Intersectional Differences: Exploring Healthcare Barriers and Access Using the All of Us Researcher Workbench

Miguel Fudolig, Ph.D., Kavita Batra, Ph.D., Ravi Batra, Jennifer Pharr, Ph.D., Emylia Terry, Brisa Rodriguez Alcantar, Liliana Davalos, James Navalta, Ph.D., Christopher Johansen, Ph.D.

Outline

  • Acknowledgments
  • Background
  • All of Us Data Set
  • Objectives
  • Methods
  • Results
  • Discussion/Implications

Funding Acknowledgment

The authors would like to acknowledge the support received for this project:

Funding Disclosure

All of Us Research Academy Institutional Champion Award Sponsor: RTI International Award Number: 19-312-0217703-67774L

Guidance and Technical Assistance

We’d like to thank Drs. Stefanee Tillman, Hunter McGuire, and Barrett Montgomery for their technical assistance over the last year on this project.

Background

Access to Care

To provide quality care to our community, we must first address the barriers they face in accessing health care services.

Important

Underserved and disadvantaged communities experience these barriers more often, leading to delay of care and poorer health outcomes.

Avoidance of Health Care

Sexual and Gender Minority Groups

Avoidance of necessary health care was prevalent in sexual and gender minority (SGM) adults. This could be attributed to identity discordance between patient and provider and previous experiences of discrimination from health care providers.

Racial and Ethnic Minority Groups

Avoidance of necessary health care was prevalent in racial and ethnic minority groups. This could be attributed to historical events, potential of cultural misunderstanding between patient and provider, and previous experiences of discrimination from health care providers.

Intersectionality

What is Intersectionality?

The overlap of multiple social identities—such as race, gender, class, and sexuality—shapes with oppression and discrimination.

Why is Intersectionality Important?

Important

Different forms of social inequality such as systemic racism, sexism, classism, and ableism work together to influence health.

Important

Rather than looking at factors like race or gender in isolation, intersectionality theory emphasizes how these identities interact and produce unique health risks, challenges, and opportunities for individuals from marginalized groups.

Barriers to Health Care

Cost Barriers

There are economic reasons to delay care such as high copay costs, high insurance plan deductibles, and high out-of-pocket costs.

Non-Cost Barriers

Aside from cost-related barriers, there are also sociocultural barriers that can delay care. Examples are patient-provider identity discordance, availability of care for children and older adults, rurality of residence, lack of transportation, anxiety, or nervousness.

All of Us Research Program

What is the All of Us Research Program?

All of Us Research Program: The Drive for Diversity in Health Data

The All of Us Research Program is an initiative by the National Institutes of Health, Office of the Director.

The main goal of the All of Us Research Program is to recruit and follow 1 million participants that include individuals from underrepresented communities.

The program partners with academic institutions, health care organizations, and community partners to accelerate advances in biomedical research and precision medicine for everyone.

What’s in the All of Us Data Set?

Included Data

  • Electronic Health Records
  • Participant Surveys
  • Genomics Data
  • Physical Measurements
  • Wearable/Mobile Tech Data

Objectives

Main Aims

The aims of this exploratory, cross-sectional study are:

  • Identify barriers to healthcare access faced by groups with intersectional identities.
  • Compare the likelihood of reporting barriers for different intersectional identities.

Methods

Study Design

This study was a cross-sectional study utilizing a secondary analysis of available data through the All of Us Researcher Workbench.

All of Us Researcher Workbench

The All of Us Researcher Workbench is a secure cloud-based platform for data analysis and collaboration.

Note

The Researcher Workbench includes dataset and cohort builders to create cohorts for analysis.

The Research Workbench also includes different interfaces for Python (Jupyter), R (Jupyter, RStudio), and SAS (SAS Studio) implementation.

Cohort Builder - Inclusion Criteria

The cohort was created using the Cohort Builder in the All of Us Researcher Workbench.

