1 Description of the dataset

This analysis uses data from the TransPop study, investigated by Ilan H. Meyer, Williams Institute, UCLA. TransPop is the first national probability sample of transgender adults in the United States. The study examines a variety of health-relevant domains including health outcomes and health behaviors, experiences with interpersonal and institutional discrimination, identity, transition-related experiences, and basic demographic characteristics (age, race/ethnicity, religion, political party affiliation, marital status, employment, income, location, sex, gender, and education). The public file combines two recruitment periods (2016 and 2017-2018) for a final sample of 274 respondents.

In the face of an influx of anti-trans legislation and narratives, I am curious to explore how trans people are impacted by various socio-political factors, and how we can utilize the insights from this database to identify ways to improve their quality of life. My analysis focuses on the gender identity groups within the transgender sample: trans women, trans men, and gender nonbinary (GNB) respondents.

My analysis is organized around three questions:

  1. How do healthcare satisfaction and access relate to trans people’s happiness? Healthcare satisfaction (q84), cost barrier to seeing a doctor (q88), hormone treatment (q65), healthcare stereotype threat (hcthreat_i, range 1-5), and life satisfaction (lifesat_i, range 1-7).
  2. How are internalized transphobia and experiences of discrimination related? Internalized transphobia (internalized_i, range 1-5) and everyday discrimination (everyday_i, range 1-4).
  3. How is perceived social support related to well-being? Social support (socsupport_i, range 1-7), social well-being (socialwb_i, range 1-7), life satisfaction (lifesat_i), and psychological distress (kessler6_i, range 0-23, higher = more distress).

Respondent characteristics are age (age), race/ethnicity (race_recode_cat5), household income (hinc_i), and education (geduc1). The scale scores that end in _i include imputation for missing item responses; I drop the _i from the names I use in the code.

2 Import the data

The ICPSR file is an R data file (.rda), so it is read with load() rather than read_csv(). The object is assigned to trans_pop, converted to a tibble, and all column names are made lowercase so they are easy to type.

# load tidyverse & dplyr packages
library(tidyverse)
library(dplyr)

# load the TransPop data; this is my personal local file path
load("/Users/JPOcket/Downloads/ICPSR_37938 3/DS0001/37938-0001-Data.rda")

# assign the dataset to a shorter name, make a tibble, and lowercase the current column names
trans_pop <- as_tibble(da37938.0001) %>%
  rename_with(tolower)

3 Select and clean variables

The survey has several hundred columns, so I only selected the variables relevant to my analysis. The labeled factors in the ICPSR file carry numeric prefixes (ex. (6) White), which I stripped so the labels are readable.

After selecting these variables, I mutated/recoded the variables based on their different metrics (labeled categories, scale scores, and numeric codes), and cleaned up the data.

This output demonstrates the variables of interest in columns.

## # A tibble: 274 × 19
##    gender              age race     household_income education life_satisfaction
##    <chr>             <dbl> <fct>    <fct>            <fct>                 <dbl>
##  1 Trans woman (MTF)    65 White    $75,000 to $99,… Some col…              2.20
##  2 Trans man (FTM)      38 White    $30,000 to $39,… College                6.20
##  3 Trans man (FTM)      25 White    $50,000 to $59,… Some col…              4.20
##  4 Trans man (FTM)      18 White    $75,000 to $99,… HS less                1.80
##  5 Man                  30 Latino   $40,000 to $49,… Some col…              5.40
##  6 Man                  64 Multira… $15,000 to $19,… Some col…              6.40
##  7 Woman                61 White    $75,000 to $99,… More tha…              6.80
##  8 Man                  57 Multira… $1 to $4,999     Some col…              4   
##  9 Trans woman (MTF)    54 Black    $1 to $4,999     Some col…              2.40
## 10 Woman                21 White    $15,000 to $19,… Some col…              2.20
## # ℹ 264 more rows
## # ℹ 13 more variables: social_wellbeing <dbl>, kessler6 <dbl>,
## #   social_support <dbl>, internalized_transphobia <dbl>,
## #   everyday_discrimination <dbl>, healthcare_stereotype_threat <dbl>,
## #   hormone_treatment <dbl>, healthcare_satisfaction <dbl>,
## #   healthcare_access <dbl>, q171a_code <dbl>, q65 <fct>, q84 <fct>, q88 <fct>

4 Group observations

My three gender identity groups of interest are trans men, trans women, and GNB respondents. Respondents who identify simply as a man or a woman are excluded from the analysis sample. One thing to note is that some trans individuals could simply identify as man/woman without disclosing their trans identity in effort to remain stealth. This exclusion of data could be a confounding factor.

