Reducing racial disparities in health care requires changes at every level, from how clinical tools are designed to how insurance programs reach different communities. The gaps are stark: Black mothers die from pregnancy-related causes at more than three times the rate of white mothers (43.3 versus 13.8 per 100,000 live births, per CDC provisional data through September 2025). Closing these gaps isn’t a single fix. It’s a combination of policy reform, clinical redesign, workforce changes, and better data collection working together.
Remove Race From Clinical Algorithms
For decades, medical formulas baked race into their calculations in ways that directly harmed patients of color. The most well-known example is the kidney function test (eGFR), which inflated estimated kidney function in Black patients by roughly 16%. The practical result: Black patients were referred to specialists later, excluded from clinical trials, and placed on kidney transplant lists at more advanced stages of disease.
The National Kidney Foundation and the American Society of Nephrology have since recommended dropping race from these equations entirely. Newer race-neutral formulas that combine two blood markers (creatinine and cystatin C) are more accurate and produce smaller gaps between Black and non-Black patients than the old approach. Similar corrections have happened across other specialties. Lung function testing now uses equations from the European Respiratory Society and American Thoracic Society that average results across racial groups rather than applying separate race-based benchmarks. A calculator used to predict successful vaginal birth after cesarean once assigned Black and Hispanic women a 15 to 18 percent lower likelihood of success with no clear biological basis. A 2021 revision replaced race with clinical factors like chronic hypertension and performed just as well. In cardiovascular risk assessment, the American Heart Association’s PREVENT model, introduced in 2020, replaced race with biomarkers, medication history, and measures of social deprivation. It matched the predictive accuracy of the older race-adjusted model across racial groups.
If you work in a clinical setting, auditing which algorithms still use race as a variable is one of the most concrete steps available. When race can be replaced with actual physiological or social data, accuracy holds up and equity improves.
Expand Insurance Coverage Strategically
Medicaid expansion under the Affordable Care Act increased health insurance coverage among low-income adults by about 7 percentage points overall, with the largest gains among white adults (11 percentage points). But the benefits didn’t reach everyone equally. Hispanic adults in expansion states saw virtually no coverage increase (a statistically insignificant change of negative 1.2 percentage points), which actually widened the disparity for that group.
The reasons are structural. Immigration status restrictions, language barriers in enrollment systems, and fear of public charge consequences all limit uptake. Expanding coverage on paper doesn’t translate to expanded access unless enrollment outreach is designed with these barriers in mind. States that paired Medicaid expansion with multilingual application assistance, community-based enrollment navigators, and protections for mixed-status families saw more equitable results. The lesson is that coverage policy alone is necessary but insufficient. The enrollment infrastructure has to be built for the populations that face the most barriers.
Rethink Implicit Bias Training
Implicit bias training has become a standard intervention in hospitals and medical schools, but the evidence for its effectiveness is thin. A systematic review of 77 studies on implicit bias training for health care providers found that only five (6.5%) actually measured behavioral changes among providers. Not a single study examined whether the training improved patient outcomes. Of the handful of studies that did look at attitudes or behaviors, fewer than half found a significant change. There is no evidence that these trainings produce long-term behavioral change.
This doesn’t mean awareness of bias is irrelevant. It means that one-off workshops are not enough to change how providers interact with patients. More promising approaches embed equity into clinical workflows: standardized treatment protocols that reduce the role of individual discretion, checklists that ensure all patients receive the same screening recommendations, and structured communication tools that don’t rely on a provider’s gut feeling about who “looks like” they need a particular test. Systems-level changes tend to outperform individual attitude interventions because they don’t depend on every clinician sustaining a personal shift.
Diversify the Health Care Workforce
Racial concordance between patients and providers, where both share the same racial background, has a complicated evidence base. A review of 27 studies found that concordance was associated with positive health outcomes for minority patients in only 9 (33%). Eight studies found no association, and 10 showed mixed results. The picture isn’t as simple as “matching” patients and doctors by race.
That said, specific studies reveal meaningful effects in targeted areas. Black patients with Black providers received certain HIV medications significantly earlier than Black patients with white providers. In some settings, Black patients using public health clinics were more likely to take prescribed medications when their physician was Black. The overall evidence suggests concordance matters most in contexts where trust affects whether patients follow through on treatment, particularly for chronic disease management and preventive care. Increasing the pipeline of providers from underrepresented backgrounds addresses this, not because every patient needs a same-race provider, but because a more diverse workforce broadens trust across the system and brings firsthand understanding of the barriers patients face.
Collect Better Data
You can’t fix disparities you can’t measure. The federal government updated its standards for collecting race and ethnicity data in 2024 through a revised statistical directive that now requires more detailed categories, multiple checkboxes, and write-in options rather than broad groupings. Federal agencies must also disclose whether the data was self-reported or collected by proxy, and describe any coding procedures used.
For Medicaid and the Children’s Health Insurance Program, states are required to use standardized applications that capture language preference and other demographic information. But for most health care organizations outside the federal system, these standards remain non-binding suggestions. Hospitals and health systems that voluntarily adopt detailed race, ethnicity, and language data collection can stratify their own quality metrics and identify where gaps exist in their patient populations. Without that data broken down by group, a hospital might report excellent average outcomes while specific communities within its service area experience far worse care.
Close the Telehealth Access Gap
Telehealth expanded rapidly during the pandemic and has real potential to reach underserved communities, but only if digital access barriers are addressed head-on. The CDC identifies three key strategies that work. First, assessing patients’ access to devices, internet, and digital literacy as part of routine social needs screening, ideally built into existing intake workflows rather than added as a separate step. Second, adapting telehealth materials to reflect patients’ languages, cultural contexts, and literacy levels. Generic patient portal instructions in English don’t help a patient whose primary language is Spanish and whose digital literacy is limited.
Third, and most effective in practice, is using digital navigators and community health workers to walk patients through the technology. These aren’t IT support lines. They’re trusted individuals, often from the same communities they serve, who help patients set up accounts, troubleshoot video visits, and build comfort with digital tools over time. Health systems that deployed community health workers as telehealth facilitators saw higher adoption rates among patients who would otherwise have been left behind by the shift to virtual care.
Address Maternal Health Specifically
Maternal mortality is one of the sharpest illustrations of racial disparity in American health care. Black mothers die at 43.3 per 100,000 live births compared to 13.8 for white mothers. This gap persists across income and education levels, which means it cannot be explained by socioeconomic factors alone.
Effective interventions include state-level perinatal quality collaboratives that standardize how hospitals respond to obstetric emergencies like hemorrhage and preeclampsia. When every patient receives the same evidence-based bundle of care regardless of race, outcomes improve. Doula support programs, particularly those using community-based doulas who share patients’ cultural backgrounds, have been associated with lower rates of cesarean delivery and higher patient satisfaction. Several states now cover doula services through Medicaid. Extending postpartum Medicaid coverage from 60 days to 12 months, which more than 40 states have now done or are implementing, addresses the fact that a significant portion of maternal deaths occur in the weeks and months after delivery, precisely when many low-income mothers previously lost their insurance.