question. There was a concern that there would be a significant hang-up rate with these questions, but, in fact, these questions posed no more of a problem than other questions in the survey. An issue that did arise was that respondents would often answer the first question by stating their sexual identity, which created problems for the interviewers and slowed down the interview process. An analysis of the 2001 survey data high-lighted the need for sexual behavior questions, and these were added in 2003. Beginning in 2003, they have asked a single sexual identity question that reads, “Do you think of yourself as straight or heterosexual; as gay, lesbian, or homosexual; or bisexual?” There is also the option of opening a text box and inputting an answer verbatim. In addition, there are two questions on sexual behavior. In cycles since 2003, these questions have been only asked of adults under the age of 70 because of a high rate of misclassification concerning the word “bisexual” in adults over age 70.
Discussing data from CHIS 2009, Grant noted that about 3 percent of men reported that they had exclusively male sex partners in the preceding 12 months, and about 1.6 percent of women reported having exclusively female sex partners in the preceding 12 months. Less than 0.5 percent of both men and women reported having sex partners of both sexes. From the sexual identity questions, about 2.5 percent of men identified as gay or homosexual, and 1.1 percent of women identified as gay, lesbian, or homosexual. One percent of men and 1.7 percent of women identified as being bisexual. Non-response to these questions was low, about 2-3 percent. In contrast, 23 percent of the sample refused to answer questions on household income. Non-response, which combines “don’t know” and refusals, was highest among Cantonese and Spanish speakers. Women who spoke Asian languages—Cantonese, Korean, Mandarin, and Vietnamese—were two to six times less likely to respond than men, which Grant said was unexpected.
In summarizing the lessons learned from over a decade of surveys, Grant said that it is clearly important to collect data on sexual identity in EHRs. “Electronic health records have great potential, especially for trans-gender populations because even in a large survey like CHIS, we’re likely to pick up very few or too few to make much statistical inference about that population,” he explained. He added that the CHIS results show that it is feasible to collect relatively accurate information about sexual identity and behavior in a diverse population, but that the wording of questions is important, as is the setting in which the questions are asked.