Physical Activity, Social Isolation, and Depression in Older Adults during COVID-19: A Path Analysis (TOP-SCORING POSTER)

Purpose: Imposed social isolation during COVID-19 (sheltering in place) provided an opportunity to expand our understanding of factors associated with depression in older adults. The purpose of this study was to identify relationships between physical activity, social isolation, and depression among community-dwelling older adults during COVID-19.

Background: Physical inactivity and social isolation have been shown to have negative consequences for mental health in older adults. During COVID-19, public policies to shelter in place were accompanied by a new phenomenon of imposed social isolation and may have resulted in physical inactivity. Researchers posited that pandemic-related sheltering in place would result in increased feelings of social isolation among older adults. In order to promote optimal mental health in this population, research was needed to identify important factors associated with depression during a pandemic or times of imposed social isolation. We hypothesized that 1) higher social isolation would predict higher depression; 2) lower physical activity would predict higher depression; 3) higher social isolation would predict lower physical activity; 4) personal factors (older age, lower income, and lower education) would be directly associated with reduced physical activity; 5) personal factors would be directly associated with increased social isolation; and 6) physical activity would mediate the relationship between social isolation and depression.

Methods: A cross-sectional, correlational design was used. Survey links were sent to older adults aged 65 or older. The physical activity scale for elders, geriatric depression scale, and PROMIS social isolation scale were administered. Path analysis was used to test the hypothesized model. Standardized regression coefficients were calculated, and R2 measures for each exogenous variable were estimated. Goodness-of-fit criteria were calculated to assess model fit with comparative fit index ≥0.90, Tucker-Lewis index ≥0.90, root mean square error of approximation