STATISTICAL MODELLING FOR THE SURVIVAL OF HIV/AIDS INFECTED PATIENTS TAKING ANTIRETROVIRAL TREATMENT IN NTCHEU: A RETROSPECTIVE COHORT STUDY

Abstract

The Cox proportional Hazard model has been broadly used and most popular technique in survival analysis. Under certain situations parametric models may offer advantages over Cox proportional hazard model. In this study five parametric models and the Cox proportional hazard model were fitted. The main objective in this study was to compare the performance of five models: exponential, loglogistic, Weibull, lognormal and Gompertz as well as the semi-parametric Cox model on the survival among HIV/AIDS infected patients taking ART in Ntcheu district. Five parametric models as well as Cox proportional model were fitted to 6670 HIV/AIDS patients who registered for ART from 2007-2012.The results of the analysis using Cox proportional hazard model and the Gompertz model were comparable. In both models, WHO clinical stage 4, body mass index and being male were found to be significantly associated with the survival of HIV/AIDS patients taking ART in Ntcheu district. In the multivariable analysis all the parametric models fit better than Cox model with respect to AIC and the Gompertz model was found to be the best model for modelling the survival among the HIV/AIDS infected patients taking ART in Ntcheu district.