Development and validation of a nomogram clinical prediction model for osteoporosis in elderly malnourished patients: A diagnostic accuracy study.

The aim of this study was to establish a nomogram model for predicting the incidence of osteoporosis (OP) in elderly malnourished patients and to verify its predictive effect. We conducted a retrospective analysis of elderly malnourished patients hospitalized at the Affiliated Hospital of Chengdu University of Traditional Chinese Medicine between December 2023 and June 2024. The cohort was randomly divided into a training set and a validation set in a 7:3 ratio. Optimal factors were identified using the least absolute shrinkage and selection operator (LASSO) regression, which were then incorporated into a multifactorial logistic regression model to ascertain independent predictors. The Hosmer-Lemeshow test, area under the curve (AUC), calibration curve, decision curve analysis, and clinical impact curve (CIC) were used to assess the model's goodness of fit, discrimination, calibration, and clinical impact, respectively. A total of 381 patients were included in the analysis. Independent predictors of OP in this population included: geriatric nutritional risk index (odds ratios (OR) = 0.520, 95% confidence intervals (CI): 0.282-0.958), activity situation (OR = 0.590, 95% CI: 0.353-0.987), hypertension (OR = 2.833, 95% CI: 1.384-5.798), type 2 diabetes mellitus (OR = 4.314, 95% CI: 1.971-9.439), serum calcium (OR = 0.012, 95% CI: 0.001-0.180), total cholesterol (OR = 4.185, 95% CI: 2.571-6.809), triglycerides (OR = 2.003, 95% CI: 1.217-3.297), albumin (OR = 0.804, 95% CI: 0.683-0.946), overall hip joint bone mineral density (BMD) (OR = 0.015, 95% CI: 0.001-0.225), overall lumbar spine BMD (OR = 0.029, 95% CI: 0.005-0.188), and alkaline phosphatase (OR = 1.022, 95% CI: 1.011-1.034). The AUC for the training and validation sets were 0.946 (95% CI: 0.920-0.972) and 0.963 (95% CI: 0.936-0.990), respectively, indicating great discriminatory ability. The nomogram model developed in this study exhibits good discrimination and accuracy, facilitating the identification of OP risk in elderly malnourished patients in a simple and efficient manner. This model supports early clinical decision-making and intervention, serving as a vital tool for improving patient prognosis. It is anticipated that larger, multicenter studies will be conducted to further validate, enhance, and update the model.