Factors associated with survival of tuberculosis patients in Southeast Iran: comparison of stepwise cox regression, survival tree, and random survival forest

Understanding the factors associated with survival of tuberculosis (TB) patients is crucial for effective control and prevention efforts. This study aimed to identify the survival factors of TB patients in Iran. A historical cohort study was conducted using data from the Iranshahr (Southeast Iran) Tuberculosis Control Program. Stepwise Cox regression, survival tree analysis, and Random Survival Forest (RSF) models were employed to identify survival predictors. The C-index was used for model comparison. Of the 3429 TB patients (2016-2021), 292 (8.5%) died. Stepwise Cox regression revealed that positive smears after 2 and 3 months of treatment significantly increased mortality risk [HR: 15.44 and HR: 38.28, respectively]. Other factors contributing to a higher risk of death included being a prisoner [HR: 3.8], chest pain [HR: 1.35], and older age [HR: 1.01] (P < 0.05). In contrast, having extrapulmonary TB [HR: 0.16], nationality (Afghan or Pakistani) [HR: 0.36], and increased weight [HR: 0.98] reduced the risk of death (P < 0.05). RSF analysis identified a positive smear after two months as the most critical variable, followed by age, smear results after three months, weight, and TB type. The stepwise Cox model demonstrated a higher C-index than the RSF model. This study highlights key factors associated with the TB patient survival in the region, urging health policymakers to focus on early diagnosis and treatment optimization, especially for high-risk groups.