Nonlinear relationship between nonresidential destinations and body mass index across a wide range of development

Destination accessibility is an important measure of the built environment that is associated with active transport and body mass index (BMI). In higher density settings, an inverse association has been consistently found, but in lower density settings, findings are limited. We previously found a positive relationship between the density of nonresidential destinations (NRD) and BMI in a low-density state. We sought to test the generalizability of this unexpected finding using data from six other states that include a broader range of settlement densities.

We obtained the address, height, and weight of 16.9 million residents with a driver's license or state identification cards, as well as the location of 3.8 million NRDs in Washington, Oregon, Texas, Illinois, Michigan, and Maine from Dun & Bradstreet. We tested the association between NRDs∙ha^-1 within 1 km of the home address, and self-reported BMI (kg∙m^-2). Visualization by locally-weighted smoothing curves (LOWESS) revealed an inverted U-shape. A multivariable piecewise regression with a random intercept for state was used to assess the relationship.

After accounting for age, sex, year of issue, and census tract social and economic variables, BMI correlated positively with NRDs in the low-to-mid density stratum (β = +0.005 kg∙m^-2/nonresidential building∙ha^-1; 95% CI: +0.004,+0.006) and negatively in the mid-to-high density stratum (β = -0.002; 95% CI: -0.004,-0.0003); a significant difference in slopes (P < 0.001).

BMI peaked in the middle density, with lower values in both the low and high-density extremes. These results suggest that the mechanisms by which NRDs are associated with obesity may differ by density level.