(https://www.academia.

edu)

Error: 404

Oops! It looks like you're in the wrong

aisle.
We're sorry for the confusion. We're currently working on it. In the meantime, feel free to browse
the restricted section.

Premium
Readers (https://www.academia.edu/upgrade?feature=searchm)
Mentions (https://www.academia.edu/upgrade?feature=readers)
Advanced Search (https://www.academia.edu/upgrade?feature=search)
Pro le Visitors (https://www.academia.edu/upgrade?feature=search)
Personal Website (https://www.academia.edu/upgrade?feature=analytics)
Grants (https://www.academia.edu/upgrade?feature=analytics)

Company
About (https://www.academia.edu/about)
Press (https://www.academia.edu/press)
Blog (https://medium.com/@academia)

Security
Privacy (https://www.academia.edu/privacy)
Terms (https://www.academia.edu/terms)

Career
Job Board (https://www.academia.edu/Jobs)
We're Hiring (https://www.academia.edu/hiring)

Support
Help Center (https://support.academia.edu/)
Contact Us (mailto:premium-support@academia.edu)
PREVENTIVE MEDICINE 25, 442–454 (1996)
ARTICLE NO. 0076

The Effects of the Child and Adolescent Trial for Cardiovascular Health
upon Psychosocial Determinants of Diet and Physical Activity Behavior
ELIZABETH EDMUNDSON, PH.D.,* GUY S. PARCEL, PH.D.,† HENRY A. FELDMAN, PH.D.,‡ JOHN ELDER, PH.D.,§
CHERYL L. PERRY, PH.D.,Ø CAROLYN C. JOHNSON, PH.D.,\ B. J. WILLISTON, M.ED.,§
ELAINE J. STONE, PH.D., M.P.H.,** MINHUA YANG, M.S.,‡
LESLIE LYTLE, R.D., PH.D.,Ø AND LARRY WEBBER, PH.D.\
*University of Texas at Austin, Austin, Texas 78712; †University of Texas Health Science Center, Houston, Texas 77225; ‡New England
Research Institutes, Inc., Watertown, Massachusetts 02172; §University of California at San Diego, La Jolla, California 92093;
Ø
University of Minnesota, Minneapolis, Minnesota 55454; \Tulane University School of Public Health and Tropical Medicine,
New Orleans, Louisiana 70118; and **National Heart, Lung, and Blood Institute, Bethesda, Maryland 20892

point to a need for greater understanding of adoles-

Background. The Child and Adolescent Trial for Car- cent developmental issues and the role of community
diovascular Health is a multisite study of a school- environment (particularly social support) in creating
based intervention to reduce or prevent the develop- effective curricula. q 1996 Academic Press, Inc.
ment of risk factors for cardiovascular disease. The Key Words: CVD; children; prevention; social cogni-
purpose of this paper is to present the evaluation re- tive theory.
sults of the 3-year intervention, focusing upon the psy-
chosocial variables conceptualized as determinants of
dietary and physical activity behaviors.
Methods. A total of 96 schools across four study sites INTRODUCTION
(California, Louisiana, Minnesota, and Texas) were
randomized to two treatment conditions: intervention The Child and Adolescent Trial for Cardiovascular
and control. Pre- and postmeasurements on the health Health (CATCH) is a school-based, comprehensive, lon-
behavior questionnaire were collected from over 6,000 gitudinal cardiovascular disease risk factor interven-
students. The data analyses utilized a nested design tion that was implemented for 3 years among children
approach in which schools served as the primary unit
in 96 schools, from the beginning of third grade through
of analysis. Repeated-measures multivariate analyses
the fifth grade.1 – 4 The conceptual framework for the
were applied to investigate effect sizes for each deter-
CATCH intervention assumes a causal model in which
minant and to explore theoretical relationships among
the determinants over time. psychosocial risk factors influence behavioral risk fac-
Results. The findings indicated sustained significant tors (dietary, physical activity, and smoking) which in-
effects in improved knowledge, intentions, self-effi- fluence physiological risk factors (blood lipids and blood
cacy, usual behavior, and perceived social reinforce- pressure), which finally determine morbidity and mor-
ment for healthy food choices (P õ 0.0001 for these tality (heart attack and stroke).1 In this model, psy-
five variables) after 3 years. Intermittent effects were chosocial risk factors became the most proximal targets
observed for perceived support and self-efficacy for for change by the CATCH intervention. The first step
physical activity. No gender by determinant interac- of the process included identification of the behaviors
tion effects were observed, and girls reported signifi- we wanted the children to adopt, specifically, to reduce
cantly greater perceived reinforcement for healthy fat and sodium intake, to increase participation in mod-
eating than did boys. erate to vigorous physical activity, and to resist the
Conclusion. The CATCH program was effective in initiation of smoking. Next, based upon Social Cogni-
changing the psychosocial variables likely to influence tive Theory5 and limited empirical data, we selected
a reduction in behavior for cardiovascular disease. those psychosocial factors considered most likely to pre-
The study is significant in that it demonstrates the via- pare the children to perform the identified behaviors.
bility and effectiveness of a sustained multifaceted in- Finally, we designed the intervention components to
tervention in a preadolescent population. The results address the personal, environmental, and behavioral
factors identified as potential determinants of the risk-
Address correspondence and reprint requests to Elizabeth Ed-
reducing behaviors.
mundson, Ph.D., University of Texas at Austin, Department of Kine- The theoretical underpinnings selected from Social
siology—BEL 222, Austin, TX 78712. Fax: (512) 471-8914. Cognitive Theory (SCT) and targeted in the CATCH

