Comparing in-person and online recruited respondents in the national transgender discrimination survey: Implications of data collection methods for transgender health research in the u.s

Background: One of the most critical decisions in research design is how to gather data in order to accurately characterize a group of people and the health issues that affect them. In the absence of probability samples of transgender people, methods are needed that yield convenience samples that reflect the diversity of the transgender population; however, research in transgender health has not explicitly considered setting where people are sampled and survey modality. Method: The National Transgender Discrimination Survey (n=6,456) is a large, non-probability sample of transgender adults age 18-89. We examined characteristics of respondents who were purposively sampled in 2008 and completed a one-time survey through community-based linkages via in-person (n=435) or online (n=6,021) methods. Missing data were multiply imputed and multivariable logistic regression models tested for differences in sociodemographics and health indicators by data collection mode. Results: A higher proportion of in-person respondents were young, male-to-female vs female-to-male, racial/ethnic minority, lower income, lower educational attainment, and publicly insured than online respondents (all p<0.05). In-person respondents were more likely to be current daily smokers, endorse substance use to cope with mistreatment, and self-report as HIV-infected than online respondents (all p<0.05). Conclusions: Findings indicate the need to match data collection methods with relevant research questions and health indicators. Achieving diverse samples of transgender adults requires diverse settings and multiple survey modalities. Research employing probability sampling methods is needed to ascertain the distribution of high/low risks among the transgender population, allowing investigators to inform their outreach strategies and recruit representative convenience samples.