Self-Organizing Maps to Multidimensionally Characterize Physical Profiles in Older Adults

Area of Science:

• Gerontology
• Biomedical Engineering
• Data Science in Healthcare

Background:

• Community-dwelling older adults require tailored health assessments.
• Understanding the interplay between physical performance and body composition is crucial for healthy aging.
• Mexican older adults present unique demographic and health characteristics.

Purpose of the Study:

• To automatically analyze, characterize, and classify physical performance and body composition data.
• To identify distinct profiles within a cohort of Mexican community-dwelling older adults.
• To explore the utility of Self-Organizing Maps (SOM) for analyzing complex health data.

Main Methods:

• Utilized Self-Organizing Maps (SOM), a type of neural network, for multidimensional data analysis.
• Analyzed data from 562 Mexican older adults, including demographics, health status, physical performance, and body composition.
• Included variables such as gait speed, grip strength, one-legged stance, lean mass, and fat percentage, stratified by sex.

Main Results:

• Identified seven distinct profile types for older men and women based on physical performance and body composition.
• Generated visual maps illustrating clusters of older adults with similar health profiles.
• Demonstrated a non-linear relationship between physical performance and body composition variables.

Conclusions:

• Self-Organizing Maps (SOM) are effective tools for analyzing complex, multidimensional healthcare data.
• The identified profiles provide a nuanced understanding of physical performance and body composition variations in Mexican older adults.
• This approach facilitates data interpretability and the characterization of distinct aging phenotypes.