0091-7435/96 $18.00 442

Copyright q 1996 by Academic Press, Inc.
All rights of reproduction in any form reserved.
 DETERMINANTS OF DIET AND PHYSICAL ACTIVITY 443

intervention include expectancies (the values placed on sion of the operational design and quality control proce-
a particular outcome); self-control and performance dures implemented during the CATCH trial is pre-
(decisions made to achieve self-control, such as the set- sented by Stone and colleagues.12
ting of a performance criterion); behavioral capability, In this paper, we present the effects of the interven-
or the necessary knowledge to understand the behavior tion on the personal determinants of behaviors related
and skills to perform the behavior; environment and to dietary behavior and physical activity. The effects
situation (the actual and perceived physical and social of the intervention presented and discussed in this pa-
environments); observational learning; and self-effi- per are the combined results of all the intervention
cacy.1,6 components. The only attempt to compare the effects
Self-efficacy is especially critical to the initiation of of individual components is an analysis of the additive
skill acquisition; this finding has been reported for chil- effect of the family component to the combined school-
dren as well as adults.5,7 – 9 Beliefs of personal self-effi- based components of the intervention.
cacy operate in their own right upon behavior and they The primary evaluation questions are as follows.
act upon other SCT determinants.5,8 According to First, was there an effect at any point during the longi-
Bandura,7 sustained high self-efficacy is contingent tudinal intervention upon the personal determinants
upon several sources, the most authentic influence be- of dietary and physical activity behavior? Second, were
ing successful mastery of increasingly challenging ex- any of the changes in the determinants sustained after
periences. As an individual begins to exercise control the 2.5-year period of intervention? The student out-
over his/her environment, other forces act in concert comes investigated here are self-reported behaviors,
with self-efficacy to affect behavior. These efficacy-acti- knowledge, self-efficacy, perceived social reinforcement
vated processes include cognitive (such as personal goal and support, and intentions. The data are also ana-
setting and problem solving), motivational (anticipa- lyzed for gender differences.
tory beliefs about the likely outcomes and the value
placed on those outcomes), affective (e.g., perceptions METHOD
of potential threats and coping response processes,
such as perceived social reinforcement and support), Design
and selection (choices made regarding activities and
environments).5,8 The CATCH curricula were designed A total of 96 elementary schools participated in the
to influence self-efficacy and selected efficacy-activated CATCH randomized intervention trial, which collected
processes during skill acquisition. Self-regulatory pro- data from an intervention group and a control group.
cesses such as self-monitoring and goal-setting activi- The intervention group was divided into two sub-
ties were added during the fourth and fifth grades.1,2 groups, each receiving different levels of CATCH inter-
Drawing from the concept of reciprocal determinism vention. The first subgroup’s intervention consisted of
from SCT,5,6 the intervention model addressed individ- CATCH health education curricula, a physical educa-
ual cognitive processes and behavioral processes and tion program, a campus no-smoking policy, and a school
influences, as well as the physical and social environ- food service intervention program. The second sub-
ment. The interventions included a series of skill-based group’s school-based intervention was enhanced by a
classroom curricula, modifications of the school envi- home-based intervention. The control group partici-
ronment, and home-based activities—all designed to pated in the schools’ standard health education curricu-
help children exercise more control over their choices lum, but did not receive any CATCH intervention.
for diet and physical activity behavior. The strategies The independent variable for this study was the
utilized in all three components of the reciprocal triad CATCH intervention program. The dependent vari-
included behavioral modeling, skills training, practice, ables are psychosocial determinants of behavior tai-
reinforcement, eliciting social support, goal setting, so- lored to measure specific antecedents of cardiovascular
cial norm setting, and improved access to the resources disease (CVD) risk reduction behavior. These anteced-
needed for the performance of the behaviors (e.g., suf- ents were dietary intention, usual food choices, dietary
ficient physical education equipment and low-fat food knowledge, perceived support for physical activity, per-
options in the school cafeteria). ceived social reinforcement for healthy food choices,
The main effects of the CATCH trial on school-level dietary self-efficacy, and physical activity self-efficacy.
environmental determinants (such as the fat content Using a nested design, several variables were covaried
of foods provided by the school food service) and the with the independent variable, such as school type and
individual level outcomes (such as cholesterol values) baseline scores on each scale score. The study site was
have been reported elsewhere.10 The effects of the inter- also covaried as part of the SAS mixed-models proce-
vention upon psychosocial determinants related to dure. This strategy is discussed further in the paper.
smoking are presented elsewhere in this issue of Pre- Measurements of the psychosocial determinants
ventive Medicine.11 Also in this issue, a detailed discus- were made prior to intervention in the third grade and
444 EDMUNDSON ET AL.

