|Home | About | Journals | Submit | Contact Us | Français|
With longitudinal data we traced how race, ethnic, and immigrant disparities in body mass index (BMI) change over time as adolescents (ages 11–19) transition to young adulthood (ages 20–28).
We used growth curve modeling to estimate the pattern of change in BMI from adolescence through the transition to adulthood.
All participants in the study were residents of the United States enrolled in high school or junior high school during the 1994–95 school year.
We used nationally representative data on 20,000+ adolescents interviewed at Wave I (1994–95) of Add Health, followed in Wave II (1996) and Wave III (2001–02) when the sample was in early adulthood.
Exposures of interest include race-ethnicity, immigrant generation, and sex.
Our main outcome measure is BMI.
Findings indicate significant differences in both the level and change in BMI across age by sex, race-ethnicity, and immigrant generation. Females, second and third generation immigrants, and Hispanics and blacks experience more rapidly increasing BMI as adolescents age into young adulthood. Increases in BMI are relatively lower for males, first generation immigrants, and whites and Asians.
Disparities in BMI and percent overweight and obese widen with age as adolescents leave home and begin independent lives as young adults in their 20s.
Obesity has reached nationwide epidemic proportions and is considered to be the most pressing public health problem in the U.S. today.1,2 Fueled by a modern lifestyle of poor diet and physical inactivity, the rising trend of obesity has affected Americans of all ages, but is perhaps most disconcerting among young people.3–9 The percentage of overweight children (body mass index [BMI]>95th percentile for age and sex) has tripled in the last two decades to the current estimate of more than 15% among those 6 to 19 years of age10 and over 17% among adolescents aged 12–19.11 The public health implications of earlier onset of obesity among American children are cogently expressed in the Surgeon General’s Report that the health problems resulting from obesity threaten to reverse the major improvements in health accomplished in the 20th century.1
Overweight children are likely to become overweight adults,12–15 and are at increased risk for a number of negative health outcomes, including hypertension, high cholesterol, abnormal glucose tolerance, type 2 diabetes, metabolic syndrome, kidney disease, coronary heart disease, congestive heart failure, stroke, osteoarthritis, some cancers, and death.12, 15–20 Overweight adolescents also complete fewer years of education, are less likely to marry, and have lower household incomes as adults.21, 22 Moreover, obesity-attributable medical expenditures reached $75 billion in 2003 and taxpayers financed about half of these through Medicare and Medicaid.23
Because the dramatic rise in overweight and obesity among children is relatively recent, fewer studies focus on the young ages, and research on obesity among young people is typically cross-sectional.24–29 The cross-sectional picture does not allow the study of developmental trends in obesity as people age and experience different life stages in which lifestyle and health habits change.30 We are therefore limited in our understanding of individual trajectories in overweight and obesity and disparities in these trajectories over time.
In addition, although much is known about race differences in obesity among adults, less is known about race and ethnic differences among young people. Research that does examine race differences focuses mainly on black-white differences,10, 31–34 failing to capture the increasing diversity of the US population fostered by high rates of Latin American and Asian immigration.24,25 The studies that include more diverse ethnic populations are based largely on cross-sectional or non-representative local samples,3,25,27,35–40 precluding a national perspective on race and ethnic variations in obesity trajectories as children age. Our research addresses this void by examining race-ethnicity disparities in BMI among young people aged 11–19, and tracing how disparities change as this adolescent cohort ages into young adulthood (ages 20–28). We use national representative longitudinal data from Add Health that permit the examination of Latino and Asian youth and the identification of immigrant generation.
