Development and Validation of a Nomogram for Preoperative Prediction of Localization of Neonatal Gastrointestinal Perforation

Background: Information regarding the localization of gastrointestinal perforation is crucial for the following surgical procedure. This study was to determine the key indicators and develop a prediction model for the localization in neonates with gastrointestinal perforation. Methods: A nomogram to predict the location of neonatal gastrointestinal perforation was developed using a cohort of patients who underwent surgery between July 2009 and May 2021. Baseline variables were analyzed using logistics regression and nomogram developed using significant predictors. The predictive performance of the nomogram was assessed by the concordance index (C-index), calibration curve, and area under the receiver operating characteristic (ROC) curve (AUC). The nomogram was further validated in an integrated external cohort. Results: We investigated the data of 201 patients, of which 65 (32.3%) were confirmed with upper gastrointestinal perforation by surgery. Multivariate logistic regression analysis identified the following as independent predictors: preterm [OR: 5.014 (1.492–18.922)], time of onset [OR: 0.705 (0.582–0.829)], preoperative hemoglobin [OR:1.017 (1.001–1.033)], bloody stool: No [OR: 4.860 (1.270–23.588)], shock [OR: 5.790 (1.683–22.455)] and sepsis: No [OR 3.044 (1.124–8.581)]. Furthermore, the nomogram was effective in predicting the perforation site, with an AUC of 0.876 [95% confidence interval (CI): 0.830–0.923]. Internal validation showed that the average AUC was 0.861. Additionally, the model achieved satisfactory discrimination (AUC, 0.900; 95% CI, 0.826–0.974) and calibration (Hosmer-Lemeshow test, P = 0.4802) in external validation. Conclusions: The nomogram based on the six factors revealed good discrimination and calibration, suggesting good clinical utility. The nomogram could help surgeons predict the location of gastrointestinal perforation before surgery to make a surgical plan.