Participants and procedures
In this cross-sectional study, multiple community–based outreach strategies were used to recruit a sample of TGW (n = 92) between the ages of 18 and 65 years who reside in Atlanta, GA. Venues that serve TGW and word–of–mouth recommendations from transgender advocates provided the primary methods of recruitment. These venues offer HIV prevention and care, housing, and counseling services to the TGW. The study was known as the Transgender Atlanta Personal Survey. Transgender advocates notified the study project director when they located a woman who was willing to be screened for study participation. In addition, the project was advertised through formal and informal communication channels via advocacy groups and Lesbian, Gay, Bisexual, and Transgender (LGBT) service organizations. The project director used print materials to provide their contact information. Data were collected from August 2014 through June 2015.
TGW were screened to determine eligibility. The inclusion criteria were: (1) 18 to 65 years of age, (2) male sex assigned at birth, and (3) self–identifying as either female or transgender. All participants who were screened, except one individual who identified as “other” were eligible and consented to participate in the study. After providing written informed consent, women engaged in a face–to–face structured interview with a trained graduate research assistant. The training involved cultural-competency and the use of non-judgmental statements. Interview responses were recorded on a portable electronic tablet, using Qualtrics© software (Provo, Utah). The Institutional Review Board of Georgia State University approved study protocols following a full board review.
Measures
The survey assessed sociodemographic characteristics, a broad range of theoretical contextual factors, and self-reported HIV status. In addtion, we assessed the prevalence of several trauma exposures, such as “ever experienced physical abuse by an intimate partner,” “ever being a victim of sexual abuse,” “ever experienced childhood sexual abuse,” and psychosocial factors.
Perceived stigma was assessed by using four items adapted for TGW from the original scale developed for gay individuals [44, 45]. These four items were a subscale that measured the TGW’s perceptions of society’s stigma or negative attitudes toward TGW. The items were: (1) “Society still punishes people for being transgender”; (2) “Most people have negative reactions to transgender people”; (3) “Discrimination against transgender people is still common”; and (4) “Only a few people discriminate against transgender people.” Response options were presented on a 5-point Likert scale, ranging from 1= strongly disagree to 5 =strongly agree. Item 4 was reverse coded. The mean of response scores for the four items were used for the analysis. Inter-item reliability was adequate (Cronbach’s alpha = 0.73).
The psychosocial impact of gender minority status was assessed using three items from a 4-item subscale developed by Sjoberg and colleagues [46]. The 4-item subscale is part of the longer Transgender Adaptation and Integration Measure and assessed four aspects of mental health related to transgender status. We used this subscale to measure psychosocial distress related to the unique experiences of TGW. The items were: (1) “I get depressed about my gender status”; (2) “My gender status interferes with my quality of life”; (3) “I have thought about suicide because of my gender status”; and (4) “Being transgender causes me relationship problems.” Response options were provided on a 5-point Likert scale, ranging from 1 = strongly disagree to 5 = strongly agree. Given that our outcome was suicidal ideation, the third item was not included in the analyses. We used the average of response scores for the three items. Inter-item reliability was adequate (Cronbach’s alpha = 0.71).
Suicidal ideation was assessed by combining two items that measured suicidal thoughts. The items were: (1) “In the past 12 months, have you considered attempting suicide?” for which the response options were Yes/No; and (2) “I have thought about suicide because of my gender status,” for which response options were provided on a 5-point Likert scale, ranging from 1 = strongly disagree to 5 = strongly agree. We dichotomized Item 2 by collapsing the responses of 4 (agree) and 5 (strongly agree) as “Yes” and all other responses as “No.” Then, we created a new variable, “suicidal ideation,” for the participants who responded “Yes” to either of the two items; these participants were considered as experiencing suicidal ideation and other participants, as not experiencing suicidal ideation.
Depression was measured using six items from the Brief Symptom Inventory [47]. This subscale is widely used as a psychological self-report symptom scale to measure depression. The items were: (1) “Feeling not interested in things”; (2) “Feeling lonely”; (3) “Feeling blue”; (4) “Feeling worthlessness”; (5) “Feeling hopeless about the future”; and (6) “Thoughts of ending your life.” Response options for all items were on a 5-point Likert scale, ranging from 1 = not at all to 5 = extremely. We calculated the mean of these six items as the depression score.
Anxiety was measured using the 3-item subscale from the Brief Symptom Inventory. The items were: (1) “Experienced nervousness or shakiness inside”; (2) “Feeling tense or keyed up”; and (3) “Feeling so restless you couldn’t sit still.” Response options for all items were on a 5-point Likert scale, ranging from 1 = not at all to 5 =extremely. For the anxiety score, we calculated the mean of the three items.
