Abstract
BACKGROUND: The study objective was to develop and validate a clinical decision support system (CDSS) to guide clinicians through the diagnostic evaluation of hospitalized individuals with suspected pulmonary tuberculosis (TB) in low-prevalence settings. METHODS: The "TBorNotTB" CDSS was developed using a modified Delphi method. The CDSS assigns points based on epidemiologic risk factors, TB history, symptoms, chest imaging, and sputum/bronchoscopy results. Below a set point threshold, airborne isolation precautions are automatically discontinued; otherwise, additional evaluation, including infection control review, is recommended. The model was validated through retrospective application of the CDSS to all individuals hospitalized in the Mass General Brigham system from July 2016 to December 2022 with culture-confirmed pulmonary TB (cases) and equal numbers of age and date of testing-matched controls with three negative respiratory mycobacterial cultures. RESULTS: 104 individuals with TB (cases) and 104 controls were identified. Prior residence in a highly endemic country, positive interferon release assay, weight loss, absence of symptom resolution with treatment for alternative diagnoses, and findings concerning for TB on chest imaging were significant predictors of TB (all P < 0.05). CDSS contents and scoring were refined based on the case-control analysis. The final CDSS demonstrated 100% sensitivity and 27% specificity for TB with an AUC of 0.87. CONCLUSIONS: The TBorNotTB CDSS demonstrated modest specificity and high sensitivity to detect TB even when AFB smears were negative. This CDSS, embedded into the electronic medical record system, could help reduce risks of nosocomial TB transmission, patient-time in airborne isolation, and person-time spent reviewing individuals with suspected TB.