The sample (N=249) for this cross-sectional study was recruited using an anonymous online questionnaire. The form link was distributed via social media platforms (‘WhatsApp’, ‘Instagram’, ‘Twitter/X’), using snowball sampling method. The study was introduced to potential participants via flyers and descriptive texts, which stated its aim as investigating the relationships between alcohol and drug use, sleep behavior, and sexual practices among LGBTQ+ individuals. 251 participants completed the survey of whom 2 were excluded. Exclusion criteria for those cases were not filling out the survey completely and not self-identifying as LGBTQ+. Answers for the form were accepted between 29.8.2024 and 29.9.2024. The questionnaire was repeatedly shared on different social media platforms and over the e-mail accounts of LGBTQ+ organizations, to reach a diverse study population. Inclusion criteria were as follows: (i) self-identifying as LGBTQ+, (ii) being at least 18 years old, (iii) living in Turkey, and (iv) reading and writing in Turkish.

The survey included demographic questionnaire to gather information on participants age (years), living arrangement (in a dormitory, with my spouse/partner/family, alone), being out or closeted about their LGBTQ+ identity, and regular usage of dating apps. Questions about working arrangements included education status, employment status, if working in shifts (full time mornings, full/part time evenings), and monthly income (no regular income, ranging from the minimum wage to more than three times the current minimum wage).

Participants were asked to indicate which gender they identify the most from the following options: cis-woman, cis-man, trans-man/transmasculine nonbinary, trans-woman/transfeminine nonbinary, and nonbinary/genderfluid/genderqueer. Additionally, they were asked to indicate their sexual orientation, with options including heterosexual, gay or lesbian, bisexual/pansexual/queer, and asexual. Cisgender participants were required to complete the Arizona Sexual Experience Scale (ASEX). Participants who engage in penetrative sex were asked about their sexual role (always insertive, mostly insertive, both insertive and receptive, mostly receptive, and always receptive).

To measure substance use, drug categories were classified into: chemsex-related substances (amphetamine, methylphenidate, methamphetamine, cocaine, MDMA, LSD, GHB, ketamine, magic mushroom), depressants (sleep medicines, opioids, heroin, benzodiazepines), and cannabinoids (Marijuana/Cannabis/THC). Participants were asked to indicate all the drugs they had used without a prescription or outside of their prescribed medical purpose within the last 3 months. The question was close-ended and presented in a multiple-choice format.

Alcohol use was measured using the AUDIT-C questionnaire, scored between 0 and 12, with higher scores indicating greater consumption (23). A score of 3+ in women, transgender, and gender-expansive participants, and 4+ in men, was considered positive for high alcohol use (24).

Pittsburgh Sleep Quality Index (PSQI) is a self rated questionnaire, that evaluates sleep quality and disturbances over the past month using 24 questions (likert type and open ended questions). The component scores are summed to generate a global score ranging from 0 to 21, with higher scores indicating poorer sleep quality. A score below 5 indicates good sleep quality (25). The PSQI was validated in 1996 in a Turkish sample, demonstrating reliable clinical and clinimetric properties over a 12-month period in healthy individuals, patients with depression, and those with sleep disorders, supporting its use in both clinical practice and psychiatric research (26).

The Index of Sexual Risk-Taking (ISRT) measures an individual’s engagement in risky sexual behaviors over the past 12 months and throughout their lifetime (27). The score ranges between 0 and 10, with higher scores indicating more risky sexual behaviors (28). Following consultations with the developer of the index and feedback from participants, particularly gender-diverse and lesbian women participants, three items related to condom use and anal sex were excluded to enhance validity. The modified version demonstrated a Cronbach’s alpha of 0.67. Certain ISRT questions are only relevant to participants whose sexual orientation matches descriptions on the Kinsey Scale; therefore, only these participants were included in the sample. Consequently, 25 participants were excluded in analyses involving ISRT.

Psychological and physiological arousal, orgasm, feelings of satisfaction, and sexual urges were measured using Arizona Sexual Experiences Scale (ASEX) (29). Comprising of 6 questions, the total score ranges between 5 and 30 (30). Lower scores indicate stronger and easier arousability and more satisfying sexual responses, while higher scores suggest sexual dysfunction. As certain questions on the ASEX specifically address biological sex functions and are designed exclusively for cis-male and cis-female individuals, only cisgender participants were asked to complete this section to ensure reliability and avoid potential distress among gender-diverse participants. Therefore, 102 participants were excluded in analyses involving ASEX.

All analyses were conducted using IBM SPSS version 22, except for the mediation analysis, which was performed using R version 4.4.0. Characteristics of the sample data were represented with mean values, standard deviation (SD), and frequencies.