Although immigrants constitute an increasing proportion of the US population each year,41 national data on the health of immigrants are hard to come by, and even less is known about children in immigrant families and their health prospects as they enter adulthood in America. Some research documents differences in obesity among immigrant adults24,25 and immigrant adolescents,42,43 but none has traced trajectories in overweight and obesity status through time among adolescents in immigrant families. The only national longitudinal study currently available focus on immigrant children ages 5–8 but does not provide estimations that differ by race-ethnicity or distinguish between first- (i.e. foreign-born with foreign-born parents), second- (U.S-born with foreign-born parents) and third- (U.S.-born with U.S.-born parents) generation children.44 Although America is leading the upward trend in obesity among industrialized countries, some rapidly developing countries, such as China and Mexico, are witnessing increases in obesity as they adopt western ways.45–47 For the most part, immigrants to the US come from countries in which average body mass is lower, and thus they are less likely to be overweight and obese than native-born Americans.42 However, with acculturation both across generations and time in the US for first-generation youth, the adoption of American diet and lifestyle increases risks of overweight and obesity among immigrant youth.42 Thus, it is important to examine immigrant disparities in body mass index over time, as adolescents age into young adulthood, to better understand acculturation processes that may occur with increasing assimilation of American lifestyle norms.
We use growth curve modeling to estimate the pattern of change in BMI across age beginning in adolescence and extending through the transition to adulthood, examining differences by race-ethnicity and immigrant generation, as well as by sex. These data provide the first evidence of increasing disparities in BMI as immigrant and U.S.-born youth transition into adulthood.
Data come from the National Longitudinal Study of Adolescent Health (Add Health), an ongoing nationally representative, school-based study of adolescents in grades 7 to 12 that began in 1994 and follows respondents into young adulthood. Add Health was designed to explore the causes of health-related behaviors, with an emphasis on the influence of social context. In 1994–95 Add Health administered an In-School Questionnaire to every student attending school from a nationally representative sample of schools. Over 70% of the schools originally selected for the survey participated.
Additionally, a random sample of adolescents and a parent was selected for in-home interviews in 1995, constituting Wave I data. Furthermore, oversamples of various ethnic groups were selected on the basis of in-school responses. As a result of high immigration to the US during the 1990s and the Add Health design that oversampled relatively rare ethnic groups (e.g., Cuban, Puerto Rican, and Chinese), Add Health contains a large number of adolescents in immigrant families—one out of four adolescents lived in an immigrant family (first and second generation). Of the adolescents selected for the in-home interviews, 79% participated in Wave I resulting in a sample size of 20,745 adolescents aged 11–19.
All adolescents in grades 7 through 11 in Wave I were followed up one year later for the Wave II in-home interview in 1996, with a response rate of 88.2%. In 2001–02 a third in-home interview was conducted with the original respondents from Wave I as they were now aged 18–28 and experiencing the transition to adulthood. Over 15,000 Add Health respondents were reinterviewed at Wave III (77.4% response rate) with longitudinal data over the various waves of interviews. See Harris et al. 200346 for more details on the Add Health design.
We use data from all three in-home interviews by pooling observations across the waves, resulting in a sample size of 48,737. Thus, respondents contribute anywhere from one to three observations on BMI to our analysis sample for growth curve models. Nonresponse analysis indicates no significant bias to Add Health estimates from attrition across waves.49
We measure raw BMI at Wave I using self-reported data on height and weight. We measure raw BMI at Waves II and III using measured height and weight. Add Health analysis (not shown) and previous research indicates a very high correlation (.99) between self-reported and measured BMI among young people.50,51 Even though BMI z-scores have been identified as optimal measures of adiposity at a single-point in time, raw BMI scores are best used to evaluate adiposity change in growing children.52
Race and ethnicity are self-reported at Wave I. We use a four-category classification: non-Hispanic white, non-Hispanic black, Non-Hispanic Asian, and Hispanic. We drop the small number of adolescents who listed race as “other” (largely Native American or unknown). Because Add Health oversampled selected ethnic groups, there are sufficient numbers of Hispanic (N=3,466) and Asian (N=1,578) youth to analyze their BMI.
We defined immigrant generation with a three-category variable signifying that the adolescent is foreign-born to foreign-born parents (1st generation), U.S.-born to foreign-born parents (2nd generation) and U.S.-born to U.S-born parents (3rd+ generation or native).42 Foreign-born adolescents with foreign-born parents are those children who were not born in the United States nor were they born U.S. citizens abroad. They migrated to the U.S. as children (in most cases with their immigrant parents).