Inclusion Criteria

The inclusion criteria are as follows:

  • Ethnicity reported as “Hispanic”, “Not Hispanic”, or “None of These”
  • Race reported as “White”, “Black or African-American”, “Asian”, “More than One Population”, “Middle Eastern or North African”, “Native Hawaiian or Pacific Islander”
  • Gender identity reported as “Woman”, “Man”, “Transgender (Man/Woman)”, “Non Binary”, “Additional Options”, “Unknown”
  • Sex at birth reported as “Female”, “Male”, “Unknown”, “None”, “Intersex”

Cohort Builder - Exclusion Criteria

Exclusion Criteria

The exclusion criteria are as follows:

  • Reported more than one sexual orientation (“Gay”, “Lesbian”,“Bisexual”,“None”, “Straight”)
  • Answered “None of These” or “Prefer Not to Answer” for race and ethnicity questions.

The total sample size of the cohort was N=405,307 participants.

Data Management

Missing Data

In the formal statistical analysis, listwise deletion was implemented in the case of missing responses.

Data Aggregation

Due to data sparsity, some intersectional groups are aggregated/excluded in the formal statistical analysis.

  • Asian Americans, Native Hawaiian or Pacific Islander \(\to\) AANHPI
  • Multiracial, Middle Eastern and North African \(\to\) Other (low representation in the data set)
  • Transgender Men, Transgender Women \(\to\) Transgender

Statistical Analysis

Binomial generalized linear models (GLM) were used to estimate the probability of reporting reasons for delay of care from the Health Care Access and Utilization survey. These reasons include the following

  • Lack of Transportation
  • Rurality of Residence
  • Nervousness
  • Could not get Time Off from Work
  • Could not get Child Care
  • Need to care for an older adult in the household
  • Copay Issues
  • Deductible Issues
  • Out-of-Pocket costs
  • Provider Identity Discordance

Statistical Analysis

Interaction analysis between race, ethnicity, sexual orientation, and gender identity was performed using R through the Jupyter notebook interface in the All of Us Researcher Workbench.

The model included the following two-way interactions: race/ethnicity and sexual orientation and race/ethnicity and gender identity

Adjusted odds ratios were calculated between different intersectional identities. The reported confidence intervals were adjusted using the Tukey-Kramer adjustment to account for multiplicity.

Warning

Due to data sparsity, the model could not include a three-way interaction term defined by race and ethnicity, sexual orientation, and gender identity.

Warning

Due to some intersectional groups having lower sample sizes compared to others, formulating conclusions based on some contrasts may lead to misleading results.

Results

Demographic Variables (N=405,307)

  • Median Age: 25
  • IQR: 29
  • AANHPI: 14,288 (3.51%)
  • Non-Hispanic Black: 76,988 (18.99%)
  • Hispanic: 74,001 (18.26%)
  • Non-Hispanic White: 222,170 (54.82%)
  • Other Racial Groups:9,159 (2.26%)
  • Bisexual: 14,893 (3.67%)
  • Gay: 9,323 (2.30%)
  • Heterosexual: 354,187 (87.39%)
  • Lesbian: 5,053 (1.25%)
  • None (Queer): 8,411 (2.08%)
  • Man: 150,798 (37.21%)
  • Non Binary: 1,149 (0.28%)
  • Transgender: 1,451 (0.36%)
  • Woman: 242,779 (59.90%)
  • Other options: 368 (0.09%)

Demographic Variables

  • Advanced Degree: 87,787 (21.66%)
  • College Graduate: 91,749 (22.64%)
  • College Undergraduate: 101,958 (25.16%)
  • Twelve or GED: 75,385 (18.60%)
  • Never Attended: 543 (0.13%)

90.07% of participants reported to have health insurance.

  • More than $200k: 26,539 (6.55%)
  • Less than $10k: 54,240 (13.38%)
  • $100k-$200k: 60,757 (14.99%)
  • $10k-$100k: 183,215 (45.2%)

Barriers to Healthcare Access

Important

Among those who answered the survey (excluding missing data), the three most reported barriers were out-of-pocket costs (18.18%), nervousness (13.72%), and patient-provider identity discordance (race/religion) (13.01%).

