Objective
To use ethnically diverse, national data to examine longitudinal trends in race/ethnic disparities in 20 leading health indicators from Healthy People 2010 across multiple domains from adolescence to young adulthood. Much of what is known about health disparities is based on cross-sectional measures collected at a single time point.
Design, Setting, and Participants
Nationally representative data for more than 14 000 adolescents enrolled in wave I (1994-1995) or wave II (1996) of the National Longitudinal Study of Adolescent Health (Add Health) and followed up into adulthood (wave III; 2001-2002). We fit longitudinal regression models to assess and contrast the trend in health indicators among racial/ethnic groups of adolescents as they transition into adulthood.
Main Outcome Measures
Diet, inactivity, obesity, tobacco use, substance use, binge drinking, violence, sexually transmitted diseases, mental health, and health care access.
Results
Diet, inactivity, obesity, health care access, substance use, and reproductive health worsened with age. Perceived health, mental health, and exposure to violence improved with age. On most health indicators, white and Asian subjects were at lowest and Native American subjects at highest risk. Although white subjects had more favorable health in adolescence, they experienced greatest declines by young adulthood. No single race/ethnic group consistently leads or falters in health across all indicators.
Conclusions
Longitudinal data indicate that, for 15 of 20 indicators, health risk increased and access to health care decreased from the teen and adult years for most US race/ethnic groups. Relative rankings on a diverse range of health indicators (and patterns of change over time) vary by sex and race/ethnicity, causing disparities to fluctuate over time.
In the United States, there are large and persistent racial, ethnic, and socioeconomic disparities in health across the life course,1-3 with increasing economic and racial inequality.4,5 The reduction and ultimate elimination of health disparities has been identified as a major public health goal,6,7 exemplified by Healthy People 2010, the nation’s promotion and disease prevention strategy for the first decade of the 21st century.8
McGuiness and Foege9 have described the major external (nongenetic) modifiable factors contributing to death in the United States. Tobacco use and poor diet/inactivity emerged as the 2 leading causes of preventable death in the United States, with alcohol abuse, motor vehicle crashes, sexually transmitted diseases (STDs), firearms, and illicit drug use also important.9,10 Leading health indicators from Healthy People 2010 with known disparities in prevalence8 include sexually transmitted disease,10-12 illicit drug use,10,13-15 asthma,16,17 lack of access to medical care,18-20 and dental care.21
Health disparities research largely provides snapshots of health disparities at a single point in time or for a series of cross-sectional comparisons of different cohorts born at different times. The major contribution of our research is that we follow up the same cohort of ethnically diverse adolescents for 6 years into young adulthood to track change in disparities in health risk over time. We specifically examine 8 of 10 leading health indicators identified in Healthy People 2010 that are particularly relevant to young people.7,8 Measures include health indicators related to diet and inactivity, obesity, tobacco use, substance use, STDs, mental health, violence, and access to health care.
The National Longitudinal Study of Adolescent Health (Add Health) cohort is a nationally representative school-based study of youths (grades 7-12), followed up with multiple interview waves into young adulthood. The study used a multistage, stratified, school-based, clustered sampling design. A stratified sample of 80 high schools (and feeder middle schools) was selected with probability proportional to size. Survey procedures were approved by the Institutional Review Board of The University of North Carolina at Chapel Hill and have been previously described.22 Wave I (1994-1995) included 20 745 adolescents (aged 12-19 years) and their parents. Wave II (1996) included 14 738 wave I adolescents (including school dropouts and excluding graduating seniors). Wave III (2001-2002) included 15 170 wave I adolescents, now aged 18 to 26 years and entering the transition to adulthood (76% response rate).23 Total bias owing to attrition rarely exceeds 1%, which is small relative to the 20% to 80% prevalence rates for most of our measures.23-26
We used repeated measures of health indicators, restricting the sample to respondents interviewed at wave I or wave II and at wave III (depending on the measure) with valid data at both time points. When race/ethnic disparity in each measure is tracked from waves I through III, more than 14 000 subjects (7538 females and 6725 males) participated; from waves II through III, more than 10 000 subjects (5719 females and 5069 males) participated. In all analyses, we used wave III sampling weights and survey analysis techniques to adjust for unequal probability of selection, clustered sampling design, and attrition to ensure national representation of US teens on enrollment rosters for grades 7 through 12 in 1994-1995.23,27,28
To provide a comprehensive picture of the major health concerns of our nation, we include a range of measures that capture 8 of 10 leading health indicators of Healthy People 20108 and that were particularly relevant to young people.
