing of the report. The corresponding author had full access to all the data in the study and had final responsibility for the decision to submit for publication.