throughout the intervention at the end of the school overall (x2 Å 0.10, P ú 0.70) or in any single semester,
year in grades 3, 4, and 5, between 1991 and 1994. gender, or ethnic subgroup. Participation was higher
Outcome comparisons were made at the level of treat- by 2.4% in intervention schools than in control schools
ment assigned, using school as the unit of analysis. at the Louisiana site, and lower by 1.7% in intervention
The CATCH team recruited 24 schools at each of its schools than in control schools at the California site;
four field sites (Austin, Minneapolis, New Orleans, and with sample sizes of over 9,000, these differences were
San Diego), consisting of 10 control and 14 intervention statistically significant but too small to represent a
schools. The Coordinating Center, the New England meaningful bias.
Research Institute, performed the randomization of Among all students enrolled during any of the four
schools to the control or intervention condition. Seven semesters when it was administered, 39% completed
of the 14 intervention schools at each site were ran- the questionnaire on all four occasions. Sixteen percent
domly assigned to receive only the school based inter- participated on three occasions, 20% twice, 20% once,
vention. The remaining 7 schools received the home- and 6% not at all. These rates did not differ signifi-
based intervention in addition to the school-based pro- cantly between control and intervention schools (x2 Å
grams. 0.36, df Å 4, P ú 0.98)

Subjects Measurement

The focus of this paper is on school-wide averages, The HBQ was developed and piloted during Phase I
in order to present the outcomes for all students ex- of CATCH (1988–1991), prior to the main field trials,
posed, to the intervention not merely the CATCH co- to measure psychosocial constructs related to diet,
hort. The subjects began the study as third-grade stu- physical activity, and intentions to smoke (fifth grade)
dents (n Å 7,795) enrolled in the 96 CATCH schools among elementary schoolchildren. The constructs mea-
during the 1991–1992 academic year. The mean age sured are behaviors and considered determinants of
of the students at baseline was 8.75 years (SD Å 0.487). behaviors underlying CVD risk factors. The scales were
There were no significant differences among study sites based on SCT: some had been utilized in previous work,
or school type (i.e., control vs intervention) for age. The including Hearty Heart and Friends,13 the Minnesota
numbers of male and female subjects were virtually Heart Health Program,14 and GO for Health,15 while
equal, with the exception of the Louisiana site which others were developed specifically for CATCH. Several
had a slightly greater number of males (54% vs 46%). SCT constructs were considered for the assessment.
During the fifth-grade administration, 7,944 students Ultimately, the constructs selected for measurement
completed the HBQ. A x2 analysis revealed no statisti- were those that guided the intervention.
cally significant gender or race differences in the sam- The constructs measured by the HBQ include dietary
ple distributions between the third-, fourth-, and fifth- intention, usual food consumption, dietary knowledge,
grade administrations. positive support for physical activity, negative support
The distribution of racial and ethnic representation for physical activity, perceived social reinforcement for
varied among the four sites. Minnesota had a predomi- healthy food choices, dietary self-efficacy, physical ac-
nately Caucasian sample (approximately 87%), with tivity self-efficacy, and smoking acquisition. The smok-
African-Americans being the largest minority group ing acquisition scale is the focus of another paper, thus
represented (3.5%). In California, Caucasians were also those findings are not reported here.
the predominant racial/ethnic group (about 70%), with The procedures undertaken to establish the reliabil-
Hispanic subjects being the largest minority group (ap- ity and validity of the scales that measured dietary
proximately 16%). Louisiana had the largest represen- intention, usual food choices, and dietary knowledge
tation of African-American subjects (about 30%), with have been reported elsewhere.13,14 The scales that mea-
Caucasians making up 67% of the subject population. sured self-efficacy, for dietary choices and for physical
The Texas site had the largest representation of His- activity, were first described by Parcel et al.15 Addi-
panic subjects (38%), with African-Americans consti- tional psychometric evidence for the dietary self-effi-
tuting 16%, while approximately 44% of the study sam- cacy scale and usual food choices scale was obtained
ple was Caucasian. during the CATCH pilot study.16 Scales which mea-
In the baseline administration (Fall 1991, grade 3), sured perceived social reinforcement for healthful food
7,795 of the 8,565 enrolled students (91%) completed choices and perceived support for physical activity were
the health behavior questionnaire (HBQ). The partici- developed and evaluated during the CATCH pilot.
pation rate remained at 91% in the subsequent two Those scales were evaluated for internal consistency
spring administrations (grades 3 and 4) and at 89% (coefficient a), content validity, and factorial validity.
in the spring of grade 5. Participation did not differ Perceived support for physical activity was conceptual-
significantly between control and intervention schools ized as two constructs: positive social support and neg-
 DETERMINANTS OF DIET AND PHYSICAL ACTIVITY 445

