Prediction of late-onset depression in the elderly Korean population using machine learning algorithms

Area of Science:

• Gerontology
• Psychiatry
• Computational Medicine

Background:

• Late-onset depression (LOD) is a significant concern in elderly populations.
• Early prediction of LOD is vital for preventing severe depressive episodes.
• Identifying at-risk individuals can improve clinical outcomes and quality of life.

Purpose of the Study:

• To develop and validate machine learning (ML) models for predicting future late-onset depression.
• To identify key predictive factors for LOD in the elderly Korean population.
• To establish novel ML-based prediction programs for clinical decision support.

Main Methods:

• Utilized data from the 'Korean Longitudinal Study of Aging' (KLOSA) nationwide panel survey.
• Employed latent growth modeling and growth mixture modeling to define depression trajectories.
• Applied binary logistic regression to select predictive variables and tested 12 ML algorithms.

Main Results:

• Identified four distinct latent classes of depression trajectories in the elderly.
• Selected 12 significant variables differentiating LOD from non-LOD groups.
• Random Forest Classifier and Gradient Boosting Classifier exhibited superior predictive accuracy.

Conclusions:

• Successfully developed ML-based prediction programs for late-onset depression.
• These models demonstrate high potential for early LOD detection in clinical settings.
• Future development could lead to self-checking online tools for primary care and health screening.