The association between microaggressions and mental health among UK trans people: a cross-sectional study

We conducted a cross-sectional survey of trans people in the United Kingdom (UK), the TRANS: Microaggressions & Mental Health survey. We invited participation from people who identified as trans, non-binary, and/or gender diverse, were aged eighteen years or older and had resided in the UK for 12 months or longer. We used recruitment strategies through social media advertising (Twitter, Facebook, and Instagram), with support from large trans charities and organisations such as Gendered Intelligence sharing survey details. The survey was delivered via Opinio, a web-based survey tool; launched in September 2021, continuing until September 2022. Participants provided written consent via the online survey programme.

Variables were selected to include key measures relevant to our hypothesis as well as other social and psychological factors hypothesised to influence the mental health of trans people, as identified in discussion in trans community members.

The survey questionnaire was co-produced with a group of trans people with lived experience of microaggressions and mental health problems, in line with recommended coproduction practices. [21] The aim of the co-production was to improve the acceptability of survey questions and comprehensiveness of the survey whilst also considering overall question burden on participants. The co-production group was comprised of five volunteers, recruited through the first author’s connections with academics and activists within trans communities. All five members were white, under the age of 35, and had experiences of depression, anxiety, suicidal ideation, and suicide attempt, as well as experiences of transphobic microaggressions. Due to funding limitations, co-production team members were not paid for their participation. They were presented with the questionnaire draft and asked to reflect on the acceptability of the survey tool, offer edits for clarification, and comment on the importance of the research questions. We piloted the instrument with the co-production group to ensure the survey programme worked as intended, for example that branching occurred appropriately, and instructions given within the survey were clear.

Our five main outcomes were:

Depression symptoms: measured using the Patient Health Questionnaire – 9 item version (PHQ-9). The PHQ-9 assesses the severity of depression symptoms, with scores ranging from 0 to 27. The PHQ has good psychometric properties [22], and good convergent validity and internal consistency [23].

Anxiety symptoms: measured using the Generalised Anxiety Disorder Scale – 7 item version (GAD-7). The GAD-7 assesses the severity of anxiety symptoms, with possible total scores ranging from 0 to 21. The GAD-7 has been well validated as a brief screening measure that is sensitive to change and acute symptom presentation [24].

Lifetime history of non-suicidal self-harm (NSSH), suicidal thoughts, and suicide attempt: measured using self-report measures from the Adult Psychiatric Morbidity Survey (APMS) [25]. We created a binary measure denoting the presence or absence of each.

Microaggressions: measured using the Gender Identity Microaggression Scale (GIMS). The GIMS is a 14-item scale with five subscales assessing lifetime occurrence of a) denial of gender identity, e.g., refusing to recognise trans people’s gender identity b) misuse of pronouns, e.g., consciously, or unconsciously, referring to a trans person with the wrong pronoun (he, she, and/or they for example) c) invasion of bodily privacy, e.g., asking inappropriate questions about a trans person’s genital configuration d) behavioural discomfort from others, e.g., acting in an uncomfortable manner with a trans person in any setting, and e) denial of societal transphobia, e.g., telling a trans person that experiences of transphobia (such as losing or being refused a job because they are trans) do not exist (see supplementary Table 1 for full list of items). Scores on the scale are summed to produce a total overall score, with higher scores indicating more experiences of gender identity microaggressions over the lifetime. The total scale ranges from 14 to 70 points. The scale has good internal consistency [26]. The GIMS has been used to examine mental health outcomes in the UK trans population, with a small sample of 39 trans people, finding associations between microaggressions and mental health outcomes over a 10 day period [18]

Perceived gender in this context describes the participants’ beliefs about how others perceive their gender. Trans people frequently report distress when other people’s perceptions of their gender conflicts with their own. Being visibly trans opens potential negativity from others, especially in transphobic and trans-hostile societies and communities. Conversely, being perceived to be the gender one identifies with may offer some protection against trans-hostility, especially in the case of binary presenting trans people (e.g. those who present as women being identified as women).

Individuals who believe they are more frequently seen as a trans person, or as the sex they were assigned at birth, experience more microaggressive events [27]. Furthermore, believing one is being perceived as trans or as the sex a trans person was assigned at birth has been linked to increased depressive symptoms [28].

