Construction and Validation of a Nomogram for Predicting Acute Kidney Injury After Pancreatoduodenectomy

Background Acute kidney injury (AKI) after pancreatoduodenectomy is common and early identification of such patients is critical. The nomogram, a visual predictive model, is commonly used to predict AKI after various types of surgery. We aimed to construct and evaluate a predictive nomogram for postoperative AKI in patients undergoing pancreaticoduodenectomy. Methods In a retrospective cohort study, we examined 844 adult patients who underwent pancreaticoduodenectomy from December 2016 to June 2020. All enrolled patients were randomly assigned to the training and validation cohorts in a 7:3 ratio. We utilized LASSO regression for feature selection. A nomogram was constructed using multivariate logistic regression. The nomogram's performance was assessed using various metrics such as the receiver operating characteristic curve, calibration curves, Hosmer-Lemeshow goodness of fit, and decision curve analysis. Results In this cohort, AKI was observed in 98 out of 844 patients, representing an incidence rate of 11.6%. Multivariate logistic analysis showed that direct bilirubin (DBIL), blood loss, urine output, intensive care unit (ICU) transfer were independent influencing factors of postoperative AKI. The nomogram, incorporating the four identified factors, demonstrated moderate discrimination in both the training and validation cohorts, achieving AUC values of 0.720 and 0.772, respectively. The Hosmer-Lemeshow goodness of fit test and the calibration curve demonstrate good agreement between predicted and observed values. The decision curve analysis (DCA) indicated a positive net clinical benefit. Conclusions We developed and validated a nomogram model that could help identify individuals at risk of AKI following pancreaticoduodenectomy. This model may help clinicians optimize perioperative management for these patients.