Recording Sexual Orientation in the UK: Pooling Data for Statistical Power

We know that sexual minority health disparities exist, but in the United Kingdom, the research demonstrating disparities in sexual minority health has been dominated by small convenience samples that do not represent clearly defined populations. Recently, UK population health surveys began to include a question on sexual orientation identity that makes available high-quality data. However, very few studies collect sexual orientation within their demographic data.¹ There need to be more, as it is this important, high-quality evidence that can be used to make a political impact and determine policy change.

Studies that collect data on sexual orientation and on health outcomes or behaviors and therefore allow prevalence of to be captured are the United Kingdom national longitudinal cohort study called “Understanding Society” (bit.ly/259UCLb) and several population cross-sectional studies. Data sets can be accessed through the UK Data Service (bit.ly/1Nz5cl3). Participant recruitment by the surveys is through random or stratified random sampling of their target population, which establishes generalizability of findings.

IDENTITY, ATTRACTION, BEHAVIOR

Sexual orientation was recorded in all of these included health surveys, using the standardized wording to capture sexual orientation identity that has been developed by the UK Office of National Statistics.² The sexual orientation identity question asks, “Which of the following options best describes how you think of yourself?” Participants can respond “heterosexual or straight,” “gay or lesbian,” “bisexual,” or “other,” or they can refuse to respond. This question does not measure sexual attraction or sexual behavior. These are different concepts well described in other literature.³ A test of the impact of including the sexual orientation identity question in the Integrated Household Survey (2009–2010), which had a sample size of around 250 000 people, found that overall survey response rates were not affected, demonstrating acceptance of the question.² Moreover, refusal rates in current surveys collecting sexual orientation are very low, indicating that people are happy to answer the question.³ Thus, with the addition of a single question, the health status of lesbian, gay, bisexual, or other (LGBO) people in the United Kingdom is available.

There are national surveys in the United Kingdom that collect data based on sexual behavior, and these find that greater numbers of people select same-sex behavior than a nonheterosexual identity. Ideally, surveys that measure populations across behavior, attraction, and identity would give the best data.⁴

POOLING DATA

By adopting the same standardized sexual orientation identity question in a range of surveys, it is possible to combine data across years or across data collection efforts. This is important, because the proportion of people that respond as LGBO in the surveys that collect sexual orientation is relatively small. In a recent pooling of 12 UK surveys, with a resulting sample of 94 818 participants, 1.1% identified as lesbian or gay, 0.9% as bisexual, 0.8% as “other,” and 97.2% as heterosexual¹; these are similar to rates found by the Integrated Household Survey.² Pooling data across surveys allows for greater numbers of participants in each sexual orientation identity category.⁵ Moreover, it is possible, through larger sample sizes, to carry out subgroup analysis. The statistical technique adopted—that is, logistic regression with a random effect—acknowledges that data are drawn from different studies. The analytic sample for pooled meta-analysis comprises the study population, with available data on health outcome or behavior (harmonized across studies to ensure comparability), sexual orientation, and covariates associated with sexual orientation identity. Similar analyses have been carried out, for example, on data pooled from the California Health Interview Survey (2001, 2003, 2005, and 2007) and from the Youth Risk Behavior Survey (2005 and 2007). This technique of pooling different study samples addresses the problems of low statistical power frequently seen in studies using sexual orientation identity categories.

Larger sample sizes can increase reliability in estimate reporting and allow greater power in analyses, making it possible to look at subgroup comparisons among sexual minorities and intersections within the data set, such as differences by ethnicity, socioeconomic status, educational attainment, or age. Commonly, because of low response rates, studies are forced to combine response categories, forming a “nonheterosexual” grouping to provide greater statistical power to detect differences compared with a heterosexual reference group. However, LGBO people are not one homogenous group, and a nonheterosexual grouping can conceal meaningful differences in these groups. We know that gay and bisexual men have a higher risk of suicide than lesbian and bisexual women,⁶ that lesbian and bisexual women have higher substance misuse than gay and bisexual men,⁶ and that bisexual women have higher depressive symptoms and perceived stress than women reporting only same-sex attraction. Moreover, this grouping eradicates any separate understanding of LGBO health and prevents any analysis of the category of sex in this group. Finally, we do not know if “other” is selected as a gender identity category or as a political rejection of sexual orientation categories.

COLLECTING SEXUAL ORIENTATION IDENTITY

Nondisclosure of sexual orientation within health care settings is well documented, resulting in reduced well-being and delayed presentation for treatment. Disclosure has been shown to be related to improved well-being. Population surveys that treat all respondents as heterosexual will not only tell us nothing about LGBO people but will be less likely to engage LGBO respondents. Low response rates to this question reflect understandable anxieties about misuse of data and fear of homophobia as a result of disclosure. The routine collection of sexual orientation identity in population surveys would make such data available as part of existing, regularly collected demographics; this would lead to increased confidence and engagement by participants in sexual orientation identity data provision and would provide a more representative data set.

BEYOND SMALL CONVENIENCE SAMPLES

The importance of recording sexual orientation cannot be overstated. Mental and physical health disparities continue to be experienced by this population, often manifested by internalization of chronic stress as a result of structural stigma, prejudice, and discrimination.⁷ The need to monitor health outcomes and health (risk) behaviors in this population is crucial as health disparities persist over time, which is demonstrated by recent studies showing the same effects as older research. Moreover, such disparities not only emerge early in adolescence, but they persist into adulthood. Routine data collection allows the possibility of early detection of health issues, a fulsome understanding of health disparities in this population, a better-informed, deeper engagement with assessment of life-course risks, and the possibility for health patterns to be explored.

The inclusion of a sexual orientation question in some national health surveys was a huge advance. To gain a comprehensive understanding of LGBO health, it is important that collection of sexual orientation be continued. Furthermore, the measurement of gender identity also needs to be adopted within these studies. Inclusion of sexual orientation and gender identity across the research and health surveillance landscape would allow significant improvement in our understanding, increase our knowledge about the health needs of sexual and gender minorities, and allow targeted interventions to address the health disparities experienced. Efforts to remove sexual orientation from surveys that currently collect it, such as the 2017 National Survey of Older Americans Act Participants (NSOAAP), would be a significant backward step. Indeed, to remove sexual orientation data from the few health surveys that collect it would be oppression by omission.