Evaluating the accuracy of a geographic closed-ended approach to ethnicity measurement, a practical alternative

Measuring ethnicity accurately is important for identifying ethnicity variations in disease risk. We evaluated the degree of agreement and accuracy of maternal ethnicity measured using the new standardized closed-ended geographically based ethnicity question and geographic reclassification of open-ended ethnicity questions from the Canadian census.

A prospectively designed study of respondent agreement of mothers of healthy children aged 1-5 years recruited through the TARGet Kids! practice-based research network. For the primary analysis, the degree of agreement between geographic reclassification of the Canadian census maternal ethnicity variables and the new geographically based closed-ended maternal ethnicity variable completed by the same respondent was evaluated using a kappa analysis.

Eight hundred sixty-two mothers who completed both measures of ethnicity were included in the analysis. The kappa agreement statistic for the two definitions of maternal ethnicity was 0.87 (95% confidence interval, 0.84-0.90) indicating good agreement. Overall accuracy of the measurement was 93%. Sensitivity and specificity ranged from 83% to 100% and 96% to 100%, respectively.

The new standardized closed-ended geographically based ethnicity question represents a practical alternative to widely used open-ended ethnicity questions. It may reduce risk of misinterpretation of ethnicity by respondents, simplify analysis, and improve the accuracy of ethnicity measurement.