Supplementary tables.

<div><p>Background</p><p>Sexual orientation has been measured in a wide variety of ways which reflect both theoretical and practical considerations. However, choice of sexual orientation measure and recoding strategy can impact analytic sample, as well as demographic and health profiles, in analyses of sexual minority populations. We aimed to examine how choice of sexual orientation dimension and recoding decisions impact estimates in the sexual minority population in two population-based studies in the UK.</p><p>Methods</p><p>We used data collected at age 17 (2018) in the UK Millennium Cohort Study and at sweep six (2012–13) and eight (2017–18) of the English Longitudinal Study of Ageing. Descriptive statistics were used to examine the impact of choice of sexual orientation dimension (i.e. identity, attraction and experience) and recoding decisions on achieved analytic sample and composition by selected demographic and health measures within and between datasets.</p><p>Results</p><p>Dimension choice and recoding decisions resulted in variation in analytic sample. For example, more respondents reported some same-sex sexual attraction than reported a non-heterosexual identity (adolescents: 20.77% vs 8.97%, older adults: 4.77% vs 1.04%). Demographic distributions varied, but not substantially by dimension choice or recoding strategy. Overall, in both datasets sexual minority respondents were more likely to be White and in the highest quintiles for income and education than heterosexual respondents. Health status did not vary substantially by dimension choice or recoding strategy, however sexual minority respondents reported worse health than their heterosexual peers.</p><p>Conclusions</p><p>This study explores a range of practical and theoretical considerations when analysing sexual minority respondents using survey data. We highlight the impact recoding decisions may have on the numbers of sexual minority respondents identified within a dataset and demographic and health distributions in this understudied population. We also demonstrate the benefits of including multiple dimensions for capturing mechanisms of interest in elucidating ambiguous responses and exploring sexual diversity.</p></div>