Socioeconomic and demographic disparities in fitness center visits

Socioeconomic and demographic disparities in physical activity participation are strong predictors of obesity prevalence, underscoring the critical need to understand and address unequal participation in fitness centers across communities. In this work, we integrate large-scale visit data, which include visits to fitness centers and other types of locations, with census-derived socioeconomic and demographic attributes, to investigate patterns related to physical activity. Specifically, we analyze over 50 million daily visits from more than 15 million users and construct socioeconomic profiles for more than 150,000 neighborhoods, including median household income, educational attainment, age distributions, and gender ratios. We develop machine learning models that predict the proportion of residents visiting fitness centers, and show that demographic and socioeconomic features alone explain over half of the observed variation in visit rates (R2 ranging from 0.536 to 0.572 across three months). Explainability analyses via linear regression and SHapley Additive exPlanations (SHAP) of LightGBM models reveal that income and education are the strongest predictors of fitness center visits, while a higher female population proportion is positively associated with fitness center visits. We further show that residents aged 18-24, and those with higher education, are more likely to visit local fitness facilities, whereas income has no effect on choosing facilities closer to home. Our findings offer crucial insights for designing targeted interventions and policies to alleviate physical activity disparities and promote equitable access to fitness facilities