Table 3.

Multivariable GEE Regression Models for Repeated Binary Outcomes.

	Alcohol			Tobacco			Marijuana
	OR	95% CI	p-value	OR	95% CI	p-value	OR	95% CI	p-value
Year
Freshman (vs. Junior)	0.57	0.39–0.83	0.0031	1.23	0.97–1.56	0.0881	0.79	0.63–0.99	0.0457
Sophomore (vs. Junior)	0.69	0.48–1.03	0.0670	1.16	0.91–1.48	0.2237	1.01	0.84–1.21	0.9407
Rural/Urban Status	2.21	1.23–3.95	0.0080	0.86	0.54–1.38	0.5310	1.43	0.94–2.19	0.0964
Age	1.41	0.88–2.25	0.1511	0.89	0.67–1.21	0.4704	0.89	0.71–1.13	0.3572
Female	1.16	0.68–1.98	0.5925	0.67	0.45–0.99	0.0466	0.67	0.48–0.95	0.0227
White	2.42	1.33–4.39	0.0040	0.19	0.71–2.03	0.5055	0.92	0.61–1.41	0.7123

Note. Urban/rural x time interaction effects were not statistically significant. OR = Odds Ratio. A generalized estimating equations (GEE) marginal model (with exchangeable covariance structure) was applied for repeated binary measures in SAS PROC GLIMMIX to estimate rural/urban status, time, and their interaction effects while adjusting for demographics. This approach allowed us to make use of all available observations over time. All associations were considered significant at the alpha level of 0.05.