Conflicting race/ethnicity reports: lessons for improvement in data quality

To learn the frequency of conflicting race/ethnicity reports, to examine patterns of conflicting reports, and to identify possible avenues for data quality improvement.

As part of the Data Improvement Project on Patient Ethnicity and Race (DIPPER), an analysis of conflicting race/ethnicity reports for cancer cases was conducted. Using matched hospital discharge data and central cancer registry data from 2009, the race/ethnicity of patients in the 2 datasets were compared. Those with conflicting reports ("mismatched") were examined more closely. From a sample of 2,356 patients, 187 had conflicting reports for their race (7.9%) and 357 had conflicting reports for their ethnicity (15% was thus developed).

In the 2009 hospital discharge data, an unknown response occurred nearly twice as often for Hispanic ethnicity as for race. Almost 85% of the mismatched race cases were classified as non-white in the hospital discharge data and white in the central cancer registry data. The most common ethnicity mismatch was coded unknown by the hospital but non-Hispanic by the registry.

Hospital cancer registrars occasionally lack easy access to race and, more often, ethnicity data. More attention should be given to discrepancies (including allowing staff to flag and verify existing data), and staff training should improve both perceived and real data accuracy. In the future, hospitals and registries would be better served by pairing race and ethnicity together in the electronic medical record. This would ensure quick, easy access for cancer registrars. Perhaps standard setters should add ethnicity to the gold standard criteria for registries.