MODELING LENGTH OF HOSPITAL STAY FOR TUBERCULOSIS TREATED IN-PATIENTS AT QUEEN ELIZABETH CENTRAL HOSPITAL: A COMPETING RISK PERSPECTIVE

Abstract

A retrospective cohort study was done on adult TB in-patients database from Queen Elizabeth Central Hospital (QECH) SPINE database to identify factors explaining time to discharge from hospital while accounting for a competing event: death. The study aimed to apply and compare competing risk models on TB data. Semi-parametric Cause-specific hazards (CSH) and Sub-distribution hazard (SDH) models were applied to model the effect of HIV status, age, and Sex in relation to death or discharge from hospital. Test for model assumptions and diagnostics were conducted. Findings showed that the SDH explained best the effect of the covariates to the probability of a patient being discharged or dying. Further the main factors affecting length of hospital stay among TB in-patients were age and HIV Status. HIV positive patients were 17.6% less likely to be discharged from hospital compared to HIV negative patients (p=0.048) and an increase in age, resulted in 2% decrease of chances of discharge. It is important to use the cumulative incidence function for calculating probability of an event. The SDH model was a better model when studying data that involves competing risks. To meet the objective of identifying prognostic factors of discharge in the presence of competing risks, the sub distribution hazard model explained better the covariate effects on event discharge than the CSH model. The findings emphasize the importance to use competing methods which best meet the study objectives.