Effectiveness of a large-scale, sustained and comprehensive community health worker program in improving population health: the experience of an urban health district in South Africa

Introduction: South Africa is an upper middle-income country with wide wealth inequality. It faces a quadruple burden of disease and poor health outcomes, with access to appropriate and adequate health care a challenge for millions of South Africans. The introduction of large-scale, comprehensive community health worker (CHW) programs in the country, within the context of implementing universal health coverage, was anticipated to improve population health outcomes. However, there is inadequate local (or global) evidence on whether such programs are effective, especially in urban settings. Methods: This study is part of a multi-method, quasi-experimental intervention study measuring effectiveness of a large-scale CHW program in a health district in an urban province of South Africa, where CHWs now support approximately one million people in 280,000 households. Using interviewer administered questionnaires, a 2019 cross-sectional survey of 417 vulnerable households with long-term CHW support (intervention households) are compared to 417 households with no CHW support (control households). Households were selected from similar vulnerable areas from all sub-levels of the Ekurhuleni health district. Results: The 417 intervention and control households each had good health knowledge. Compared to controls, intervention households with long-term comprehensive CHW support were more likely to access early care, get diagnosed for a chronic condition, be put on treatment and be well controlled on chronic treatment. They were also more likely to receive a social grant, and have a birth certificate or identity document. The differences were statistically significant for social support, health seeking behavior, and health outcomes for maternal, child health and chronic care. Conclusion: A large-scale and sustained comprehensive CHW program in an urban setting improved access to social support, chronic and minor acute health services at household and population level through better health-seeking behavior and adherence to treatment. Direct evidence from households illustrated that such community health worker programs are therefore effective and should be part of health systems in low- and middle-income countries. This study determined if CHWs were effective in improving household (HH) health. We explored measures of CHW effectiveness by looking at social and health outputs and outcomes. The study is based on an analysis comparing 417 intervention households exposed to CHWs with a control group of 417 households not exposed to CHWs, in Ekurhuleni, South Africa. Household interviews were conducted between April and June 2019. Health and social support output measures were explored through reported perceptions of access to social services, early antenatal and child health services and chronic disease care (HIV, TB, hypertension and diabetes), improvements in care and knowledge of common health conditions and on healthy behavior (diet, exercise, hand washing). Self-reported health outcome measures for maternal, child health issues and chronic disease were also explored. For the purposes of this study, the social support activities are defined as support in getting a birth certificate, an identity document for adults, social grants and food parcels. The district is divided into East, North and South sub-districts and vulnerable households were selected from all three. In Ekurhuleni, properties valued at less than USD 9415 were considered indigent/vulnerable. The six district intervention sites in Ekurhuleni that had more than 60% CHW team coverage, two from each sub-district, with approximately 56 000 households in total were identified. Intervention households were proportionately selected from these areas. Assuming a 95% confidence level, 5% margin of error and 50% response distribution, using Raosoft sample size calculator,1 a sample size of 381 households was representative of this population. A 5% buffer was added to cover for replacement households or incomplete surveys, so the sample size was approximated to 400 intervention households/HH. A map divided into small areas was used and each day, a random spot on the map was selected as a starting point. The field workers went to every fifth household supported by CHWs until the required number of households were reached. 400 control households in Ekurhuleni were chosen randomly in the same way; situated in similar areas, but where CHW teams were not working. Eleven retired nurses proficient in the locally used language were trained as fieldworkers to undertake the household interviews. The questionnaires were piloted in an area in the same district in a site not included in the study and fieldworkers were asked to provide feedback on their experiences in the field. The questions in the tools were easily translated and understood, so no major changes had to be made. Inclusion criteria included that the household member interviewed had resided in the area for 18 months or longer and that each household had at least one vulnerable member (a pregnant woman, child under five, an elderly person, or a household member with a chronic disease). Intervention households had to have an allocated CHW, and controls not. The household head or the person who knew about other members was interviewed. 417 intervention and control households each were interviewed, but due to some missing data, approximately 400 households in each group were available for analysis. Frequencies of the socio-economic and demographic variables were conducted. For the categorical data, we mitigated for possible confounders by correcting the data for socio-economic and demographic differences between the two groups (Tables ​(Tables11 and ​and2).2). In order to compare equitable access to social services, we excluded non-South African and non-indigent households from the analyses; we also corrected where relevant for gender and age in HIV status, family planning, immunization and chronic diseases (Table ​(Table3).3). We conducted bivariate analysis using Pearson Chi-squared tests. 2 × 2 tables were also used to compare the effect of the exposure to CHW teams in intervention households (Table ​(Table4).4). In cases where the Pearson’s Chi-square test was significant, the cell standardized residuals, expected and observed values were investigated to establish which cells in the cross-tables contributed most to the significant associations between two variables. Socio-economic and demographic measures of ALL study households Estimated value of property (indigent households =  < USD 9715) USD: United States dollar Statistically significant p values (< 0.05) were also in bold Corrected socio-economic and age variables in South African indigent households p < 0.001 Significantly more SA citizen households in intervention group p < 0.0001 Significantly fewer indigent households in intervention group Statistically significant p values (< 0.05) were also in bold Health and social outputs and outcomes in South African indigent households* *Shaded areas represent more than 5% differences between the two groups of households ** Respondents were asked a specific question Comparison of health and social outputs and outcomes in South African indigent households Statistically significant p values (< 0.05) were also in bold *Where odds ratios > 1 (in bold), it is in favour of intervention households; where odds ratios < 1, control households did better