Abstract: Objective To construct and verify the predictive model for a prolonged length of stay（LOS）in patients with non-severe acute pancreatitis, and provide reference for clinical management and nursing guidance. Methods A total of 980 patients with non-severe acute pancreatitis from January 2018 to December 2021 in a tertiary grade-A hospital in Dali were enrolled in this research and were divided into non-prolonged LOS group and prolonged LOS group. Duality logistic regression analysis was used to construct a predictive model. Hosmer-Lemeshow test, receiver operating characteristic curve (ROC) and Youden index were used to verify the differentiation, consistency and clinical usefulness of the model. Results Six independent influencing factors of education background, dietary habit (unhealthy diet category), white blood cells, alanine, aspartic and Morse Fall score were included in this research. The final predictive model was Z=-2.357-0.276×education background+2.248×dietary habit (unhealthy diet category)+0.095×WBC+0.014×alanine-0.009×aspartic+0.018×Morse score; the area under the ROC curve(AUC)=0.789,95%CI(0.733,0.845); the maximum Youden index was 0.512; the sensitivity 0.656, and the specificity 0.856. Verification model showed that the sensitivity was 0.775; the specificity 0.937, and the accuracy 93.9%. Conclusion This model could predict the risk of prolonged LOS in patients with non-severe acute pancreatitis, which could provide reference for clinical management and nursing guidance.

Key words: acute pancreatitis, non-severe, prolonged length of stay, risk prediction model