Table 3. Logistic regression analyses for HIV testing and high-risk behavior, Monroe County, New York, 2012.

Model	OR	95% CI
Model 1—Ever tested for HIV
Race/ethnicity
Latino (vs White)	1.8	1.0 – 3.3
Black (vs White)	3.8	2.3 – 6.4
Model 2—Offered HIV test in past year
Race/ethnicity
Latino (vs White)	3.1	1.8 – 5.4
Black (vs White)	2.4	1.5 – 3.8
Age 18-34 (vs 35-64)	3.0	2.0 – 4.5
Female (vs male)	2.7	1.7 – 4.0
Routine health checkup in past year	2.5	1.5 – 4.1
Model 3—Any HIV high-risk behavior
Age 18-34 (vs 35-64)	1.7	1.1 – 2.4

OR, odds ratio; CI, confidence interval. Variable selection algorithm based on Bayesian information criterion using the add-on “bicdrop1” module in Stata version 12 (College Station, Texas, 2011). For each model, potential variables included: race/ethnicity; age; sex; residence; educational attainment; marital status; sexual orientation; household poverty level; current employment status; current health insurance status; health insurance status over past two years; routine health checkup in past year; needed but did not seek health care due to cost; self-rated health status; and frequent mental distress in past month. Final point estimates and variances were determined using survey-specific procedures (Stata version 12).