The other barriers had the following report rates:

  • inability to take time off from work (11.33%)
  • Deductible (10.64%)
  • Copay (9.28%)
  • Transportation (7.51%)
  • Rural Residence (3.30%)
  • Child Care Unavailability (3.19%)
  • Caring for an Older Adult (2.04%)

Out-of-Pocket Costs

Out-of-Pocket Costs

Some key results include:

  • Individuals who identify as bisexual were 70% (95%CI:[48%,94%]) more likely to report delaying care due to high out-of-pocket costs compared to heterosexual individuals after averaging over all race groups.

  • Gay, non-Hispanic, White participants were 89% more likely to report compared to gay, non-Hispanic, Black participants.

Some key results include:

  • For non-Hispanic White participants, non binary (OR = 2.11, 95% CI: [1.64,2.73]), transgender (OR=2.21, 95%CI:[1.57,3.11]), and women (OR=1.59, 95%CI:[1.52, 1.67]) were more likely to report delaying care due to high out-of-pocket costs compared to men.

  • For non-Hispanic Black participants, men were less likely to report than women (OR = 0.85, 95%CI: [0.75, 0.96]).

Nervousness

Nervousness

Some key results include:

  • Among non-Hispanic White participants, queer individuals (“None”) had the highest reported rate of nervousness (40.5%, 95%CI:[37.8%,43.2%]) followed by bisexual individuals (36.6%, 95%CI:[34.4%,38.8%]).
  • Compared to heterosexual, non-Hispanic, White participants, queer and bisexual participants were 3.08 [2.82, 3.36] and 3.63 [3.15,4.19] times more likely to report nervousness, respectively.
  • Among bisexual and queer participants, there were no statistically significant racial and ethnic disparities in reported rate of delay of care due to nervousness.

Some key results include:

  • Across racial groups, non binary and transgender participants reported higher rates of delay of care due to nervousness compared to cisgender men and women.
  • Non-Hispanic Black (OR=0.59, 95%CI:[0.49,0.70]) and Hispanic (OR=0.80, 95%CI:[0.49,0.70]) women were less likely to report delay of care due to nervousness compared to non-Hispanic White women.

Identity Discordance

Identity Discordance

Some key results include:

  • Among heterosexual participants, non-Hispanic White participants reported the lowest rate of delay of care due to identity discordance (12.6%, 95%CI=[11.7%,13.5%]).
  • In the Black and Hispanic participant groups, bisexuals and queer participants had the highest reported rate of delay of care.
  • For AANHPI participants, queer and lesbian participants reported the highest rate of delay of care.

Some key results include:

  • For AANHPI cisgender participants, men were less likely to delay care compared to women (OR=0.76, 95% CI:[0.63, 0.92]). Similar trends were observed in non-Hispanic White (OR=0.54, 95% CI:[0.51,0.57]) and Hispanic (OR=0.78, 95%CI:[0.70,0.87]) participants.

  • Across all racial groups, non binary (38.1%) and transgender (35.0%) reported the highest rates of delay of care after averaging across racial groups.

Discussion

Summary

  • We found disparities in the reporting rates of different barriers to health care access between intersectional identities.

  • The three most reported barriers were a mix of cost-related barriers (out-of-pocket costs) and non-cost related barriers (identity discordance, nervousness).

Implications

Stigma

Stigma experienced by minority groups from healthcare providers could have been a factor.

Sociocultural Differences

Cultural differences between provider and patient’s intersectional identities could have caused hesitation in availing health care.

Cost-Related Barriers

Economic disparities across axes of intersectional identities were also observed.

Limitations

  • Associations, not causations

  • Limited representation of intersectional identities

  • Lack of covariates due to data sparsity

  • Lack of power to detect significant statistical effects due to sparsity.

  • AANHPI had low representation of some SGM groups, hence a lot of statistically insignificant results even with high effect sizes.

Recommendations

  • Account for intersectional identities in addressing barriers to healthcare access
  • Identify clusters of reported barriers to healthcare access

Thank you!

Questions? Email me at miguel.fudolig[at]unlv.edu.

Intersectional Inequities (Fudolig et. al.)- 2025 CCPH Webinar