Prediction Model of Delayed Hemothorax in Patients with Traumatic Occult Hemothorax Using a Novel Nomogram.

Background: Delayed hemothorax (dHTX) can occur unexpectedly, even in patients who initially present without signs of hemothorax (HTX), potentially leading to death. We aimed to develop a predictive model for dHTX requiring intervention, specifically targeting those with no or occult HTX.

Methods: This retrospective study was conducted at a level 1 trauma center. The primary outcome was the occurrence of dHTX requiring intervention in patients who had no HTX or occult HTX and did not undergo closed thoracostomy post-injury. To minimize overfitting, we employed the least absolute shrinkage and selection operator (LASSO) logistic regression model for feature selection. Thereafter, we developed a multivariable logistic regression (MLR) model and a nomogram.

Results: In total, 688 patients were included in the study, with 64 cases of dHTX (9.3%). The LASSO and MLR analyses revealed that the depth of HTX (adjusted odds ratio [aOR], 3.79; 95% confidence interval [CI], 2.10-6.85; p<0.001) and the number of totally displaced rib fractures (RFX) (aOR, 1.90; 95% CI, 1.56-2.32; p<0.001) were significant predictors. Based on these parameters, we developed a nomogram to predict dHTX, with a sensitivity of 78.1%, a specificity of 76.0%, a positive predictive value of 25.0%, and a negative predictive value of 97.1% at the optimal cut-off value. The area under the receiver operating characteristic curve was 0.832.

Conclusion: The depth of HTX on initial chest computed tomography and the number of totally displaced RFX emerged as significant risk factors for dHTX. We propose a novel nomogram that is easily applicable in clinical settings.