Key takeaways from the multi-center record analysis
- Researchers from Stanford University and the University of California, Berkeley analyzed electronic health records from roughly 2.4 million U.S. patients diagnosed with acute COVID-19 between January 1 2022 and March 31 2023.
- Black patients were diagnosed with long COVID at a 6% lower rate than White patients, yielding a controlled direct effect relative risk of 0.938 with a 95% confidence interval of 0.887 to 0.992.
- No statistically significant diagnostic rate differences were detected between White patients and those who identified as Hispanic or Latino, Asian American or Pacific Islander, or American Indian or Alaska Native.
- Overall diagnostic documentation remains low across all groups, with under 1% of electronic health records showing a formal long-COVID diagnosis compared to prior research estimates where over 10% of acute cases develop post-acute sequelae.
- Pulmonary symptoms such as shortness of breath and cough showed the strongest relative correlation with receiving a diagnosis across most demographic groups, though these associations remained overall weak.
Comparative data from the Health Affairs analysis
| Variable | Study metric or finding |
|---|---|
| Study population size | Approximately 2.4 million U.S. patients |
| Acute infection window | January 1 2022 to March 31 2023 |
| Lead academic institutions | Stanford University and University of California, Berkeley |
| Diagnostic disparity for Black patients | 6% lower diagnosis rate relative to White patients (RR 0.938) |
| EHR long COVID diagnostic rate | Under 1% of acute COVID cases |
| Estimated true long COVID incidence | Greater than 10% based on prior research estimates |
| Symptoms evaluated | 222 distinct symptoms tracked within one year |
Uncovering systemic diagnostic gaps in post-acute care
A large-scale investigation published in Health Affairs in July 2026 indicates that Black patients in the United States are less likely to receive a formal diagnosis for long COVID than White patients presenting with similar health records. Long COVID diagnosis disparities can leave patients without access to targeted clinical management.
The research team, drawing from Stanford University and the University of California, Berkeley, evaluated medical data from approximately 2.4 million U.S. patients who tested positive for acute COVID-19 between January 1 2022 and March 31 2023. To qualify for the symptom analysis, individuals needed to visit a healthcare provider at least bimonthly over the course of a year following their initial infection.
The analytical models controlled for multiple background variables. Despite seeking care with equal frequency, Black patients showed a 6% lower rate of formal long-COVID diagnoses compared to White patients. The study authors reported a controlled direct effect relative risk of 0.938, with a 95% confidence interval spanning from 0.887 to 0.992.
When comparing White patients against other racial and ethnic cohorts, researchers found a different statistical picture. There was no statistically significant difference in diagnostic rates when comparing White patients to Hispanic or Latino patients, Asian American or Pacific Islander patients, or American Indian or Alaska Native patients.
Understanding long COVID and electronic health record tracking
Long COVID, technically termed post-acute sequelae of SARS-CoV-2 infection (PASC), involves persistent symptoms lasting four weeks or longer after an initial infection. These manifestations range from chronic fatigue and dyspnea to cognitive dysfunction. Because clinicians lack a single blood test or imaging scan to confirm the condition, diagnosis depends heavily on clinical judgment, subjective symptom reporting, and active provider screening.
Electronic health records serve as digital archives of a patient’s medical history. Researchers rely on these large datasets to track epidemiological patterns across millions of lives. However, medical codes only capture what a clinician explicitly diagnoses and inputs into the system.
A stark discrepancy emerged when scientists compared actual medical record codes against broad epidemiological projections. Prior research estimates that over 10% of acute COVID-19 patients develop long COVID. In contrast, this analysis showed that under 1% of patients in the electronic health record database received a formal diagnostic code for the condition.
This gap suggests that the vast majority of lingering post-viral illness goes completely unrecorded in formal clinical coding systems.
How provider decision making and diagnostic patterns interact
To understand what drives a clinician to assign a long-COVID code, the researchers tracked 222 distinct symptoms recorded within twelve months of initial infection. They examined how specific complaints influenced the likelihood of receiving an official diagnosis.
