Professional Documents
Culture Documents
Original Research
F
ew youth consume a diet that aligns with current
(vegetable and fruit, fast food, starchy food, and snack dietary guidelines, especially with respect to fruit
food). Linear regression was used to examine the rela- and vegetable intake (1). The poor dietary intake
tionship of Time 1 socioeconomic status and race (mutu- exhibited by many adolescents is of concern given the
ally adjusted) on factor scores for each dietary pattern, high prevalence of obesity and other health-related out-
and then of Time 1 socioenvironmental characteristics comes associated with diet (2). The development of
(adjusted for socioeconomic status and race) on these healthy eating habits is critical in this age group to de-
factor scores. crease disease risk in adulthood (3). Understanding char-
Results In prospective analyses, socioeconomic status, acteristics of adolescents with less-than-optimal diets is
family meal frequency, and home availability of healthy crucial for developing more targeted and effective inter-
food were positively associated with the vegetable and ventions focused on improving dietary intake.
fruit and starchy food patterns and inversely associated Adolescent eating behavior is a function of multiple
with the fast food pattern. Home availability of unhealthy levels of influence, and Social Cognitive Theory and eco-
logic models can be helpful in understanding the different
levels of influence (4). One important level of influence is
G. J. Cutler is a senior epidemiologist, Minnesota De- an adolescent’s social environment (4). Peers can have a
partment of Health, St Paul; at the time of the study, significant influence on an adolescent’s eating behavior
she was a predoctoral fellow, Division of Epidemiology (4). Adolescents spend a substantial amount of time with
and Community Health, University of Minnesota, Min- their peers and eating is often an important part of this
neapolis. A. Flood is an associate professor, P. Hannan socializing (4). In addition, adolescents eat the majority of
is a senior research fellow, and D. Neumark-Sztainer is their meals and snacks at home, and families influence
a professor, Division of Epidemiology and Community eating behavior by being the provider of food while also
Health, University of Minnesota, Minneapolis. influencing attitudes, preferences, and values surround-
Address correspondence to: Gretchen J. Cutler, PhD, ing food (4). Sociodemographic factors such as income,
MPH, PO Box 64975, St Paul, MN 55164. E-mail: education level, and race can influence all of the above.
cutle007@umn.edu Adolescent dietary intake has customarily been exam-
Manuscript accepted: August 24, 2010. ined in terms of single foods or nutrients, but dietary
Copyright © 2011 by the American Dietetic pattern analysis has recently emerged as an important
Association. alternative to these traditional methods (5-7). The di-
0002-8223/$36.00 etary pattern approach has intuitive appeal because the
doi: 10.1016/j.jada.2010.10.052 human diet does not consist of a single nutrient or food,
230 Journal of the AMERICAN DIETETIC ASSOCIATION © 2011 by the American Dietetic Association
Table 1. Study characteristics of the Time 1 Project EAT (Eating Among Teens) cohort by age
and sex subgroup
High School Middle School
Girls Boys Girls Boys
Characteristic (nⴝ1,154) (nⴝ1,168) (nⴝ465) (nⴝ453)
but instead represents a complex set of highly correlated acteristics and habitual dietary patterns identified at
dietary exposures (8,9). Statistical methods such as factor baseline and 5 years later. Understanding the character-
analysis are often used to derive empirical dietary pat- istics of adolescents consuming specific dietary patterns
terns, and can be a useful technique to summarize di- is crucial. This knowledge will help determine what sub-
etary intake and relate that intake to specific character- groups and specific modifiable characteristics should be
istics of a research population (3,10). Few studies have targeted in dietary interventions, ultimately increasing
examined dietary patterns in adolescents, especially in the influence and efficiency of these interventions.
