Analysis of prognostic factors and construction of a prediction model for patients with initially treated severe pulmonary tuberculosis

Background Tuberculosis is a chronic infectious disease caused by Mycobacterium tuberculosis infection and is the leading cause of death among single infectious diseases worldwide. The mortality rate of patients with severe pulmonary tuberculosis (PTB) is relatively high. This study aimed to explore the prognostic factors of patients with initially treated severe PTB and construct a predictive model. Methods A retrospective analysis was conducted on 189 patients with initially treated severe PTB admitted to Beijing Chest Hospital from January 2024 to October 2024. The patients were randomly assigned to a modeling group (n=107) and a validation group (n=82). Patients in the modeling group were further assigned to a survival group (n=73) and a mortality group (n=34) based on their 60-day outcome after admission. Multivariable least absolute shrinkage and selection operator (LASSO) regression analysis was used to assess the prognostic factors for 60-day outcomes in the modeling group. A nomogram prediction model was constructed using R software and validated with the validation group data through receiver operating characteristic (ROC) curves, decision curve analysis (DCA), and calibration curves. Results LASSO regression analysis indicated that C-reactive protein (CRP), sodium (Na), albumin (ALB), blood urea nitrogen (BUN), lymphocyte count (LY), neutrophil-to-lymphocyte ratio (NLR), respiratory failure, and consciousness disorder were independent risk factors for 60-day mortality in patients with initially treated severe PTB. ROC curve analysis showed that the area under the curve (AUC) of the nomogram model was 0.8774 in the modeling group and 0.8341 in the validation group. DCA curve results demonstrated that the clinical benefit of this model was superior to both the "treat all" and "treat none" strategies in both the modeling and validation groups. The calibration curve results indicated good fit between the actual and corrected curves of the model, which were close to the ideal curve. Conclusions The predictive model constructed based on CRP, Na, ALB, BUN, LY, NLR, respiratory failure, and consciousness disorder has good predictive value for the prognosis of patients with initially treated severe PTB.