Determinant factors associated with time to death of HIV/AIDS patients: An application of parametric shared frailty models

Abstract

Globally, 37.9 million people were living with HIV at the end of 2018. Among them, 36.2 million were adults and 1.7 million children (<15 years), and 23.3 million people living with HIV were accessing treatment, called antiretroviral therapy (ART). This study used the Accelerated Failure Time (AFT) model with gamma and inverse Gaussian frailty distributions to estimate the effect of different factors on time to death of HIV/AIDS patients with ART at Debre Tabor referral hospital, Ethiopia.This investigation was conducted on 351 HIV-positive adults who initiated their treatment between January 2015 and December 2019. The accelerated failure time model with shared frailty distributions has been compared using Akaike Information Criterion (AIC) and Bayesian Information Criterion (BIC) to select the best model for our study. The risk factors age, marital status, education level, TB, WHO clinical stage, OIs, occupation and weight are statistically significant for the survival of HIV/AIDS patients who are in ART. The Gamma shared frailty model with Weibull as baseline distribution is found to be the best model for HIV/AIDS patients' data. The model also suggested that there is strong evidence of heterogeneity between residences of HIV/AIDS patients. From different candidate models, Gamma shared frailty model with the Weibull baseline is an appropriate approach for analyzing this HIV/AIDS data. The study has shown high mortality of the cohort in the earlier months of treatment. About 57.8% of deaths occurred within ten months of ART initiation and tended to decrease later. There was a frailty effect on the time-to-death of patients that arises due to differences in the distribution of residence of patients.