Optimizing Data on Race and Ethnicity for Veterans Affairs Patients

Area of Science:

• Health Informatics
• Data Science
• Public Health

Background:

• Accurate race and ethnicity data is crucial for equitable healthcare delivery within the Veterans Affairs (VA) system.
• Historically, maintaining precise demographic information in the VA has presented significant challenges.

Purpose of the Study:

• To enhance the accuracy of race and ethnicity data for patients within the VA healthcare system.
• To explore and optimize race- and ethnicity-specific data collation algorithms by integrating multiple VA data sources.

Main Methods:

• Linked VA patient data (2000-2018) with race/ethnicity information from four sources: VA Medical SAS files (MedSAS), Corporate Data Warehouse (CDW), VA Centers for Medicare extracts (CMS), and VA Defense Identity Repository Data (VADIR).
• Assessed data source accuracy by comparing values to self-reported data from the Survey of Health Experiences of Patients (SHEP).
• Utilized Cohen's Kappa for overall source agreement and positive predictive values (PPV) for race/ethnicity-specific accuracy.

Main Results:

• All sources demonstrated excellent overall agreement (Kappa > 0.80) with self-reported data.
• Race- and ethnicity-specific agreement varied, with all sources performing best for White and Black patients (average PPV = 0.94 and 0.93, respectively).
• Combined algorithms substantially reduced unknown race and ethnicity values compared to single-source methods.

Conclusions:

• Combining multiple data sources significantly improves the accuracy of race and ethnicity data for VA patients.
• Recommended prioritized order for filling race values: SHEP, CMS, CDW, MedSAS, VADIR.
• Recommended prioritized order for filling ethnicity values: SHEP, CDW, MedSAS, VADIR, CMS.