Diet indicators include the number of days the respondent ate breakfast and fast foods in the past 7 days (waves II and III). Physical activity is reported using a standard physical activity behavior recall that is similar, although not identical, to other self-report questionnaires that have been used and validated in other large-scale epidemiological studies.29,30 Wave III measures were scaled to be equivalent to those at wave II.31 Lack of exercise is defined by self-report of no bouts of moderate to vigorous physical activity (5-8 metabolic equivalents) per week (waves II and III).
Obesity is defined using body mass index (BMI; calculated as weight in kilograms divided by the square of height in meters) from the measured height and weight at waves II and III. To deal with the discrepant obesity definitions for adolescents and adults, the International Obesity Task Force reference is used. This reference links childhood and adolescent BMI centiles to the adult BMI cut point of 30 to determine obesity prevalence at wave II.32 For the young adult population at wave III (aged 19-26 years), we consistently use the adult BMI cut point. Measures have been described in detail elsewhere.33
The following health care measures (waves I and III) are dichotomous indicators reflecting lack of care: (1) whether the respondent had current health insurance; (2) whether the respondent needed but did not receive medical care in the last year; (3) whether the respondent needed but could not afford dental or medical care last year; (4) whether the respondent visited a dentist in the last year; and (5) whether the respondent had a physical examination within 2 years.
Self-reported health is measured as a dichotomous indicator indicating fair or poor relative to good or excellent health. Asthma (parental report in wave I and self-reported asthma diagnosis in wave III) is measured as a dichotomous indicator.
All measures of substance abuse are self-reported from waves I and II and reflect (1) smoking at least 1 whole cigarette daily during the last 30 days; (2) any use of marijuana in the past 30 days; (3) any use of hard drugs such as cocaine, inhalants, or injected or other illegal drugs in the past 30 days in wave I, and any use of cocaine, injected or other illegal drugs, or crystal methamphetamine in wave III; and (4) binge drinking, defined as 5 or more drinks (ie, glass of wine, can of beer, wine cooler, shot glass of liquor, or mixed drink) on 1 occasion during the past year.
Our only measure of reproductive health is self-reported STD diagnosis (chlamydia, syphilis, gonorrhea, human immunodeficiency virus/AIDS, genital herpes, genital warts, and trichomoniasis for both sexes, and bacterial vaginosis and vaginitis for female subjects) at waves II and III.
We have the following 2 indicators of mental health at waves I and III: feelings of depression (feeling depressed a lot, most, or all of the time last week) and suicide ideation (the respondent has seriously thought about suicide in the past 12 months).
Exposure to violence in the past year is measured across the following items at waves I and III: (1) victim of violence who has been shot, stabbed, beat up, jumped, or threatened with a gun or a knife, and (2) perpetrator of a violent act who has pulled a knife or gun on someone or shot or stabbed someone.
Five race and ethnic groups were used based on self-identified race and Hispanic origin at wave III. In the largest sample, tracked from waves I through III, the sample includes non-Hispanic white (n = 7728), non-Hispanic black (n = 3038), non-Hispanic Asian (n = 1021), Hispanic (n = 2340), and Native American (n = 136) subjects.
We selected a marginal (or “population-average”) longitudinal regression model to estimate temporal change in each health indicator; the hypotheses under investigation were best answered by examining the average experience of each race/ethnic group rather than by examining person-specific variables. Separate models were run for male and female subjects. Covariates used to predict each health indicator in the model included race/ethnic group, time, and the interaction between race/ethnic group and time. Nonsignificant interaction terms were omitted from the final model. We hypothesized that (1) each race/ethnic group experienced no change in the health indicator over time, and (2) the rate of change in the health indicator for each nonwhite race/ethnic group was equal to the rate of change in the health indicator for the white group. To determine if the significant differences in the trend in health over time could be explained by socioeconomic status (SES), median tract income and maximum parental education were included in each of the final models, but had minimal influence on estimates of variables and made no difference in the tests of the hypotheses. Hence, we report results for the models that omitted these 2 covariates.