the United States. This approach has been used success-
fully in multiple adult populations and additional re-
search in adolescent populations is warranted (3). METHODS
To our knowledge, only four other studies have exam- Study Design and Population
ined characteristics associated with dietary patterns in Project EAT (Eating Among Teens) is an observational
adolescent populations, and all took place outside of the study of the socioenvironmental, personal, and behav-
United States (10-13). These studies examined sociode- ioral determinants of dietary intake and weight status
mographic (10-13) and lifestyle characteristics (12,13), among a large and ethnically diverse adolescent popula-
but none examined the association between socioenviron- tion. In Project EAT-I (Time 1), 4,746 middle school and
mental characteristics and dietary patterns. Our study high school students in 31 Minnesota schools completed
contributes to the literature by examining both sociode- in-class surveys and anthropometric measures during the
mographic and modifiable socioenvironmental character- 1998-1999 academic year. Trained Project EAT staff ad-
istics associated with specific habitual dietary patterns of ministered surveys and measurements. Project EAT-II
adolescents. It is one of only a handful of studies to (Time 2) aimed to resurvey all original participants 5
identify dietary patterns in adolescents in the United years later, in 2003-2004. During this follow-up, 2,516
States, and has the benefit of a large and ethnically and participants completed mailed surveys, representing 69%
socioeconomically diverse cohort. of those who could be contacted, and 53% of the original
The primary aim of this study was to investigate the cohort. Study characteristics of the Project EAT cohort
cross-sectional and prospective relationship between can be found in Table 1. The University of Minnesota’s
baseline sociodemographic and socioenvironmental char- Institutional Review Board Human Subjects Committee
Socioenvironmental
Family meal frequency During the past 7 days, how many times did all, or most, of your family living in your house eat a meal
together: Never, 1-2 times, 3-4 times, 5-6 times, 7 times, or More than 7 times (44).
Parental support for Based on four items assessing perceptions of how strongly one’s mother and father care about and
healthy eating encourage healthy eating. These questions included: My mother/father cares about eating healthy food, and
My mother/father encourages me to eat healthy food. Response options included: Not at all, A little bit,
Somewhat, or Very much (44). Cronbach’s .79.
Peer support for Based on agreement with the statement: Many of my friends care about healthy eating. Response options
healthy eating included: Not at all, A little bit, Somewhat, Very much, or I don’t know.
Healthy home food Created by combining the following scale variables: Fruits and vegetables are available in my home,
availability Vegetables are served at dinner in my home, We have fruit juice in my home, Milk is served at meals in
my home. Response options included: Strongly disagree, Disagree, Agree, and Strongly agree (44,45).
Cronbach’s .63.
Unhealthy home food Created by combining the following scale variables: We have “junk food” in my home, Potato chips or other
availability salty snack foods are available in my home, Chocolate or other candy is available in my home, and soda
pop is available in my home. Response options included: Strongly disagree, Disagree, Agree, and Strongly
agree (45). Cronbach’s .80.
Sociodemographic
Socioeconomic status Based on adolescent self-report. The prime determinant was parental education level, defined by the higher
level of educational attainment of either parent. An algorithm was developed that also took into account
family eligibility for public assistance, eligibility for free or reduced-cost school meals, and employment
status of the mother or father. Five categories were created (Low, Low Middle, Middle, Upper Middle, and
High). Validity was supported by parental assessment in a substudy (46). Spearman correlation 0.67,
weighted .79.
Race/ethnicity Based on the question: Do you think of yourself as: white; black or African American; Hispanic or Latino;
Asian American; Hawaiian or Pacific Islander; or American Indian or Native American. Adolescents were
given the option to choose more than one category, and those with multiple responses were coded as
Mixed/Other for analyses.
Figure. Description of measures used at Time 1 of Project EAT (Eating Among Teens) to assess socioenvironmental and sociodemographic
characteristics of participants. NOTE: Information from this figure is available online at www.adajournal.org as part of a PowerPoint presentation.
approved all study protocols. Passive or active consent be within acceptable ranges for dietary assessment tools
procedures were followed in Project EAT based on the (16,17). The average correlation coefficient from the val-
requirements of the participating schools’ research idation study comparing the YAQ to three 24-hour di-
boards. All participants signed an assent form before etary recalls was 0.54 (15,16). Some studies have sug-
survey completion. gested that there may be higher levels of measurement
error in low socioeconomic status (SES) minority popula-
tions (18,19). Before using the YAQ for Project EAT, the
Participant Survey questionnaire was pretested among a group of very low-
The Project EAT Time 1 survey assessed a range of so- income, ethnically diverse middle school students and
cioenvironmental, personal, and behavioral factors of po- was found to be acceptable in terms of the students over-
tential relevance to weight status, weight control, dietary all comprehension and ability to complete the question-
intake, and physical activity in adolescents. Development naire within one class period (20). In another study, Xie
of the Project EAT survey was guided by 21 focus groups and colleagues (21) addressed the use of the YAQ in a
with 141 adolescents, Social Cognitive Theory (14), and a multi-ethnic, multi-income, and multi-educated popula-
literature review. The survey was pilot-tested with 252 tion of adolescents and found it to be acceptable.
multi-ethnic 7th- and 10th-grade students.
