New model for predicting the development of pancreatic pseudocyst secondary to acute pancreatitis

Abstract

Pancreatic pseudocyst (PPC) increases the risk of a poor prognosis in in patients with acute pancreatitis (AP). Currently, an efficient tool is not available for predicting the risk of PPC in patients with AP. Therefore, this research aimed to explore the risk factors associated with PPC secondary to AP and to develop a model based on clinical information for predicting PPC secondary to AP. This study included 400 patients with acute pancreatitis and pancreatic pseudocyst secondary to acute pancreatitis admitted to the emergency department and gastroenterology department of The First Affiliated Hospital of the University of Science and Technology of China from January 2019 to June 2022. Participants were divided into no PPCs (321 cases) and PPCs (79 cases). Independent factors of PPC secondary to AP were analyzed using univariate and multivariate logistic regression. The nomogram model was constructed based on multivariate logistic regression analyses, which included all risk factors, and evaluated using R. We enrolled 400 eligible patients and allocated 280 and 120 to the training and test sets, respectively. Clinical features, including severe pancreatitis history [odds ratio (OR) = 4.757; 95% confidence interval (CI): 1.758–12.871], diabetes mellitus (OR = 6.919; 95% CI: 2.084–22.967), history of biliary surgery (OR = 9.232; 95% CI: 3.022–28.203), hemoglobin (OR = 0.974; 95% CI: 0.955–0.994), albumin (OR = 0.888; 95% CI: 0.825–0.957), and body mass index (OR = 0.851; 95% CI: 0.753–0.962), were significantly associated with the incidence of PPC after AP in the training sets. Additionally, the individualized nomogram demonstrated good discrimination in the training and validation samples with good calibration, The area under the curve and 95% CI of the nomogram were 0.883 (0.839–0.927) in the training dataset and 0.839 (0.752–0.925) in the validation set. We developed a nomogram model of PPC secondary to AP using R Studio. This model has a good predictive value for PPC in patients with AP and can help improve clinical decision-making.