Exploratory analysis using machine learning of predictive factors for falls in type 2 diabetes

Area of Science:

• Gerontology
• Endocrinology
• Neurology

Background:

• Falls pose a significant risk to individuals with type 2 diabetes (T2D), impacting their quality of life and healthcare costs.
• Previous research has primarily focused on elderly populations, potentially overlooking risks in non-elderly individuals with T2D.

Purpose of the Study:

• To investigate the prevalence of falls in persons with T2D, including non-elderly individuals.
• To identify critical risk factors associated with falls in this population using advanced statistical methods.

Main Methods:

• A cohort of 316 individuals with T2D underwent assessment of medical history, laboratory data, and physical capabilities.
• A questionnaire on falls was administered one year post-discharge, with a 72% response rate.
• Logistic regression and random forest classification models were employed to identify fall predictors.

Main Results:

• The fall rate within the first year post-discharge was 19%.
• Logistic regression identified knee extension strength, fasting C-peptide (F-CPR) levels, and dorsiflexion strength as independent predictors.
• Random forest analysis highlighted grip strength, F-CPR, knee extension strength, dorsiflexion strength, and proliferative diabetic retinopathy as key variables.

Conclusions:

• Lower extremity muscle weakness, elevated F-CPR levels, and reduced grip strength are significant risk factors for falls in individuals with T2D.
• The random forest model demonstrated utility in identifying novel risk factors beyond traditional logistic regression analysis.