A Clinical Prediction Model for Atypical Tuberculosis Manifestations Among Older Adults

Background and Objectives: Active pulmonary tuberculosis (aPTB) in the Older Adults (≥75 years) is frequently under-recognized in non-pulmonology settings due to atypical symptoms and multiple comorbidities. This study aimed to develop and validate a TRIPOD-compliant clinical prediction model for early identification of atypical aPTB in this vulnerable population. Materials and Methods: We retrospectively analyzed 5651 patients aged ≥75 years with culture-confirmed aPTB and World Health Organization (WHO) symptom scores < 5. Patients were stratified into Group a (Ga, Patients with aPTB not initially suspected by non-pulmonologists (atypical presentation, WHO/CDC 7-point score < 5, n = 1155) and Group b (Gb, Patients without aPTB within the first 24 h (non-TB comparators), n = 4496). Multivariate logistic regression identified independent predictors of delayed diagnosis. A weighted scoring system was derived from β-coefficients and validated in independent derivation (2000-2020) and temporal validation (2021-2023) cohorts. Model discrimination, calibration, and decision curve analysis (DCA) were assessed following TRIPOD standards. Results: Five independent predictors-age > 85 years (OR = 6.31, 95% CI = 5.31-8.72), hypoalbuminemia (OR = 4.10, 95% CI = 3.92-7.26), cardiovascular disease (OR = 3.32, 95% CI = 1.23-5.27), diabetes mellitus (OR = 2.03, 95% CI = 1.32-4.07), and predominant lower-lung field involvement (OR = 1.25,95% CI = 1.03-2.44)-were incorporated into the scoring model. Using a cutoff ≥ 7, the model achieved excellent performance across all cohorts (AUC 0.95-0.96; sensitivity 91-94%; specificity 97-99%). Calibration plots and DCA confirmed strong agreement and high net clinical benefit. Nearly 70% of atypical cases had symptom scores ≤ 1, lacking typical signs such as fever or cough. Conclusions: Oldest-old (>85 years) emerged as the strongest independent predictor of atypical TB, surpassing conventional frailty indicators such as sarcopenia or osteoporosis. The proposed score provides a simple, accurate, and validated tool for early detection of aPTB in non-pulmonology settings. Its integration into electronic medical records may reduce diagnostic delays and improve outcomes in this high-risk, late-elderly population.