Background Little is known aboutthe mental health of gay men and lesbians living in Europe.
Aims To compare psychological status, quality of life and use of mental health services by lesbians and gay men with heterosexual people.
Method Cross-sectional study in England and Wales using ‘ snowball’ sampling.
Results Participants: 656 gay men, 505 heterosexual men, 430 lesbians and 588 heterosexual women. Gay men were more likely than heterosexual men to score above threshold on the Clinical Interview Schedule, indicating greater levels of psychological distress (RR 1.24, 95% CI 1.07–1.43), as were lesbians compared with heterosexual women (RR 1.30, 95% CI 1.11–1.52). Gay men and lesbians were more likely than heterosexuals to have consulted a mental health professional in the past, deliberately harmed themselves and used recreational drugs. Lesbians were more likely to have experienced verbal and physical intimidation and to consume more alcohol than heterosexual women.
Conclusions Awareness of mental health issues for gay men and lesbians should become a standard part of training for mental health professionals, who need to be aware of the potential for substance misuse and self-harm in this group and of the discrimination experienced by many lesbians.
Lesbians and gay men experience intolerance of their sexuality, discrimination and victimisation (Mays & Cochran, 2000) and appear to have higher rates of anxiety, depression, substance use disorders and suicidal behaviour than heterosexual men and women (Hershberger & D'Augelli, 1995; Fergusson et al, 1999; Lock & Steiner, 1999; Cochran & Mays, 2000). Almost all research is North American and there are few data for Europe. Furthermore, many studies did not include a comparison group (Atkinson et al, 1988; Coyle, 1993; Pillard, 1998), employed convenience samples, applied unusual definitions of homosexuality (Fergusson et al, 1999), or failed to differentiate between findings in gay and bisexual people.
Five per cent of the British population is estimated to be gay or lesbian (Johnson et al, 2001), but recruiting a probabilistic sample is unlikely to achieve sufficient numbers (Gilman et al, 2001) or candid responses. Probabilistic sampling in predominantly gay and lesbian neighbourhoods has been used in the USA, but it is limited because most gay people do not live in such neighbourhoods. In England and Wales they live more commonly in urban areas (Johnson et al, 2001) but gay neighbourhoods are unusual. ‘Snowball’ sampling is a useful strategy when no adequate sampling frame exists and the target population is dispersed (Gilbert, 1993).
Our hypothesis was that gay men and lesbians in England and Wales differed from heterosexual men and women in terms of mental health, quality of life and experiences of mental health services.
We conducted a cross-sectional survey of gay and lesbian and heterosexual people living in England and Wales, identified by snowball sampling.
Identification of gay and lesbian people
We asked participants to indicate their sexual identity as gay or lesbian (homosexual), bisexual or straight (heterosexual), a choice that determined the group in which they were analysed. We also used a six-point scale based on the Kinsey scales to assess same-sex attraction (McWhirter et al, 1990). A score of 1 indicated attraction only to the opposite sex; 2, mostly to the opposite sex; 3, both sexes equally; 4, mostly same sex; 5, same sex only; 6 indicated where respondents were unsure.
Snowball sampling begins by identifying participants through advertising or direct contacts and asking each participant to recruit others. It is able to reach individuals who do not visit gay or lesbian venues or who might not respond to advertisements. In order to reach a wide range of first contacts we recruited participants by:
placing posters and notices in health clubs and public libraries;
advertising in the national, local and gay press;
leaving postcards about the study in gay venues;
mailing gay and lesbian societies;
placing advertisements on gay and lesbian websites.
The research was described as the ‘National Well Being Study’ or ‘Sexuality and Well Being Study’ and included sexuality as one of the factors being explored. We gave three postcards advertising the study to each first-wave participant to pass on to other people who would be prepared to take part. No requirement was made about sexuality and thus participants identified their sexual orientation only during the interview. The only inclusion criterion was age 16 years and over. Our main outcome was psychiatric status as measured by the revised Clinical Interview Schedule (Lewis et al, 1988), a structured assessment used to assess mental health in two successive national surveys in Britain (Meltzer et al, 1995; Singleton et al, 2000). The interview enquires about the presence and severity of 14 non-psychotic psychiatric symptoms during the week prior to interview: somatic complaints associated with low mood or anxiety; fatigue; problems with memory and/or concentration; sleep disturbance; irritability; worry about physical health; depressed mood; depressive thoughts; non-health-related worry; generalised anxiety; phobic anxiety; panic attacks; compulsive behaviours; and obsessional thoughts. It may be analysed as a continuous score, along a single continuum of severity, or as a dichotomous variable (case threshold ≥12) (Lewis et al, 1988).