Across all sex and race/ethnic groups, there were significant increases in health risk and health disparities from adolescence into young adulthood. There were statistically significant increases over time in the proportion who get no exercise or report asthma, significant declines in breakfast consumption, and significant increases for almost every sex and race/ethnic group in the proportion of respondents who were obese, needed but could not afford medical care, used cigarettes and marijuana, had any STDs, or reported binge drinking, all indicating increasing health risk during the transition from adolescence to adulthood (Table 1). Reductions in risk were found in the significant declines over time across almost all sex and race/ethnic groups in the proportion who perceived their health as poor to fair, had feelings of depression or suicidal ideation, and were a victim of violence. All groups (except black and Native American male subjects) had significant declines in perpetrating violent acts.
The race/ethnic groups shift in their position as occupying the best or worst ranking across the various health indicators over time. No single race/ethnic group consistently occupies the best or worst ranking. The tendency was for Asian and non-Hispanic white subjects to generally rank best across time and across multiple domains of health risk, whereas the other minority groups ranked worst (Table 2 and Table 3). However, white subjects lose their favored position more often than other groups as they age into adulthood. Native American subjects tended to appear in the worst rankings most often, followed by non-Hispanic black subjects, with Hispanic subjects in the middle range across most indicators.
Selected illustrative results are presented in Figure 1 and Figure 2 by category of health indicator. The full set of figures for male and female subjects in all 20 leading health indicators can be found at http://www.cpc.unc.edu/projects/addhealth/findings.
Health habits and health care
There was a significant increase in the proportion of overweight/obese adolescents and those engaging in no weekly physical activity (female subjects shown in Figure 1A and B). The trend lines fan out, indicating increasing disparity in young adulthood, highest among black vs white subjects. The proportion of respondents with no health insurance coverage increased significantly for most groups (male subjects shown in Figure 1C). The proportion with the last physical examination more than 2 years ago (Figure 1D) also increased significantly for most male subjects, with increases for Hispanic and Native American subjects proportionately greater than those for white subjects.
General and mental health, substance use, reproductive health, and violence
Self-reports of poor health (male subjects in Figure 2A) and feelings of depression (female subjects in Figure 2B) declined significantly as adolescents aged into young adulthood, and race/ethnic disparities decreased as well. Smoking (Figure 2C) and binge drinking (Figure 2D) increased significantly for all groups, with increasing disparity over time and greatest disadvantage among white subjects. Diagnosis of STDs increased significantly in the transition to adulthood for all groups, with increasing race/ethnic disparity over time (Figure 2E). Violence declined significantly over time for all groups, with decreasing race/ethnic disparity (Figure 2F).
The overall patterns of health disparities include significantly different trends for black relative to white female subjects across almost all domains. Among male subjects, the black-white disparity in trends is significant for more than half of the indicators. For both sexes, the white-Hispanic disparity in health trends is significant for 35% of the indicators. Patterns of change in health indicators are similar for white and Asian subjects, and greater among male than female Native American compared with white subjects.
To our knowledge, this research is the first to present longitudinal, nationally representative data on race/ethnic disparities in leading health indicators across the transition to adulthood. The fact that we can follow up the same individuals longitudinally to examine changes in these health risk behaviors, specifically the placement of these changes in the context of health disparities, is a major contribution of this research.
For 15 of 20 indicators, health risk increased across all race/ethnic groups during the transition to adulthood. Relative rankings on a diverse range of health indicators (and patterns of change over time) varied by sex and race/ethnicity, causing disparities to fluctuate over time. Asian and white subjects generally scored highest across the range of health indicators; Hispanic subjects scored within the moderate range; and Native American and non-Hispanic black subjects scored the lowest. Although white subjects had more favorable health risk in adolescence, they experienced the greatest decline by young adulthood.
