Development and external validation of risk prediction models for depression in patients with osteoporosis

Patients with osteoporosis have a significantly increased risk of depression compared with those without osteoporosis. The aim of this study was to develop and validate predictive models of depression risk for the osteoporosis population.

Data from the National Health and Nutrition Examination Survey (NHANES) 2005-2010, 2013-2014, and 2017-2018 were used in this study. Data from the 2017-2018 survey cycle was used as the external validation set. For the remaining years of data, 70 % of the subjects were randomly assigned to the training set and the remaining subjects formed the testing set. Based on literature reference, variables were collected and SVM-RFE algorithms were used for feature selection to screen predictor variables. The prediction model was constructed via multivariate logistic regression; the nomogram was established based on the results. The model performance was evaluated using the area under the receiver operating characteristic curve (AUC), calibration, and decision curve analysis (DCA).

Sleep disorder, education, HDL-C, 25-hydroxyvitamin D, BMI, age, triglycerides, Bilirubin, smoke, C-reactive protein, gamma glutamyl transferase, alkaline phosphatase, diabetes mellitus and HbA1c were included in the nomogram after filtering predictive variables. The AUCs of the nomogram for the training set, testing set and external validation set were 0.831 (95 % confidence intervals (CI) = 0.784-0.878), 0.710 (95 % CI = 0.623-0.796), and 0.714 (95 % CI = 0.623-0.804), respectively. The brier scores in the training set, testing set and external validation dataset were 0.080, 0.101, and 0.114 respectively. DCA revealed that the nomogram model in the training, testing set, and external validation set had a net benefit when the risk thresholds were 0-0.38, 0-0.20, and 0-0.22 respectively.

We have developed a depression risk prediction model for clinical application in patients with osteoporosis, which aids in identifying individuals at an early stage who are at risk of developing depression.