Abstract: Using a Directed Acyclic Graph to Support Instrumental Variable Estimation of Discrimination's Effect on Psychological Distress (Society for Social Work and Research 27th Annual Conference - Social Work Science and Complex Problems: Battling Inequities + Building Solutions)

All in-person and virtual presentations are in Mountain Standard Time Zone (MST).

SSWR 2023 Poster Gallery: as a registered in-person and virtual attendee, you have access to the virtual Poster Gallery which includes only the posters that elected to present virtually. The rest of the posters are presented in-person in the Poster/Exhibit Hall located in Phoenix A/B, 3rd floor. The access to the Poster Gallery will be available via the virtual conference platform the week of January 9. You will receive an email with instructions how to access the virtual conference platform.

Using a Directed Acyclic Graph to Support Instrumental Variable Estimation of Discrimination's Effect on Psychological Distress

Friday, January 13, 2023
Desert Sky, 3rd Level (Sheraton Phoenix Downtown)
* noted as presenting author
Tural Mammadli, MSW, PhD Student, University of Maryland Baltimore, Baltimore, MD
Sarah Tanveer, Doctoral Student, University of Maryland at Baltimore
Background and purpose: Transgender people experience discrimination in multiple aspects of their lives. Literature supports a relationship between discriminatory events experienced by transgender people and adverse mental health outcomes. A randomized control trial with a potentially harmful exposure like discrimination is unethical, making it a challenge to understand causal impacts. An underused resource in social work research, Instrumental Variable Estimation (IVE) is a promising approach to obtaining causal estimates when working with observational data. IVE requires that the data meet three assumptions: relevance (association of instrument and exposure), independence (no confounding), and exclusion (the effect of the instrument on the outcome is through the exposure only). Directed Acyclic graphs (DAGs) work well with IVE; they visually represent confounding paths and help identify instruments that satisfy the assumptions of IVE. In the current study, we demonstrate the use of DAGs and IVE to obtain a causal estimate of experiences of discrimination on psychological distress, an application of the third step of using a DAG in social work research.

Methods: Data were obtained from the TransPop U.S. Transgender Population Health Survey, a cross-sectional survey administered between 2015 and 2018 (N=274). Data include respondent’s demographic characteristics, mental health outcomes, and experiences with discrimination and victimization. Using the Minority Stress Model and empirical evidence, a DAG was constructed to identify instruments for discrimination. An ordinary least squares model was run first. Then, sexual orientation, race, and gender of the participants were used as instruments in unconditional two-stage least squares (2SLS) regressions to examine the causal impact of exposure to discrimination (modified Everyday Discrimination Scale-9) on the psychological distress (Kessler-6) levels of transgender people. The DAG suggested that the independence assumption could not be met without conditioning on income, education, and disability, so conditional 2SLS was used as well.

Results: A majority of the participants were White (68%) and non-heterosexual (79%), and a plurality identified as transgender men (44%). The average participant had a psychological distress score of 9.3 and a discrimination score of 18.7. OLS regression results showed that exposure to discrimination is a significant predictor of levels of psychological distress. The IVE findings suggest that higher exposure to everyday discrimination causes higher psychological distress in transgender participants, with the impact higher in the unconditional model.

Conclusion: Our results show that exposure to everyday discrimination is associated with greater psychological distress among the transgender participants. Comparison of the findings from multivariate regression and IVE showcase how biased estimates can result from incorrectly assuming that exposure to discrimination is random. While recognizing limitations of IVE, especially as they pertain to its assumptions, this study demonstrates the utility of robust causal inference methods, such as IVE in a discipline like social work. Use of DAGs clearly highlight casual assumptions made in the research design, and if the DAG was correctly specified, these assumptions were met. This study demonstrates the value of combining IVE with DAGs to support the use of this method, and address its limitations related to independence, relevance, and exclusion assumptions.