Imputational modeling of spatial context and social environmental predictors of walking in an underserved community: The PATH trial

BackgroundThis study examined imputational modeling effects of spatial proximity and social factors of walking in African American adults.PurposeModels were compared that examined relationships between household proximity to a walking trail and social factors in determining walking status.MethodsParticipants (N = 133; 66% female; mean age = 55 years) were recruited to a police-supported walking and social marketing intervention. Bayesian modeling was used to identify predictors of walking at 12 months.ResultsSensitivity analysis using different imputation approaches, and spatial contextual effects, were compared. All the imputation methods showed social life and income were significant predictors of walking, however, the complete data approach was the best model indicating Age (1.04, 95% OR: 1.00, 1.08), Social Life (0.83, 95% OR: 0.69, 0.98) and Income <$10,000 (0.10, 95% OR: 0.01, 0.97) were all predictors of walking.ConclusionsThe complete data approach was the best model of predictors of walking in African Americans.

• Imputational modeling tested spatial proximity, crime and social life on walking.
• Hierarchical Bayesian modeling across all the imputation methods were analyzed.
• The analyses of complete data showed age, social life and income predicted walking.
• Lower social life and lower income led to more trail use in underserved communities.