Virtual Booth

Background: Tuberculosis (TB) recurrence following successful treatment contributes to ongoing transmission and repeated healthcare utilisation. Identifying patients at increased risk may support targeted post-treatment surveillance. Objective: To develop and validate the R-CUR Score for assessing TB recurrence risk among successfully treated pulmonary TB patients in Malaysia. Methods: Routinely collected National Tuberculosis Registry data were used to develop a prognostic model and a points-based score incorporating 10 routine predictors. The score classifies patients into low, intermediate and high-recurrence risk groups. The analysis included 52,176 patients: 27,076 in development and 25,100 in external validation. Internal validation used 1,000 bootstrap resamples of the development cohort. Discrimination of model-derived probabilities was evaluated using inverse probability-of-censoring-weighted time-dependent receiver operating characteristic curves at 1, 2 and 3 years. Calibration was assessed using the calibration intercept and calibration slope with ideal values of zero and one, respectively. Results: There were 435 recurrences in the development cohort and 310 in the validation cohort. Development AUCs at 1, 2 and 3 years were 0.666 (95% CI 0.633–0.696), 0.642 (0.611–0.671) and 0.642 (0.615–0.669), respectively. Corresponding validation AUCs were 0.638 (0.599–0.676), 0.641 (0.608–0.674) and 0.638 (0.606–0.670). Across 1,000 bootstrap resamples, the two assessed risk-group coefficients had estimated biases of −0.002 and −0.047, supporting the stability of the risk-group associations within the development cohort. Calibration slopes were 1.022 in development and 0.968 in validation, indicating satisfactory calibration. Conclusion: The TB R-CUR model demonstrated modest discrimination and bootstrap-supported stability of risk-group associations. Calibration slopes were close to 1, although overall risk estimates showed systematic deviations. The score offers a practical approach to recurrence risk stratification using routine registry information, further evaluation of score-specific performance, recalibration and clinical usefulness is required before implementation.