Menu
Menu
Menu
Menu
N/A
Introduction: In Ethiopia, more than half of newborn babies do not have access to Emergency Obstetric and Neonatal Care (EmONC) services. Understanding the effect of distance to health facilities on service use and neonatal survival is crucial to recommend policymakers and improving resource distribution. We aimed to investigate the effect of distance to health services on maternal service use and neonatal mortality. Methods: We implemented a data integration method based on geographic coordinates. We calculated straight-line (Euclidean) distances from the Ethiopian 2016 demographic and health survey (EDHS) clusters to the closest health facility. We computed the distance in ESRI ArcGIS Version 10.3 using the geographic coordinates of DHS clusters and health facilities. Generalised Structural Equation Modelling (GSEM) was used to estimate the effect of distance on neonatal mortality. Results: Poor geographic accessibility to health facilities affects maternal service usage and increases the risk of newborn mortality. For every ten kilometres (km) increase in distance to a health facility, the odds of neonatal mortality increased by 1.33% (95% CI: 1.06% to 1.67%). Distance also negatively affected antenatal care, facility delivery and postnatal counselling service use. Conclusions: A lack of geographical access to health facilities decreases the likelihood of newborns surviving their first month of life and affects health services use during pregnancy and immediately after birth. The study also showed that antenatal care use was positively associated with facility delivery service use and that both positively influenced postnatal care use, demonstrating the interconnectedness of the components of continuum of care for maternal and neonatal care services. Policymakers can leverage the findings from this study to improve accessibility barriers to health services.
We included neonates aged ≤ 28 days from women’s most recent deliveries in the prior five years from the 2016 Ethiopian EDHS. We included the most recent births since some data on maternal health services use, such as antenatal and postnatal care, were only available for this population. The primary outcome variable was neonatal mortality, and the primary exposure variable was distance to health facilities that provide maternal and newborn health services. We considered antenatal care, facility delivery, postnatal care and postnatal counselling services as mediator variables. Term of pregnancy, application of substances on the umbilical cord, place of residence, gender of the neonate and the type of pregnancy (twin/singleton) were considered as covariates. The data for this analysis were obtained from two health-related surveys: the Ethiopian 2016 Emergency Obstetric and Neonatal Care (EmONC) survey [23] and the Ethiopian 2016 EDHS [14]. The outcome variable, mediating variables, socio-economic covariates and the corresponding Geographic Positioning System (GPS) coordinates were extracted from the DHS data, while the distance variable and facility location GPS coordinates were extracted from EmONC dataset. The EmONC assessment survey was a national cross-sectional census of health facilities providing maternal and newborn health services. A total of 3,804 geo-referenced health facilities were included in the survey. A detailed description of the EmONC survey procedure is available in the main report [23]. The EDHS is a cross-sectional survey of nationally representative samples. In the DHS surveys, samples were selected using a stratified, two-stage cluster design, using enumeration areas (clusters) as primary sampling unit and households as secondary sampling unit. The detailed methodology is found in the final EDHS 2016 report [14]. Based on the DHS recommendations [24], sample weighting was applied to compute frequencies and percentages of neonatal mortality and maternal health service use variables. The analysis was carried out in four steps. where, (x1,y1) are the coordinates of one point (e.g., the centre of the cluster) (x2,y2) are the coordinates of the other point (the location of the health facility) and d is the distance between (x1,y1) and (x2,y2) In order to check clustering effects, we fit and compared multiple models: 1) a model without clustering effect, 2) considering administrative region as a clustering effect, 3) considering enumeration area as a clustering effect, 4) considering nesting of clusters in regions as a clustering effect and we used the fourth step as a final model. We have checked for collinearity between variables using variance inflation factor, and no collinearity was detected.