Age has been identified as a factor associated with microaggressions, with a tendency for the experience of microaggressions to decrease as age increases [27]. A USA-based study involving 223 trans individuals found that those aged 25–34 had a mean score of 0.75 points lower on microaggressions compared to those aged 18–35 [29]. Similarly, individuals aged 35–49 scored a mean of 1.32 points lower than those aged 18–24. Additionally, evidence suggests that age is associated with depressive symptoms, anxiety symptoms, NSSH, suicidal thoughts, and suicide attempts [30‐33]. Population-based studies in the UK indicate that individuals aged 45–59 have a significantly higher prevalence of probable depressive disorder compared to those aged 16–29 [34]. Similarly, there are age-related associations with anxiety symptoms and NSSH, suicidal thoughts, and suicide attempts, with a decrease in suicidal thoughts, plans, and attempts as age increases [33]. Moreover, anxiety symptoms tend to be higher in younger age groups compared to older age groups [35]. Given these findings, it is plausible that age may act as a confounding factor in the relationship between microaggressions and depressive symptoms within our sample. We asked participants to choose an age category from the following: 18–24, 25–34, 35–44, 45–54, 55–64, 65 + .

Education was selected as an indicator of socioeconomic status (SES). SES is defined as a person’s work experience and of an individual’s economic access to resources and social position in relation to others [36]. There is evidence that education level may predict microaggressions [37] and that education is also a predictor of mental health distress [37]. Participants were asked to report their highest educational attainment with the following options: No qualifications, GSCEs or equivalent, A-levels/Scottish Highers, University Degree (e.g., BSc, BA), Master’s Degree or equivalence, Doctorate (e.g., MD or PhD), or Vocational Qualifications.

Lesbian, Gay, and Bisexual (LGB) people are at a higher risk of adverse mental health conditions such as depressive symptoms, anxiety symptoms, and suicidal thoughts and behaviours when compared to heterosexual people [38, 39]. There is also evidence that LGB individuals experience specific forms of microaggressions relating to their sexuality [40, 41]. These interpersonal and environmental microaggressions are manifested in the form of hostile messages towards sexual minorities [42]. There is evidence that LGB people are at increased risk of microaggressions [43] and that these microaggressions may increase risk of depressive symptoms and anxiety symptoms [43]. Participants were able to select or self-define their sexuality. For the purposes of analysis, these were combined into a binary variable with the two categories representing LGB/Queer/Questioning/Asexual/Aromantic, and heterosexual.

We selected disability as a confounder because people with disabilities, including physical disability and neurodiversity, are at an increased risk of experiencing disability-specific microaggressions [44]. Examples of microaggressions specific to those with disabilities include patronisation, desexualisation, and second-class citizenship [41]. People with disabilities also show an increased risk of developing poor mental health when compared to able-bodied people [45, 46]. People with disabilities who experienced ableist microaggressions had higher scores on depressive symptoms and anxiety symptoms measures compared to those who did not experience ableist microaggressions [47]. Participants were asked to report whether they had a disability lasting 12 months or longer, and to indicate whether this reduced their ability to carry out day to day with the following option, “Not at all”, “Yes a little”, and “Yes a lot”.

There is evidence that supports a relationship between membership of ethnic minority communities and experiences of microaggressions [37, 48]. Furthermore, there is evidence supporting a relationship between ethnicity and mental health distress, with ethnicity/race-related stressors increasing susceptibility to depressive symptoms, anxiety symptoms, NSSH, suicidal thoughts, and suicide attempt [49]. Ethnicity questions were derived from the Office for National Statistics Census and presented alphabetically [50].

Physical transition was taken from the Trans Mental Health Study (2012) where participants are asked to state at which point they are currently on their physical transition. Participants can answer in the following ways: “No, I have not undergone or propose to undergo any part of a process”, “Yes, I have undergone a process”, “Yes, I am currently undergoing a process”, “Yes, I am proposing to undergo a process”, and “Unsure”, “Prefer not to say”, and “Other”. Stage of social transition was a separate question and utilised the same response format.

Regression analyses: We transformed the GIMS exposure variable into standard deviation units as the unit increase was small relative to the range, to aid interpretation [51]. We used linear regression models with depressive symptoms (PHQ-9) and anxiety symptoms (GAD-7) scores as separate continuous outcomes and the transformed microaggressions scale (GIMS) as a continuous exposure. We used logistic regression models with NSSH, suicidal thoughts, and suicide attempt as separate binary outcomes and the transformed microaggressions scale as a continuous exposure. All analyses were reported before and after adjustment.

