Menu
Menu
Menu
Menu
Background: An effective continuum of maternal care ensures that mothers receive essential health packages from pre-pregnancy to delivery, and postnatally, reducing the risk of maternal death. However, across Africa, coverage of skilled birth attendance is lower than coverage for antenatal care, indicating mothers are not retained in the continuum between antenatal care and delivery. This paper explores predictors of retention of antenatal care clients in skilled birth attendance across Africa, including sociodemographic factors and quality of antenatal care received. Methods: We pooled nationally representative data from Demographic and Health Surveys conducted in 28 African countries between 2006 and 2015. For the 115,374 births in our sample, we estimated logistic multilevel models of retention in skilled birth attendance (SBA) among clients that received skilled antenatal care (ANC). Results: Among ANC clients in the study sample, 66% received SBA. Adjusting for all demographic covariates and country indicators, the odds of retention in SBA were higher among ANC clients that had their blood pressure checked, received information about pregnancy complications, had blood tests conducted, received at least one tetanus injection, and had urine tests conducted. Conclusions: Higher quality of ANC predicts retention in SBA in Africa. Improving quality of skilled care received prenatally may increase client retention during delivery, reducing maternal mortality.
The study sample was drawn from the births recode data files of the latest Standard DHS conducted in each sub-Saharan African country between 2000 and 2016, where the full complement of variables for the study was collected. The DHS samples were based on a stratified two-stage cluster design. In the first stage, clusters are drawn from census files. In the second stage, a sample of households is drawn from each selected cluster. The birth recode data files of the nationally representative Demographic and Health Surveys include the full birth histories over the 3–5 preceding years of women in these households including information on pregnancy, postnatal care, immunization, and child health. The final sample covers surveys from 28 countries with unrestricted data access and that include the full complement of variables explored in the study. This sample represents a population of 740 million or 70% of the total population in sub-Saharan Africa in 2015. The following surveys were included: Benin, 2011–2012; Burkina Faso, 2010; Burundi, 2010; Cameroon, 2011; Chad, 2014–2015; Comoros, 2012; Congo, 2011–2012; Democratic Republic of Congo/DRC, 2013–2014; Ethiopia, 2011; Gabon, 2012; Gambia, 2013; Ghana, 2014; Ivory Coast, 2011–2012; Kenya, 2014; Lesotho, 2014; Liberia, 2013; Madagascar, 2008–2009; Malawi, 2010; Mali, 2012–2013; Mozambique, 2011; Namibia, 2013; Niger, 2012; Nigeria, 2013; Sierra Leone, 2013; Swaziland, 2006–2007; Tanzania, 2010; Togo, 2013–2014; Zambia, 2013–2014; and Zimbabwe, 2010–2011. The dependent variable in this study is retention in skilled birth attendance (SBA) among skilled antenatal care (ANC) clients. This variable is coded as ‘1’ if the respondent received any ANC (that is attended ANC at least once) and SBA in the index pregnancy, and ‘0’ if the respondent did not receive SBA, but had received any ANC in the index pregnancy. We defined skilled care as care provided by a doctor, nurse, or midwife, in line with the World Health Organization policy guidelines, as several countries did not have standardized definitions for skilled maternal care providers [6]. To fit a model of retention in SBA for ANC clients, we drew on the framework for health care access by Penchansky and Thomas [7]. The framework captures demand and supply-side determinants of care access along five dimensions (availability, accessibility, accommodation, affordability, and acceptability). We conducted a review of the literature on factors demonstrated to be associated with the use of maternal health care [8], [9]. We then included covariates, collected consistently across the 28 countries that represented at least one dimension of access within the framework. The availability dimension refers to the adequacy of the supply of skilled health workers, facilities, and services, and provides information on the quality of care received during ANC, where good quality of care corresponds to the recommended model by the World Health Organization of focused ANC based on at least four goal-oriented-visits [2]. We included indicators for the following variables: location of care in the facility, the conduct of any urine test, the conduct of any blood test, having had a blood pressure check, receiving at least one tetanus injection, attending up to 4 visits, and receiving any information on potential pregnancy complications. The accessibility dimension accounts for client transportation resources, distance and travel time to care. We thus included an indicator for living in an urban area, as poor physical access to social services correlates with rural dwelling across Africa [10]. Under the affordability dimension, that is the ability to pay and financial protection during care-seeking, we included indicators for having health insurance, possessing any primary education or higher, having a partner who has any primary education or higher and belonging to the richest two wealth quintiles. The acceptability dimension refers to the influences of personal characteristics of the provider and client on care-seeking. We thus included indicators for parity (primiparous for the first birth and grand multiparous for more than five previous births, so that women with 1 to 4 previous births were considered the reference category). We also included indicators for women’s age. Women below 18 years and those above 35 years were collapsed into one category and considered as the reference category (compared with women between 18 and 35 years old), as young and older maternal age has been shown to influence both maternal decisions to initiate care-seeking and the interaction with health care providers during pregnancy [11]. We also included an indicator variable for each country included in the study as a proxy for the national context. For each included country, we calculated the mean levels of ANC, SBA, and the gap in coverage between ANC and SBA (calculated as the difference between mean ANC and mean SBA levels). For the observations with the complete set of covariates (the analytic sample), we estimated the means and standard errors for the study dependent and independent variables, weighted based on client sampling weights. On the analytic sample, we then estimated a two-level logistic regression model of SBA retention, nesting each birth (individual-level) within a cluster. As several mothers reported only one birth over the survey period, we did not construct a three-level model that included random effects at the maternal level. The empirical model included random intercepts for the cluster, fixed effects for each country, and was weighted using respondent sample weights to ensure representativeness at the national level. We categorized the covariates into three blocks: country indicators (binary variables indicating the country in which the survey was conducted), ANC characteristics (corresponding to the availability dimension of the access to care framework) and demographic characteristics. We progressively added these blocks of covariates into the empirical model and computed the intraclass correlation (ICC), that is the DHS cluster-level correlation, to estimate the extent to which the individual probability of retention in SBA for ANC clients in the same DHS cluster was similar compared to individuals from other DHS clusters. The ICC expresses the proportion of the total variance that is at the DHS cluster level. We estimated the ICC using the latent variable method [12] as follows: Where Var DHS Cluster is the variance between DHS clusters and π2/3 is the variance between individuals. We then estimated the proportion of the cluster-level variance that is explained by different blocks of covariates as follows: Where Var 0 is the variance in the initial or empty model, and Var 1 is the second-level variance in the models with various blocks of covariates. For each covariate, we reported the odds ratio (OR) and 95% confidence interval (CI). As Benin had the highest percentage of ANC clients retained in SBA in the fully-adjusted models, we considered this the reference category in our multilevel models. All analyses were conducted using STATA 14.2.
N/A
