Sampling procedure and participants
Data were collected as part of a larger mixed-methods research project (‘UNderstanding the MEchanisms, NAture, MAgnitude and Impact of Sexual violence in Belgium’; UN-MENAMAIS) that included a cross-sectional online survey administered to a nationally representative sample of persons aged 16 to 69 years in Belgium. The Belgian National Register (BNR), containing demographic information (but not about sexual orientation) on all Belgian residents, served as the sampling frame for two periods of data collection. A random disproportionate stratified sample was drawn from the BNR with the aim to reach an equal number of male and female legal Belgian inhabitants equally divided into three age groups (i.e., 16–24 years old, 25–49 years old, and 50–69 years old). Overrepresentation of certain subgroups (e.g., male and female participants), was post hoc corrected using quota based sampling to obtain estimates representative of the population residing in Belgium (see  for more details).
The online survey was started by 6504 respondents. Respondents were excluded because they either did not give informed consent (n = 706), did not complete the survey (n = 909), did not meet criteria regarding age (i.e., between 16 and 69 years old; n = 6), completed the survey multiple times (n = 37), and because there were concerns about the quality of the responses (n = 1). Respondents who had missing values in key variables (e.g., items on sexual orientation) for this study were excluded as well (n = 213). The total final sample consisted of n = 4632, which corresponds to a response rate of 11.16%.
Questionnaire development and validation
The UN-MENAMAIS survey included questions regarding sexual victimization and perpetration, but also questions on sociodemographic information, on sexuality and gender, mental health, quality of life and resilience, and minority identity which were analyzed for this paper. The initial version of the survey was developed in English by a multidisciplinary research consortium with a background in Health Sciences, Sociology, Psychology, Psychiatry, Criminology, Human Sexuality Studies, and Anthropology. Information about the generation and validation of all measures can be found elsewhere (see [40,41,42]).
The final version of the survey was translated into the three most commonly spoken languages in Belgium (i.e., Dutch, French, and English), and into Arabic, Farsi, and Pashtu which were at the time the three most spoken languages among refugees and applicants for international protection residing in Belgium (see ). The survey was completed 2886 times in Dutch, 1578 times in French, 154 times in English, nine times in Arabic and five times in Farsi. No one completed it in Pashtu.
Assessment of sex, gender, and sexual orientation
Following guidelines on collecting data on sexual orientation and gender identity [4, 44, 45], we used multiple-step questions to assess these variables. First, sex was measured by asking participants to name the sex they were assigned to at birth (male/female; the two only legal possibilities in Belgium). The second step entailed a multiple choice question “how do you describe yourself” allowing to answers as a man/as a woman/as transman/as transwoman/other, namely as …. . When participants chose the option “other, namely as”, they could write down their gender description of preference. Participants who self-identified as trans or other and participants who indicated a sex at birth different from their gender identity, were considered as non-cisgender participants. In this paper we compare findings based on the sex assigned at birth. Analysis based on gender identity falls beyond the scope of this study.
Sexual orientation was measured using multiple items: we asked participants to whom they felt sexually attracted, how they label their sexual orientation, and the gender of their sexual partners. This paper focuses on self-identifying LGB+ persons. The exploration of overlap between sexual attraction, self-labelling and sexual behavior is the focus of another study. To select the relevant subgroups in our sample, we asked to indicate which description applied to them: heterosexual; bisexual; gay/lesbian; pan−/omnisexual; asexual; other, namely …. The options pansexual and omnisexual were combined to limit the number of answer possibilities and the received feedback during the survey validation phase that both terms can be used as synonyms in our local context. Choosing “other, namely …” meant that they could complete their answer with their preferred sexual orientation label. Sexual orientation was recoded into a dummy variable LGB+/heterosexual. Hence, all participants who chose ‘heterosexual’ were labelled ‘heterosexual’. All others were grouped together into ‘LGB+’.