Imputation: To assess the potential influence of missing data when modelling the association between microaggressions and our five outcomes we investigated differences between participants with complete data and those with missing data[52]. Auxiliary variables were selected if they were associated with the exposure and outcomes and were not included in the analysis. The role of the variables was to improve imputation estimates and reduce coefficient bias in the multiple imputation [53] Auxiliary variables were loneliness, and whether the participant was living in their affirmed gender. We used Multiple Imputation by Chained Equations (MICE) to impute twenty-five datasets, which were then combined using Rubin’s rules. We imputed on the exposure (GIMS), outcomes, and confounders. In our analyses we first restricted the sample to those with complete cases on the exposure. Next, we used the imputed data on the exposure, outcomes, and confounders, and restricted to complete case of exposure.

Loneliness as a putative mediator: After discussing our initial findings, we decided to explore the potential role of loneliness in our models. We therefore added a post hoc sensitivity analysis to assess whether poor social support might help explain the association between microaggressions and our mental health outcomes, based on the possibility that this might lie on the causal pathway from microaggressions to mental distress. This was based on our theory that experiencing more frequent microaggressions might lead to low perceived social connectedness, which in turn would worsen mental health. We used loneliness as a proxy for poor social support, measured using the 3-item UCLA Loneliness Scale, a validated measure capturing the subjective experience of loneliness [54]. To test for evidence of a potential mediating role, we added loneliness to our final adjusted models and compared the coefficients from these sensitivity analyses to those in our main analyses.

All analyses were conducted using Stata version 17.1 [55].

A total of 787 participants responded to the online advertisement about the TRANS: Microaggressions & Mental Health Survey and took part in the cross-sectional study, of whom 574 (79%) provided complete data on exposure, outcomes, and confounders, comprising our complete case analytical sample (Fig. 1).

Most participants were under the age of 35 (n = 428, 73%), white (n = 525, 91.5%), and resided in England (n = 483, 84%). The mean GIMS score was 42.46 (SD 13.28), and 97.6% (n = 560) endorsed at least one form of lifetime microaggression experience. The most common specific microaggression experience was misuse of pronouns (n = 544, 94.8%), and the least endorsed was behavioural discomfort from others (n = 389, 67.8%).

After splitting participants into those with a high versus low number of microaggression experiences, based on a median split of the GIMS (median = 42; IQR = 33– 52; see Table 1), those who experienced high microaggressions had a mean score of 13.52 (SD 6.46) on the PHQ-9 and a mean score of 11.49 (SD 5.90) on the GAD-7. The high microaggression (HM) group, compared with the low microaggression (LM) group, had a significantly higher proportion of those with a lifetime history of NSSH (HM: 86.6% vs LM: 68.3%), suicidal thoughts (HM: 95.4% vs LM: 86.9%), and suicide attempts (HM: 49.0% vs LM: 30.2%).

Depressive symptoms: We found evidence of an association between microaggressions and depressive symptoms, whereby as microaggression experiences increased by one standard deviation (13.28-points) of the GIMS scale, this was associated with an increase in depressive symptom scores (unadjusted coefficient: 2.09, 95%CI = 1.59– 2.60; adjusted coefficient: 1.86, 95%CI = 1.35 –2.36; see Table 2).

Anxiety symptoms: We found evidence of an association between microaggressions and anxiety symptoms, whereby when scores on microaggression experiences increased by one standard deviation of the GIMS scale, this was associated with an increase in anxiety symptom scores (unadjusted coefficient: 1.76, 95%CI = 1.28–2.23; adjusted coefficient: 1.57, 95%CI = 1.09–2.05).

Lifetime NSSH, suicidal thoughts, and suicide attempt: We found evidence of an association between microaggression experiences and lifetime NSSH (OR_crude 1.95, 95% CI = 1.57–2.42; OR_adj 1.83, 95%CI = 1.45 –2.30), lifetime suicidal thoughts (odds ratio [OR]_crude 2.59, 95%CI = 1.85–3.62; OR_adj 2.18, 95%CI = 1.52–3.13), and lifetime suicide attempt (OR_crude 1.66 95%CI = 1.39–1.99; OR_adj, 1.59, 95%CI = 1.32–1.92).

Correlation coefficient estimates for GIMS subscales ranged from 0.31 to 0.56 (see Table 3), suggesting weak to moderate correlations between each subscale. This justified use of the subscales as exposures in multivariable models.

In multivariable models for GIMS subscales, we found evidence to support a positive association between four subscales and specific mental health outcomes: behavioural discomfort from others, denial of societal transphobia, misuse of pronouns, and denial of gender identity, (see Table 4). Findings for each of these subscales were as follows, each with estimates from fully adjusted multivariable models, including mutual adjustment for other subscales Table 5.

Misuse of pronouns: As scores on the misuse of pronouns subscale increased by one standard deviation this was associated with increased odds of NSSH (OR_crude 1.95, 95%CI = 1.57–2.42; OR_adj 1.83, 95%CI = 1.45– 2.30), and with an increased odds of suicidal thoughts (OR_adj 1.49 95%CI = 1.10– 2.03).

