Development and Validation of a Risk Prediction Model for New-Onset Diabetes After Pancreatectomy

Overview

New-onset diabetes is one of the major complications after pancreatectomy. Without timely diagnosis, treatment and management, it can lead to adverse outcomes such as cardiovascular and cerebrovascular events and severe infections, increasing mortality. Moving the management threshold forward, early risk identification, and precise and effective intervention are the keys to its management. Existing risk prediction studies show high heterogeneity across different populations and surgical procedures. Risk factors available for early identification remain unclear, and most included factors are non-interventional. There are limitations in study design, modeling and validation methods, and the research results provide limited value for guiding risk prediction, prevention and management. This study adopts a mixed-methods design. We retrospectively analyze the general and clinical data of patients who underwent pancreatectomy at two institutions, follow up the subjects to determine the incidence of postoperative diabetes and postoperative self-management status, and use regression analysis to identify independent risk factors for new-onset diabetes after surgery. We separately establish a nomogram model, a random forest model, and a deep learning model, perform internal and external validation, compare the performance of the three models, and select the model with optimal clinical performance. Based on the prediction model and evidence-based medicine, we formulate intervention strategies, evaluate their feasibility through clinical trials, refine the intervention items, and finally establish a precise intervention strategy for new-onset diabetes after pancreatectomy to achieve comprehensive, dynamic, efficient and precise management. Implementation of this project can reduce the incidence and slow the progression of diabetes after pancreatectomy, improve clinical outcomes, lower readmission and mortality rates, save social medical resources, and provide decision-making guidance and practical evidence for the prevention and control of chronic diseases.

Conditions