Assessment of minority identity
Participants were asked to indicate whether they considered themselves as belonging to a minority group in Belgium (yes/no) and if so, to indicate in a grid which characteristics (i.e., sexual orientation, gender identity, intersex or DSD condition, religion or life philosophy, skin color, ethnicity, disability, age or another characteristic) defined their minority status. Multiple answers were possible. In this study, we focused on LGB+ participants and their identification with a minority group based on sexual orientation related characteristics. The LGB+ participants were grouped in either the ‘sexual minority’ or the ‘non-sexual minority’ group.
Participants who indicated belonging to any minority group (e.g., sexual minority subgroup), received a binary follow-up question to assess the importance (i.e., important/not important) of each indicated characteristic for their identity.
Social support, substance use, mental health, and well-being
As a global measure of well-being, all participants were asked to rate their quality of life on a five-point Likert scale ranging from 1 = ‘very poor’ to 5 = ‘very good’. Specific mental health aspects were measured in all participants by validated scales from the international literature. Depression was assessed using the 9-item Patient Health Questionnaire (PHQ-9) . Responses were made on a 4-point likert scale ranging from ‘not at all (0)’ to ‘nearly every day (3)’. All items were summed in a final score ranging from 0 to 27, Cronbach’s Alpha = .872. Anxiety was measured by the General Anxiety Disorder (GAD)-7 . The scale had seven items, and responses were made on a four point likert scale ranging from ‘not at all (0)’ to ‘nearly every day (3)’, Cronbach’s Alpha = .890. All items were summed in a final score ranging from 0 to 21 to yield a total anxiety score. Both scales assessed symptoms in the 2 weeks prior to filling in the survey and both used a cut-off score of five as a positive screening for depression and/or anxiety [46, 47].
Posttraumatic Stress Disorder (PTSD) was measured using the PC-PTSD-5, which questioned symptoms in the month before the interview . On this scale with five items with a response format of ‘yes (1)/no (0)’ answers, a score of three of a maximum of five was regarded as an indication for PTSD . Resilience was assessed using the 6-item 5-point-Likert Brief Resilience Scale (BRS) (Cronbach’s Alpha = .814. All six items were averaged in a final score ranging from 0 to 5 .
Hazardous alcohol use was screened for using the AUDIT-C [50, 51]. The AUDIT-C consists of three questions, being ‘How often do you have a drink containing alcohol?’ ranging from ‘Never (0)’ to ‘4 or more times a week (4)’ (the screening ends with a score of 0 for respondents that indicated ‘Never’ in this first item), ‘How many standard drinks containing alcohol do you have on a typical day’ ranging from ‘1 or 2 (0)’ to ‘10 or more (4)’ and ‘How often do you have six or more drinks on one occasion?’ ranging from ‘Never (0)’ to ‘Daily or almost daily (4)’. In accordance to the guidelines of ‘Vlaamse Expertisecentrum voor Alcohol en andere Drugs (VAD)’, a cut-off score of four for females and five for males was used on this 3-item scale with a total score between zero and 12 . In addition to the validated scales, participants were asked using yes-no questions about sedative use, cannabis use, illegal drug use, self-harm and suicide attempts, both during their lifetime and in the past 12-months. These questions were then combined into a variable per coping mechanism with categories ‘No (0)’, ‘Yes, during the lifetime, but not in the past 12-months (1) and ‘Yes, during the past 12 months (2)’.
