Development and validation of a predictive scoring model for risk stratification of tuberculosis treatment interruption.

Background: Tuberculosis (TB) treatment interruption poses risks of antimicrobial resistance, potentially leading to treatment failure and mortality. Addressing the risk of early treatment interruption is crucial in tuberculosis care and management to improve treatment outcomes and curb disease transmission.

Objectives: This study aimed to identify risk factors of TB treatment interruption and construct a predictive scoring model that enables objective risk stratification for better prediction of treatment interruption.

Methods: A multicentre retrospective cohort study was conducted at public health clinics in Sarawak, Malaysia over 11 months from March 2022 to January 2023, involving adult patients aged ≥18 years with drug-susceptible TB diagnosed between 2018 and 2021. Cumulative missed doses or discontinuation of TB medications for ≥2 weeks, either consecutive or non-consecutive, was considered as treatment interruption. The model was developed and internally validated using the split-sample method. Multiple logistic regression analysed 18 pre-defined variables to identify the predictors of TB treatment interruption. The Hosmer-Lemeshow test and area under the receiver operating characteristic curve (AUC) were employed to evaluate model performance.

Results: Of 2953 cases, two-thirds (1969) were assigned to the derivation cohort, and one-third (984) formed the validation cohort. Positive predictors included smoking, previously treated cases, and adverse drug reactions, while concurrent diabetes was protective. Based on the validation dataset, the model demonstrated good calibration (P = 0.143) with acceptable discriminative ability (AUC = 0.775). A cutoff score of 2.5 out of 11 achieved a sensitivity of 81 % and a specificity of 64.4 %. Risk stratification into low (0-2), medium (3-5), and high-risk (≥6) categories showed ascending interruption rates of 5.3 %, 18.1 %, and 41.3 %, respectively (P < 0.001).

Conclusion: The predictive scoring model aids in risk assessment for TB treatment interruption, enabling focused monitoring and personalized intervention plans for higher-risk groups in the early treatment phase.