Establishment and validation of a model for predicting depression risk in stroke patients.

Abstract

Objectives: This study aimed to develop and validate a clinically applicable nomogram to predict depression risk in stroke patients by integrating multidimensional predictors from rehabilitation assessments, biochemical markers, and lifestyle metrics.

Methods: Using data from 767 stroke patients (training/testing: 363/242; external validation: 162) in the CHARLS database and the First Hospital of Changsha, the Least Absolute Shrinkage and Selection Operator (LASSO) regression identified five predictors: Activities of Daily Living (ADL), Instrumental Activities of Daily Living (IADL), sleep (optimal: 6-8 h), uric acid, and Triglyceride-Glucose-Body Mass Index (TyG-BMI). Multivariable logistic regression constructed the nomogram, validated through ROC analysis (AUC), calibration curves, decision curve analysis (DCA), and SHapley Additive exPlanations (SHAP).

Results: The nomogram demonstrated moderate to strong discrimination, with AUC values of 0.731 (training), 0.663 (testing), and 0.748 (external validation). Calibration plots confirmed high predictive accuracy, while DCA revealed substantial clinical utility. SHAP analysis ranked sleep (protective) and ADL (risk) as top contributors. Lower uric acid and TyG-BMI correlated with higher depression risk, contrasting prior studies on TyG-BMI.

Conclusions: This model enables rapid, cost-effective depression risk stratification using routine clinical data, prioritizing high-risk stroke patients for early intervention. Despite limitations (single-country data, unaddressed stroke subtypes), it bridges predictive analytics and clinical workflows, emphasizing sleep hygiene, metabolic monitoring, and functional rehabilitation.