Abstract:
The survival time of HIV positive patients depends not only on the quality of care and highly active anti-retroviral therapy, but also on a host of other baseline demographic factors and clinical characteristics of the patients. The use of the appropriate model will certainly lead to identify more reliable and precise predictors of patients' survival time after initiation of ART and thereby help to have a more effective treatment program. Objective: To model the time-to-death of HIV positive patients under highly active antiretroviral therapy using Cox PHs and AFT models. Methods: The study consists of 440 HIV positive patients who were aged 18 years or above and who were placed under HAART any time in between 1st January 2010 to 30th June 2015 in JUSH. The data were analyzed using Cox PHs a Semi-Parametric model and AFT a parametric models. The performances of these models were compared using AIC criteria. For nested models using LR test and for non-nested models using AIC Results: We fit a cox proportional model and check the proportionality assumption. For all variables except baseline CD4 count the assumption holds. To overcome the violation of proportional hazards we fit two alternative models; time-dependent cox PH and AFT model and compare their goodness of fit. The result shows the lognormal AFT model was the ―best‖ fitting model for this dataset. Using the selected model Functional status, WHO clinical disease stage, pre-TB positive test, regimen type at ART initiation, body mass index and baseline CD4 count were found to be significantly associated with survival time of patients on HAART. Conclusion: The AFT model is a more valuable and realistic alternative to the Cox PH model in situation when the PH assumption does not hold and therefore should be considered as an alternative to the Cox PH in modeling time-to-death of patients under HAART. Patients with low CD4 count, WHO stage III and IV, being underweight, pre-TB positive test and being ambulatory and bedridden are factors that accelerate survival time on HAART.