For most of the health indicators showing increased health risk over time, race/ethnic disparities in these indicators also increased. Although the disparities increased, different race/ethnic groups occupied more and less favored positions over time, indicating that no single common approach to reduce disparities is possible given the race/ethnic variation in change and rankings in health risk across time. Documenting race/ethnic disparities in health indicators is essential in understanding risk and formulating preventive strategies to tackle these important behavioral/health concerns.34-37 Furthermore, research needs to move beyond race/ethnicity to understand the underlying causes of health disparities in the United States.38
The extent to which socioeconomic differences may account for observed disparities in health outcomes across race/ethnic groups varies across studies.2,39 Our analytical models suggest no changes in race/ethnic disparities over time with controls for SES. Thus, factors beyond income and education may play a role in the cause of these health trends. Furthermore, SES may be related to the starting point on our health indicators in adolescence, but not to the trends in these indicators over time. It is also possible that we do not capture the complexity of the SES-health relationship across the transition from adolescence to young adulthood. For example, as teens leave their parental homes, they are likely to lose health insurance coverage owing to aging out of Medicaid coverage or leaving their parents’ health insurance coverage.
The top 3 leading actual causes of death in the year 2000 were (1) tobacco use, (2) poor diet and physical inactivity, and (3) alcohol consumption.10 We found dramatic increases in behaviors related to these 3 leading causes of death. As adolescents become young adults, they are more likely to eat fast food, get no exercise, be obese, and smoke cigarettes. Binge drinking (the most common pattern of alcohol abuse40) increased among male and female subjects (except black and Native American subjects). These findings underscore the importance of ongoing preventive efforts related to smoking, poor diet and physical inactivity, and alcohol consumption early in the life course.
As adolescents aged into young adulthood, an increasing proportion had no current health insurance, did not receive health care when they needed it, and did not get regular dental or physical examinations (male but not female subjects, likely due to the entrance into prime childbearing age). Race/ethnic disparities increased in the need for but inability to afford medical care, further exacerbating race/ethnic disparities in health19,20 and potentially leading to increased mortality.18
Similarly, the levels and disparities in asthma increased from adolescence to young adulthood, affecting Native American and black subjects most and confirming research that shows increasing prevalence of asthma, particularly among minorities.16,17 Diagnosis of STDs increased over time with substantial race/ethnic disparity (highest for black subjects). Similarly, the proportion using marijuana and hard drugs (male subjects) increased over time, with increasing race/ethnic disparity in marijuana use (highest for white subjects). The trends in prevalence of STDs and illicit drug use prompts concern owing to the association of these health indicators with increased mortality risk.9
On the positive side, feelings of depression and suicidal ideation decreased over time (particularly for black subjects), and differences among the race/ethnic groups declined. It is possible that this co-occurrence could, as the literature shows, influence other health risk behaviors, such as a range of STD/human immunodeficiency virus–associated sexual activity,41 substance abuse,15 tobacco use,42 and obesity.43 Similarly positive was the reduction in victimization and perpetration of violence. The improvement in self-reported health indicates an interesting dichotomy as these young adults consider themselves to be in good health, despite engaging in a range of health risk behaviors with negative current (and future) health implications.
We included measures of 8 of the 10 leading health indicators highlighted in Healthy People 20108 salient for young people (not included are measures of environmental quality and immunization). Although these are admittedly a broad array of health indicators that represent several different health constructs, they are important indicators reflecting major health concerns in the United States at the beginning of the 21st century.8 Furthermore, the meaning of health risk may change from adolescence to young adulthood (eg, sexual behavior may be considered more risky in adolescence than young adulthood). Another limitation is our use of the standard race/ethnic categories, which may mask important race/ethnic heterogeneity within these groups.
