Application of logistic regression model and decision tree model in the analysis of the recurrence of acute pancreatitis / 中华肝胆外科杂志

ABSTRACT

Objective:To study the logistic regression model and Chi-square automatic interaction detection decision tree model in the prediction of the recurrence of acute pancreatitis (AP).Methods:Clinical data of 364 patients with AP admitted to the Affiliated Hospital of Jiangnan University from June 2021 to June 2022 were retrospectively analyzed, including 219 males and 145 females, aged 53 (19-91) years. The patients were divided into the recurrence group ( n=63), those who experienced a second or more episodes of AP, and the initial group ( n=301), those who were diagnosed of AP for the first time. Univariate and multivariate logistic regression analyses were performed to identify the factors associated with recurrence of AP, and the decision tree model was used to analyze those factors. Receiver operating characteristic (ROC) curve were plotted to analyze the predictive performance of the two models. Results:Multivariate logistic regression analysis showed that age ( OR=0.969, 95% CI: 0.949-0.990, P=0.004), body mass index ( OR=1.142, 95% CI: 1.059-1.232, P=0.001), and hyperlipidemia ( OR=3.034, 95% CI: 1.543-5.964, P=0.001) were independent factors influencing the recurrence of AP. The accuracy of the model in predicting recurrence was 83.2% (303/364). The decision tree model showed that hyperlipidemia and body mass index were factors influencing the recurrence of AP, with an accuracy of 82.7% (301/364) in predicting recurrence. The area under the ROC curve was larger in the logistic regression model compared to that in the decision tree model (0.776 vs 0.730, Z=2.02, P=0.043). Conclusion:The logistic regression model and the Chi-square automatic interaction detection decision tree model can help predict the recurrence of AP. It is recommended to combine the two models to better guide clinical practice.