Bivariate logistic regression analyses were performed to examine the association between the PSQ and all predictor variables. Multivariable logistic regression analyses were conducted to examine the association between PSQ and predictor variables AUDIT-C, chemsex-related substance use (CRSU), ISRT, and ASEX, while adjusting for living arrangement, employment, monthly income, gender, and sexual orientation. Additionally, linear regression analyses were conducted to evaluate associations between CRSU, AUDIT-C, and ISRT, and bivariate logistic regression analyses were performed to assess the relationships between CRSU, AUDIT-C, and ASEX. Odds ratios (ORs) for bivariate analyses, adjusted odds ratios (aORs) for multivariable models, and 95% confidence intervals (CIs) are reported for all relevant findings.

The mediating effect was evaluated using structural equation modeling (SEM) to decompose the total and indirect effects of the mediator. The mediation analyses were conducted in R version 4.4.0 using the lavaan package (31). The model specified three pathways (a, b, and c) to investigate the effect of the mediator (M) in the relationship between exposure (X) and outcome (Y). Pathway a assessed the relationship between X and M, while controlling for covariates including monthly income, employment status, living arrangement, gender, and sexual orientation. Pathway b evaluated the association between M and Y, with the same covariates. Pathway c (direct effect) analyzed the relationship between X and Y, adjusting for M and all covariates. The indirect effect, representing the mediation pathway through M, was calculated as a×b. The total effect was derived as the sum of the direct and indirect effects (c+a×b). The model was estimated using a weighted least squares (WLS) estimator. Statistical significance was set at α=0.05.

Participation to the survey was anonymous. Each participant was asked to read and digitally sign an informed consent form before answering the survey. Ethics approval was granted by ATADEK-’Acıbadem Mehmet Ali Aydınlar University Ethics committee’ (number 2024-12/526).

The final cohort comprised 249 adult individuals who identified as LGBTQ+ and living in Turkey. The mean age was 25 years (SD = 5.89), with a range of 18 to 50 years. Transgender and nonbinary individuals constituted 40.1% of the cohort, while cisgender participants accounted for 59.8%. Regarding sexual orientation, 51.4% identified as bisexual, 40.2% as gay, 4.8% as heterosexual, and 3.6% as asexual.

For education and socioeconomic status, 49.4% of participants were students, and 91.6% had received an undergraduate-level education or higher. Present employment was reported by 39.0%, with most employed participants (68.1%) working morning shifts only. 68.3% had a stable monthly income, but 42.6% earned less than twice the minimum wage (MW). During the data collection period, the MW was 17,002 Turkish Liras (TRY), equivalent to approximately 500 USD.

Regarding social and lifestyle factors, 32.5% of participants lived with their families, and 26.5% lived alone. 70.7% identified as closeted in their daily lives. Regular use of dating applications and alcohol use disorder were each reported by 40.6% of participants. Substance use was indicated by 37.3% of the cohort. The most used substance was marijuana (21.3%), followed by chemsex-related substances (18.9%).

Scaled-based outcomes revealed that 25.9% of participants experienced sexual dysfunction, while 80.7% reported poor sleep quality. The mean ISRT score was 3.0 (SD = 1.85), with a range of 0 to 7. Detailed cohort characteristics are presented in Table 1 .

Figure 1 summarizes the mediation model of the effect of CRSU on the association between PSQ and SRT. The mediation analysis examined the relationship between PSQ and SRT, with CRSU as a mediator. Pathway a (PSQ → CRSU) demonstrated a statistically significant positive association (β=0.608, p=.046). Similarly, Pathway b (CRSU → SRT) showed a significant positive relationship (β=1.719, p<.001). In contrast, the direct effect of PSQ on SRT (Path c) was not statistically significant (β=−0.163, p=.777). The indirect effect of PSQ on SRT through CRSU was significant (β=1.045, p=.047), indicating the mediating role of CRSU. The total effect, combining direct and indirect pathways, was also significant (β=0.881, p=.007), suggesting that the influence of PSQ on SRT primarily operates through its impact on CRSU among LGBTQ+ individuals.

When the relationship between PSQ and SRT was examined with AUDIT-C serving as the mediator, there was no statistically significant indirect pathway (β=0.163, p=.137), suggesting no evidence that alcohol use mediates the relationship between PSQ and SRT. There was also no statistically significant mediation between PSQ and sexual dysfunction via either CRSU (β=-0.173, p=.445) or alcohol use (β=-0.008, p=.942).

To assess the directionality of associations, two alternative mediation models were tested. In the first model, CRSU was specified as the predictor, PSQ as the mediator, and SRT as the outcome. CRSU was significantly associated with worse PSQ (β = 0.247, p = .025), and PSQ was significantly associated with higher SRT (β = 0.218, p = .010); however, the indirect effect of this pathway did not reach statistical significance (β = 0.251, p = .090). In the second model, SRT was specified as the predictor CRSU as the mediator, and PSQ as the outcome. While SRT was strongly associated with CRSU (β = 0.461, p <.001), neither the direct effect on PSQ (β = 0.154, p = .175) nor the indirect effect via CRSU (β = 0.057, p = .152) were statistically significant.

The models in this study were just-identified (degrees of freedom = 0). In such cases, the model necessarily fits the data perfectly by definition, and fit indices are not informative. For this reason, model fit statistics were not reported.