Across nearly every racial and ethnic category, pulmonary complaints provided the strongest push toward a diagnosis. Symptoms like persistent cough and shortness of breath correlated most closely with clinician coding, acting like a direct signal that prompted doctors to log post-acute viral complications. But even for breathing issues, the overall strength of that correlation remained weak.
One demographic exception stood out. For Asian American and Pacific Islander patients, loss of taste or smell acted as the single strongest symptom associated with getting a formal long-COVID diagnosis.
Why do these diagnostic gaps persist? The study authors observed that objective diagnostic tools alone cannot fully erase healthcare inequities. Implicit bias can alter provider choices at every stage, influencing who gets screened, how effectively symptoms are investigated, and how diagnostic results are interpreted.
Broad population impact and health equity consequences
Black Americans account for approximately 13% of the total United States population. When a diagnostic system misses 6% of cases in a group that size, the absolute numbers add up fast. Hundreds of thousands of individuals may be living with chronic post-viral symptoms without receiving an official designation in their charts.
A missing code is not just a clerical oversight. Without an accurate diagnosis in their electronic health records, patients frequently miss out on critical specialty referrals, physical rehabilitation programs, and emerging post-viral therapies.
Under-diagnosis also distorts public health planning. When administrative databases undercount post-acute illness among Black communities, health systems risk underestimating local resource demands, leading to fewer dedicated recovery clinics and inadequate community support services.
The authors noted that targeted outreach, implicit bias education for medical providers, and standard diagnostic tools represent necessary steps toward closing these equity gaps.
Structural coverage barriers and insurance disparities
Diagnostic disparities rarely happen in isolation; they intersect directly with structural healthcare access. Data from health policy analyses show that insurance status heavily dictates how quickly patients can obtain specialist evaluations for complex chronic conditions like long COVID.
In states that expanded Medicaid under the Affordable Care Act, income eligibility extends up to 138% of the Federal Poverty Level, covering individuals earning approximately $22,025 in 2026. However, in ten states that have not expanded Medicaid, roughly 1.2 million uninsured adults fall into a coverage gap, earning too much to qualify for traditional Medicaid but too little to afford private plans.
Black and Hispanic adults represent 56% of the population trapped in this coverage gap. Furthermore, 39% of uninsured adults report delaying or forgoing medical care due to cost concerns, compared to just 17% of insured adults. When financial barriers prevent patients from seeing a doctor regularly, the odds of securing a complex diagnosis like long COVID drop even further.
Key caveats in electronic health record analysis
Evaluating real-world medical data carries specific limits that readers should weigh alongside the headline findings. Electronic health record analyses observe recorded clinical behavior rather than controlled laboratory environments.
First, under-coding affected the entire dataset. Because less than 1% of the total sample received a formal diagnostic code, the overall sample of diagnosed patients represents a tiny fraction of the estimated true disease burden. This broad under-reporting could introduce unmeasured noise into relative risk estimates.
Second, symptom-to-diagnosis correlations proved weak across all racial groups. While pulmonary symptoms were the top driver for most patients, their modest statistical predictive power indicates that unmeasured factors outside of documented symptoms strongly influence whether a clinician enters a diagnostic code.
Finally, the study brief did not outline specific statistical adjustments for individual socioeconomic status, local healthcare facility quality, or detailed pre-existing medical conditions. Further research will need to untangle how these overlapping variables contribute to diagnostic variation across diverse patient communities.
Related Coverage
- A cheap old antidepressant eased long COVID fatigue. The benefit faded once people stopped taking it
Sources:
- CIDRAP, “Many Black US patients with long COVID being missed, analysis reveals” (https://www.cidrap.umn.edu/covid-19/many-black-us-patients-long-covid-being-missed-analysis-reveals)
- KFF, “How Many Uninsured Are in the Coverage Gap and How Many Could Be Eligible If All States Adopted the Medicaid Expansion?” (https://www.kff.org/medicaid/how-many-uninsured-are-in-the-coverage-gap-and-how-many-could-be-eligible-if-all-states-adopted-the-medicaid-expansion/)
Disclaimer: This article is for general information only and does not constitute medical advice, diagnosis, or treatment. Always consult a qualified healthcare provider about any medical condition or before making health decisions.