Identifying Health Disparities When Race and Ethnicity are Missing

Project Summary

The project team aims to address and overcome lack of access to high-quality race and ethnicity data in Electronic Health Records (EHRs) by analyzing discrepancies in race reporting and investigating factors behind hesitancy to disclose race.

Research Questions/Aims

  1. How much discordance exists between reported race/ethnicity in health and administrative records, and how does this discordance impact our understanding of racial health disparities?
  2. How well do statistical methods recover race/ethnicity information, and how does missingness impact our understanding of racial/ethnic health disparities?
  3. How well do statistical methods disaggregate race in healthcare records, and how does disaggregation impact our understanding of racial health disparities?

Actionability

  • Develop and encourage use of best practices in clinical settings related to race reporting and identification of health inequities;
  • Empower communities and community-based organizations with evidence to advocate for health equity; and
  • Inform local, state, and federal decision makers of the knowledge transfer necessary to improve racial inequity assessments through partnerships with government agencies.

Methodology

The team will first quantify and characterize the degree to which race and ethnicity are reported in EHRs and Census Bureau micro data and then analyze concordance and discordance rates by race/ethnicity. By investigating factors that explain hesitancy to report race and ethnicity, the team will then engage in rigorous evaluation of statistical methods to impute race/ethnicity when it is never reported. In advance of the implementation of new federal standards for disaggregated race/ethnicity reporting, the team will then develop and test methods to impute race/ethnicity at a disaggregated level.

Outcomes

Primary: Discordance rate between EHR and Census Bureau reported race/ethnicity, difference in observed disparity between racial and ethnic groups on diagnosis prevalence; patient- and practice-level factors that are predictive of hesitancy to report race/ethnicity; accuracy metrics for race/ethnicity imputation and disaggregation methods. Secondary: Difference in means between concordant and discordant patients across various attributes; patient-level factors that are predictive of imputation accuracy, with statistical uncertainty estimates; difference in observed disparity between disaggregated groups on diagnosis prevalences; exploration of method performance on other race subgroups.


Illustration of a doctor next to a tablet that says "EHR system" and paper files
Grantee and Partner organizations

Leland Stanford Junior University
American Board of Family Medicine
Enhancing Health Data Program

Grant status
In Progress
Project Director(s)
Daniel Ho, PhD
Start date
Award amount
$546,250
Duration
36 months

Stay Connected