Female sex is coded as 1, with males the reference category.
To determine differences in BMI by race-ethnicity, immigrant generation, and sex, we first examined the BMI scores by age group. We then explored mean differences in age-grouped BMI scores by race-ethnicity, immigrant generation, and sex.
To evaluate demographic differences in the risk of becoming overweight and developing obesity, we fit both linear and non-linear growth curve models with BMI as the continuous outcome and the categorical variables race-ethnicity, immigrant generation, and sex as the primary independent variables.
Growth curve models allow us to evaluate changes over time (i.e. age) for individuals. The model fits a developmental trajectory for changes in BMI as youth age into adulthood (Level 1 model) and allows race-ethnicity, immigrant generation, and sex to shift that trajectory (Level 2 model).53 In the Level 1 model, BMI is purely a function of an individual intercept and age for each individual (i) at time (t). With three waves of data, we have up to three time observations for each person in the sample. The intercept in level 1 of the growth curve model gives the expected level of BMI at the earliest observed age (11 years old). The slope of the level 1 model provides the expected change in BMI with a 1-year increase in age. In the second level of the growth curve model the level 1 intercept and the level 1 slope coefficient are estimated as functions of gender, immigrant generation, and race-ethnicity. Thus, the estimated coefficients in the level 2 model provide information on how each individual characteristic affects the intercept and slope parameters, respectively. The growth curve model presented here was estimated as a linear growth curve model. In sensitivity analyses, we estimated quadratic and log-linear models. Results were consistent across all specifications.
Using the estimated growth curve model, we then predicted each individual’s BMI from adolescence (age 11–19) through early adulthood (age 20–28), and examine disparities in the BMI trajectories by sex, immigrant generation, and race-ethnicity. Based on these predictions and using the international cutoff points for BMI by sex for overweight and obese,54 we estimate the % of our sample that would be considered overweight or obese at each age. The work of the International Obesity Task force54 provides cutoff points for overweight and obesity in children that are linked to adult cutoff points.
All analyses were conducted using mixed models in SAS.55 Bivariate analyses were weighted to reflect the national population estimates. To maximize the power of our analysis, growth curve model were estimated without weights. All growth curve models were recalculated using the weighted data. There were no substantial differences in adjusted coefficients when the weighted data were used but the results were less precise (i.e. estimated with larger confidence intervals).
Table 1 presents weighted descriptive statistics on the central variables of our analysis. These results are based on pooled observations across the three interview waves. Results show that BMI increases with age, reflecting developmental growth during adolescence up to age 20. Similar proportions of males and females are present in the data. The majority of observations are native-born with native-born parents in the third and higher immigrant generation (84%), followed by 11% who are native-born with foreign-born parents in the second generation, and 5% foreign-born in the first generation. The race and ethnic distribution reflects sufficient proportions for analysis across the four groups, with the majority non-Hispanic white (68%), followed by 16% Non-Hispanic black, 12% Hispanic, and 4% Asian.
Table 2 presents the results from our linear growth curve model. The results in the first panel (“Intercept” model) indicate that the first generation has a significantly lower BMI at the initial age (age 11) compared to the 3rd+ generation. Females also have lower BMIs than males in early adolescence. Hispanic youth and black youth have higher BMIs than whites at the initial age. The constant reports a BMI of 19.9 at the initial age across all groups.
The second panel (“Slope” model) presents results in the change in BMI across age for each of the groups. Results indicate that the growth rate in BMI is lower for both the 1st and 2nd generation compared to the 3rd+ generation, implying a protective effect of immigrant status from increasing BMI during the transition to adulthood. Females experience a greater increase in BMI than males over these ages, as do Hispanics and blacks compared to whites. The constant term indicates that BMI increases at an average rate of .55 across all ages. The model fits the data well (AIC=270111; BIC=270127; ChiSq=27765, p<.001).
