Clinical Prediction of Intra-Abdominal Infection in Patients with Severe Acute Pancreatitis: Logistic Regression and Nomogram

Objective: The goal is to create a clinical prediction model for individuals suffering from severe acute pancreatitis (SAP) who may have an intra-abdominal infection (IAI). Methods: Patients with acute severe pancreatitis at our institution from January 2020 to December 2023 were retrospectively analyzed. The study population was carefully split into two groups: a training set and a validation set, using a 70:30 ratio. This division was designed to facilitate a thorough development and assessment of the predictive model. In the training set, we identified predictive features utilizing LASSO regression, a method known for its ability to enhance model accuracy by selecting the most relevant variables. Following this, we established both a prediction model and a nomogram through multivariable logistic regression analysis, allowing for a comprehensive assessment of the identified risk factors. To assess the diagnostic performance of our model, we utilized receiver operating characteristic (ROC) curves for both the training and validation cohorts. This analysis yielded valuable information regarding the sensitivity and specificity of our predictive model. Furthermore, we conducted decision curve analysis (DCA) and created calibration plots to enhance our evaluation of the model's accuracy and its practical relevance in clinical settings. Results: A total of 415 participants were included in the analysis, with baseline demographic and clinical characteristics documented. The cohort consisted of 291 individuals in the training set and 124 in the validation set. LASSO regression identified four significant predictors with non-zero coefficients (HCT, PCT, APACHE II, NLR) for subsequent modeling. The prediction model's AUC was 0.853 (95% CI: 0.804-0.901) in the training set and 0.858 (95% CI: 0.786-0.930) in the validation set, according to ROC curve analysis. The calibration curve closely resembled the ideal line, and calibration plots demonstrated a strong alignment between the observed instances of IAI and the predicted values. The DCA demonstrated substantial net benefits for clinical application. Conclusion: The clinical prediction model integrating HCT, PCT, APACHE II, and NLR effectively predicts the risk of IAI in patients with SAP, thereby enhancing patient management strategies.