Related Experiment Videos
Development and Internal Validation of Machine Learning-Based Risk Prediction Models for Depression in Older Adults
Yumeng Zhang1,2, Mei Yuan1,3, Minzhu Chen4
1Department of Nursing, Affiliated Hospital of Zunyi Medical University, Zunyi, Guizhou, People's Republic of China.
Clinical Interventions in Aging
|August 3, 2026
Summary
Depressive symptoms affect over half of older patients on maintenance hemodialysis (MHD). Vision impairment, hearing loss, and poor self-management are key predictors identified by new machine learning models for early screening.
Area of Science:
- Nephrology
- Geriatrics
- Psychiatry
Background:
- Older patients undergoing maintenance hemodialysis (MHD) exhibit a high prevalence of depressive symptoms.
- Early identification and screening for depression are crucial in this vulnerable population.
Purpose of the Study:
- To identify core predictors of depressive symptoms in older MHD patients.
- To develop and internally validate risk prediction models for depression in this cohort.
Main Methods:
- 226 older MHD patients were recruited, with data collected on demographics, self-management, symptoms (Dialysis Symptom Index - DSI), physical function, nutritional status (Mini Nutritional Assessment-Short Form - MNA-SF), and depression (Geriatric Depression Scale - GDS-15).
- Machine learning algorithms, including Naive Bayes, were employed for variable selection and model development, utilizing 10-fold cross-validation and a train-test split.
- Model performance was assessed using AUC, calibration metrics, F1-scores, and decision curve analysis.
Main Results:
- The prevalence of depressive symptoms (GDS-15) was 54.4% among the study participants.
- All developed models showed strong discriminative performance (AUC range: 0.851-0.865).
- The Naive Bayes model demonstrated optimal performance, with key predictors including vision impairment, hearing loss, DSI, MNA-SF, and self-management.
Conclusions:
- Five key predictors were identified, leading to the development of internally validated machine learning models for depression risk screening in older MHD patients.
- These models show promise for clinical application but require external validation before widespread use.