Denial of gender identity: As scores on the denial of gender identity subscale increased by one standard deviation, this was associated with increased odds of suicide attempt (OR_adj 1.40 95%CI = 1.11– 1.77).

Our comparison of samples with complete and missing data (see Supplementary Table 2) identified differences on seven variables. Comparing models before and after imputation, we found similar estimates of the coefficients and odds ratios in the main analysis and in our two imputed models, with some attenuation in the coefficients as sample sizes increased for our imputed models (see Supplementary Table 3 for comparison of main models and MICE models in the main analysis). Similarly, in our exploratory analyses, we observed minimal attenuation when comparing the complete case analysis and the two imputation models (see Supplementary Table 4, 5, and 6 for comparison of main models and MICE models in the hypothesis-generating analyses).

After adding loneliness to our five main fully adjusted models, we found no significant attenuation of any of the associations (supplementary Table 7), providing no evidence to support loneliness as a putative mediator of these associations.

Analysing data from a sample of trans people from across the UK, we found that experiences of microaggressions were common and were associated with increased severity of depressive symptoms and anxiety symptoms, and increased odds of lifetime NSSH, suicidal thoughts, and suicide attempts. This supported our hypothesis that experiencing more microaggressions would be associated with greater mental health symptoms in trans people compared to those who experienced fewer microaggressions, tested for the first time using validated measures. In exploratory analyses we also found evidence to support associations between specific types of microaggression and selected mental health outcomes: behavioural discomfort with depressive symptoms, denial of societal transphobia with anxiety symptoms, misuse of pronouns with NSSH and suicidal thoughts, and denial of gender identity with suicide attempts. These exploratory findings may allow more specific hypotheses as to the role specific microaggression experiences might play in causal pathways to the development of depressive symptoms, anxiety symptoms, NSSH, suicidal ideation, and suicide attempts. Our findings were unchanged when testing the potential biases introduced by missing data.

Our findings are generally consistent with other empirical evidence on microaggressions describing associations with poorer mental health and wellbeing in other minoritised and marginalised communities, such as minority ethnic communities and LGB communities [56, 57]. They are also consistent with evidence supporting an association between microaggressions and general distress and wellbeing measures among trans people [16, 18, 27, 58, 59]. The findings are also in agreement with the only other study of microaggressions and mental health in the UK trans population, namely our finding between denial of societal transphobia and anxiety symptoms [18]. However, this study improves on previous research [27, 58, 59] by using validated measures of gender identity microaggressions and adding to broad findings on general distress and wellbeing [16] by investigating common mental health problems in the general population. The current study has also generated further hypotheses to test regarding specific microaggression experiences and their associations with specific mental health outcomes.

Several mechanisms may explain associations between the specific microaggression experiences and specific mental health outcomes. First, experiencing behavioural discomfort from others may increase depressive symptoms by reinforcing feelings of being socially excluded, which in turn increases social isolation and loneliness [60]. Evidence from general population samples identifies the role of loneliness in the development of depressive symptoms [61]. Experiences of discrimination can compound loneliness. Whilst we found no evidence that loneliness helped explain the association between trans microaggressions and poor mental health, there is a need for longitudinal work in large, representative samples to explore the relationships between these variables, and specifically the role of loneliness as a mediator. Second, denial of societal transphobia fosters an unsafe environment for trans people to express their concerns about transphobic experiences they may have had. This may increase worry and rumination around lived experiences of transphobia, contributing to anxiety[62]. Third, misuse of pronouns may increase NSSH and suicidal thoughts by signaling to a trans person that they do not belong, giving rise to a sense of thwarted belongingness that drives (according to the Interpersonal Theory of Suicide) self-harm and suicidal ideation [63]. Finally, denial of gender identity may single trans people out and also contribute to a sense of thwarted belongingness, thereby increasing the risk of self-harm and suicidal ideation [63].

Strengths of this study include a sample size that is larger than other studies examining microaggressions using the GIMS (N = 292), the use of a measure of microaggressions psychometrically validated with trans community members, and the use of including validated measures of mental health [16]. We used robust statistical models, with confounders chosen a priori, and used a range of sensitivity analyses to test for the influence of missing data on estimates.

