Table 3.
Regression of weight status on individual and contextual predictors (N=714,054).
| Variable | Outcome: BMI (kg/m2) Model I β (95% CI) | Outcome: obese, BMI≥30 Model II OR (95% CI) |
|---|---|---|
| Individual predictors | ||
| Age (years) | 0.32 (0.31–0.33)*** | 1.12 (1.11–1.12)*** |
| Female | −1.03 (−1.10 to −0.96)*** | 0.92 (0.89–0.94)*** |
| Race/ethnicity (Ref: White) | ||
| Black | 1.93 (1.82–2.04)*** | 1.83 (1.75–1.93)*** |
| Hispanic | 0.75 (0.57–0.93)*** | 1.22 (1.14–1.30)*** |
| Other | −0.72 (−0.91 to −0.53)*** | 0.74 (0.68–0.81)*** |
| Education (Ref: < HS diploma) | ||
| HS graduate, no college | −0.24 (−0.34 to −0.15)*** | 0.90 (0.85–0.95)*** |
| Some college+ | −0.84 (−0.93 to −0.74)*** | 0.71 (0.68–0.75)*** |
| Income (Ref: Quartile 1) | ||
| Quartile 2 | −0.12 (−0.19 to −0.04)** | 0.95 (0.92–0.98)** |
| Quartile 3 | −0.29 (−0.37 to −0.21)*** | 0.87 (0.85–0.90)*** |
| Quartile 4 | −0.77 (−0.87 to −0.68)*** | 0.72 (0.69–0.75)*** |
| Smoking (Ref: never smoked) | ||
| Current smoker | −0.71 (−0.80 to −0.61)*** | 0.77 (0.74–0.81)*** |
| Former smoker | 0.39 (0.34–0.44)*** | 1.13 (1.11–1.17)*** |
| County-level predictors | ||
| Population size (log) | −0.18 (−0.25 to −0.10)*** | 0.93 (0.91–0.96)*** |
| Median household income | −0.13 (−0.20 to −0.05)** | 0.95 (0.92–0.98)** |
| % HS diploma+ | −0.14 (−0.23 to −0.05)** | 0.96 (0.93–0.99)** |
| Region (Ref: South) | ||
| Northeast | 0.05 (−0.10–0.20) | 1.00 (0.94–1.07) |
| Midwest | 0.45 (0.34–0.57)*** | 1.17 (1.12–1.22)*** |
| West | −0.09 (−0.28–0.11) | 0.96 (0.90–1.03) |
| Intercept | 27.06 (27.00–27.12)*** | 0.31 (0.30–0.31)*** |
| Intercept variance component | 0.18*** | 0.02*** |
*
p<0.05;
**
p<0.01;
***
p<0.001
Note: Model I is a linear model predicting BMI as a continuous measure. Model II is a logistic model predicting the odds of being obese (BMI≥30). All county-level variables are measured continuously and scaled by the value of their inter-quartile range. For the county-level variables, β indicates the predicted change in BMI, kg/m2, (Model I) and the odds ratio for obesity (Model II) between the 25th and 75th percentile of the indicated county-level variable. All models control for BRFSS sample year and a squared-term for age.