Household income, active travel, and their interacting impact on body mass index in a sample of urban Canadians: a Bayesian spatial analysis.

BACKGROUND

Active travel for utilitarian purposes contributes to total physical activity and may help counter the obesity epidemic. However, the evidence linking active travel and individual-level body weight is equivocal. Statistical modeling that accounts for spatial autocorrelation and unmeasured spatial predictors has not yet used to explore whether the health benefits of active travel are shared equally across socioeconomic groups.

METHODS

Bayesian hierarchical models with spatial random effects were developed using travel survey data from Saskatoon, Canada (N = 4625). Differences in log-transformed body mass index (BMI) were estimated for levels of active travel use (vehicular travel only, mixed vehicular/active travel, and active travel only), household income, and neighbourhood deprivation after controlling for sociodemographic and physical activity variables. The modifying effect of household income on the association between active travel and BMI was also evaluated.

RESULTS

Significant and meaningful decreases in BMI were observed for mixed (β = - 0.02, CrI - 0.036 to - 0.004) and active only (β = - 0.043, CrI - 0.06 to - 0.025) compared to vehicular only travelers. BMI was significantly associated with levels of household income and neighbourhood deprivation. Accounting for the interaction between travel mode and household income, decreases in BMI were observed for active only compared to vehicular only travellers in the highest income category (β = - 0.061, CrI - 0.115 to - 0.007).

CONCLUSION

Strategies to increase active travel use can support healthy weight loss and maintenance, but the opportunity to benefit from active travel use may be limited by low income. Considerations should be given to how interventions to increase active transportation might exacerbate social inequalities in BMI. Spatial statistical models are needed to account for unmeasured but spatially structured neighbourhood factors.