BATES ET AL. RESPOND

Abstract

We appreciate Oza-Frank and Narayan’s interest in our article. We are pleased they agree with our conclusions regarding the need for data and analyses that are adequately powered and disaggregated to account for the heterogeneity of US immigrant populations. Their comments provide an opportunity to highlight some of the important challenges in immigrant health research that we raised in our article. We used several approaches to deal with the issue of immigrant heterogeneity. First, we stratified by country of origin to the extent sample sizes permitted; the size and composition of the National Latino and Asian American Survey allowed the disaggregation of a nationally representative sample to a unique degree. Unfortunately, we could not stratify all groups (e.g., only 154 and 122 respondents reported Indian and Japanese ancestry, respectively), and as Oza-Frank and Narayan note, we highlighted this as a major limitation. To mitigate this problem, we provided information on the national origin composition of the other panethnic (e.g., Asian American) groups to show readers who was actually represented in these analyses. In addition, Oza-Frank and Narayan argued that our results might not apply to recent immigrant streams. By definition, our focus on generational differences allowed us to present estimates only for national origin groups with sufficient numbers in the second and third generations. This indeed limited generalizability. Second, we disaggregated the US born into second and third generations to improve comparability of body mass index and obesity trends across national origin groups. “US born” may mean very different things for groups with different immigration histories; for example, someone who is born in the United States may be from a family with multiple generations in the United States or may have parents who arrived in adulthood. A more substantively interesting but related question is whether the longevity of the national origin group matters for the outcomes of individuals independent of their own ancestry. This is a separate, worthwhile question that we did not attempt to address; we hope it will receive attention in future research. Third, we controlled for national origin in our multivariable analyses. An important validity issue arises in analyses of generational differences using panethnic categories if in comparing generations we are in effect comparing national origin groups because of different average immigration tenures. Regarding Oza-Frank and Narayan’s second point, studies of the effect of duration of residence within one generation complement investigations of generational differences, but they are also problematic methodologically unless based on longitudinal data. With cross-sectional data, any observed effects of years in the United States may be attributable to cohort differences, period effects, age at arrival, and selective out-migration. Unfortunately, without the appropriate data, neither of these study approaches allows an adequate examination of differences across successive immigrant cohorts. Our article and Oza-Frank and Narayan’s comments highlighted the need for the types of data and analysis that would enable us to better understand immigrant health trajectories and account for their complexity. Funding is needed for new data collection that addresses the heterogeneity of Latino and Asian populations. We also recommend that existing large-scale routine data collection and data analysis efforts include, at a minimum, information on nativity, subethnic composition, and age at arrival.