Combination of radiomic and clinical characteristics to predict mortality in patients with colorectal perforation

We aim to construct and verify a model combining radiomic and clinical data to predict early mortality in patients with colorectal perforation in a two-center study.

Data from 147 patients at Xiaogan Central Hospital (2014-2024) and 52 patients at Southern Hospital of Southern Medical University (2021-2023) were collected for model training and validation. Univariate and multivariate analyses were performed to identify risk factors associated with mortality. Radiomic characteristics from CT scans were extracted via least absolute shrinkage and selection operator (LASSO) regression to construct an imaging score. A nomogram was developed by integrating the findings from the multivariate analysis. Predictive performance was evaluated via the area under the receiver operating characteristic curve (AUC), and clinical utility was assessed via decision curve analysis (DCA).

Univariate analysis highlighted age, ASA classification, shock index, rad-score, white blood cell (WBC) count, neutrophil (N) and lymphocyte (L) counts, sodium (Na⁺), creatinine (Cr), and procalcitonin (PCT) as significant prognostic indicators for mortality (p < 0.05). Multivariate analysis confirmed age, ASA classification, PCT, and rad-score as independent prognostic factors. The radiomic combined with clinical characteristics nomogram (RCCCN) includes four variables: the patient's age, ASA classification, PCT level, and rad-score. The RCCCN model demonstrated excellent predictive performance for mortality risk in the validation cohort (AUC: 0.92, 95% CI: 0.84-0.99) with good calibration.

A nomogram combining radiomic features and clinical characteristics effectively predicts mortality in patients with colorectal perforation, providing a valuable tool for clinical decision-making and patient management.