Methodologic Concerns in Defining Lesbian for Health Research

Brogan, Donna1; Frank, Erica2; Elon, Lisa1; O’Hanlan, Katherine A.3

Original Articles

A recent report from the Institute of Medicine recommends more methodologic and substantive research on the health of lesbians. This study addresses one methodologic topic identified in the Institute of Medicine report and by a subsequent scientific workshop on lesbian health: the definition and assessment of sexual orientation among women. Data are from the Women Physicians’ Health Study, a questionnaire-based U.S. probability sample survey (N = 4,501). The two items on sexual orientation (current self-identity and current sexual behavior) had a high response rate (96%), and cross-tabulation of responses indicated several combinations of identity and behavior. Three conceptually different definitions of “lesbian” are compared on the basis of (1) identity only, (2) sexual behavior only, and (3) both identity and sexual behavior. Suggestions and cautions are given to researchers who will add items on sexual orientation to new or ongoing research on women’s health.

Author Information

From the 1Rollins School of Public Health and the 2School of Medicine, Emory University, Atlanta, GA, and 3Stanford Medical Center, Palo Alto, CA.

Address correspondence to: Donna Brogan, Biostatistics Department, Rollins School of Public Health, 1518 Clifton Road NE, Emory University, Atlanta, GA 30322.

Supported by the Lesbian Health Fund of the Gay and Lesbian Medical Association (to E. F.).

Submitted December 6, 1999; final version accepted June 6, 2000.

Article Outline

Research on the health of lesbians is starting to receive recognition among mainstream health scientists. The recently published Institute of Medicine (IOM) report, “Lesbian Health: Current Assessment and Directions for the Future,”1 suggests that lesbians may have a higher prevalence of some health risks and disorders. One of the issues highlighted by the IOM report is methodologic: the definition and measurement of “lesbian.”

Lesbian health research has used varied and sometimes unclear definitions of the construct “sexual orientation.”2,3 Earlier studies simply assumed that any subject who participated in a study about lesbians was a lesbian. 4,5 Lesbians also have been defined as women whose emotional, social, and/or sexual relationships are primarily with other women (regardless of public identity) or women who partner with women. 1,6 Consensus is building that sexual orientation can be defined along three dimensions 1,7 : self-identification, sexual behavior, and sexual attraction or desire. Each of these dimensions can be assessed for a specific time frame, for example, current/recent or lifetime, because each dimension may vary over time. 6,7

First and foremost, sexual orientation questions should be determined by the study outcome(s). Current or recent sexual behavior is a crucial dimension if the study outcomes include sexually transmitted diseases or pregnancy risk; in that case, self-identification and attraction or desire are irrelevant or secondary. On the other hand, current, recent, or past self-identification can be important for physical and mental health outcomes that relate to social factors, such as access to and utilization of health services.

Studies restricted to lesbian participants have used a variety of subject-recruitment techniques, with questionnaires distributed at women’s festivals or bookstores or through gay/lesbian newspapers and journals as well as through lesbian social or political organizations. 1,4–6 Some recent studies of women’s health have used specific recruitment activities to include lesbian subjects, for example, the Women’s Health Initiative (WHI) 8 (D. Bowen, Cancer and lesbians, presented at the IOM Workshop on Lesbian Health Research Priorities, Washington, DC, 1997). Some ongoing longitudinal studies of women’s health have added a question on sexual orientation, for example, the Nurses’ Health Study II 9 (NHS-2) (P. Case, Disclosure of sexual orientation in the Nurses’ Health Study II, presented at the IOM Workshop on Lesbian Health Research Priorities, 1997). Large-scale population-based health surveys typically have not included questions on sexual orientation, although such questions appeared on the National Health and Nutrition Examination Survey (NHANES) in 1999 and 2000 (R. Kington, The experience of NCHS/CDC with sexual orientation survey questions, presented at the Scientific Workshop on Lesbian Health, Washington, DC, 2000, which was cosponsored by the Department of Health and Human Services with other federal agencies and organizations).

Data-collection mechanisms for sexual orientation have only occasionally recognized that women may require some degree of confidentiality to disclose a lesbian orientation. 10 Recent survey methodologic research on disclosure of sensitive information has concentrated on sexual behavior and substance abuse 11,12 and has not focused on lesbians. Because of social stigmatization, misclassification errors in research seem likely to favor heterosexual assignment of lesbians, rather than vice versa.

This paper presents the response rate and congruence for two dimensions of sexual orientation assessed in the Women Physicians’ Health Study 13–15 (WPHS) and compares three definitions of lesbian. We give suggestions for including sexual orientation items in women’s health research so that lesbians can be compared with other women on health status, behavior, and risk factors.

