Predicting dyslipidemia in Chinese elderly adults using dietary behaviours and machine learning algorithms

Area of Science:

• Gerontology
• Public Health
• Computational Medicine

Background:

• Dyslipidemia poses a significant health risk to the elderly population.
• Predictive models for dyslipidemia in elderly Chinese adults are crucial for public health interventions.
• Integrating lifestyle factors, such as diet, can improve risk prediction.

Purpose of the Study:

• To predict dyslipidemia risk in elderly Chinese adults using machine learning (ML) and dietary analysis.
• To identify novel predictors of dyslipidemia in this demographic.
• To inform public health strategies for dyslipidemia prevention and control.

Main Methods:

• A cross-sectional study of 13,668 Chinese adults aged 65+ from the 2018 Chinese Longitudinal Healthy Longevity Survey.
• Utilized various ML algorithms (Light Gradient Boosting Machine, Support Vector Machine, XGBoost, etc.) and logistic regression for dyslipidemia prediction.
• Analyzed dietary behaviors and other lifestyle factors as potential predictors.

Main Results:

• The prevalence of dyslipidemia was 5.4% among the study participants.
• Light Gradient Boosting Machine (LGBM) demonstrated the highest predictive accuracy (AUC > 0.70), outperforming other ML models and logistic regression.
• Frequency of nut consumption, childhood water source, and housing type emerged as significant predictors of dyslipidemia.

Conclusions:

• ML models integrating dietary habits effectively predicted dyslipidemia in elderly Chinese adults.
• Identified novel risk factors, including nut consumption frequency, childhood water source, and housing type.
• These findings offer potential targets for developing new strategies to prevent and manage dyslipidemia in older adults.