How We Adapt: Analyzing Physical Activity, Sedentary Time, Sleep & Depression in a Post-Pandemic Environment Using Penalized Regression.

There is an increasing need for advanced statistical methodologies, like machine learning, to explore complex relationships between behavior and mental health disorders. This study investigates the complex interplay between sedentary time, physical activity (both moderate and vigorous), sleep quality, and contextual factors; specifically contrasting pre-pandemic (NHANES 2017–2020) and post-pandemic (NHANES 2021–2023) cohorts, while accounting for key sociodemographic covariates. Motivated by prior research that highlights gaps in understanding the relative influence of these factors on depression, the study employs an Elastic Net variable selection approach to identify critical predictors, which are then validated through two survey-weighted ordinal regression models that handle missing data in distinct ways: Complete Case Analysis (CCA) and a “Not Missing Completely At Random” (NOMCAR) method. The results reveal that while variables such as income ratio (OR ≈ 0.83), age (OR ≈ 0.99), and BMI (OR ≈ 1.02) consistently emerge as strong predictors of depression severity. Total sedentary time, despite a modest effect (OR ≈ 1.07) per hour increase, exerts a compounding influence on depression, particularly when interacting with insufficient sleep (OR ≈ 2.22) and environmental factors observed in the post-pandemic dataset (post-COVID cohort OR ≈ 2). When all of these terms coincide, the relationship between sedentary time and more severe depression is heightened (OR ≈ 1.08) and compounds upon the previous effects of sleep habits and environmental context. These findings suggest that interventions targeting reductions in sedentary behavior and improvements in sleep may have significant public health benefits, providing a nuanced perspective that challenges conventional emphasis solely on recreational physical activity.