Back to Top | Article Outline
Subjects and Methods

WPHS 13–15 used a stratified random sample (by decade of medical school graduation) of women 30–70 years of age with a doctor of medicine degree awarded during 1950–1989, not in residency training, and residing in the United States or its territories in September 1993. The sampling frame was based on the Physician Masterfile of the American Medical Association. A self-administered questionnaire was mailed up to four times, and responses were accepted until October 1994 (N = 4,501). There was a 59% response rate among physicians eligible to participate.

The 73 questions covered medical training and practice and health status, history, and behavior. The reproductive history section assessed two of the three IOM-identified 1 dimensions of sexual orientation, sexual behavior and self-identification:

Question 48. Are you now sexually active with: men, women, both, or neither?

Question 49. Do you now self-identify as: heterosexual, bisexual, lesbian/gay/homosexual, or other?

Respondents gave their own interpretation to the phrase “now sexually active” in question 48 and to the self-identify options in question 49.

We compare three definitions of “lesbian” and “heterosexual” (comparison group). The first two definitions, identity only (question 49) and sexual behavior only (question 48), are based on the response to each item. The third definition uses both identity and behavior. Lesbian is defined as self-identified as such in question 49 or having reported sexual activity with “women” in question 48; heterosexual is defined as: (1) self-identified as heterosexual in question 49 and did not report sexual activity with “women” or with “both” men and women in question 48 or (2) self-identified as “other” or nonrespondent to question 49 and reported sexual activity with “men” in question 48.

We used unweighted analyses to describe the sample’s responses to the two items as well as to compare definitions of lesbian. “Coupled” is defined as married or a member of an unmarried couple; “uncoupled” is defined as never married, widowed, or separated/divorced.

Back to Top | Article Outline


The item nonresponse rate was 3.6% for both sexual orientation items (Table 1). Almost all respondents (95.5%) answered both items; 2.7% answered neither item, and 1.8% answered one item but not the other. This rate is consistent with nonresponse to other WPHS items (Table 2): approximately 2% for nonsensitive items, 3% for slightly personal questions, 5% for human immunodeficiency virus-related items, but 20% for personal income.

Most respondents (92.6%) self-identified as heterosexual, with 2.0% as lesbian and 1.2% as bisexual (Table 1). Sexual activity was reported with men only by 71.8%, with women only by 2.1% and with neither men nor women by 22.3%. Three-fourths of the cells in Table 1 contain respondents, indicating several combinations of answers to the two items and a potential limitation of measuring only one dimension of sexual orientation. About three-fourths (77%) of self-identified heterosexual respondents reported sexual activity with men only; a similar percentage (79%) of self-identified lesbian respondents reported sexual activity with women only. Virtually all respondents who self-identified as lesbian reported sexual activity with women only or with neither men nor women (89/90). Among respondents who reported sexual activity with women only, 74% self-identified as lesbian, 15% self-identified as bisexual, and 11% self-identified as heterosexual.

Sample sizes vary for comparative analyses of lesbians and heterosexual women across the three definitions of lesbian/heterosexual (Table 3). The sample size for the lesbian group ranges from 90 to 115, with the larger sample size for the combined definition based on identity and behavior. One-fourth of respondents cannot be classified using only sexual behavior, because a current or recent time frame was used. A few subjects cannot be classified because of missing data for one or both items.

The 25 additional respondents classified as lesbian by using the combined identity and behavior definition, compared with the identity definition, reported current sex with women but did not self-identify as lesbian (14 as bisexual and 11 as heterosexual). These 25 respondents differ from the 90 self-identified lesbian respondents on marital status (Table 4): they are more likely to be ever married or currently married. These two respondent groups of lesbians do not differ much by age, race, ever having been pregnant, parity, being currently coupled, number of children (biological and other), political affiliation, or self-rated health (Table 4).

Back to Top | Article Outline


The rate of subject nonresponse to the WPHS survey appears unaffected by sexual orientation questions, because it is comparable with that of other physician surveys without such questions. 13 Low item-nonresponse rates on sexual orientation dimensions, as observed in WPHS, have also been reported in two large longitudinal studies of women’s health that use self-administered questionnaires: the NHS-2 (N = 117,000) and the WHI (N = 170,000).

In the NHS-2 in 1995, more than 99% of female respondents reported that they were heterosexual, bisexual, or lesbian in response to the following item (P. Case, 1997), without implying current sexual activity:

Whether you are currently sexually active or not, what is your sexual identity or orientation? Please choose one answer: (a) heterosexual; (b) bisexual; (c) lesbian, gay, or homosexual; (d) none of the above; or (e) prefer not to answer.

