Predictors of Depression Among People with Disabilities: A SHAP-Based Machine Learning Analysis

Key Points

• This research aims to explore the predictors of depression among individuals with disabilities and their relative importance.
• Utilized data from 1,977 individuals with disabilities from the 2022 5th Panel Survey on Life of Disabled Persons.
• Applied machine learning-based random forest regression model to analyze the data.
• Examined non-linear predictive structures of various factors related to depression.
• Life satisfaction was the most significant predictor of depression, with a sharp decline in predicted depression levels at certain thresholds.
• Factors like perceived disability impact and overall health status contributed to lower predicted depression levels.
• Conversely, stress from daily life and the degree of limitations increased predicted depression, indicating a complex interplay of factors.

Abstract