Sociodemographic
Socioeconomic statusb .06 ⬍0.0001 .06 ⬍0.0001 ⫺.03 NS .14 ⬍0.0001
Race/ethnicity
Other ⫺.01 NSc .18 0.008 .05 NS ⫺.30 ⬍0.0001
Asian .71 ⬍0.0001 .67 ⬍0.0001 ⫺.03 NS ⫺1.33 ⬍0.0001
Black .24 0.0009 .04 NS .01 NS ⫺.42 ⬍0.0001
Hispanic ⫺.23 ⬍0.0001 .29 ⬍0.0001 .34 ⬍0.0001 ⫺.52 ⬍0.0001
White (reference group)
Socioenvironmental
Frequency of family meals .05 ⬍0.0001 .05 ⬍0.0001 ⫺.03 0.02 .05 ⬍0.0001
Healthy food availability .07 ⬍0.0001 .02 0.003 ⫺.04 ⬍0.0001 .07 ⬍0.0001
Unhealthy food availability ⫺.07 ⬍0.0001 ⫺.07 ⬍0.0001 .07 ⬍0.0001 ⫺.02 0.0002
Maternal support .14 ⬍0.0001 .08 ⬍0.0001 ⫺.04 0.01 .10 ⬍0.0001
Paternal support .12 ⬍0.0001 .06 0.0002 ⫺.03 NS .08 ⬍0.0001
Peer support .15 ⬍0.0001 .09 ⬍0.0001 ⫺.02 NS .07 ⬍0.0001
a
The change in factor score for a 1-unit change in variable (values for race/ethnicity⫽difference in score in comparison to white adolescents).
b
All models adjusted for socioeconomic status, race/ethnicity (socioeconomic status was adjusted for race/ethnicity and vice versa), age, and sex.
c
NS⫽not significant.
dietary pattern was identified that we labeled fast food, vegetable and fruit pattern and negatively associated
characterized by foods one would likely eat at a fast-food with scores for the fast food pattern. High availability of
restaurant such as hamburgers, french fries, fried food, healthy foods in the home was positively associated with
and non-diet soda. Participants received a factor score for factor scores for the vegetable and fruit pattern and the
each of the four dietary patterns identified at Time 2 starchy food pattern, and negatively associated with fac-
(vegetable and fruit, snack food, starchy food, fast food) tor scores for the fast food pattern. In contrast, high
and a higher factor score for an indentified pattern cor- availability of unhealthy foods in the home was nega-
responded to a higher consumption of foods represented tively associated with scores for the vegetable and fruit
by that dietary pattern. In addition, the identified dietary and starchy food patterns, and positively associated with
patterns were uncorrelated, and a participant could score scores for the fast food and snack food patterns. Maternal,
high on the vegetable and fruit patterns while also scor- paternal, and peer support for healthy eating was posi-
ing high on the fast food pattern. tively associated with scores for the vegetable and fruit
pattern and negatively associated with scores for the fast
food pattern.
Time 2 Dietary Patterns: Prospective Associations
In general, prospective findings were similar to associa- Diet Profile of Identified Dietary Patterns
tions found in cross-sectional analyses. Multivariable re-
The mean intake of multiple dietary variables is shown by
gression coefficients linking Time 1 sociodemographic
highest vs lowest quintile of factor score in Table 6 for each
and socioenvironmental characteristics to factor scores
of the four dietary patterns identified at Time 1 and Time 2.
for the Time 2 dietary patterns are shown in Table 5.
As expected, high scorers for the vegetable, fruit, and veg-
When looking prospectively, SES was positively associ-
etable and fruit pattern had higher mean servings per day
ated with the vegetable and fruit and starchy food pat-
of fruits and vegetables. High scorers for the starchy food
terns, and negatively associated with the fast food
pattern had higher mean servings per day of grains at Time
pattern. When compared to white adolescents, Asian-
1 and Time 2. High scorers for the snack food pattern at
American and black adolescents had higher scores on the
Time 1 and Time 2 and for the fast food patterns at Time 2
vegetable and fruit pattern. In addition, Asian-American
had the lowest mean daily servings of fruits and vegetables.
and Hispanic adolescents had higher scores on the fast
High scorers for the fast food pattern at Time 2 had the
food pattern, and lower scores on the starchy food pattern
overall highest daily percentage of energy from fat.
when compared to white adolescents.
