A Nomogram Prediction Model and Validation of Hypoglycemia Risk in Patients With Decompensated Cirrhosis and Type 2 Diabetes

Abstract

Purpose: This research sought to develop and evaluate a predictive nomogram for determining hypoglycemia likelihood in cases of concurrent decompensated cirrhosis and type 2 diabetes. Methods: This investigation included a cohort of 830 individuals with decompensated cirrhosis and type 2 diabetes in a retrospective design. A random 7:3 split was carried out to create model training and validation sets. Key predictors were selected using LASSO regression followed by multivariate logistic regression to construct the nomogram. Evaluation of the model encompassed both discrimination, quantified by the area under the receiver operating characteristic (ROC) curve (AUC), and calibration, assessed via calibration curves. The model’s generalizability was further confirmed using the held-out validation set. Results: Among the 830 participants included in the analysis, hypoglycemia occurred in 118 cases, corresponding to an incidence rate of 14.22%. Eight variables were significantly associated with hypoglycemia: diabetic peripheral neuropathy, creatinine (Cr), indirect bilirubin (IB), triglycerides (TGs), use of antibiotics within 48 h of admission, duration of diabetes, insulin dosage, and body mass index (BMI). In the training set, the model’s discriminative ability was substantial, with an AUC of 0.736 (95% CI: 0.680–0.792). Notably, this performance was well-maintained in the validation set (AUC = 0.709,95% CI: 0.614–0.803), with reliable calibration demonstrated through strong concordance between model forecasts and actual results. Conclusions: We have constructed a nomogram that incorporates eight predictors to estimate hypoglycemia risk in patients with decompensated cirrhosis and type 2 diabetes. It serves as a practical instrument for hypoglycemia risk assessment, providing valuable support for clinical decisions aimed at prevention and management strategies.