We show these disparities in BMI trajectories in a series of graphs in Figure 1 that plot predicted BMI by age using the model presented in Table 2. The top panel shows predicted BMI by sex. Females have a lower BMI at the initial age 11, but their increase in BMI is more rapid across age than it is for males, surpassing males in BMI at around age 25. Though these differences are slight, they are significant (Table 2).
The second panel in Figure 1 shows predicted BMI by immigrant generation. Most noticeable is the lower BMI among the first generation. First generation youth have lower BMIs in adolescence, and the increase in BMI as they age into young adulthood is slower than both the second and third generations. Although native-born youth in both the second and third generations have similar BMIs in early adolescence, as they age into young adulthood, BMI among the third generation increases more rapidly than the second generation, and this difference is significant (slope coefficient in Table 2). Thus, not only does immigrant status protect youth against higher BMIs at a point in time (i.e., age), but the increase in BMI as young people age into young adulthood is also slower among immigrants, especially those in the first generation.
The third panel shows predicted BMI by race-ethnicity. Hispanic and black adolescents have higher BMIs than white and Asian youth in adolescence, and their trajectories indicate increasing disparities over time, especially among Hispanics. Whites and Asians experience the same pattern of change in BMI as they develop from adolescence into young adulthood.
The bottom panel shows the generation disparities in BMI within race and ethnic groups. Although the levels of BMI are somewhat higher for Hispanic and black youth, the pattern of change by immigrant generation is identical across race and ethnic groups, with first generation experiencing a slower rate of increase in BMI over time compared to second and third generation youth. Tests for interactions between immigrant generation and race-ethnicity in the age patterns of change in obesity did not substantially improve model fit (AIC=263751; BIC=263783; ChiSq=34107, p<.001). Thus, over this age range, the immigrant generation slopes are the same across race-ethnicity.
Although sex differences in trajectories of obesity were not the primary focus of this paper, we also tested for interactions between race-ethnicity and sex and between immigrant-generation and sex. Though first- and second-generation females had slower growth in BMI than third-generation females, these interactions were not significant enough to substantially improve the model’s fit (AIC=263727; BIC=263759; ChiSq=34106, p<.001). Similarly, BMI grew more rapidly for black females and more slowly for Asian females than for white females. But, these differences were not substantial enough to greatly improve the model’s fit (AIC=263634; BIC=263665; ChiSq=34004, p<.001).
The next set of graphs in Figure 2 displays our transformation of raw BMI into standardized percentile distributions using the international cutoff points for BMI by sex for overweight and obese. These results are useful because they provide an additional perspective on these disparities according to standardized definitions for overweight and obese, by showing the percent of our sample that would be considered overweight or obese at each age by sex, immigrant generation, and race-ethnicity. The top left panel shows the percent overweight and the top right panel shows the percent obese by sex. The sex disparity in the percent overweight is minor until after age 22 when men outnumber women in the percent overweight. Sex differences in the percent obese do not appear until age 20 when women begin to show equal or slightly higher percentages obese.
Disparities in overweight and obese by immigrant generation shown in the middle panel are consistent with trends in raw BMI. The first generation is less likely to be overweight or obese at any age from early adolescence into young adulthood. Differences between the second and third generations, however, are not as evident in the percentile distributions, and second generation even surpasses the third generation in the percent obese in the later 20s.
Finally, overweight and obesity disparities by race-ethnicity are shown in the bottom panel. Throughout these ages, Hispanics and blacks are more likely to be overweight than whites, and Asians are the least likely. Disparities in obesity, however, emerge mainly after adolescence, when the percent obese among Hispanic and black young adults begins to increase, while the percent obese among whites increases only slightly, and the percent among Asians even declines in some ages.
