Thumbnail Image

Subnational Analysis of Birth Weight in Ghana using Bayesian Spatial Regression Models

Child mortality in sub-Saharan Africa is reducing but the levels remain high with subnational within-country variations. Birth weight is a key predictor of child survival and monitoring birth weight outcomes, in particular, prevalence of low birth weights, is important for resource allocation to improve child survival outcomes. Past research in sub-Saharan Africa has found that different individual-level factors are associated with birth weight including BMI of mother, sex of baby, educational level of mother, and wealth index of household. Some environmental factors are found to be associated with birth outcomes. However, past findings regarding the association of birth weight with household air pollution (HAP) resulting from cooking fuels are non-conclusive. In this study, we analyze variability in birth weights subnationally for Ghana and assess its association with household air pollution resulting from cooking fuels, accounting for variation due to other factors including maternal and household predictors, as well as geographical location. The analysis was based on birth weights for 1310 births, obtained from data collected in 2014 in the Demographic and Health Survey (DHS). We use Bayesian spatial regression models to estimate associations and capture spatial variation. Spatial variation was captured with a conditional autoregressive (CAR) model. Based on various models, we do not find evidence to suggest that cooking fuel is associated with birth weight. After accounting for covariates, the average birth weights per district ranged from 2823g (95% CI: 2613g, 3171g) in Ketu district to 3243g (95% CI: 3083g, 3358g) in Ashanti Akim North district. Across Ghana, difference in birth weight attributable to district spatial effects range from -33g in Lawra district in Upper West region of Ghana to 11g in Ho in the Volta region.
Research Projects
Organizational Units
Journal Issue
Publisher Version
Embedded videos