Who's Falling Through the Cracks?
An analysis of healthcare access barriers across race, income, and education — and how the pandemic temporarily narrowed gaps that are now at risk of widening again.
An analysis of healthcare access barriers across race, income, and education — and how the pandemic temporarily narrowed gaps that are now at risk of widening again.
Every year, millions of American adults skip medical care, delay treatment, or go without mental health support — not because they don't need it, but because they can't afford it. The CDC's National Health Interview Survey tracks these barriers across demographics, giving us a window into who the healthcare system serves and who it leaves behind.
We analyzed 23,609 data points spanning 54 health conditions and access metrics across six years (2019–2024), broken down by race, income, education, insurance status, and more. What we found is a story of progress, fragility, and persistent structural gaps.
Who faces the steepest barriers to healthcare? The answer depends on which lens you use — race, income, or education all reveal different dimensions of the same problem.
Looking at trends over time reveals a striking pattern: all four barrier types dropped sharply in 2020-2021 as pandemic-era policies took effect, then began rising again as those protections expired. By 2024, we're approaching pre-pandemic levels — and in some cases, exceeding them.
The heatmap reveals a stark pattern: uninsured adults face barriers 3-5× higher than those with private insurance across all four access metrics. Bisexual adults report the highest mental health care barriers at 28.5%, while those with disabilities consistently face elevated barriers across all categories.
These aren't abstract statistics — the gap between the best-off and worst-off subgroups translates to real health outcomes: earlier diagnoses, preventive care, chronic disease management, and ultimately, life expectancy differences.
The data shows a striking pattern: reported cost-related access barriers dropped sharply in 2020-2021, and gaps between high- and low-income groups narrowed. But the story is more complex than it appears.
Two forces were at play: expanded safety net policies (Medicaid continuous enrollment, enhanced unemployment, ACA subsidies) and universal healthcare avoidance due to lockdowns, fear of infection, and disrupted care systems. When everyone delays care — whether by choice or circumstance — gaps in reported access barriers can narrow even as actual healthcare utilization plummets.
The poverty gap in missed medical care shrank from 10.3 percentage points in 2019 to 5.0 in 2024 — a halving of the disparity. The trajectory bounced as pandemic protections expired (2022–2023), then dropped again. But this doesn't necessarily mean access improved for low-income groups — it may mean fewer people across the board sought care during peak pandemic years.
While self-reported cost barriers fluctuated, insurance coverage disparities barely budged. The racial gap in uninsured rates has hovered between 20–25 percentage points for six years straight. Emergency policies may have helped people afford care within the system — but they didn't bring excluded populations into the system, and they didn't address the fact that many people simply stopped accessing healthcare entirely.
Healthcare access barriers aren't evenly distributed across the country. At the county level, uninsured rates range from less than 2% to over 28% — a 14-fold difference. Border counties in Texas, rural Appalachia, and Native American territories face the steepest barriers, while coastal urban areas typically have the lowest rates. These disparities reflect differences in Medicaid expansion, local economies, and insurance market dynamics.
Using machine learning, we forecast which demographic groups will face the highest healthcare access barriers in the coming years. Choose between two complementary models:
This analysis is open. Download the cleaned data, explore the full methodology in our Colab notebook, fork the code, or plug the dataset into Tableau or Power BI.
Download our Tableau-ready CSV exports: full dataset, time series, disparities, and ML forecasts.
Download for TableauView the full analysis notebooks with all code, ML models, methodology, and commentary.
Open ColabClone the repo to extend our pipeline, add new models, or integrate new data sources.
View RepoData source: CDC National Health Interview Survey (NHIS) Adult Summary Health Statistics, 2019–2024. 26,208 records spanning 54 health topics across demographic, geographic, and socioeconomic classifications.
Cleaning: Removed 2,599 rows flagged as statistically unreliable (*, **, ***, ****, ---, -). Final dataset: 23,609 rows, 17 columns. No imputation was performed on missing estimates.
Derived metrics: Year-over-year change calculated per topic/subgroup combination. Confidence interval width used as a reliability indicator. Disparity gap computed as the difference between the highest and lowest estimate within each demographic group per year.
Supplemental data: U.S. Census Bureau's American Community Survey (ACS) 2022 county-level uninsured rates used for geographic visualizations. Fetched dynamically via Census API (848 counties with population >65,000).
Tools: Python (Pandas, Plotly, NumPy), Google Colab, Streamlit, GitHub Pages.
Centers for Disease Control and Prevention (CDC). (2024). National Health Interview Survey (NHIS): Adult Summary Health Statistics, 2019-2024. National Center for Health Statistics. https://www.cdc.gov/nchs/nhis/
U.S. Census Bureau. (2022). American Community Survey (ACS) 1-Year Estimates: Health Insurance Coverage (Table S2701). U.S. Department of Commerce. data.census.gov
DubsTech. (2026). DubsTech Datathon 2026: Healthcare Track. University of Washington. https://datathon-2026.webflow.io/
West Health. (2025, April 2). Inability to Pay for Healthcare Reaches Record High in U.S. West Health Policy Center. https://westhealth.org/news/inability-to-pay-for-healthcare-reaches-record-high-in-u-s/
Park, J., & Fung, V. (2024). Health Care Affordability Problems by Income Level and Subsidy Eligibility in Medicare. PubMed Central (PMC12455370). https://pmc.ncbi.nlm.nih.gov/articles/PMC12455370/
The Commonwealth Fund. (2024). Hispanic/Latino Adults Lack Adequate, Affordable Health Insurance Coverage. Commonwealth Fund Blog. https://www.commonwealthfund.org/blog/2024/hispaniclatino-adults-lack-adequate-affordable-health-insurance-coverage
Python Software Foundation. Python Programming Language (version 3.8+). Libraries: Pandas, NumPy, scikit-learn, XGBoost.
Plotly Technologies Inc. Plotly.js: Open-source JavaScript Graphing Library. https://plotly.com/javascript/
GitHub, Inc. GitHub Pages: Static Site Hosting. https://pages.github.com/
Google LLC. Google Colaboratory: Cloud-based Jupyter Notebook Environment. https://colab.research.google.com/