After collecting standard demographic information, we asked participants to complete the following measures:
the 12-item General Health Questionnaire (Goldberg & Williams, 1988); possible scores range from 0 to 12 and a threshold indicating significant psychological distress of 3/4 was used;
the 12-item Short Form measure of quality of life (Ware et al, 1996), which provides scores for physical and emotional health (higher scores indicate poorer quality of life);
a brief version of the Social Support Questionnaire (Sarason et al, 1987), in which we used scores for satisfaction with support ranging from 1 (very satisfied) to 6 (very dissatisfied); a threshold score of 2 or more indicated dissatisfaction with social support;
the Alcohol Use Disorders Identification Test (AUDIT), which was validated for use in the community by the World Health Organization (Barbor et al, 1989); we used the recommended threshold of over 8 to identify hazardous drinking;
other questions covered recreational drug use in the preceding month and lifetime; cigarette smoking; previous contacts with mental health professionals and general practitioners; experiences of verbal or physical violence and bullying at school.
The interview was programmed into a laptop computer and took approximately 60 min to complete. Computerised questions have advantages over face-to-face interviews for the collection of potentially embarrassing data (Millsten & Irwin, 1983; Turner et al, 1998). Informed consent was obtained in writing and each participant received £ 10 to offset the costs of taking part. The study was approved by the Royal Free Hospital ethical practices subcommittee.
Sample size and analysis
Although our main outcome was the threshold score on the Clinical Interview Schedule, we focused on the General Health Questionnaire in our power calculations as a precaution against the possibility that many participants might agree to complete a postal questionnaire rather than a computerised interview. We expected 25% of heterosexual participants to score above the threshold of the General Health Questionnaire (Goldberg & Huxley, 1992). To detect a difference of at least ±5% (i.e. 20% or 30%) with the gay and lesbian population at 80% power and α set at 0.5, required 1250 participants in each group. A sample of 1250 gay and lesbian participants provides an estimate of prevalence of psychiatric disorder within 95% confidence limits of ±2.5%.
Data for men and women were kept separate in the analysis. We report medians and interdecile ranges for scores on the rating scales. We used the chi-squared statistic to compare proportions, and the Mann–Whitney U-test and the t-test for analysis of non-parametric and parametric continuous variables. Where each test is applied is indicated by reporting of means (s.d.) or medians (interdecile range). We first determined relative risks for scoring above the usual threshold of 11/12 on the Clinical Interview Schedule in the gay and lesbian group compared with the heterosexual group and adjusted them for age. We then used odds ratios to adjust for groups of variables that might confound or mediate this relationship. Adjusting for such factors requires use of logistic regression and odds ratios. The blocks of variables entered were as follows: demographic variables were age, employment v. unemployment, White v. other ethnicity, having a partner, living alone or with others, and recruitment in the first v. subsequent waves; health and lifestyle factors were the Short Form 12 physical scale score, scoring 2 or more on the Social Support Questionnaire (indicating dissatisfaction), scoring 8 or more on the AUDIT questionnaire (indicating hazardous drinking) and having used recreational drugs in the preceding month; discrimination factors were reports of physical attack, property damage or verbal insults in the preceding 5 years, and verbal or physical bullying at school. The data were analysed using Stata version 7.
We collected data between September 2000 and July 2002. Response rates cannot be reported with snowball sampling. In total 2430 people – 1268 (52%) men, 1149 (47%) women and 13 (0.5%) transgendered people – agreed to participate. There were 505 heterosexual men, 656 gay men, 588 heterosexual women, 430 lesbians, 85 bisexual men and 113 bisexual women; 24 men and 29 women could not describe their sexuality. The results of the 251 participants who identified themselves as bisexual, transgendered or ‘other’ will be reported in a future paper. The following results relate to the main sample of 2179 participants.
Self-identification as gay, lesbian or straight (heterosexual) accorded closely with reported sexual attraction on the adapted Kinsey scale. Almost all the heterosexual men (494; 99%) were attracted principally to women, 596 (95%) of gay men to other men, 573 (98%) of heterosexual women to men, and 408 (96%) of lesbians to other women. (Small variations in totals are due to missing data for 4 heterosexual men, 27 gay men, 2 heterosexual women and 6 lesbians.)
Six snowball waves of recruitment were obtained, with 841 (40%) recruited in the first and 1239 (60%) in the subsequent waves. Gay men and lesbians were more likely than heterosexual men and women to be recruited in the first wave, as were participants who reported they were White and those aged 16–24 years (Table 1). Participants came from a wide area of England and Wales (Table 2).