Excessive drinking was measured by three items: (1) “In the past 30 days, on how many days did you drink any alcohol?”; (2) “On the days when you drank alcohol in the past 30 days, about how many drinks did you have on average?”; and (3) “In the past 30 days, how many times did you have 5 or more alcoholic drinks in one sitting?” Based on the Dietary Guidelines for Americans, 2015–2020 [48], participants who consumed 15 or more drinks during the prior week or consumed more than 5 or more drinks in one sitting were considered to evidence excessive drinking.
Non-injection drug use was measured by one item: “In the past 12 months, have you used any non-injection drugs, other than those prescribed for you?” Response options were Yes/No.
Injection drug use was measured by one item: “Have you ever in your life shot up or injected any drugs other than those prescribed for you? By drugs, I am referring to drugs such as heroin, meth – not hormones or silicone? By shooting up, we mean anytime you might have used drugs with a needle, either by mainlining, skin popping, or muscling.” Response options were Yes/No.
Intimate partner violence is the experience of physical and emotional violence by a romantic or sexual partner in one's lifetime. This variable was measured by three items: (1) “In your lifetime, have you ever been physically abused by a romantic or sexual partner? By physical abuse, we mean a range of behaviors, from slapping, pushing, or shoving, to severe acts, such as being beaten, burned, or choked”; (2) “In your lifetime, have you ever been emotionally abused by a romantic or sexual partner? By l emotional abuse, we mean name-calling, or humiliating you, or trying to monitor and control or threaten you”; and (3) “Have you ever been physically abused or beaten by a romantic or sexual partner because of your gender identity or presentation?” Response options were Yes/No.
Sexual abuse is the experience of forced oral/anal sex in one's lifetime. This variable was measured by three items: (1) “In your lifetime, has someone ever made you perform oral sex?”; (2) “In your lifetime, has someone ever made you receive anal sex? By receiving anal sex, we mean they put their penis in your anus (you were the bottom)”; and (3) “In your lifetime, has someone ever made you perform anal sex? By performing anal sex, we mean they made you put your penis in their anus (you were the top)”. Response options were Yes/No.
Child sexual abuse was measured by one item: “As a child (less than 16 years old), were you ever sexually abused?” Response options were Yes/No/I do not remember. HIV status was measured by one item, “What was the result of your most recent HIV test?” for which the response options were “negative”/ “positive”/ “I do not know my status”.
Data analysis
Descriptive statistics were computed among TGW who reported suicidal ideation and those who reported no suicidal ideation. To estimate bivariate associations, TGW with suicidal ideation were compared to TGW without suicidal ideation, using chi-square analyses and Fisher’s exact test for categorical variables (i.e., education, race, homelessness, excessive drinking, non-injection, injection drug use, and HIV status). Wilcoxon rank-sum tests were used for continuous variables (i.e., age, perceived stigma, anxiety, depression, and psychosocial impact of gender minority status). Control variables were selected based on the literature and the presence of statistically significant differences in our bivariate analyses (alpha = 0.10). Age, race, and education were statistically significantly different between both groups, and homelessness was associated with suicidal ideation among TGW in prior studies. To estimate the association of substance abuse behaviors, violence, abuse, HIV status, and other psychosocial factors with suicidal ideation, we conducted separate multivariable logistic regression models, adjusting for age, race, education, and homelessness.
We also examined the impact of perceived stigma on suicidal ideation through mediation analyses. A mediator is a variable that explains, or accounts for, the effect of the independent variable on the dependent variable [49, 50]. To examine the underlying mechanism between perceived stigma and suicidal ideation, we tested the role of six psychosocial factors as potential mediators to explain the effect of perceived stigma (Variable X) on suicidal ideation (Variable Y) (Fig. 1): anxiety, depression, the psychosocial impact of gender minority status, excessive drinking, injection drug use, and non-injection drug use (Variables M). Six separate mediation models were tested, one for each psychosocial factor. The effect of X on Y is the total effect (path c); the effect of X on M is indicated by path a; the effect of M on Y controlling for X, is indicated by path b; and the direct effect of X on Y, controlling for M, is path c’. The indirect effect is the product of path a and path b, which is path ab. The equation (c = c’ + ab), which indicates the total effect is equal to direct and indirect effects, does not hold true due to the use of logistic regression. By standardizing the coefficients expressed on a log-odds metric (multiplied by the standard deviation of the predictor variable and divided by standard deviation of outcome variable), however, c would be approximately equal to c’ + ab [49]. Using PROCESS macros v3.3 by Andrew F. Hayes [51], we tested the psychosocial impact of gender minority status, anxiety, and depression as mediators, and the point estimates for path a, path b, path c’, and path c were generated. For testing the dichotomous mediators (excessive drinking, injection drug use and non-injection drug use) we used the INDIRECT macro [50]. Bootstrapping (N= 5000) was used to construct confidence intervals (CIs) for the indirect effect (path ab) to determine statistically significant mediators. The Statistical Package for Social Sciences (SPSS), version 25.0, (IBM, Chicago, IL), was used for all analyses.