An early postoperative physiological model for clinically significant postoperative pulmonary complications after lung resection: development and external validation

Current prediction tools, largely preoperative, may not reflect early physiological changes after lung resection. This study therefore aimed to develop and externally validate a postoperative day-1 prediction model for clinically significant postoperative pulmonary complications (PPCs) using routinely available physiological data. This retrospective study included a single-center development cohort of 7,395 adults who underwent lung resection and an independent cohort of 3,762 patients from the INSPIRE dataset for external validation. The primary endpoint was clinically significant PPCs occurring after a 24-h postoperative landmark and within 30 days after surgery, operationalized as early respiratory deterioration requiring escalation of respiratory support. Candidate predictors were screened using LASSO penalization and Boruta. Five candidate algorithms were evaluated using fivefold cross-validation, and logistic regression was selected as the final model. Model performance was assessed by discrimination, calibration, and decision curve analysis, with external validation performed using the locked model. Clinically significant PPCs occurred in 132 of 7,395 patients (1.8%) in the development cohort and 245 of 3,762 patients (6.5%) in the external validation cohort. The final model included 10 predictors: age, sex, body mass index, surgical approach, and six physiological markers obtained within the first 24 postoperative hours (SaO2, PaCO2, blood urea nitrogen, albumin, hematocrit, and white blood cell count). Discrimination remained stable in internal validation (AUROC 0.806; 95% CI 0.768–0.844) and external validation (AUROC 0.805; 95% CI 0.774–0.836). A postoperative day-1 model using routine physiological measurements may facilitate postoperative risk reassessment and identify patients needing intensive surveillance after lung resection. However, prospective validation is a prerequisite for its translation into routine practice.