Multiple Time 1 socioenvironmental characteristics
were also statistically significantly associated with fac- DISCUSSION
tors scores for the Time 2 dietary patterns. Frequency of The goal of this study was to investigate the cross-sec-
family meals was positively associated with scores for the tional and prospective relationships between sociodemo-
Sociodemographic
Socioeconomic statusb .05 0.001 ⫺.13 ⬍0.0001 ⫺.01 NS .10 ⬍0.0001
Race/ethnicity
Other .05 NSc .27 0.005 .19 0.05 ⫺.40 ⬍0.0001
Asian .96 ⬍0.0001 .21 0.0003 ⫺.09 NS ⫺1.50 ⬍0.0001
Black .22 0.02 .18 NS .14 NS ⫺.15 NS
Hispanic .05 NS .28 ⬍0.0001 .55 ⬍0.0001 ⫺.61 ⬍0.0001
White
Socioenvironmental
Frequency of family meals .06 ⬍0.0001 ⫺.07 ⬍0.0001 .01 NS ⫺.004 NS
Healthy food availability .03 0.003 ⫺.06 ⬍0.0001 ⫺.01 NS .02 0.007
Unhealthy food availability ⫺.06 ⬍0.0001 .04 ⬍0.0001 .04 ⬍0.0001 .02 0.02
Maternal support .07 0.0002 ⫺.08 ⬍0.0001 ⫺.03 NS .02 NS
Paternal support .06 0.0009 ⫺.07 0.0002 ⫺.003 NS ⫺.03 0.02
Peer support .05 0.03 ⫺.14 ⬍0.0001 .004 NS ⫺.03 NS
a
The change in factor score for a 1-unit change in variable (values for race/ethnicity⫽difference in score in comparison to white adolescents).
b
All models adjusted for socioeconomic status, race/ethnicity (socioeconomic status was adjusted for race/ethnicity and vice versa), age, and sex.
c
NS⫽not significant.
found this result somewhat counterintuitive. On closer that a Western dietary pattern was inversely associated
inspection, we found that the SES association was limited with family income, whereas a healthy pattern was asso-
to the white adolescents; in the other ethnic groups, SES ciated with greater maternal education. Another study
was not associated with scores on these two factors. The examined correlates of dietary patterns identified in a
relationship between SES and dietary patterns found in cohort of Spanish children, adolescents, and young adults
our study may be driven by the large sample of white (10). Aranceta and colleagues (10) showed that low level
adolescents in our cohort, and may not be as strong in of maternal education was positively associated with
other racial/ethnic groups. In fact, Asian-American ado- scores for a snacky pattern rich in sweets, high-fat baked
lescents consistently scored much higher on the fruit and goods, and salty snacks, and inversely associated with
vegetable patterns when compared to white adolescents, scores for a healthy pattern rich in fruits, vegetables, and
and may be an interesting population for future research fish in children (aged 2 to 13 years) and young people
aimed at identifying factors that increase the likelihood (aged 14 to 24 years). Finally, Kourlaba and colleagues
of meeting dietary recommendations for these foods. We (13) examined correlates of dietary patterns in a popula-
did find that Asian-American and Hispanic adolescents tion of Greek adolescents and found no association be-
scored higher on the unhealthy fast food pattern, indicat- tween socioeconomic indicators (eg, number of cars in
ing that SES should still be a consideration when plan- family, existence of separate bedrooms for each child) and
ning dietary interventions. Our counterintuitive findings dietary patterns they identified as representing healthy
may have also been related to the fact that a higher and unhealthy food behaviors. We found that SES was
number of students in non-white ethnic groups in our positively associated with the starchy food pattern in
adolescent population indicated that they were eligible each subgroup of our population, and that it was also
for free or reduced-cost lunch, which additional analyses associated in expected directions (eg, higher SES associ-
showed was positively associated with higher scores for ated with healthier patterns and vice versa) with the
the vegetable pattern. other identified patterns in specific subgroups. The use of
Four other studies have examined correlates of dietary different indicators of socioeconomic status may be a pos-
patterns in adolescent populations (10-13). These studies sible explanation for the differences in our study and
focused on sociodemographic and/or lifestyle correlates, others in adolescent populations. In addition, as seen in
and unlike our study, did not examine socioenvironmen- our results regarding race/ethnicity, it is possible that the
tal characteristics. McNaughton and colleagues (11) ex- relationship between SES and dietary patterns may vary
amined an Australian cohort of adolescents, and observed in different cultures, or in rural vs urban regions of res-
no association between dietary pattern scores and mea- idence.
