Are Surname Telephone Oversamples an Efficient Way to Better Understand the Health and Healthcare of Minority Group Members?

Abstract

Surname oversamples are commonly used in health research to increase the number of persons from minority racial and ethnic groups represented in general population surveys. This article considers the sample design efficiency in the use of Hispanic, Hmong, and Asian surname samples. The study uses 3 state surveys (Alabama, Missouri, and Minnesota) that used surname oversamples to increase the proportion of Hispanic, Hmong, or Asian respondents included in the studies. We examine whether surname oversamples lead to more completed surveys with the targeted minority groups than would have been achieved if surname oversamples had not been used. We also assess gains in terms of effective sample sizes from the use of surname oversampling. The sensitivities of the Hispanic surname list ranged from 46% to 63% across the 3 surveys. The sensitivity of the Asian survey was 34%, and the sensitivity of the Hmong was 38%. Although the use of surname increased the number of targeted minority group members in the final study, the increased number had a very minimal impact on the effective sample size of the minority populations for the key survey estimates of interest in the 3 health surveys. The use of surname samples achieved the goal of having more persons who identify as Hispanic, Hmong, or Asian in the final sample. However, the use of surname oversamples is inefficient when considering the statistical power gained for minority group estimates.