Social support was assessed via four items analyzed as two variables. The first item inquired about with how many people one feels comfortable with to discuss secrets or private matters (i.e., variable: ‘number of trusted persons’). Every participant received this question and added the respective number in an open format. The three other items were only presented to those participants who indicated to belong to a minority group in Belgium because of their sexual orientation, gender identity, intersex or DSD condition, religion or life philosophy, skin color, and/or ethnicity. They received the Othering-Based Stress Scale (OBS-S) - which is an adapted version of the minority stress measure - relevant to the characteristic they had indicated. The OBS-S (see Additional file 1) was used to assess minority stress experienced in relation to either ‘sexual orientation and gender identity-related’ characteristics (i.e., sexual orientation and gender identity) or ‘cultural-related’ characteristics (i.e., religion or life philosophy, skin color, and/or ethnicity) and consisted of six subscales: identity concealment (3 items), micro-aggressions (3 items), rejection anticipation (3 items), victimization events (10 items), internalized stigma (3 items), and community connectedness (3 items). The community connectedness scale (i.e., the second variable) also served as a proxy to observe social support in these participants. Responses were made on a five-point scale ranging from ‘Strongly disagree (1)’ to ‘Strongly agree (5)’. The items from the last subscale community connectedness were rescaled from ‘Strongly disagree (5)’ to ‘Strongly agree (1)’ before creating a mean across all 25 items (Cronbach’s Alpha = 0.794) where ‘1’ equals ‘low othering-based stress’ and every value higher than four means high othering-based stress.
Ethical considerations and procedure
This study was approved by the Commission for Medical Ethics of Ghent University Hospital/Ghent University (B670201837542). It was designed and performed in line with the principles of the Declaration of Helsinki. This study only included participants of 16 years and older given ethical and practical regulations related to the legal age of consenting to sex, which is 16 years old in Belgium. All participants gave informed consent before initiating the online survey.
To limit self-selection bias, the study was presented as a broader survey about health, sexuality, and well-being. The sample size calculations based on the design of the UN-MENAMAIS study led to a required sample size of 5190 participants with a targeted 864 participants per subgroup. To reach this target while considering potential non-response and refusals to participate, four times the estimated required sample size was invited for participation (i.e., N = 41,520). Between 10/10/2019 and 01/01/2021 two independent waves of data collection took place. The second wave of data collection was meant to increase the sample size and quota based sampling was applied to balance the first wave of data collection and to reach a sufficient sample size per subgroup of interest. The sample comprised 2018 participants from the first wave and 2614 participants from the second wave of data collection.
The online survey was administered via the survey software Qualtrics (Qualtrics, Provo, UT, USA). Participants could access the self-administered survey using either a link or a Quick Response (QR) code, that could be scanned using a smartphone, as indicated in the letter sent by the BNR. Before participation, potential participants received online additional information on the study and an online informed consent form. Only upon informed consent were respondents able to proceed in the survey. To increase response rates, sampled potential participants received one reminder letter sent out again by the BNR 2 weeks after their initial invitation and all invitees were informed about the possibility to receive a raffled voucher worth 30 EUR upon participation. To take part in the latter, participants were directed to a separate short questionnaire after completing the main survey to ensure that survey answers could not be linked to personal contact information.
All analysis were run in R4.1.1. Descriptive statistics (means, standard deviations, counts, and percentages) were computed for all variables figuring across all tables. Significant differences in the distribution of nominal variables between 1) participants who self-identified as heterosexual and participants who self-identified as LGB+, between 2) LGB+ participants who self-identified as being part of a minority group because of sexual orientation related characteristics (sexual minority) and LGB+ that did not self-identify as being part of a sexual minority group (Non-sexual-minority), as well as between 3) sexual minority participants who find their sexual orientation related characteristics important for their identity and sexual minority participants that do not find these characteristics important for their identity were computed using chi-square-tests. Chi2 tests going beyond 2 × 2 tables were followed up by post-hoc Chi2 tests to facilitate pairwise comparisons between categories. Effect sizes were explored by comparing the Cramer’s V coefficient (V). If the assumptions of a Chi2 test were not met, a Fisher’s Exact test was used. To compare the means of the continuous variables, the independent samples t-test was used. All assumptions were checked. The Levene’s Test was used to check for homogeneity of variance, which led to the use of the Welch t Test statistic if equal variances could not be assumed. Effect sizes were determined by calculating the Cohen’s d coefficient (D) if the sample size of the two groups were approximately the same or by using Hedges’ correction (G) if the sample size of the two groups were too different.