Heterosexual people were older and less likely to describe their ethnicity as White than their gay and lesbian counterparts, whereas the latter were more likely to report being in employment (Table 3). Although 35–43% of respondents were economically inactive (Table 3), only 111 (22%) heterosexual men, 112 (19.4%) gay men, 68 (11.6%) heterosexual women and 65 (15.3%) lesbians were unemployed, seeking work. Gay men and lesbians were less likely to have a partner and were more likely to live alone than the heterosexual participants.
Psychological and social measures
Homosexual men recorded significantly higher scores (indicating psychological distress) than heterosexual men on the Clinical Interview Schedule, General Health Questionnaire and the mental sub-scale of the Short Form 12 (Table 4). Lesbians recorded higher scores than heterosexual women on the Clinical Interview Schedule, Short Form 12 mental sub-scale and the AUDIT questionnaire (Table 4). Gay and lesbian participants were more likely than heterosexual participants to have used recreational drugs (Table 4). Lesbians were more likely than heterosexual women to report having experienced verbal and physical harassment for whatever reason (Table 4). Reported levels of violence, verbal insults and bullying at school were similar in both groups of men, whereas property offences were reported more often by heterosexual men. Nevertheless, the latter commonly regarded their sexuality as the provocation for such experiences (Table 5). Lesbians were much less likely than gay men to attribute verbal harassment in adulthood or intimidation at school to their sexuality. More than a quarter of gay men and almost a third of lesbians reported that they had ever harmed themselves deliberately, compared with one in seven heterosexual participants; 65% of gay men and 48% of lesbians who reported having harmed themselves cited their sexual orientation as wholly or partly the motive.
Sexuality and psychological distress
The Clinical Interview Schedule was completed by 1134 men (98%) and 999 women (98%). Prevalence of scoring above the standard threshold of 11/12 for common mental disorder was 44% (277) in gay men and 35% (178) in heterosexual men (χ2(1)=8.28, P=0.004). Figures for women were 44% (184) in lesbians and 34% (197) in heterosexual women (χ2(1)=10.34, P=0.001). Thus, gay men and lesbians were at greatest risk of scoring in the higher range on this instrument (Table 6). After stratification for age, this was significant for gay men and lesbians aged 25–34 years. When we explored the relationship between sexual orientation and common mental disorder by adjusting for groups of variables that might confound it, the increased odds for gay and lesbian participants to score above the Clinical Interview Schedule threshold were not affected (Table 7).
Gay men were more likely than heterosexual men (OR 2.9, 95% CI 2.2–3.7) and lesbians were more likely than heterosexual women (OR 2.8, 95% CI 2.1–3.6) to have consulted a mental health professional (see Table 4). This difference was not affected by adjustment for Clinical Interview Schedule score. Gay men were also more likely than their heterosexual counterparts to have sought advice from their general practitioners for emotional difficulties (Table 4).
Sixty-three gay men (10%) and 14 lesbians (3%) had considered seeking treatment to change their sexual orientation, of whom 15 men and 2 women had actually received it. Sixty-two per cent of gay men who had considered seeking help to change their sexual orientation scored above the threshold of the Clinical Interview Schedule compared with 39% of those who had not done so (χ2(1)=9.58, P=0.002; data missing for 3 men). The figures for women were 54% and 42% respectively (P=0.41; data missing for 1 woman).
Gay men and lesbians reported more psychological distress than heterosexual men and women, despite similar levels of social support and quality of physical health. They were also more likely to have used recreational drugs, and lesbians were more likely than heterosexual women to drink excessively. Violence and bullying were more commonly reported by lesbians than heterosexual women, but there were few differences on these factors between the gay and heterosexual men. Bullying at school was reported no more often in gay than heterosexual men, but the gay men who had been bullied regarded their sexual orientation as the main provocation. Gay and lesbian participants were more likely than heterosexual participants to have consulted a mental health professional in the past, regardless of current mental state.
Strengths and limitations of the study
The main strength of our research is the power of the study. No European study in mental health has recruited over a thousand gay and lesbian participants. The main limitation, however, is the method of recruitment. Although there is little alternative to snowball sampling to obtain these numbers, the prevalence of mental disorders was much higher than expected. In the two national studies in Britain that used the Clinical Interview Schedule, prevalence rates were approximately 12% in men and 20% in women (Meltzer et al, 1995; Singleton et al, 2000). Although the most important finding in our study concerns the differential in rates rather than the prevalence in each group, why common mental disorder was more common than expected in the heterosexual participants is not clear. Although providing respondents with a small sum of money for their expenses might have played a part by attracting people experiencing difficult circumstances, the amount given was small. Our sample contained fewer people aged over 55 years than the national population and since this older group has the lowest rate of psychiatric disorder (Meltzer et al, 1995) this might have contributed to the higher rate. It is difficult to see how snowball sampling would have differentially affected prevalence of psychiatric disorder in the gay and heterosexual groups.