Our findings indicate a strong and unprecedented decline in health during the transition to young adulthood in the current generation. Although we have no point of reference to compare these findings with those for other generations, the within-cohort trends indicate deterioration in health behaviors. As this cohort is followed up in adulthood, the assessment of morbidity and mortality outcomes associated with this increased health risk will be possible. Shifts in health profiles are part of our epidemiological history when, for example, during the demographic transition a mortality pattern of malnutrition and infectious disease shifted to one of high prevalence of chronic and degenerative diseases associated with urban-industrial lifestyles44 and into a period of delayed degenerative diseases, a major shift in age-specific mortality patterns, and a consequent increase in life expectancy.45 Furthermore, evidence suggests that life expectancy gains have dropped in recent decades.46
One finding of major concern is the fact that young US adults are engaging in a range of health risk behaviors that puts them at increased risk of future adverse health outcomes. Our findings have serious implications for the present and future health of this generation. We find substantial race/ethnic disparities in health that vary depending on the measure. Because our findings show great variability in the levels, relative rankings, and changes in race/ethnic disparities, policies to reduce race/ethnic disparities in health require completely different actions for each health indicator. Attention must be paid to national efforts to reduce and ameliorate these preventable conditions, particularly in light of national health objectives for reducing race/ethnic disparities in leading health indicators.
Correspondence: Penny Gordon-Larsen, PhD, The University of North Carolina at Chapel Hill, Carolina Population Center, University Square, 123 West Franklin St, Chapel Hill, NC 27516-3997 (gordon_larsen@unc.edu).
Accepted for Publication: July 8, 2005.
Funding/Support: This study was supported by grant P01 HD31921 (Drs Harris and Udry) as part of the National Longitudinal Study of Adolescent Health (Add Health)program project, grant U01 HD37558 (Dr Harris) as part of the National Institute of Child Health and Human Development (NICHD) Family and Child Well-Being Research Network, and grant K01 HD044263 (Dr Gordon-Larsen) from the NICHD, Rockville, Md.
Additional Information: This research uses data from Add Health, a program project designed by J.
Richard Udry, PhD, Peter S. Bearman, PhD, and Kathleen Mullan Harris, PhD, with funding from the NICHD and cooperative funding
from 17 other agencies. Persons interested in obtaining data files from Add Health should contact Add Health, Carolina Population
Center, 123 W Franklin St, Chapel Hill, NC 27516-2524 (www.cpc.unc.edu/projects/addhealth/data/contract).
Acknowledgment: We thank Carol Ford, MD, and Parker Frisbee, PhD, for their helpful comments on an earlier draft and Lani Cartier and Sally Middleton for their assistance in preparation of the manuscript. Special acknowledgment is due Ronald R. Rindfuss, PhD, and Barbara Entwisle, PhD, for assistance in the original design.
1.National Center for Health Statistics, Health, United States, 2002 Hyattsville, Md National Center for Health Statistics2002;
2.Williams
DRCollins
C Racial residential segregation: a fundamental cause of racial disparities in health
Public Health Rep 2001;116404- 416
PubMedGoogle ScholarCrossref 3.Keppel
KGPearcy
JNWagener
DK Trends in racial and ethnic-specific rates for the health status indicators: United States, 1990-98 Hyattsville, Md: National Center for Health Statistics; 2002. Healthy People 2000 Statistical Notes, No. 23
5.Atkinson
A-BBourguignon
F Handbooks in Economics Vol 16 New York, NY Elsevier Science Inc2000;
6.Satcher
D Eliminating racial and ethnic disparities in health: the role of the ten leading health indicators
J Natl Med Assoc 2000;92315- 318
PubMedGoogle Scholar 7.US Department of Health and Human Services, National Institutes of Health strategic research plan and budget to reduce and ultimately eliminate health disparities, Vol I: fiscal years 2002-2006 Available at
http://www.ncmhd.nih.govAccessed June 17, 2005
8.US Department of Health and Human Services, Healthy People 2010: Understanding and Improving Health Washington, DC US Dept of Health and Human Services, Government Printing Office2000;
