Spatial Modeling of Risk Factors of Maternal Mortality in Kenya

Statistics and Actuarial Science

Spatial modeling is important for conducting geospatial analysis to understand the world and guide decision-making. This study conducts spatial modeling of risk factors of maternal mortality. Spatial means that each data item has a geographical reference. Perinatal and maternal mortality are significant health concern among nations. The specific objectives are to estimate the national and county prevalence of maternal mortality in Kenya, to identify the most influential risk factors associated with maternal mortality in Kenya, and to model spatial variations of maternal mortality and produce a Kenyan atlas of maternal mortality by county. Secondary data was sourced from KNBS and Africa open Data for the 2019 KPHC and Kenyan Counties shape files, respectively. The study used maternal death as the dependent variable. The study considered four models: the logistic regression, the normal unstructured heterogeneity (UH) random effects, ICAR Spatial random effects, and the convolution model. Best subset selection was achieved using the forward stepwise selection method, where the best model was determined using AIC. The study compared the spatial models using the DIC. The study estimated the models using the Bayesian approach. The descriptive results revealed that countrywide, maternal mortality has a prevalence of 10.7%. The comparison results showed that the convolution model performed better than the logistic regression, normal UH random effects, and ICAR spatial random effects models. The study concluded that Wajir, Mandera, Laikipia, Nyandarua, Nyeri, Tharaka Nithi, Elgeyo Marakwet, Siaya, and Migori had the greatest prevalence indices. Isiolo, Embu, and Machakos had the lowest Maternal Mortality prevalence indices. The coun-ties with considerably low risks of maternal mortality included Isiolo, Embu, and Machakos, The counties with moderate risks of maternal mortality included West Pokot, Narok, and Lamu.

This work is licensed under a Creative Commons Attribution-ShareAlike 4.0 International License.