TABLE 3—
Adjusted Multivariate Generalized Estimating Equations Models With Blocked Models Fit to Examine Variables Across Each Domain: Virginia Transgender Health Initiative Study, September 2005–July 2006
| Model | AOR (95% CI) | P |
| 1: Sociodemographics | ||
| Rural | 0.63 (0.32, 1.25) | .186 |
| Suburban | 0.63 (0.54, 0.72) | < .001 |
| FTM spectrum | 1.40 (1.17, 1.68) | < .001 |
| Racial/ethnic minority | 1.15 (1.08, 1.22) | < .001 |
| Low education | 1.55 (1.12, 2.15) | .008 |
| Low income (100% low-income level < $16 999) | 1.36 (1.31, 1.41) | < .001 |
| No health insurance | 1.31 (1.22, 1.39) | < .001 |
| 2: Gender transition | ||
| Live full-time | 1.85 (1.70, 2.02) | < .001 |
| Age first aware transgender | 0.97 (0.94, 0.99) | .004 |
| Hormones (estrogen or testosterone) | 2.18 (1.72, 2.76) | < .001 |
| 3: Health care needs | ||
| Hormonal therapy | 1.72 (1.66, 1.78) | < .001 |
| Transgender-related surgery | 1.10 (0.95, 1.28) | .201 |
| Counseling or psychotherapy | 1.54 (1.00, 2.37) | .049 |
| 4: Violence | ||
| Sexual violence | 2.22 (1.75, 2.81) | < .001 |
| Physical violence | 3.20 (1.83, 5.60) | < .001 |
| 5: Substance use health behaviors | ||
| Tobacco problem ever | 1.33 (1.04, 1.71) | .025 |
| Past or current drinking problem (alcohol) | 1.30 (1.24, 1.37) | < .001 |
| 6: Interpersonal factors | ||
| Family not supportive | 1.45 (1.22, 1.71) | < .001 |
| Experienced hostility in high school | 2.05 (1.53, 2.76) | < .001 |
| Community connectedness | 1.27 (1.19, 1.36) | < .001 |
Note. AOR = adjusted odds ratio; CI = confidence interval; FTM = female-to-male. We used generalized estimating equations to fit multivariate logistic regression models, adjusted for clustering induced by study design (design covariate: Web vs paper-and-pencil survey). Models 2–6 included the following sociodemographic covariates: geographic context (rural, suburban, with urban as the referent), gender vector (FTM vs male-to-female), race/ethnicity (racial/ethnic minority vs not), education (low vs high education), income (100% low-income level vs not), and health insurance (no health insurance vs insured). In models 2–6, sociodemographic factors remained statistically significant as follows: Model 2, gender transition: suburban, FTM, racial/ethnic minority, education, income, no health insurance. Model 3, health care needs: suburban, FTM, racial/ethnic minority, education, and income. Model 4, violence: suburban, FTM, racial/ethnic minority, income, and no health insurance. Model 5, health behaviors: suburban, FTM, racial/ethnic minority, education, and income. Model 6, family, school, community: suburban, racial/ethnic minority, education, income, and no health insurance.