It could also be argued that our study might have been more relevant to gay and lesbian participants. However, sexuality was only one of several factors mentioned in our advertisements about the research and thus first-wave gay and lesbian participants should not have been more motivated to take part. We cannot know whether this possible bias occurred in subsequent recruitment waves. However, we had no difficulty recruiting heterosexual participants and thus salience of the study did not seem to affect entry to the study.
Our definition of homosexual or heterosexual is also important. Same-sex attraction and behaviour occur in a social context and we would argue that participants' own view of their sexuality is the most valid construct. In making comparisons between gay and heterosexual people, we were not assuming that the latter constitute a standard of normality. Rather, placing the mental health of gay men and lesbians within the context of the wider population is useful for the purposes of comparison and for service planning.
The increased risk of contemplating suicide and actually harming oneself in gay and bisexual people has previously been reported in the USA (Fergusson et al, 1999; Herrell et al, 1999; Russell & Joyner, 2001) and requires much greater attention, particularly in adolescents (Muehrer, 1995). Although our data cannot identify the reasons for this susceptibility, it would seem to have less to do with confusion about sexuality than confusion about how to express it openly in society (Herdt & Boxer, 1993). No study has examined whether gay and lesbian people have elevated rates of completed suicide, but there are indications from medical examiners' reports of suicides in males that this may be the case (Bagley, 1992).
Experiences of violence and verbal abuse as an adult and intimidation at school were reported frequently by both groups of men and women, a finding that underscores the need for comparison groups in studying these risks in gay and lesbian people (Kessler et al, 1999). Reports that gay and lesbian people are vulnerable to such experiences because of their sexuality are often taken at face value and are not judged in the context of the prevalence of such events in society. Nevertheless, our data show that lesbians are at greater risk of verbal or physical violence than heterosexual women. Contrary to other reports, perceived discrimination did not attenuate the association between psychological distress and sexual orientation.
Use of services
Gay men and lesbians are greater users of mental health services in primary and secondary care than heterosexual men and women. We need to know more about the quality of treatment they receive, particularly because mental health professionals may be insensitive or even hostile to their needs (Golding, 1997; Project for Advice Counselling and Education, 1998). Psychoanalysts in particular may continue to hold outdated views on homosexuality and pathology (Bartlett et al, 2001; Friedman & Lilling, 2001). We shall report later on a qualitative sub-study, in which 23 of the gay and lesbian participants provided accounts of their experiences of mental health services.
There are several potential explanations for our findings. It may be that prejudice in society against gay men and lesbians leads to greater psychological distress and higher use of services than in the heterosexual population (Savin-Williams, 1994). It is also possible that gay people might place greater emphasis on the value of psychological therapy in helping them through these hardships. Conversely, gay men and lesbians may have lifestyles that make them vulnerable to psychological disorder. Such lifestyles may include increased use of drugs and alcohol. Until the past 25 years, homosexuality was considered to be a deviation from normal development that was accompanied by psychological symptoms (King & Bartlett, 1999). There is no evidence, however, for other developmental or physical abnormalities in gay men and lesbians that would lend support to the view that it is a developmental error (Bailey, 1999). In a further stage of our study we shall examine predictors of psychological distress within the gay and lesbian sample.
Clinical Implications and Limitations
Gay men and lesbians report more psychological symptoms than heterosexual report people.
Recreational drug use is more common in gay and lesbian people than in the heterosexual population, and lesbians are at greater risk of alcohol problems than heterosexual women.
Gay and lesbian people are greater users of mental health services in primary and secondary care than heterosexual people.
Fewer older people were recruited than we would have expected.
The representativeness of our gay and lesbian sample cannot be estimated as there are no statistics for gay men and lesbians living in England and Wales.
It is not clear why mental disorders were more common in the heterosexual participants than expected.
We thank all the men and women who participated in the study. We also thank Robert Cabaj and Graham Hart, who advised us on the original protocol, and Catherine Chin, Jeanette Copperman, Sophie Corlett, Julienne Dickie, Rita Glynn, Fiona Hill, Frank Keating, Alan McNaught, Margaret Pedlar, Helen Schoenberg, Amy Sheehan and Melba Wilson, who advised on the study and/or served as members of the Study Steering Group. We also thank Athina Bakalexi for her help in data collection.
The study was funded by the Community Fund in collaboration with Mind, the mental health charity.
- Received April 3, 2003.
- Revision received June 26, 2003.
- Accepted July 22, 2003.
- © 2003 Royal College of Psychiatrists