Development and External Validation of a Multimodal Clinical Prediction Model for Early Stratification of Treatment Response Among Patients with Tuberculosis

This study developed and externally validated a TRIPOD-compliant multimodal LASSO machine learning prediction model integrating rifampicin resistance status and multi-domain clinical indicators to accurately stratify the risk of unfavourable treatment outcomes for pulmonary tuberculosis patients at treatment initiation, with robust discrimination, excellent calibration and satisfactory clinical utility for triage management in drug-resistant tuberculosis programmes.