Defaulters in a cohort of HIV infected patients

Abstract

The advent of antiretroviral therapy (ART) has transformed HIV/AIDS from a primary deadly disease into a chronic disease characterized by enhanced quality of life and increased life expectancy. Once started, the antiretroviral treatment should be continued lifelong and adherence to this treatment should be nearly perfect to enable long-term efficacy. Despite the improvements in management in Europe, HIV infected persons still remain vulnerable to drop out/loss to follow up from care and treatment. Therefore, this research tries to review the defaulter rate during the last five years (2002-2006) at the HIV outpatient clinic and how it is evolving during this period. Data was explored using Kaplan-Meier curve to know whether or not the groups are proportional through the assumption of proportional hazards i.e. if the estimated survival functions for two groups of survival data are approximately parallel (do not cross) and a Logrank and Gehan-Wilcoxon tests were used to compare the survival estimates between two or more groups. The Survival data is modeled using Cox’s proportional Hazard model to explore the relationship between survival and explanatory variables thereby analyzing for the effect of several risk factors on survival. Collett’s approach criterion was applied to select the best model ignoring the missingness mechanism in the data. In order to assess the adequacy of the fitted model, residual plots and some formal tests (time-dependent covariate) were used to check for the assumption of proportional hazard. A stratified analysis was carried out to compare the Cox’s proportional hazard model and the stratified model which tells us if the fitted model is good or not. The use of single and multiple imputation methods were used to investigate the nature of the missingness mechanism and its effect, observing the possibility of Missing at Random (MAR). Based on this study, there was an association between defaulters and gender, risk group, clinical stage, sex preference, origin group, ART, viral load and age group, also an evolution over time shows how the patients default through an increasing trend. Our analysis showed that 13.97% of 1167 patients defaulted while 8.94% were lost to follow-up but out of 163 defaulters, more than half of them defaulted in 2006 alone (53.99%) while 3.68% defaulted in the first year of ART. Of all the patients, 67.15% were male but in proportion, female defaulted more than male with 19.84% and 13.05% of female were lost to follow-up. It was observed that the model does not satisfy the proportional hazard assumption even with time dependent covariates, the global goodness of fit test shows that the model is fitted at a borderline significant level but correcting for missingness, the PH assumptions hold in both responses.