Menu
Menu
Menu
Menu
Introduction Measurement of effective coverage (quality-corrected coverage) of essential health services is critical to monitoring progress towards the Sustainable Development Goal for health. We combine facility and household surveys from eight low-income and middleincome countries to examine effective coverage of maternal and child health services. Methods We developed indices of essential clinical actions for antenatal care, family planning and care for sick children from existing guidelines and used data from direct observations of clinical visits conducted in Haiti, Kenya, Malawi, Namibia, Rwanda, Senegal, Tanzania and Uganda between 2007 and 2015 to measure quality of care delivered. We calculated healthcare coverage for each service from nationally representative household surveys and combined quality with utilisation estimates at the subnational level to quantify effective coverage. Results Health facility and household surveys yielded over 40 000 direct clinical observations and over 100 000 individual reports of healthcare utilisation. Coverage varied between services, with much greater use of any antenatal care than family planning or sick-child care, as well as within countries. Quality of care was poor, with few regions demonstrating more than 60% average performance of basic clinical practices in any service. Effective coverage across all eight countries averaged 28% for antenatal care, 26% for family planning and 21% for sick-child care. Coverage and quality were not strongly correlated at the subnational level; effective coverage varied by as much as 20% between regions within a country. Conclusion Effective coverage of three primary care services for women and children in eight countries was substantially lower than crude service coverage due to major deficiencies in care quality. Better performing regions can serve as examples for improvement. Systematic increases in the quality of care delivered- not just utilisation gains-will be necessary to progress towards truly beneficial universal health coverage.
To calculate the effective coverage of three primary care services across multiple countries, we identified countries with standardised information on population utilisation of care and primary care performance. Countries were eligible for inclusion if a Service Provision Assessment (SPA) survey that included direct observation of primary care services had taken place in the past decade, providing a standardised assessment of the content of care throughout the country. Eight countries met this criterion: Haiti, Kenya, Malawi, Namibia, Senegal, Rwanda, Tanzania and Uganda. To define population in need and utilisation of care, we identified the population-representative survey conducted closest in time to the health facility survey for each country, using Demographic and Health Survey (DHS) household surveys for all countries except Malawi, where we used the Multiple Indicator Cluster Survey (MICS). The Malawi MICS was completed in the same year as the SPA; estimates of utilisation are quite comparable between the 2014 MICS and the 2015–2016 DHS survey. We estimated total population size using the census or census-based projections of total population cited in each DHS or MICS report and applied to these the proportion of the population in the relevant age groups (women 15–49, children under 5) based on the population survey results. Reports for two countries—Malawi and Uganda—did not include total population size; we used the World Development Indicators population estimate for the year of the population survey in these countries. SPA surveys of the health system are conducted by the DHS Program in collaboration with a national statistics office. Three of the eight countries included in this study elected to conduct a complete census of health facilities: Haiti in 2013, Malawi in 2013 and Namibia in 2009. The SPA in Rwanda in 2007 was a census of public health facilities and large private facilities plus a representative sample of small private facilities. The other four countries (Kenya in 2010, Senegal in 2013–2014, Tanzania in 2015, Uganda in 2007) sampled health facilities from a a master facility list stratified by facility type and subnational region, with deliberate oversampling of hospitals. All SPA surveys included both public and private facilities. The survey includes a facility audit and direct observation of ANC, family planning and curative care for children under 5. Within each sampled health facility, patients presenting for these services were sampled using systematic random sampling; trained observers assessed the care provided according to a checklist of possible provider actions. The data include sampling weights for client visit calculated to account for probability of sampling the client and the facility. All population and health facility surveys were designed to be representative at a subnational level; we extracted the regional boundaries used in each survey from DHS or the GADM database of global administrative areas for surveys where subnational regions matched national administrative units. Boundaries were not always consistent between the health system and population surveys due to differences in survey design or changes in administrative boundaries over time. We used QGIS mapping software to identify the smallest possible identical spatial areas to which findings could be generalised from both surveys as our units of analyses; we calculated the area of each region and estimated regional population using LandScan (2010) High-Resolution Global Population Data Set.25 For each primary care service with available data—ANC, family planning and curative care for children—we used global standards to define the population in need as well as metrics of population coverage of care and technical quality of care delivered. The population in need of ANC was defined as women aged 15–49 with a live birth in the past 2 years; coverage was defined in two ways: attending at least one and the recommended minimum four ANC visits during the most recent pregnancy. Women with contraceptive need were those 15–49 who are married or in a union and wish to space or limit childbearing; women using a modern contraceptive method are considered covered. Children under 5 who had experienced diarrhoea, fever or acute respiratory illness in the prior 2 weeks had a need for curative health services; coverage was calculated as an interaction with a health facility or formal provider. To quantify the number of people in need, we multiplied the relevant population total (eg, women 15–49) by the proportion in need from the household survey (eg, women reporting live birth in the past 2 years). We defined technical quality of care in each service by identifying key domains of care and the essential clinical actions within each domain from international guidelines.26–29 These domains include history, exam and counselling; ANC and sick-child care also include items on testing and management, respectively (full lists for all services are in online supplementary table 1). Example items include the provider asking expectant mothers if they experienced danger signs, providing counselling on the family planning method prescribed and taking a child’s temperature. For each directly observed clinical visit, we calculated the quality score as the per cent of actions completed out of items assessed per country. Actions in follow-up ANC visits were weighted to reflect the number of times they should be performed. For example, provision of tetanus toxoid vaccination contributed one-third of one action since this service should be provided in one of the three follow-up visits. bmjgh-2017-000424supp001.pdf We summarised population in need, proportion seeking care and average quality of care at the subnational and national levels, weighting individual observations in all population and facility surveys with the appropriate sampling weight. For each subnational region and each country, we multiplied use of healthcare by average quality to yield effective coverage; in the case of ANC we used four visits as the coverage indicator to capture full utilisation. We quantified aggregate primary care coverage, quality and effective coverage measures by averaging across the three services. In this case we used any ANC visit as the coverage indicator for ANC to capture any access to care, matching the other two services. We report descriptive statistics of the population and health system surveys as well as national summaries for need, coverage, quality and effective coverage. To quantify uncertainty around each estimate, we calculated the standard error (SE) of the mean for proportion in need, proportion seeking care and average quality by country, accounting for repeated sampling by cluster and health facility in population and facility surveys, respectively. Uncertainty estimates were not available for total population sizes; we multiplied the SE of the estimated proportion of the population in need by total population to quantify uncertainty in units of thousands of people. To calculate uncertainty in effective coverage, we used the formula for variance of a product of independent variables,30 treating utilisation and quality at the country level as independent due to the separate sampling sources. We present national summaries of coverage and quality to identify priority deficits in primary care services in study countries. We quantified variation in quality of care by estimating the difference between best and worst effective coverage per country, as well as calculating the intraclass correlation (ICC) by country for each quality metric; we repeated this test on the 5th–95th percentile of the sample by population size to exclude outlying observations. We mapped primary care coverage and primary care quality by subnational region, assessed the correlation between coverage and quality at the subnational level, and plotted coverage versus effective coverage by service. The original survey implementers obtained ethical approvals for data collection; the Harvard University Human Research Protection Program approved this secondary analysis as exempt from human subjects review.
N/A
