Unveiling Fall Triggers in Older Adults: A Machine Learning Graphical Model Analysis.

Area of Science:

• Gerontology
• Biostatistics
• Public Health

Background:

• Falls are a major concern for adults aged 60 and older.
• Existing research identifies numerous fall risk factors but often lacks comprehensive analysis of their interrelationships.
• Understanding these complex interactions is crucial for effective fall prevention.

Purpose of the Study:

• To investigate the intricate relationships between diverse fall risk factors in older adults.
• To apply a novel mixed undirected graphical model (MUGM) for analyzing high-dimensional, heterogeneous data.
• To identify key factors and their dependencies contributing to fall risk.

Main Methods:

• Employed a mixed undirected graphical model (MUGM) to analyze the interplay of sociodemographics, mental well-being, body composition, fall risk assessments, and physical activity.
• Utilized a parameterized joint probability density to specify higher-order dependence structures.
• Analyzed 37 fall risk factors in 120 elders from central Florida.

Main Results:

• The MUGM successfully revealed complex interrelationships among fall risk factors.
• 34 out of 37 analyzed features exhibited pairwise relationships.
• COVID-19-related factors and housing composition were found to be conditionally independent of other factors.

Conclusions:

• The MUGM provides innovative insights into fall risk factor dynamics, surpassing traditional correlation analysis.
• This foundational study highlights the need for further longitudinal research to understand fall prevention dynamics.
• Findings can inform the development of more targeted and effective fall prevention strategies for older adults.