Assessment of quality of primary care with facility surveys: a descriptive analysis in ten low-income and middle-income countries

Background: Primary care has the potential to address a large proportion of people’s health needs, promote equity, and contain costs, but only if it provides high-quality health services that people want to use. 40 years after the Declaration of Alma-Ata, little is known about the quality of primary care in low-income and middle-income countries. We assessed whether existing facility surveys capture relevant aspects of primary care performance and summarised the quality of primary care in ten low-income and middle-income countries. Methods: We used Service Provision Assessment surveys, the most comprehensive nationally representative surveys of health systems, to select indicators corresponding to three of the process quality domains (competent systems, evidence-based care, and user experience) identified by the Lancet Global Health Commission on High Quality Health Systems in the Sustainable Development Goals Era. We calculated composite and domain quality scores for first-level primary care facilities across and within ten countries with available facility assessment data (Ethiopia, Haiti, Kenya, Malawi, Namibia, Nepal, Rwanda, Senegal, Tanzania, and Uganda). Findings: Data were available for 7049 facilities and 63 869 care visits. There were gaps in measurement of important outcomes such as user experience, health outcomes, and confidence, and processes such as timely action, choice of provider, affordability, ease of use, dignity, privacy, non-discrimination, autonomy, and confidentiality. No information about care competence was available outside maternal and child health. Overall, scores for primary care quality were low (mean 0·41 on a scale of 0 to 1). At a domain level, scores were lowest for user experience, followed by evidence-based care, and then competent systems. At the subdomain level, scores for patient focus, prevention and detection, technical quality of sick-child care, and population-health management were lower than those for other subdomains. Interpretation: Facility surveys do not capture key elements of primary care quality. The available measures suggest major gaps in primary care quality. If not addressed, these gaps will limit the contribution of primary care to reaching the ambitious Sustainable Development Goals. Funding: Bill & Melinda Gates Foundation. We used the most recent Service Provision Assessment (SPA)18 facility survey data from 2007 to 2016 in our analyses. The SPA is a nationally representative health-facility assessment that includes a facility assessment, a questionnaire for health-care providers, observations of visits, and exit interviews with observed patients.18 We selected SPA surveys for this analysis because they are the most comprehensive, standardised, cross-nationally available datasets of health-system measurements.18 Other global facility surveys include WHO’s Service Availability and Readiness Assessment surveys,19 which focus mainly on infrastructure and equipment, and the World Bank’s Service Delivery Indicators surveys,20 which measure the knowledge of health providers and health facility resources. However, neither of these surveys captures the process of care that people receive. We included only the ten countries with available data: Ethiopia, Uganda, Senegal, Nepal, Kenya, Tanzania, Rwanda, Malawi, Haiti, and Namibia. SPAs, are done differently across countries. Ethiopia, Uganda, Senegal, Nepal, Kenya, and Tanzania had nationally representative survey data, whereas Rwanda, Malawi, Haiti, and Namibia had national censuses.18 Although a SPA survey was done in Bangladesh in 2014, we did not include it because it did not contain any observations of visits, which were required for many of the measures we defined. We wanted to use survey data from 2007 to 2016 only to approximate the contemporary situation, although we recognise that many changes might have occurred since some of the older surveys. In the survey countries, facility sampling weights were used to correct for oversampling of hospitals and to create health-system representative estimates, and providers and clients were randomly sampled within facilities on the day of the survey.18 Typically, SPA surveys collect data from 400–700 facilities selected at random from a comprehensive list of health facilities in a country.18 Hospitals can be oversampled because there tend to be only small numbers of hospitals in a country.18 Subsequently, the data were weighted during analysis to ensure that data were proportionally representative when presented (appendix).18 We limited our analysis to primary care facilities, which include the first level of care from health centres, clinics and polyclinics, health posts, dispensaries, and other low-level facilities. Although primary care services can be provided at hospitals, this assessment was designed to provide a cross-nationally comparable view of care quality at the first level of care. Thus, we removed hospital primary care from the analysis, but still weighted the facilities to ensure that the analyses were nationally representative of primary care facilities in the study countries. To provide country context corresponding to the SPA survey year, we obtained data for gross domestic product per person, Gini index, health expenditure per person, and number of health workers (community health workers, physicians, and nurses) per 100 000 people, land area, and proportion of urban areas for each country from the World Development Indicators.21 We adapted the Commission framework to focus on primary-health-care systems. The four Cs of primary care—continuous, coordinated, first contact, and comprehensive care—were mapped to the three main domains of the processes of care: competent systems, evidence-based care, and positive user experience (table 1).13 Competent systems were composed of the subdomains safety, prevention and detection, continuity and integration, population-health management, and timely action. Evidence-based care included systematic assessment, correct diagnosis, appropriate treatment, and counselling, and were assessed for key primary care services (antenatal care, family planning, sick-child care, non-communicable diseases, mental health, HIV, tuberculosis, and other primary-care-sensitive conditions [ie, conditions for which good primary care could prevent the need for hospital admission, or for which early intervention could prevent complications or more severe disease—eg, angina, asthma, chronic obstructive pulmonary disease, congestive heart failure, diabetes mellitus, hypertension10]). Positive user experience was composed of patient focus—which included short wait times and patient voice and values—and clear communication. Mapping of primary care indicators to Commission framework To identify the indicators, two authors (EKM and ADG) individually assessed the list of indicators from the SPA datasets and, on the basis of the HQSS framework, identified and classified indicators relevant to measurement of quality of care. Individual assessments were cross-checked through group discussion to ensure consistency of classification. In the cases of discrepancy, a third researcher (MEK) participated to corroborate the domain and subdomain of each indicator. Three types of score were calculated for each facility: subdomain scores (mean of component indicators relevant for each subdomain), domain scores (mean of the nine subdomain scores), and overall quality of primary care scores (mean of the three domain scores). All index component indicators were either binary (ie, 0 or 1) or indexes ranging from 0 (lowest) to 1 (highest). If the indicators were at the patient level, such as for evidence-based care and user experience indicators, the average was calculated to get the mean scores for each facility. For evidence-based care, the average technical quality indices were calculated by averaging the client-level scores for systematic assessment, correct diagnosis, appropriate treatment, and counselling for every visit in the primary care facility. To calculate for the technical quality indices, process indicators specific to each type of primary care service were selected from SPA.22 These binary indicators, ranging from 0 to 1, were then averaged to create a score for technical quality for each service. These indices 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.22 These domains included history, examination, and counselling. Antenatal care and sick-child care included items on testing and management.22 The list of indicators for each technical quality index is in the appendix. All patient-level analysis included SPA client survey weights.18 The overall quality scores, and scores for the domain and subdomains were calculated on the basis of facility survey weights.18 We weighted each country equally when averaging scores across countries because our goal was to generalise across countries instead of across populations. We then compared scores at national and subnational levels. Scores were mapped with the Database of Global Administrative Areas and QGIS (version 2.18). Correlations between quality scores and several national-level predictors were calculated. All analyses were done in Stata (version 15.0), which was also used to plot the figures (except for the coxcombs, which were made in Vizzlo). The study funders had no role in study design; data collection, analysis, or 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.