Development and validation of a predictive model for depression risk in older patients with multiple chronic diseases in the community: cross-sectional study based on a health ecology model

Guided by the health ecology model, which posits that an individual's health is shaped by the dynamic interplay between personal and environmental factors, we investigated factors associated with depression in community-dwelling older adults with multimorbidity and developed a nomogram-based risk prediction model. While previous research has predominantly focused on depression in the context of single chronic diseases, the psychosocial and clinical complexities inherent to multimorbidity remain largely overlooked. This study addresses this gap by constructing a tailored prediction model that integrates the multidimensional determinants of depression in this vulnerable population.

Using convenience sampling, a questionnaire survey was conducted among 679 older patients with chronic diseases from 12 community health institutions in China between March and August 2023. Participants were randomly divided into a training group (n = 475, 70%) and a validation group (n = 204, 30%). Depressive status was assessed using the Geriatric Depression Scale-15 (GDS-15). Logistic regression analysis identified factors associated with depression, based on which a nomogram prediction model was constructed. The model was internally validated using the Bootstrap method with 1,000 resamples. Its predictive performance was comprehensively evaluated using receiver operating characteristic (ROC) curve analysis, calibration curves, and decision curve analysis.

The prevalence rate of depression among older community residents with comorbid chronic conditions was 29.90%. We identified specific predictors for depression in this population: age ≥ 80 years, excess body weight, types of medication, self-management (the ability to actively manage one's health conditions), self-efficacy (confidence in one's ability to perform health-related actions), and educational level. For the training group, the area under the receiver operating characteristic (ROC) curve was 0.815, indicating a model accuracy of 74%, a sensitivity of 79%, and a specificity of 72%. Hosmer-Lemeshow fitting testing results in a χ² value of 8.801 (p = 0.359).

Our new predictive model for the risk of depression in older patients in the community with multiple chronic diseases exhibited good discrimination, calibration and clinical practicability, serving as a valuable reference for the early detection of depression among this population.