Fig. 1.

Adjusted predicted probabilities of diet, food insecurity, and physical activity outcomes among Baltimore public housing residents with and without access to a personal vehicle (full model). Predicted probabilities calculated from results of full logistic regression models, which were adjusted for age, gender, comorbidity risk score, and neighborhood. In models examining the diet outcomes, we also adjusted for food insecurity status and perceived easy transport to healthy food stores. In models examining the physical activity outcome, we also adjusted for perceived easy transport to recreation. Dietary outcomes were estimated from the National Health Interview Survey (NHIS) 5-factor dietary screener (National Cancer Institute, n.d.), where high fruit & vegetable intake was ≥6.7 servings/day and high added sugar intake was ≥39.9 tsp/day, which represent the upper quartile of intakes reported in our sample. For food insecurity, we used the results of a validated 2-item screener (Hager et al., 2010). For physical activity, we used a validated exercise screener (Ainsworth et al., 1993) to determine whether a participant's leisure time activity level dichotomized as ‘active’ if high or moderate and ‘not active’ if low or very low levels. There was a statistically significant difference found in the predicted probabilities of food insecurity (p < 0.01), but no statistically significant difference in the predicted probabilities of high fruit & vegetable intake (p = 0.50), high added sugar intake (p = 0.43), or physical activity (p = 0.19).