Page 9 - IJPS-5-1

Van Natta M and Zimmer Z

response with different meanings into a single group. Ordered treatments are problematic if the conceptual distance
between categories differs across languages. For example, the conceptual distance English speakers perceive between
good and fair may be greater than the difference Spanish speakers perceive between Buena and regular. To address this
limitation, the present study models SAH as a nominal measure with multiple categories.
Second, our study addresses “acculturation,” which many scholars account for when considering health disparities
between immigrants and native-born populations. Acculturation supposedly signals degrees of immigrant integration into
the receiving country’s characteristic “culture” – including language, behaviors, and social norms. However, the concept
has been considered atheoretical and vague (Abraído-Lanza, Armbrister, Flórez et al., 2006), and a systematic review by
Hunt, Schneider, and Comer (2004) concludes that its application is highly variable since the constituents of a culture,
differences between ethnic and “mainstream” culture, and what cultural adaptation means have not been adequately
described. Rather than engaging with ill-defined notions of culture, we follow DuBard and Gizlice (2008) and Kimbro,
Gorman, and Schachter (2012) in assessing acculturation by time spent in the U.S.
Third, we examine the phenomenon using a large nationally representative, inclusive and contemporary, dataset
associated with the CDC, the NHANES, providing a higher level of statistical power than any such study to date.
Earlier studies considered small samples and samples that were generally geographically limited. The large number of
observations in our analysis, which spans the years 2003-2016, allows us to divide our sample by survey language, foreign
nativity, immigrant status, and years lived in the U.S. to account for the multifaceted nature of immigration and health
while maintaining statistical power. Using this dataset, our study asks: (1) Is there a significant association between survey
language and SAH? (2) Is the association mediated by citizenship status? (3) Do associations hold when accounting for
ethnic variation?
2. Data and Method

2.1 Data
The National Center for Health Statistics conducts NHANES biennially to assess the health and nutrition status of the
civilian non-institutionalized U.S. population. Sampling follows a cross-sectional, multistage, stratified, and clustered
design. To increase the number of surveys conducted in Spanish and among immigrants, our analysis aggregates data from
the 2003 to 2016 surveys. The sample size across these surveys is 39,221 observations age 20 and older, which is reduced
to 39,107 valid observations after deleting cases where citizenship status and SAH is unknown. This valid N includes
over 3600 interviews conducted in Spanish and over 3400 among non-U.S. citizens. Weights included in NHANES data
are applied in this analysis.

2.2 Measures
SAH is registered in NHANES as a response to the question, “would you say your health in general is … ?/¿Diría usted
que su salud en general es …?” Responses are scored along the following (English/Spanish) scale: Poor/mala, fair/regular,
good/buena, very good/muy buena, and excellent/excelente. Missing responses (including “do not know” or “refused”)
represented <0.1% of total responses and were therefore excluded. For the remainder of this paper, we use the English
terms to describe these categories; however, the Spanish categories are used when the survey language is Spanish.
Across survey waves, NHANES includes four standard responses to the question, “are you a citizen of the United
States?”: “Citizen by birth or naturalization,” “not a citizen of the U.S.,” “do not know,” and “refused.” Those in the
latter two categories represented <0.2% of total responses and were therefore removed from the study. Data on country
of birth were then used to separate the U.S.-born citizens from naturalized while the non-citizen category remained
intact. Unweighted N’s for these categories are 28,474 (82.9%); 5012 (7.9%); and 5621 (9.2%), respectively. Citizenship
categories were converted into dichotomous variables, with U.S.-born used as the comparison category in multivariate
models. Across all survey years examined here, ethnicity/race (NHANES does not distinguish between these constructs)
is categorized as: Hispanic, Non-Hispanic white, non-Hispanic black, and other race – including multi-racial. The survey
includes two different categories for Hispanic respondents: “Mexican American” and “other Hispanic.” Our analysis
indicated that results do not differ significantly across these groups; therefore, we consider them together. Survey language
is dichotomized into English and Spanish. Time lived in the U.S. is measured categorically in NHANES. For this study,
four categories were created: <5 years, 5-10 years, 10-20 years, and 20 or more years.
Models adjust for other demographic and socioeconomic variables. Age was converted into decades to capture any
nonlinear association with SAH (20-29, 30-39, 40-49, 50-59, 60-69, and 70+) and treated as a categorical variable. Sex

International Journal of Population Studies | 2019, Volume 5, Issue 1 3