Common EHR inclusion criteria quietly reshape who counts as a long COVID patient

Electronic health records have become one of the main tools for studying long COVID, but the criteria researchers use to build patient cohorts are far from neutral. A new study in BMC Medicine shows that routinely applied inclusion rules can substantially change the size, demographic composition, and healthcare utilization patterns of long COVID cohorts identified through the U09.9 diagnosis code.

The research drew on harmonized electronic health record data from the National COVID Cohort Collaborative, known as N3C. The study population included nondeceased individuals with at least one U09.9 long COVID diagnosis recorded between October 2021 and October 2023. Within that cohort, the team evaluated three additional inclusion criteria that researchers commonly apply: a documented COVID-19 diagnosis, a likely COVID-19-related hospitalization, and the availability of at least two years of prior medical history.

To measure how these criteria shaped cohort composition, the investigators used mixed-effects logistic and multinomial regression models to estimate associations between each selection criterion and sex, race and ethnicity, area-level poverty, and health insurance coverage. Zero-inflated negative binomial models were used to assess estimates of healthcare utilization. The data-contributing partner was included as a random effect in all models, and sensitivity analyses examined regional variation and compared the findings with a COVID-19-only cohort.

The results showed that cohort size, demographic makeup, and healthcare utilization varied substantially depending on which inclusion criteria were applied. Requiring at least two years of medical history increased the representation of women and of individuals with higher outpatient utilization, but it reduced the inclusion of people from areas with low insurance coverage.

Requiring a COVID-19-related hospitalization produced a different profile. This criterion was associated with higher inpatient and non-inpatient utilization, including acute care clinic visits and emergency room visits, along with increased representation of non-Hispanic Black individuals but lower representation of women.

Having a documented COVID-19 diagnosis enriched cohorts with patients from higher-poverty and lower-insurance-coverage areas and with greater healthcare utilization overall. Meanwhile, multinomial models revealed that medical history requirements were negatively associated with the inclusion of all non-White racial and ethnic groups, a finding with direct implications for how disparities in long COVID outcomes are measured and interpreted.

The researchers report that their findings were largely consistent across geographic regions and aligned with the patterns observed in the COVID-19-only cohort, suggesting that the effects of inclusion criteria are not merely artifacts of one particular region or coding strategy.

The authors conclude that routinely used EHR-based inclusion criteria are not analytically neutral. In the N3C setting, cohort composition reflects site participation, data harmonization, common data model implementation, and data quality processes. Although these features are specific to N3C, the team notes that similar design considerations may affect disparities research and generalizability across EHR-based long COVID studies more broadly.

For researchers and clinicians interpreting long COVID evidence, the study is a caution: the way a cohort is assembled can shape who is counted, whose utilization patterns are measured, and to whom the findings may generalize.

Subject of Research: How electronic health record inclusion criteria affect the representativeness of long COVID research cohorts identified with the U09.9 diagnosis code.

Article Title: Assessing the impact of additional inclusion criteria on the representativeness of EHR-based long COVID cohorts identified using U09.9

Article References: Madlock-Brown, C., McGrath, B. M., Wilkins, K. J., Seltzer, J., Anderson, A. R., Sharathkumar, A., Davis, H., & Weiskopf, N. (2026). Assessing the impact of additional inclusion criteria on the representativeness of EHR-based long COVID cohorts identified using U09.9. BMC Medicine. https://doi.org/10.1186/s12916-026-05326-5

Image Credits: AI Generated

DOI: 10.1186/s12916-026-05326-5

Keywords: long COVID, electronic health records, U09.9, N3C, cohort representativeness, healthcare utilization, health disparities, inclusion criteria, BMC Medicine, regression models, data harmonization, Assessing