According to our model, 50% (10%) of first-generation, 72% (30%) of second-generation, and 70% (21%) of third-generation youth are likely to be overweight (obese) by age 25. Though astounding, these high prevalence rates are quite consistent with previous research on adult obesity.56 Overall, disparities in both raw BMI and the percent overweight and obese tend to increase with age. In particular, disparities tend to widen when adolescents leave the home and begin independent lives as young adults in their 20s. This suggests lifestyle changes may accompany this change in BMI, and may especially explain the generation and race and ethnic differences over time. Most at risk for increasing obesity during the transition into young adulthood are black and Hispanics and young people in native-born families. Much has been written about the lure of new technology in American culture that entices young people to replace outdoor activities and exercise with indoor video and internet use,6 fostering a sedentary lifestyle that evidently begins when adolescents leave home and set up their own households and schedules. In addition, diet choices that include fast food and fewer fruits and vegetables further exacerbate an unhealthy lifestyle among young people living on their own.57
Poor nutrition and lack of exercise are especially likely among minority groups in America, largely because of the social environments in which minorities live. For example poor neighborhoods often do not have large grocery stores, which force people, particularly those without cars, to shop at local convenience stores that stock small amounts of fresh fruits or vegetables, but large amounts of high fat, high starch processed food. As a result, high diet quality foods can be sometimes more expensive and less accessible for those residing in economically deprived neighborhoods.58,59 Food that is nonperishable and inexpensive is more likely to be energy-dense food that often contains fat, sugar or starch, a diet that increases the risks of weight gain. Young adults have not yet acquired sophisticated cooking skills and frequent fast food take out, another source of unhealthy diets among the poor because of the low cost. Not only do predominately white neighborhoods have significantly greater access to low-cost, quality food sources than minority neighborhoods,60 but white neighborhoods have greater access to recreational facilities and opportunities for physical exercise including neighborhood parks, safe running trails, and sidewalks.61 High crime and social disorganization within poor neighborhoods make outdoor exercise less safe or accessible.61 These spatial disparities in recreational access and cost of healthy food choices may explain the trajectories we see for blacks and Hispanics within the developmental stage of the transition to adulthood.
The young adult lifestyle described above may be less salient for young people in immigrant families who do not experience the traditional form of leaving home and striking out on one’s own in the American version of the transition to adulthood.62 High school graduation is less of a transitional marker for immigrant youth who either leave home when they marry, which can be at an early age, or remain in the parental home until they can support themselves.63,64 Given their continue presence in their parents’ home, they are more likely to retain their cultural and ethnic diets and lifestyles as they move into their early 20s.65,66 Thus, we see a protective effect of immigrant status in overweight and obesity trajectories that slows increasing weight gain with age, consistent with much of the emerging research on health disparities among immigrants in America.42, 43
The purpose of our research was to document the first nationally available prevalence data on trajectories of obesity by immigrant generation, race-ethnicity, and gender during the transition to adulthood. Future research should turn to the investigation of factors that might explain these trajectories by exploring socioeconomic, environmental, diet, lifestyle, and other health behavioral factors that may underlie the immigrant differentials we document here.
These trends make it increasingly clear that intervention and prevention efforts should be directed toward U.S.-born Hispanic and black youth entering young adulthood and transitioning from high school to work or college, and more generally among all youth as they leave home and begin to develop their own health habits which set pathways for their future health into adulthood.
We gratefully acknowledge research support from the Russell Sage Foundation to Perreira and Harris through grant RSF #88-06-07, from the Carolina Population Center to Lee through an NICHD pre-doctoral fellowship and from the National Institute of Child Health and Human Development to Harris through grant P01 HD31921 as part of the Add Health program project and grant U01 HD37558 as part of the NICHD Family and Child Well-being Research Network. This research uses data from Add Health, a program project designed by J. Richard Udry, Peter S. Bearman, and Kathleen Mullan Harris, and funded by a grant P01-HD31921 from the National Institute of Child Health and Human Development, with cooperative funding from 17 other agencies. Special acknowledgment is due Ronald R. Rindfuss and Barbara Entwisle for assistance in the original design. Persons interested in obtaining data files from Add Health should contact Add Health, Carolina Population Center, 123 W. Franklin Street, Chapel Hill, NC 27516-2524 (www.cpc.unc.edu/addhealth/contract.html). There are no conflicts of interest for any authors of this manuscript. Harris had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.