Survival analysis of time-to-death for under-five children In Ethiopia: using parametric shared frailty models

Abstract

under-five child mortality indicates the probability of dying between birth and exactly five years of age, expressed per 1,000 live births. Under-five child mortality is considered to be one of the key health indicators in an economy. This study aimed to investigate the potential risk factors affecting time-to-death of under-five children in Ethiopia using parametric shared frailty models where region were used as a clustering effect in the model. Methods: Parametric shared frailty models have been used with three baseline hazard function (Weibull, Log-logistic, Log-normal) and two frailty distributions (Gamma, Inverse-Gaussian). From 2011 Ethiopian Demographic Health Survey (EDHS) 9433 under-five children were included from nine regional states and two city administrations. Data were analyzed using statistical software such as: R version 3.2.5 and STATA version 12.0. Results: The median death time of under-five children in Ethiopia was 12 months. The clustering effect was significant and Log-normal-Inverse Gaussian shared frailty model was preferred over Weibull and Log-logistic Gamma shared frailty models based on Akaike Information Criterion (AIC) and graphical evidence. The result showed women’s educational level, wealth index, type of births, total children ever born, preceding birth intervals and place of delivery were significant, where as sex of household head and religion were not significant covariates for under-five child mortality in Ethiopia. Conclusion: The result suggested that the timing of death of under-five children from different region had different pattern, since there was a frailty (clustering) effect on the time-to-death of under-five children among regions of Ethiopia.