Mortality and its predictors among TB patients within two years after TB treatment discharge at Kawaala and Kiswa Health Centre IVs, Kampala City, Uganda

Abstract

Background: Tuberculosis (TB) treatment completion is an important indicator of programme success; however, this does not always indicate full recovery, and some patients die after treatment has ended. Evidence on the magnitude, timing and determinants of post-treatment mortality remains limited in routine Ugandan settings. This study assessed mortality among TB patients within two years after treatment discharge at Kawaala and Kiswa Health Centre IVs, Kampala City. Objective: To determine two-year mortality, potential predictors of mortality and time to mortality among TB patients after treatment discharge Methods: An ambispective cohort study included 259 eligible participants from a parent TB-diabetes cohort. Vital status was established from cohort and facility records, telephone follow-up and verbal-autopsy interviews with next of kin. Mortality was summarized as a cumulative proportion and person-time rate. Firth penalised logistic regression assessed potential predictors, while Kaplan–Meier and exploratory Cox proportional-hazards methods assessed survival and time to death Results: Participants contributed 6,036 person-months (503 person-years). Ten deaths occurred, giving a two-year cumulative mortality of 3.9% and a crude mortality rate of 19.9 deaths per 1,000 person-years. Five deaths occurred within 0–6 months and five during 6–12 months; none occurred during 12–24 months. Overall survival was 0.981 at six months and 0.961 at 12 months, remaining unchanged through 24 months. HIV-positive status was the strongest potential predictor of mortality across the Firth models (Model A: AOR 3.34, 95% CI 0.96–11.64; Model B: AOR 3.37, 95% CI 0.97–11.71) and the exploratory Cox model (AHR 3.41, 95% CI 0.98–11.84). Dysglycaemia did not show a clear prognostic relationship with mortality (AHR 1.12, 95% CI 0.23–5.54). All adjusted estimates were imprecise because only 10 deaths occurred Conclusion: Post-treatment mortality was relatively low but was concentrated entirely within the first year. HIV-positive status showed the strongest potential predictor signal, although the estimates remained imprecise. The findings support evaluating structured first-year post-TB follow-up and mortality surveillance, with particular attention to continuity of HIV care among HIV-positive survivors. Larger multicentre prospective studies are needed to confirm predictors of mortality and improve precision and generalizability.