Development and external validation of a clinical prediction model for occult diaphragmatic rupture in patients with flail chest

Occult diaphragmatic rupture (ODR) is an uncommon but clinically important complication of blunt thoracic trauma. It is frequently missed on preoperative imaging, leading to delayed diagnosis and catastrophic outcomes. Reliable preoperative risk stratification remains challenging. We aimed to develop and externally validate a clinically interpretable prediction model to estimate the risk of ODR in patients with flail chest. This retrospective multicenter cohort study included consecutive flail chest patients undergoing surgical stabilization of rib fractures (SSRF). The development cohort comprised 911 patients from a tertiary trauma center, and the external validation cohort included 151 patients from independent trauma registries. To eliminate detection bias, ODR was confirmed via uniform intraoperative exploration during SSRF. Candidate predictors were strictly prespecified. To address sparse-data bias and quasi-complete separation inherent to rare events, the multivariable regression model was developed using Firth’s penalized logistic regression. Internal validation was performed using 1000 bootstrap resamples. Model performance was assessed via discrimination, calibration and Brier scores, followed by external validation. ODR occurred in 1.8% of the development cohort and 9.3% of the external validation cohort. The final penalized model retained three highly accessible independent predictors: Injury Severity Score (ISS), preoperative chest tube drainage volume (per 100 mL increment), and rib fractured ends oriented toward the diaphragm. The model demonstrated exceptional discriminatory performance with a bootstrap-corrected C-index of 0.971 and negligible optimism (0.0009). External validation confirmed outstanding discrimination (AUC 0.991, 95% CI 0.978–1.000) alongside satisfactory calibration slopes and low Brier scores across distinct populations. A clinical nomogram was constructed based on the penalized estimates. We successfully developed and externally validated a robust penalized prediction model for ODR in surgically treated flail chest patients. This nomogram serves as an effective preoperative risk-alert system. It particularly assists non-specialized trauma or orthopedic surgeons in identifying high-risk patients who mandate proactive thoracic consultation, preoperative material preparedness, and targeted intraoperative diaphragmatic exploration despite equivocal radiologic findings.