Modeling time to recovery from tuberculosis in south west Ethiopia: a case of Jimma University Special Hospital

Abstract

Tuberculosis is a chronic infectious disease that has long been one of the major health problem. Survival analysis is a set of methods used for analysis of the data which exist until the occurrence of an event. Many researches on tuberculosis treatment have reported varying recovery times. This research gets the average recovery time from TB in South West Ethiopia by using parametric survival models. Objective: The objective of this study is to identify the best predictors for the time to recovery from tuberculosis and to compare different parametric survival models on the time to recovery from TB. Method: The study has been used the retrospective data collected from 384 tuberculosis patients’ selected randomly from last three years in JUSH. Survival analysis was used as the population under study is changing, we only consider the individual risk to recovery for those who are still ill, but this means that many standard statistical approaches cannot be applied. Parametric survival models are statistically more powerful than non-parametric or semi-parametric models. Result: Among the potential parametric models fitted, Gompertz model was the best model to study TB data. The covariates: age, body weight, dose level of drug, type of TB, residence and HIV status were statistically significant covariates that affect the time to recovery from TB. The average time to recovery from TB was 172 days while 53.65% (206) of patients’ were recovered from TB. Conclusion: The Gompertz model is the best model to study time to recovery from TB data. From the Gompertz model result we conclude that being old, rural residence, having Extra-pulmonary TB, having HIV, lower doses and body weight at baseline prolonged the recovery time. Covariates significant in the Gompertz model are also significant in the Weibull model. Key words: TB, Gompertz, Hazard, Weibull, Deviance residuals