Temporal individual-level data on obesity, overweight, and selected socioeconomic, demographic, and behavioral characteristics were derived from the 1976 and 1991–2008 National Health Interview Surveys (NHIS) [17
]. The NHIS, which is conducted by the National Center for Health Statistics, uses a complex, multistage probability design and is representative of the civilian non-institutionalized population of the US [17
]. The household response rate for an annual NHIS generally exceeds 85%. All data are based on self-reports, including height and weight information, and obtained via in-home person interviews [1
]. Substantive and methodological details of the NHIS are described elsewhere [1
Annual trends in obesity and overweight prevalence and BMI were estimated for the overall immigrant and US-born groups and for five educational groups from 1991 to 2008. To analyze trends over time by detailed ethnic-immigrant and socioeconomic characteristics, we pooled 4 years of the NHIS data from 1992 to 1995 and 6 years of data from 2003 to 2008. Aggregating data for several years in this fashion ensured sufficient sample sizes for analyzing patterns for groups stratified by ethnicity, immigrant status, and length of immigration. We could not use the 1991 NHIS file in the pooled analyses because it lacked detailed ethnic and income groupings. The 1976 NHIS, the earliest survey to collect height and weight data, was used to provide baseline estimates for various socioeconomic groups. The 1976 NHIS did not include information on immigrant status.
Obesity and overweight differentials were analyzed for 323,627 adults in 1992–1995 and 154,649 adults aged ≥18 in 2003–2008 for whom information on BMI was available. Adult overweight was defined as a BMI ≥ 25 kg/m2
and obesity as a BMI ≥ 30 kg/m2
]. Note that the overweight category includes obese individuals.
Immigrant status was defined on the basis of adults’ place of birth [4
]. US-born were those born in one of the 50 US states or Washington, DC. Immigrants or foreign-born refer to those born outside these territories [4
]. Race/ethnicity was classified into 11 major categories: non-Hispanic whites, non-Hispanic blacks, American Indians/Alaska Natives, Chinese, Asian Indians, Filipinos, other Asian/Pacific Islanders, Mexicans, Puerto Ricans, Cubans, and Central and South Americans, including other Hispanics. The joint variable of ethnic-immigrant status included 30 categories, with each racial/ethnic group (except for American Indians/Alaska Natives who are, by definition, a native group) divided into the US-born group, the recent immigrant group, and the long-term immigrant group [9
]. Although all Puerto Ricans are US citizens, those born in Puerto Rico and abroad were classified as “immigrants” for convenience. Following a previous study and given the health and socio-behavioral profiles by duration of residence, recent immigrants were defined as those who immigrated to the US in the previous 15 years, whereas long-term immigrants were those who immigrated to the US more than previous 15 years [9
In addition to ethnic-immigrant status, we considered the following socioeconomic and demographic factors that are known to influence obesity: age, gender, marital status, region of residence, educational attainment, family income/poverty status, occupation, and physical activity (PA) [4
]. These covariates were measured as shown in Tables and .
Table 1 Observed (weighted) prevalence and adjusted odds of obesity (BMI ≥ 30) among 30 ethnic-immigrant groups aged 18+ years and by selected socioeconomic and demographic characteristics: The National Health Interview Survey, (more ...)
Table 2 Observed (weighted) prevalence and adjusted odds of overweight(BMI ≥ 25) among 30 ethnic-immigrant groups aged 18+ years and by selected socioeconomic and demographic characteristics: The National Health Interview Survey, (more ...)
Educational attainment was measured both as a categorical variable (0–8, 9–11, 12, 13–15, ≥16 years) and a continuous variable in terms of years of school completed. Annual family income was also measured both as a categorical variable and a continuous variable. The seven income strata for 2003–2006 were: <10,000, 10,000–19,999, 20,000–24,999, 25,000–34,999, 35,000–44,999, 45,000–64,999, and ≥65,000. The corresponding income strata for 1992–1995 were: <7,000, 7,000–14,999, 15,000–19,999, 20,000–24,999, 25,000–34,999, 35,000–49,999, and ≥50,000. Detailed income categories were not available for 2007–2008. The income categories were roughly comparable for the 1992–1995 and 2003–2006 periods, given an increase by a factor of about 1.3 in the consumer price index between 1995 and 2005 [20
]. Continuous income was measured in thousands of dollars for both time periods.
Occupational class was defined in terms of 5 broad categories: professional and managerial occupations, sales/clerical and technical support occupations, service, craft and repair, and laborers. These occupational groups, derived from the major occupational groups defined by the Census Bureau, are consistent with previously defined social class positions of upper white collar, lower white collar, upper blue collar, and lower blue collar jobs [20
]. “Professional and managerial” occupations included executives, managers, administrators, engineers, architects, mathematical and computer scientists, teachers, writers, artists, and other professional specialty occupations. “Sales/clerical and technical support” occupations included technicians, health technologists, sales workers, computer equipment operators, secretaries, typists, and financial records processing, mail, message distributing, and other administrative support occupations. “Service” occupations included private household, protective service, food service, health service, cleaning and building service occupations, farm and other agricultural workers. “Craft and repair” occupations included mechanics, repairers, and those in construction, extractive trades, and precision production jobs. “Laborers” included machine operators, fabricators, assemblers, motor vehicle and material moving equipment operators, construction laborers, handlers, equipment cleaners, and helpers. In addition, a residual category for the unemployed and those outside the labor force was used.
PA level was measured by the number of times/week of vigorous activities of at least 10 min that caused heavy sweating or large increases in breathing or heart rate. The variable was coded as <1, 1–2, 3–4, ≥5 times/week of activity. PA was not available in 1976 and 1992–1995 [18
Multivariate logistic regression was used to examine the association between the binary outcomes of obesity and overweight and selected socioeconomic and demographic factors. Least squares regression was used to model mean BMI. To account for the complex sample design of the NHIS, SUDAAN software was used to conduct all statistical analyses [22
The two-sample t
test was used to test the difference in prevalence between any two groups at one point in time or to test for change in prevalence between two time points for a specific group. The gamma (γ) statistic, varying between −1 and 1, was used to measure the magnitude of the association between an ordinal covariate and obesity [16
]. An index of disparity (ID), which approximated in relative terms the average deviation of the rates from the rate for the best-off ethnic-immigrant or socioeconomic group, was used to summarize disparities across all social groups [16
]. The relative mean deviation index of disparity was calculated as:
is the obesity/overweight prevalence for the ith group, Orl
is the rate for the “standard” group or group with the lowest obesity/overweight prevalence, and I is the number of groups. A simulation method was used to estimate the standard error for ID [24