Our sampling method carries a risk of selection bias and digital exclusion due to our snowball sampling methods. Selection bias may have arisen due to selective avoidance and selective sharing of the recruitment call. Our comparison of samples with missing and complete data suggests that those at an earlier stage of transition were more likely to drop out and had higher odds of NSSH, suicidal ideation, and suicide attempts, but lower scores on depressive symptoms and anxiety symptoms. There may have been overrepresentation of those with poor mental health thus inflating the reported prevalence of mental health conditions. However, our estimates of the associations of poor mental health with microaggressions should remain valid [64, 65].

Recent census data on gender diversity within the UK suggests that 16–24 year-olds were the age group most likely to indicate that their gender identity was different to their sex assigned at birth, followed by 25–34 year-olds [3, 4]. This is consistent with the 73% of respondents to our survey who were aged under 35. However, our study has an under-representation from minority ethnic communities when compared to the general population distribution [50]. This under-representation may have been influenced by the socio-demographic composition of the co-production team and the wider research team. We lacked advice from trans people of colour during the design stage and therefore could have omitted to cover issues salient to these communities within our survey, or indeed sampling considerations that would have improved ethnic minority representation. Future research should prioritise diversity within research and co-production teams and make efforts to ensure adequate sampling from minority ethnic communities to improve the diversity and representativeness of the study sample. These might include, for example, appropriate incentivisation and outreach efforts [66].

It is important to recognise the role of intersectionality between ethnicity, disability, and sexuality when considering microaggressions [67]. Multiple intersecting identities increase the risk of exposure to microaggressions, and in turn the risk for mental ill health [67]. Whilst we did collect data on ethnicity, disability, and sexuality we lacked statistical power to investigate the effects of intersectionality. Further work is needed to explore associations within intersecting identities with regards to microaggression experiences and mental health outcomes.

The cross-sectional nature of the study means we were unable to establish temporality in the associations observed. The measures we used to capture microaggression experiences, NSSH, suicidal ideation, and suicide attempts related to lifetime experiences, whereas depressive symptoms and anxiety symptoms were assessed over the previous two weeks. With data collection at one time point, we were unable to establish whether the exposure (microaggression experiences) preceded the symptoms. We therefore cannot rule out reverse causality in the association between microaggressions and mental health outcomes. There is potential for a bi-directional relationship between microaggressions and mental health, [68]. Longitudinal studies are needed to examine the directionality of these associations. Furthermore, as with any observational study, we are unable to rule out any residual confounding from unmeasured variables or confounders measured imperfectly (such as access to gender-affirming care), which may partially explain the associations observed in this study. We adjusted our final models for stage of physical transition and stage of social transition but ideally would have liked to have collected primary data on access to gender-affirming care. Similarly, our sensitivity analyses tested whether there was evidence to support social support as a mediator of these associations, by adding loneliness to final models. This was an inadequate proxy for poor social support, and ideally, we would have liked to conduct formal mediation analysis using longitudinal data and a validated measure of social support.

Public policy and education could reduce the occurrence of microaggressions. Research to develop our understanding of why microaggressions are enacted and how best to reduce their occurrence would lead to better interventions, and furthermore to better mental health outcomes for trans people. Interventions developed for ethnic minority communities include workshops that include the targeted minority alongside the wider community, and other interventions that promote social connectedness, familiarity, closeness, and management of uncomfortable feelings, and reduce social distance [69]. Findings suggest that white students randomized to a Racial Harmony Workshop were less likely to perpetrate microaggressions towards minority ethnic students [69]. However, as highlighted in critiques of the literature on the effectiveness of interventions to address microaggressions, we need more research on effectiveness and acceptability [70]. Given our study’s findings relating to GIMS subscales, a better understanding of how specific microaggression types influence specific mental health outcomes would also allow clinicians to tailor support around how these microaggressions interact with psychiatric presentations in trans people.

Further research is needed to strengthen our understanding on whether microaggressions cause changes in depressive symptoms, anxiety symptoms, NSSH, suicidal ideation, and suicide attempt. Longitudinal studies are needed to answer questions about whether experiencing microaggressions increases the risk of subsequent depressive and anxiety symptoms, NSSH, suicidal ideation, and suicide attempts. Understanding the temporal relationship between microaggressions and mental health using longitudinal designs will help researchers and clinicians better understand both the short and long-term associations of microaggressions with mental health, NSSH, suicidal ideation, and suicide attempts, and any reciprocal influences, identify mediators, and tailor intervention design around these findings. Given the complexity of microaggression experiences, we also recommend further qualitative work through focus groups and interviews to understand how each subscale may be experienced in relation to specific mental health outcomes. Qualitative data regarding microaggression experiences and depressive symptoms, anxiety symptoms, suicidal ideation, and suicide attempts will provide a more nuanced understandings of mechanisms underlying the associations described here.