12.Centers for Disease Control and Prevention, Sexually Transmitted Disease Surveillance 2000 Atlanta, Ga US Dept of Health and Human Services, Centers for Disease Control and Prevention2001;
13.Rivara
FPMueller
BASomes
GMendoza
CTRushforth
NBKellermann
AL Alcohol and illicit drug abuse and the risk of violent death in the home
JAMA 1997;278569- 575
PubMedGoogle ScholarCrossref 15.Regier
DAFarmer
MERae
DS
et al. Comorbidity of mental disorders with alcohol and other drug abuse: results from the Epidemiologic Catchment Area (ECA) Study
JAMA 1990;2642511- 2518
PubMedGoogle ScholarCrossref 16.Ma3ssey
ITMoore
TFParsons
VLTadros
W Design and Estimation for the National Health Interview Survey, 1985-1994 Hyattsville, Md: US Dept of Health and Human Services; 1989. Dept of Health and Human Services publication (PHS)89-1384
18.Eisenberg
JMPower
EJ Transforming insurance coverage into quality health care: voltage drops from potential to delivered quality
JAMA 2000;2842100- 2107
PubMedGoogle ScholarCrossref 25.Kalsbeek
WDYang
JAgans
RP Predictors of nonresponse in a longitudinal survey of adolescents Paper presented at: Annual Meeting of the American Statistical Association, Joint Statistical Meetings August 13, 2002 Alexandria, Va
26.Kalsbeek
WDMorris
CBVaughn
BJ Effects of nonresponse on the mean squared error of estimates from longitudinal study Paper presented at: Annual Meeting of the American Statistical Assocation, Joint Statistical Meetings August 5, 2001 Alexandria, Va
28.Brogan
DDaniels
DRolka
DMarsteller
FChattopadhay
MArmitage
PedColton
Ted Software for sample survey data: misuse of standard packages
Encyclopedia of Biostatistics Vol 5 New York, NY John Wiley & Sons Inc1998;4167- 4174
Google Scholar 29.Andersen
RECrespo
CHBartlett
SJCheskin
LJPratt
M Relationship of physical activity and television watching with body weight and level of fatness among children: results from the Third National Health and Nutrition Examination Survey
JAMA 1998;279938- 942
PubMedGoogle ScholarCrossref 30.Sallis
JFBuono
MJRoby
JJMicale
FGNelson
JA Seven-day recall and other physical activity self-reports in children and adolescents
Med Sci Sports Exerc 1993;2599- 108
Google ScholarCrossref 31.Gordon-Larsen
PNelson
MCPopkin
BM A longitudinal physical activity and sedentary behavior trends: adolescence to adulthood
Am J Prev Med 2004;27 277- 283
PubMedGoogle Scholar 32.Cole
TJBellizi
MCFlegal
KMDietz
WH Establishing a standard definition for child overweight and obesity worldwide: international survey
BMJ 2000;3201240- 1243
PubMedGoogle ScholarCrossref 33.Gordon-Larsen
PAdair
LSNelson
MCPopkin
BM Five-year obesity incidence in the transition period between adolescence and adulthood: the National Longitudinal Study of Adolescent Health
Am J Clin Nutr 2004;80569- 575
PubMedGoogle Scholar 35.Krieger
N Counting accountably: public health implications of new approaches to classifying race/ethnicity in the United States 2000 census: US and global perspectives
Am J Public Health 2000;901687- 1689
PubMedGoogle ScholarCrossref 37.Flores
GFuentes-Afflick
ECarter-Pokras
O
et al. Why ethnicity and race are so important in child health services research today
Arch Pediatr Adolesc Med 2001;1551178- 1179
PubMedGoogle Scholar 39.Lillie-Blanton
MParsons
PEGayle
HDievler
A Racial differences in health: not just black and white, but shades of gray
Annu Rev Public Health 1996;17411- 448
PubMedGoogle ScholarCrossref 41.DiClemente
RJWingood
GMCrosby
RASionean
CBrown
L A prospective study of psychological distress and sexual risk behavior among black adolescent females
Pediatrics 2001;108e85.Accessed June 15, 2005
Google ScholarCrossref 42.Patton
GCCarlin
JBCoffey
CWolfe
RHibbert
MBowes
G Depression, anxiety and smoking initiation: a prospective study over three years
Am J Public Health 1998;881518- 1522
PubMedGoogle ScholarCrossref 43.Pine
DSGoldstein
RBWolk
SWeissman
MM The association between childhood depression and adulthood body mass index
Pediatrics 2001;1071049- 1056
PubMedGoogle ScholarCrossref 45.Olshansky
SJAult
AB The fourth stage of the epidemiologic transition: the age of delayed degenerative diseases
Milbank Q 1986;64355- 391
Google ScholarCrossref 46.Bell
FCMiller
ML Life Tables for the United States: 1900-2050 Baltimore, Md Social Security Administration2000;Actuarial study 87