Getaneh and Findley (1) raise an important issue in response to our article (2). Grouping diverse populations into broad race/ethnicity categories (Hispanic, black, white, Asian, etc.) can mask meaningful differences within groups for cardiovascular and other disease risks. We agree with Getaneh and Finley that where possible, differences in disease prevalence should be reported by race/ethnicity subgroup, and settings like New York City provide an excellent opportunity to examine subgroup differences.
For example, using data from New York City birth certificates, we published an earlier study on gestational diabetes mellites trends in New York City from 1990 to 2001, and we demonstrated stark differences in the prevalence and trends across stratifications of Asian- and Hispanic-mother subgroups. Within Hispanic subgroups, mothers from the Dominican Republic had higher rates of gestational diabetes mellitus than Mexican or Puerto Rican mothers in 1990. Between 1990 and 2001, however, Mexican mothers showed the largest increase in the rate of gestational diabetes mellitus, and by 2001 their levels surpassed those found among mothers in other Hispanic subgroups examined (Puerto Ricans, Dominicans, and other Central and South Americans) (3).
Our recent article described findings from the New York City Health and Nutrition Examination Survey (NYC HANES). This study was designed to measure citywide levels of various health conditions, and ∼1,350 of the 1,999 study participants were randomly selected to fast before the exam for diabetes performance measures. For all participants, we collected information on Hispanic ethnicity and country of birth, but statistical power to examine diabetes prevalence by subgroup was limited. For total diabetes prevalence (diagnosed and undiagnosed), our point estimates suggested possible differences between Hispanic subgroups, but our estimates were unstable. Patterns of impaired fasting glucose (IFG) were also unstable but suggested that pre-diabetes patterns may differ from diabetes patterns. By combining data on IFG and diabetes, we were able to present stable, although imprecise, measures of abnormal glucose levels by Hispanic subgroup, adjusting for age (32.6% among Puerto Ricans, 42.0% among Dominicans, 44.7% among Mexicans, and 35.0% among other Central and South Americans). Sample size was small, and none of these differences are statistically significant.
In sum, we support disease reporting by race/ethnicity subgroup where possible. Where this is not possible, authors should recognize such inabilities in the discussion of study limitations. In our discussion of study limitations, we did mention our inability to stratify our analysis by Asian subgroups but failed to mention that we had the same problem with Hispanic subgroups.
Acknowledgments
No potential conflicts of interest relevant to this article were reported.