Biopsychosocial Factors That Classify Meeting/exceeding Optimal Physical Activity For Generalized Anxiety Disorder: A Random Forest

We have shown that physical activity doses as low as the equivalent of 50 minutes of moderate-intensity physical activity per week were associated with 47% lower odds of Generalized Anxiety Disorder (GAD) across 10 years among older adults from The Irish Longitudinal Study on Ageing (TILDA). Identifying biopsychosocial profiles that are associated with these protective doses is critical to optimization of activity prescription for mental health. PURPOSE: To identify biopsychosocial factors that classify individuals as meeting or not meeting minimally-protective physical activity dose for GAD among 4,752 older adults from TILDA with complete Wave 1-5 GAD data, physical activity data (outcome), and biopsychosocial data (predictors). METHODS: 87 biopsychosocial variables of logical, practical and/or prior empirical relation to physical activity, GAD, and/or their association were considered. Feature selection was performed using Chi-square and planned contrasts at each Wave (p ≤ 0.0001) to identify differences between groups not meeting and meeting/exceeding optimal dose. 41 variables were retained in the random forest model. Bootstrapped aggregation quantified model accuracy. The most important features were those that most reduced model accuracy when removed. Data processing and analyses were conducted in Stata 18.0 and RStudio(v4.2). RESULTS: The random forest model classified 72.84% of cases correctly, with greater misclassification among those not meeting the optimal dose (59% error) compared to those meeting/exceeding optimal dose (21% error). The top 10 features contributing to the classification were grip strength, physical limitations, timed-up-and-go performance, employment status, self-reported physical health, activities of daily living impairments, fear of falling, quality of life, sex, and depressive symptoms. CONCLUSION: Collective expression of mostly physical function-related characteristics appears to be a hallmark of those meeting/exceeding the minimally-protective activity dose identified for GAD. However, heterogeneity in these features among those not meeting optimal dose made classification more prone to error, reinforcing the need for precision health initiatives to promote physical activity for mental health among older adults. Supported by: This work was funded by the Health Research Board of Ireland (SDAP-2021-004)