This output shows the number of respondents in each gender group:

## # A tibble: 3 × 3
##   gender          n percent
##   <fct>       <int>   <dbl>
## 1 Trans woman    66    36.9
## 2 Trans man      37    20.7
## 3 GNB            76    42.5

4.1 Recode variables

Since there is some variability in the variables I selected, I used mutate() to recode them to make them more digestible.

  • Everyday discrimination is checked against one raw item (q171a, coded 1 = Often to 4 = Never). A positive correlation means the scale is also “higher = less discrimination”, so I reversed it to make the higher number correlated with more discrimination.
  • Healthcare satisfaction is also reversed so that 5 = very satisfied.
  • Hormone treatment and cost barrier (q65 and q88: 1 = yes, 2 = no) become 0/1 variables.

To check that the recoding matches the original answers, here’s an example:

tp %>% count(q84, healthcare_satisfaction)
## # A tibble: 6 × 3
##   q84                     healthcare_satisfaction     n
##   <fct>                                     <dbl> <int>
## 1 (1) Very satisfied                            5    36
## 2 (2) Mostly satisfied                          4    64
## 3 (3) Neutral                                   3    14
## 4 (4) Mostly dissatisfied                       2     7
## 5 (5) Very dissatisfied                         1     1
## 6 <NA>                                         NA    57

4.2 Identifying missing values and a summary helper

I counted the missing values in each question’s variables and used drop_na() to keep respondents with complete data on only those variables. This keeps more respondents than dropping everyone with any missing value across all variables.

  • Skip patterns. Healthcare satisfaction is only asked of people who have a usual place for care, so complete cases over-represent people already connected to healthcare, who may be better off.
  • Missingness can relate to the outcomes. People who are more distressed or who have faced more discrimination may skip sensitive items, which would make the remaining sample look healthier.
  • Small sample. Each dropped respondent reduces power and can change the group mix, so I am also reporting group sizes before and after the drop.

To summarize each question’s variables by gender group, I pivoted the data long, computed the descriptive statistics, and pivoted wide again to turn the groups into columns. Scores are shown as mean (SD) and yes/no variables as n (%).

5 Question 1: How do healthcare satisfaction and access relate to happiness?

Variables: healthcare satisfaction, healthcare stereotype threat, cost barrier, and hormone treatment (healthcare), and life satisfaction (happiness).

5.1 Missing data

I counted the missing values, then used md.pattern() from the mice package to see how they overlap. Each row of the output is a pattern (1 = observed, 0 = missing). The left margin is the number of respondents with that pattern, the right margin is the number of missing variables in it, and the bottom row is the total missing per variable.

5.2 Summary by gender identity group

Mean (SD) by gender identity group, 1 = low, 5 = high
Variable Trans woman Trans man GNB
Healthcare satisfaction 4.28 (0.96) 4.13 (0.63) 3.83 (0.81)
Healthcare stereotype threat 3.22 (1.26) 3.43 (1.01) 3.57 (1.02)
Life satisfaction 3.84 (1.64) 3.88 (1.88) 4.01 (1.63)
n (%) answering yes, by gender identity group. Cost barrier is defined as whether or not here a time in the past 12 months when you needed to see a doctor but could not because of cost.
Variable Trans woman Trans man GNB
Cost barrier 10 (23.3%) 3 (13%) 17 (32.7%)
Hormone treatment 22 (51.2%) 14 (60.9%) 10 (19.2%)

5.3 Relationship

The correlation matrix shows how each healthcare variable moves with life satisfaction. Healthcare stereotype threat is defined as whether or not an individual’s healthcare experience is negatively affected due to their gender identity/sexual orientation.