The following WHI sexual behavior questions, placed at the end of the baseline self-administered questionnaire, had an item nonresponse rate of 3.2% (D. Bowen, 1997) and assess one dimension, sexual behavior, for two time frames (lifetime and after 45 years of age):

Question 125. Regardless of whether you are currently sexually active, which response best describes who you have had sex with over your adult lifetime? (1) Have never had sex. (2) Sex with a woman or with women. (3) Sex with a man or with men. (4) Sex with both men and women. (5) Prefer not to answer.

If answer 4 is given, then the next question is asked:

Question 125.1. Which response best describes who you have had sex with after 45 years of age? [Choose from options] 2 through 5 above.

NHANES III 16 included items for adult males, but not females, to 60 years of age on gender of lifetime sexual partners. NHANES 1999 included these sexual behavior questions for two recent time frames (last 12 months and last 30 days), as well as lifetime for both male and female adults to 60 years of age, and NHANES 2000 recently added a question on current self-identification (R. Kington, 2000).

Although these particular studies are not the only ones to measure sexual orientation among women, they have substantial experience to offer and show that the inclusion of sexual orientation items is feasible both in cross-sectional population-based studies and in longitudinal studies. These studies also demonstrate that researchers must specify or imply a relevant time frame when evaluating dimensions of sexual orientation. WPHS used the word “now” to assess current or recent behavior and self-identification. It is possible, although very unlikely, that “now” could be interpreted as instantaneously to completing the questionnaire, likely precluding sexual activity with anyone.

Virtually all WPHS respondents who self-identified as lesbian reported sexual activity with women only or with no one. Three-fourths of WPHS respondents who reported sexual activity only with women, however, self-identified as lesbian, with the remainder equally likely to identify as heterosexual or bisexual. These findings are consistent with results from Laumann et al7 in their population-based study of 1,749 U.S. women 18–59 years of age and illustrate the desirability of including more than one dimension of sexual orientation in questionnaires. 6 Furthermore, given that 22% of WPHS respondents reported no sexual activity, even though almost all of them indicated an orientation identity, it becomes apparent that assessing only current or recent sexual behavior provides incomplete information to evaluate outcome variables other than sexually transmitted diseases or pregnancy risk.

The three definitions of lesbian and heterosexual in our study differ conceptually and yield somewhat different classifications of subjects for stratified analyses by sexual orientation. Other conceptual dimensions of orientation include attraction or desire 7; membership in lesbian social or political organizations 4; and degree of self-acknowledgment of sexual self-identification (“outness”) to relatives, friends, and co-workers. 4 Another approach 4 views each sexual orientation dimension on a continuum, as in the reports by Kinsey and colleagues, 17,18 contrary to the categorical questionnaire items reviewed above. Morris and Rothblum 4 have shown that several continuous dimensions of sexual orientation among a nonprobability sample of lesbians have only low or moderate correlations.

To compare lesbians with heterosexual women on health status and health behavior (excluding variables directly related to sexual behavior), we generally prefer the more inclusive definition of lesbian involving both behavior and self-identity. This definition includes women who are affected by one or both dimensions of same-sex sexual behavior and lesbian identity and does not imply that those who report sex only with women necessarily self-identify as lesbian, nor vice versa. We believe the behavior and identity dimensions define lesbians more strongly because they are overt statements or commitments to at least one other person (behavior) or to a community (identity) 19 that potentially incur more societal stigma and risk than do simple attraction or desire. We prefer not to include attraction or desire in our combined definition because few significant health or psychosocial outcomes are likely to correlate with this dimension. When investigating stigma or its effects, however, additional aspects of being lesbian may be important to assess, for example, “outness,” overtness, appearance, and social behavior.

We agree with Laumann et al, 7,pp.285–286 who state, “It makes more sense to ask about specific aspects of same-gender behavior, practice and feeling during specific periods of an individual’s life rather than a single yes-or-no question about whether a person is homosexual.” Indeed, most researchers agree that although most persons are raised as heterosexuals, a homosexual orientation can be recognized at any age from childhood on.

With the recent publication of the IOM report, 1 it is likely that more health research studies that focus on or include women, especially population-based surveys, will include items to allow analyses of health data stratified by sexual orientation. This research will confirm whether lesbians have a higher prevalence of some health risks and disorders, as well as of various protective factors. 5

Although researchers may assume that behavioral dimensions are easier to assess, Sanders and Reinisch 20 showed that a sample of college students held widely divergent opinions about the specific behaviors that defined having “had sex.” Many studies, for example, WPHS, NHS-2, WHI, and NHANES, are under pressure to limit the number of questions, so it is important to determine subjects’ interpretation of short phrases such as “had sex with” and “sexually active.” Respondent interpretation of items to assess dimensions of sexual orientation, as well as willingness to disclose a sexual orientation other than heterosexual, is best studied in a survey research cognitive laboratory. 21 The recent Scientific Workshop on Lesbian Health explicitly recommended that cognitive research on sexual orientation items in women’s health research be conducted in a timely manner.

