External Validation of a Simple Mortality Risk Prediction Score for Tuberculosis Patients

Abstract

Background: Tuberculosis (TB) remains a significant global health issue, contributing to substantial mortality annually; therefore, effective risk stratification is essential for clinical decision-making and optimal resource allocation. This study externally validates a simple mortality prediction score for TB patients that was initially developed by Saisudjarit and colleague in a different population. Methods: A retrospective cohort study was conducted using TB registry data from four Thai provinces (Nakhon Sawan, Kamphaeng Phet, Chai Nat, and Uthai Thani) collected January 1, 2019–December 31, 2020. Eligible patients were aged ≥18 years and had either completed treatment or died during that period. The risk score incorporated three predictors: Charlson Comorbidity Index scores of 1–2 and ≥3 points, and the presence of tuberculous meningitis. Model performance was assessed using the area under the receiver operating characteristic curve (AuROC). Results: Among 4680 patients, the score showed good discrimination (AuROC = 76.3 %) and good calibration (Hosmer–Lemeshow χ² = 0.76; P = 0.68). The positive likelihood ratio was 1.02 in the low-risk group (scores ≤3) and 5.31 in the high-risk group (scores ≥6). Sensitivity/specificity were 1.69 %/99.68 % in the high-risk group, whereas they were 99.17 %/3.21 % in the low-risk group. The model correctly predicted mortality in high-risk patients (87.3 %) and survival in low-risk patients (87.0 %). Conclusion: This externally validated TB mortality risk score shows acceptable calibration and predictive performance. The risk score can guide clinical risk stratification and treatment planning, thereby improving TB-related outcomes.