Analysis of Influencing Factors of Acute Pancreatitis Complicated With Persistent Inflammation and Construction of a Prediction Model

Objective: To investigate the contributing factors for the development of systemic inflammatory response syndrome (SIRS) in acute pancreatitis (AP) patients and subsequently develop a novel nomogram prediction model. Methods: A multivariate logistic regression analysis was conducted to determine independent predictors of SIRS, where the variables were chosen based on statistical significance from univariate analysis. Based on their presence, 238 AP patients were grouped into non-SIRS (n=170) and SIRS (n=68). Logistic regression analysis identified independent predictors of SIRS complications. We then developed a visual nomogram prediction model alongside a logistic regression model. The model’s predictive power cut-off was determined by receiver operating characteristic (ROC) curve analysis, providing sensitivity, specificity, and predictive accuracy. Results: The study found that in the cohort of acute pancreatitis (AP) patients, systemic inflammatory response syndrome (SIRS) incidence was 28.6%. From our analysis, we determined that red blood cell distribution width (RDW), fibrinogen (FIB), amylase (AMY), blood glucose (Glu), and lactate dehydrogenase (LDH) were independent risk factors for SIRS. In addition, we calculated the area under the ROC curve (AUC) for our prediction model of SIRS reached 0.816, which exceeded the AUCs of the individual risk indicators (RDW, FIB, AMY, Glu, and LDH) and the bedside index of severity in acute pancreatitis (BISAP) score. In addition, we conducted a correlation analysis to validate the relationships among the predictive factors and to eliminate possible multicollinearity. The calibration curve plot showed that the nomogram agreed well between the predicted SIRS and actual risks. Finally, the clinical decision curve for our model also indicated its clinical utility by guiding decision-making for timely interventions at a threshold probability range of 0.4–1. Conclusion: The model predicted non-SIRS with a critical value ≥0.332, a sensitivity of 71.3% and specificity of 87.1%, and a kappa value of 0.56. These results indicate that this prediction model is based on admission data, with recommended additional validation assessments at multiple time points (eg, 24, 48, and 72 h) to fully characterize the progression of SIRS's risk. Overall, this nomogram prediction model provides an efficient and simple means to predict SIRS for patients with AP.