An answer of 1 = strongly disagree and 5 = strongly agree (ex. “When seeking healthcare, I worry about being negatively judged because of my gender identity or sexual orientation?” An answer of 5 means they strongly agree with the negative impact of their gender identity/sexual orientation on their healthcare experience)

Here is a multiple linear regression model on life satisfaction interacting with healthcare satisfaction, stereotype threat, cost barrier, hormone treatment, and gender. The results demonstrate that stereotype threat is the strongest indicator of lower life satisfaction. In terms of the cost barrier, respondents who were not able to afford a doctor visit report around .9 points lower life satisfaction on the 1-7 scale.

## 
## Call:
## lm(formula = life_satisfaction ~ healthcare_satisfaction + healthcare_stereotype_threat + 
##     cost_barrier + hormone_treatment + gender, data = tp_q1)
## 
## Residuals:
##     Min      1Q  Median      3Q     Max 
## -2.9922 -1.0854  0.2621  1.0973  4.1878 
## 
## Coefficients:
##                              Estimate Std. Error t value Pr(>|t|)    
## (Intercept)                   5.37094    0.99873   5.378 4.23e-07 ***
## healthcare_satisfaction       0.04236    0.17691   0.239  0.81119    
## healthcare_stereotype_threat -0.55620    0.13255  -4.196 5.49e-05 ***
## cost_barrier                 -0.88619    0.33597  -2.638  0.00954 ** 
## hormone_treatment             0.54599    0.29827   1.831  0.06985 .  
## genderTrans man               0.02603    0.38357   0.068  0.94602    
## genderGNB                     0.64350    0.32381   1.987  0.04935 *  
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Residual standard error: 1.464 on 111 degrees of freedom
## Multiple R-squared:  0.2736, Adjusted R-squared:  0.2344 
## F-statistic: 6.969 on 6 and 111 DF,  p-value: 2.628e-06

The correlation graphs below demonstrated a positive correlation between overall healthcare satisfaction and life satisfaction, and a negative correlation between healthcare stereotype threat and life satisfaction.

Average life satisfaction by cost barrier for healthcare. The graph demonstrates a much higher rating in life satisfaction when cost barrier is NOT a factor in receiving healthcare. Lack of healthcare access is detrimental to everyone’s well-being, and can be amplified when one is also experiencing gender dysphoria.

Average life satisfaction correlated with access to hormone replacement therapy (HRT) treatment. The histogram demonstrates a slightly higher average in life satisfaction when trans individuals receive hormone therapy. This may be attributed to alleviated feelings of gender dysphoria experienced by many trans individuals.

8 Conclusion

Question 1: Healthcare and life satisfaction

Healthcare satisfaction was positively correlated with life satisfaction, and healthcare stereotype threat was negatively correlated with it. Respondents who could not afford a doctor visit also reported noticeably lower life satisfaction than those who could, suggesting that limited access to healthcare is linked to poorer well-being. Life satisfaction was also slightly higher among those who had received hormone treatment, which may reflect relief from gender dysphoria, though dysphoria was not measured as a variable here.

Question 2: Internalized transphobia and discrimination

More everyday discrimination was associated with somewhat higher internalized transphobia, a small-to-moderate positive relationship. This outcome may reflect the psychological distress one would experience after emotionally detrimental encounters related to their gender identity/sexual orientation and later leading to suppression or hatred of their identity.

Question 3: Social support and well-being

Higher perceived social support was associated with higher social well-being, even after accounting for life satisfaction and psychological distress. Gender identity groups did not differ significantly. Receiving more abundant and high quality social support would likely improve overall well-being due to feelings of acceptance and support.

9 Citation

Meyer, Ilan H. TransPop, United States, 2016-2018. Inter-university Consortium for Political and Social Research [distributor], 2021-06-23. https://doi.org/10.3886/ICPSR37938.v1