Evaluating Misclassification in Surrogate Methods for Assigning Hispanic/Latino Survey Respondents to a Racial/Ethnic Category, 2011 New Mexico Behavioral Risk Factor Surveillance System (NMBRFSS)

BACKGROUND:  Per federal guidelines, the NMBRFSS collects Hispanic ethnicity separately from race. However, the two are often combined into a single variable for reporting on health disparities. While self-identified assignment would be preferred, Hispanics have not had the opportunity to indicate whether ethnicity or race best describes them. As a surrogate approach, New Mexico has used a “smallest group” method that assigns respondents to the ethnic or racial group that comprises a smaller percentage of the population. To assess how accurately this approach and another method compare to self-identification, an additional question was piloted allowing Hispanic respondents to identify their preference for racial/ethnic assignment.

METHODS:  Following standard ethnicity and race survey questions, all 2011 NMBRFSS respondents who reported Hispanic ethnicity and at least one race were also asked “Which one of these groups would you say best describes you?” Respondents could select a single response from a list that included Hispanic or Latino, all standard racial categories, or “other”. Using self-identification as the gold standard for validity, comparisons were made to assignments using the “smallest group” method and an alternate “aggregate Hispanic” method (assigning all Hispanic respondents to a racial/ethnic designation of “Hispanic” regardless of racial identity). We assessed the effect of these different assignment methods on estimates of general health status, smoking, and obesity.

RESULTS:  2,429 respondents qualified for the additional preferred race/ethnicity question. Compared to self-identification, the “aggregate Hispanic” method correctly assigned more of the 162 Hispanic/American Indian respondents than the “smallest group” method (69% vs. 28%, respectively). 85% of 2,211 Hispanic/White respondents were correctly assigned with both methods. Among Hispanic/White and Hispanic/American Indian respondents, estimates of general health and obesity did not vary substantially based on method of racial/ethnic assignment. Hispanic/American Indian respondents were more likely to report being current smokers than Hispanic/White only or American Indian only respondents. Therefore, “aggregate Hispanic” assignment underestimated and “smallest group” assignment overestimated the percent of American Indians who reported being smokers. Very small numbers for other racial/ethnic combinations prohibited precise estimates.

CONCLUSIONS:  One year of pilot data suggests that neither surrogate method for assigning Hispanic NMBRFSS respondents to a single racial/ethnic category appears to consistently reflect self-identified preferences. Both methods can lead to biased estimates of important health indicators, such as smoking prevalence among American Indian adults. In the absence of self-identified preference, the “aggregate Hispanic” method appears more accurate for assigning race and ethnicity than the “smallest group” method.