[Construction and validation of a prediction model for prolonged hospitalization in patients with severe acute pancreatitis].

Abstract

Objective: To construction the risk factors associated with prolonged hospitalization in patients with severe acute pancreatitis (SAP) and develop a prediction model for assessing these risks.

Methods: SAP patients admitted to the department of emergency of Hebei Province Cangzhou Hospital of Integrated Traditional Chinese and Western Medicine from January 2015 to December 2023 were retrospectively selected as the study subjects. The 75% of hospital stay was used as the cut-off point, and the patients were categorized into a normal group and an extended group. Clinical indicators of patients were collected, and independent risk factors for prolonged hospital stay in SAP patients were analyzed using multifactor Logistic regression. A prediction model was established, and a nomogram was created. The efficiency of the prediction model was evaluated using a receiver operator characteristic curve (ROC curve). The accuracy of the model was assessed using Hosmer-Lemeshow goodness-of-fit test. Decision curve analysis (DCA) was employed to evaluate the clinical applicability of the model. Finally, internal validation of the model was conducted using Bootstrap method.

Results: A total of 510 patients with SAP were included, and the length of hospital stay was 18 (6, 44) days, including 400 cases in the normal group (<24 days) and 110 cases in the extended group (≥24 days). Multivariate Logistic regression analysis showed that abdominal effusion [odds ratio (OR) = 4.163, 95% confidence interval (95%CI) was 2.105-8.234], acute physiology and chronic health evaluation II (APACHE II; OR = 1.320, 95%CI was 1.185-1.470), C-reactive protein (CRP; OR = 1.006, 95%CI was 1.002-1.011), modified CT severity index (MCTSI; OR = 1.461, 95%CI was 1.213-1.758), procalcitonin (PCT; OR = 1.303, 95%CI was 1.095-1.550) and albumin (OR = 0.510, 95%CI was 0.419-0.622) were independent risk factors for prolonged hospital stay in SAP patients (all P < 0.01). ROC curve analysis showed that the area under the curve (AUC) of the model was 0.922 (95%CI was 0.896-0.947), the optimal cut-off value was 0.726, the sensitivity was 87.3%, and the specificity was 85.3%. Hosmer-Lemeshow test showed that χ ² = 5.79, P = 0.671. It showed that the prediction model had good prediction efficiency and fit degree. The DCA curve showed that the prediction probability of the model could bring more clinical benefits to patients at 0.1 to 0.7. Bootstrap internal verification showed that the model had a high consistency (AUC = 0.916).

Conclusions: Abdominal effusion, high APACHE II score, high CRP, high MCTSI, high PCT and low albumin level are significantly associated with prolonged hospital stay in SAP patients. The prediction model can help clinicians make more scientific clinical decisions for SAP patients.