SURVIVAL ANALYSIS AND FRAILTY MODELLING OF TIME-TO-EVENT DATA: AN APPLICATION TO INFANT MORTALITY IN MALAWI

Abstract

Infant mortality rate is one of the important health and development indicators in a country or community and that is why reduction of infant mortality has been the main target of public health policies for the past decades. Malawi, like many countries in the sub Saharan Africa is a country that suffers from the highest rates of infant mortality across the globe. Studies have been conducted to identify factors associated with infant mortality in Malawi but none of these studies used recent data. This study used the most recent survey data to identify the factors associated with infant mortality in Malawi by using survival analysis techniques and frailty modelling to control for unobserved heterogeneity using. The data used for this study was from the 2015-16 Malawi Demographic Healthy Survey (2015-16 MDHS) and was obtained from DHS program website: https://www.dhsprogram.com. Bivariate analysis was conducted to identify variables that had significant association with infant mortality in Malawi using both Kaplan-Meier and log rank test and were subsequently considered into the cox proportional hazard model analysis to estimate their strength of effect on infant mortality in Malawi. The variables were also modelled using both the semi parametric cox frailty model and parametric frailty models to find the best fit model using the maximum likelihood estimation. The results showed that sex of household head, mothers’ age group, source of drinking water, religion, type of birth and place of delivery were significantly associated with infant mortality and that there are unobservable family effects which make infant deaths to cluster in some households/families.