Risk factors and forecasting model for bleeding during bronchoscopy in patients with airway and pulmonary lesions.

BACKGROUND: The severe bleeding events during bronchoscopy in airway and pulmonary lesions may induce severe clinical outcomes. At present, the risk factors and forecasting model for hemorrhage during bronchoscopy is still unclear. The purpose of this study was to develop and validate a prediction model for severe bleeding during bronchoscopy. METHODS: This retrospective and multicenter study enrolled patients who underwent bronchoscopy with airway and pulmonary lesions. Patients in the derivation cohort (n=7,375) were recruited between January 2021 and December 2022, and patients in the external validation cohort (n=228) were recruited from another institution. In the derivation cohort, weighted points were assigned to each predictor of bleeding determined in the multivariate logistic regression analysis and a prediction model was established. Both internal and external validation of the model was conducted to analyze discrimination and calibration. RESULTS: Type of disease, type of bronchoscopy and thrombin time were identified as risk factors for bleeding during bronchoscopy. And a predictive model that comprised these three variables was established in this study, which had a good overall performance on forecasting severe bleeding during bronchoscopy. In the training cohort, the model demonstrated a Nagelkerke R² of 0.76 and a Brier score of 0.12, and displayed good discrimination, with a c-statistic of 0.89 [95% confidence interval (CI), 0.86 to 0.92]. In the external validation cohort, this model also showed good discrimination, with a c-statistic of 0.87 (95% CI, 0.78 to 0.92), and good calibration (calibration slope, 1.02). CONCLUSIONS: We identified the risk factors and further derived a prediction model for serious bleeding during bronchoscopy in airway and pulmonary lesions, which might be a good clinical decision-making support tool for predicting hemorrhage in patients with bronchoscopy.