Back to Top | Article Outline


We thank Dr. Suzanne Haynes for review of drafts of the manuscript and the referees for constructive comments.

Back to Top | Article Outline


1. Solarz AL. Lesbian Health: Current Assessment and Directions for the Future. Washington DC: National Academy Press, 1999.
2. Chung YB, Katayama M. Assessment of sexual orientation in lesbian/gay/bisexual studies. J Homosex 1996; 30: 49–62.
3. Gonsiorek JC, Sell RL, Weinrich JD. Definition and measurement of sexual orientation. Suicide Life Threat Behav 1995; 25 suppl: 40–51.
4. Morris JF, Rothblum ED. Who fills out a lesbian questionnaire? The interrelationship of sexual orientation, years “out,” disclosure of sexual orientation, sexual experience with women, and participation in the lesbian community. Psychol Women Q 1999; 23: 537–557.
5. Rothblum ED. “I only read about myself on bathroom walls”: the need for research on the mental health of lesbians and gay men. J Consult Clin Psychol 1994; 62: 213–220.
6. O’Hanlan KA. Lesbian health and homophobia: perspectives for the treating obstetrician/gynecologist. Curr Probl Obstet Gynecol Fertil 1995; 18: 93–136.
7. Laumann EO, Gagon JH, Michael RT, Michaels S. The Social Organization of Sexuality: Sexual Practices in the United States. Chicago: University of Chicago Press, 1994.
8. Mathews KA, Shumaker SA, Bowen DJ, Langer RD, Hunt JR, Kaplan RM, Klesges RC, Ritenbaugh C. Women’s health initiative: Why now? What is it? What’s new? Am Psychol 1998; 52: 101–116.
9. Colditz GA, Manson JE, Hankinson SE. The Nurses’ Health Study: 20-year contribution to the understanding of health among women. J Women’s Health 1997; 6: 49–62.
10. Bradford J, Honnold JA, Ryan C. Disclosure of sexual orientation in survey research on women. J Gay Lesbian Med Assoc 1997; 1: 169–177.
11. Tourangeau R, Smith TW. Asking sensitive questions: the impact of data collection mode, question format and question context. Public Opin Q 1996; 60: 275–304.
12. Turner CF, Forsyth BH, O’Reilly JM, Cooley PC, Smith TK, Rogers SM, Miller HG. Automated self-interviewing and the survey measurement of sensitive behaviors. In: Couper MP, Baker RP, Bethlehem J, Clark CZF, Martin J, Nicholls WIL,II O’Reilly JM, eds. Computer Assisted Survey Information Collection. New York: John Wiley and Sons, 1992.
13. Frank E, Brogan D, Mokdad A, Simoes E, Kahn H, Greenberg R. Health-related behaviors of women physicians versus other women in the United States. Arch Intern Med 1998; 158: 342–348.
14. Frank E, Rothenberg R, Brown WV, Maibach HT. Basic demographic and professional characteristics of U.S. women physicians. West J Med 1997; 166: 179–184.
15. Frank E. The Women Physicians’ Health Study: background, objectives, and methods. J Am Med Women’s Assoc 1995; 50: 64–66.
16. National Health and Nutrition Examination Survey III, 1988–1994. National Center for Health Statistics, CD-ROM, Series 11, No. 1A, ASCII version. Washington DC and Atlanta, GA: U.S. Department of Health and Human Services, Centers for Disease Control and Prevention, 1997.
17. Kinsey A, Pomeroy WB, Martin C. Sexual Behavior in the Human Male. Philadelphia: WB Saunders, 1948.
18. Kinsey AC, Pomeroy WB, Martin CE, Gebbard PH. Sexual Behavior in the Human Female. Philadelphia: WB Saunders, 1953.
19. O’Hanlan KA. Recruitment and retention of lesbians in health research trials. In: Recruitment and Retention of Women in Clinical Studies. NIH Pub. No. 95-3756. Washington DC: U.S. Department of Health and Human Services, Public Health Service, National Institutes of Health, 1995;101–104.
20. Sanders SA, Reinisch JM. Would you say you “had sex” if...? J Am Med Assoc 1999; 281: 275–277.
21. Forsyth BH, Lessler JT. Cognitive laboratory methods: a taxonomy. In: Biemer PP, Groves RM, Lyberg LE, Mathiowetz NA, Sudman S, eds. Measurement Errors in Surveys. New York: John Wiley and Sons, 1991.

homosexuality/female,; homosexuality,; surveys,; sexual orientation,; physicians,; physicians/women.

© 2001 Lippincott Williams & Wilkins, Inc.