Abstract:

Objective: To identify the risk factors associated with mortality in patients with pulmonary tuberculosis admitted to the emergency department and to develop a predictive risk model to assist clinical decision-making and improve patient outcomes. Methods: A retrospective analysis was conducted on 237 patients with pulmonary tuberculosis who were treated in the emergency department of Beijing Chest Hospital, Capital Medical University, between January 2024 and June 2024. Data collected included baseline demographic characteristics, clinical symptoms, laboratory findings, radiographic severity, and bacteriological evidence. Patients were categorized into two groups—the survival group and the mortality group—based on their clinical outcomes. Comparative analyses of clinical data between the two groups were performed. Multivariate logistic regression was employed to identify independent risk factors associated with mortality in the emergency setting. A predictive risk model was subsequently developed, and its performance in forecasting poor outcomes was evaluated using the receiver operating characteristic (ROC) curve. Results: Of the 237 patients included in the study, 189 survived and 48 succumbed, resulting in an overall mortality rate of 20.3%. Multivariate logistic regression analysis identified five independent risk factors associated with mortality in the emergency department: elevated blood urea nitrogen (BUN) levels (>8.3 mmol/L), advanced radiographic lesions, respiratory distress, respiratory failure, and consciousness disorders. The respective odds ratios (ORs) and 95%CIs for these factors were 4.550 (1.761-11.758), 4.429 (1.443-13.588), 27.507 (3.985-189.847), 2.687 (1.042-6.926), and 27.184 (6.926-106.699). The predictive risk model was expressed as: logit P=ln(P/(1-P))=-1.433+1.515×(BUN>8.3 mmol/L)+1.488×(advanced radiographic lesions)+3.314×(respiratory distress)+0.988×(respiratory failure)+3.303×(consciousness disorder). ROC curve analysis demonstrated that the area under the curve (AUC) was 0.8968 when the five risk factors were combined. The predictive model achieved a sensitivity of 87.50% and a specificity of 74.60% for identifying mortality in emergency pulmonary tuberculosis patients. Conclusion: Incorporating blood urea nitrogen levels, radiographic grading, and clinical indicators such as respiratory distress, respiratory failure, and consciousness disorders provides a robust framework for effectively predicting the prognosis of patients with pulmonary tuberculosis in the emergency setting.

Key words: Tuberculosis, pulmonary, Emergency treatment, Death, Factor analysis, statistical