SURVIVAL MODELLING OF UNDER-FIVE HIV INFECTED CHILDREN TAKING ANTIRETROVIRAL THERAPY AT CHIRADZULU DISTRICT HOSPITAL, MALAWI: AN APPLICATION OF COX AND PARAMETRIC REGRESSION MODELS

Abstract

The hazard of a child dying before reaching 5 years is highest in sub-Saharan African countries including Malawi, with elevated risk in rural areas. Chiradzulu district has a higher under-five mortality rate as compared to a national average mortality rate for Malawi. The main aim of the study was to model and determine survival patterns of under-five children living with HIV who were on antiretroviral therapy. Using retrospective secondary data for 186 cases of under five children on anti-retroviral therapy, collected from July 2011 through July 2016, both Cox and parametric models were analyzed. The effect of the following covariates were investigated: Weight, height, sex, residence, mother’s occupation, mother’s education level and mother’s marital status. Weibull, exponential, Gompertz, loglogistic and lognormal regression were performed as parametric models and Cox as a semi-parametric model. Akaike Information Criterion (AIC) was used to compare the efficiency of fitted models. Residence, mother’s education level, mothers’ occupation and weight factors influenced the survival of under-five children living with HIV in Chiradzulu District.During the study, 56 (30.0%) of the participants died. Gompertz model was found to be the best fit model for predicting survival of under-five children living with HIV on antiretroviral therapy in Chiradzulu district. It is recommended that the results should be substantiated by similar survival studies from other parts of the district to generalize the results to other individuals in the country. Researchers should check the underlying assumptions of Cox model before using it in order to use a proper model during analysis.