sures of SES based on economic resources, education, Multiple studies have examined the relationship be-
occupation, family structure, and income level. Ambrosini tween sociodemographic characteristics and dietary pat-
and colleagues (12) also examined correlates of dietary terns in adult populations (23-34), and most found an
patterns in an Australian cohort of adolescents and found association between indicators of SES and dietary pat-
Time 2
Vegetables (servings/d) 0.4⫾0.2 1.8⫾0.8 1.0⫾0.7 0.9⫾0.7 1.0⫾0.6 0.8⫾0.5 0.9⫾0.7 1.3⫾0.8
Fruit (servings/d) 0.6⫾0.5 1.3⫾0.8 1.4⫾0.7 0.5⫾0.4 0.9⫾0.6 0.8⫾0.5 1.2⫾0.9 0.8⫾0.5
Grains (servings/d) 3.0⫾0.9 3.1⫾0.8 3.2⫾0.8 2.8⫾0.8 3.0⫾0.8 2.8⫾0.7 2.7⫾0.8 3.5⫾0.8
Regular soda (servings/d) 0.9⫾0.6 0.5⫾0.4 0.4⫾0.4 0.9⫾0.6 0.7⫾0.5 0.6⫾0.4 0.7⫾0.6 0.6⫾0.5
Calcium (mg/d) 597⫾218 448⫾181 634⫾215 378⫾134 500⫾192 453⫾149 391⫾200 528⫾159
Folate (g/d) 89⫾25 144⫾45 132⫾39 87⫾30 108⫾37 98⫾29 106⫾40 119⫾39
Fiber (g/d) 6⫾2 11⫾3 9⫾3 6⫾2 8⫾3 8⫾2 7⫾3 9⫾3
Carbohydrate (g/d) 137⫾16 142⫾20 147⫾16 129⫾19 137⫾17 140⫾13 142⫾21 133⫾18
Protein (g/d) 37⫾7 41⫾9 41⫾8 36⫾7 38⫾8 34⫾5 37⫾9 41⫾8
Energy from fat (%) 31⫾5 28⫾6 26⫾5 34⫾6 30⫾6 31⫾4 29⫾6 30⫾6
a
Q1⫽Lowest quintile of factor score for defined pattern, Q5⫽highest quintile of factor score for defined pattern.
terns (23-32). Although the specific foods in the identified be useful for informing dietary interventions (7,9,40).
dietary patterns varied across countries, similar to the Findings from our study in conjunction with findings
results of this study, higher SES, higher income level, from studies examining single foods or nutrients suggest
and/or higher education level was commonly associated that there are modifiable characteristics that we could
with the more healthy pattern, whereas low levels of target in adolescent populations to encourage healthier
these socioeconomic indicators were usually associated dietary patterns, while also decreasing unhealthy dietary
with the more unhealthy pattern (23-27,30-32). To the patterns. When the goal is to ensure that adolescents eat
best of our knowledge, no studies in adult populations a healthy diet, it is likely important to know how factors
have examined socioenvironmental correlates of dietary such as family meal frequency, support for healthy eating
patterns. and home food availability influence the whole diet in
Our results are similar to several studies that have addition to how they influence intake of single foods or
examined single foods or nutrients in children and ado- nutrients.
lescents. For example, higher frequency of family meals Strengths of this study include a large, ethnically and
has been found to be positively associated with intake of socioeconomically diverse population. Data were avail-
fruits, vegetables, grains, and calcium-rich foods, and able on multiple sociodemographic and socioenviron-
negatively associated with soft drink consumption (35). mental characteristics that have not been studied previ-
Positive associations have also been found between home ously in relation to dietary patterns in youth. Several
availability of healthy food and parental encouragement, limitations of this study must be considered. Although
and children’s fruit and vegetable intake (36-38). In ad- the Project EAT study population included adolescents
dition, multiple studies have found a positive relationship from multiple ethnic and socioeconomic groups, it may
between SES and fruit and vegetable intake (36,39). Al- not be representative of all adolescents. More than 70% of
though the above studies have found similar results re- the adolescents in the Asian-American group identified
garding socioenvironmental characteristics as were found themselves as Hmong, therefore, it is difficult to general-
in the current study, results from studies using dietary ize the results found for these adolescents to other Asian-
pattern analysis provide a whole diet approach that may American adolescents. A food frequency questionnaire