ative social support (e.g., criticism). The coefficient a your health,’’ for example, whole wheat bread or white
and factor analyses supported maintaining the concep- bread. A total of 14 items were used in the scale, where
tual distinction; the result was two scales, one for each one choice was more heart healthful (lower fat, lower
type of support. The perceived social reinforcement for sodium) than the other option (third grade a Å 0.76;
healthful food choices scale also demonstrated evidence fifth grade a Å 0.78).
of reliability and validity. Additionally, in order to max-
imize the validity of the children’s responses, the proto- Perceived support for physical activity. This section
col for the administration of the HBQ was pilot-tested of the HBQ was divided into two subscales: a positive
for strategies to ensure student confidentiality. support for physical activity subscale (11 items, third
Overall, the reliability values were reasonable for grade a Å 0.68; fifth grade a Å 0.67) and a negative
third-grade children,4 and therefore were retained for support subscale (7 items, third grade a Å 0.60; fifth
the subsequent measurement periods. Cronbach’s a co- grade a Å 0.56). The dichotomously scored items at-
efficients were calculated for every scale for each of the tempted to measure perceived support (positive or neg-
four measurement periods, for the total sample, and for ative) for physical activity among family members,
the control and experimental groups separately. The teachers, and friends (e.g., ‘‘One or both of my parents
results for the two groups were essentially the same. or guardians do exercises with me like running, jog-
The largest difference in a coefficients between control ging, dancing, or skating’’). This scale was developed
and intervention, for a same measurement period com- specifically for CATCH.
parison, was 0.07 (which appeared for one scale only);
Social reinforcement for healthy food choices. This
the mode difference for all the scales was 0.02. Given
21-item scale, developed specifically for CATCH, mea-
the stability of these coefficients, a brief description of
sured perceived social reinforcement from family,
each scale with the standardized reliability coefficient
teachers, and friends for heart-healthy foods (third
values for all students, during third grade and 2 years
grade a Å 0.86; fifth grade a Å 0.89). Utilizing the
later when they completed the questionnaire in fifth
dichotomous forced-choice format, students were asked
grade, is provided below.
whether or not their family members wanted them to
Dietary intention. This scale was designed to mea- eat the lower fat or lower sodium food presented in the
sure students’ intentions to choose foods considered question (e.g., apples instead of cookies). The scale also
heart healthful, i.e., those foods with lower fat and measured students’ perception of social reinforcement
lower sodium. The scale consisted of 13 dichotomous from teachers and friends for eating the same heart-
forced-choice items in which students were asked healthy foods used in the family items. A sample item
which food they would choose, e.g., ‘‘Which food would is ‘‘Who wants you to eat lots of fruits and vegetables?
you ask for, frozen yogurt or ice cream?’’ The food prod- Your parents (yes/no)? Your teachers (yes, no)? Your
ucts presented in each item were realistic options for friends (yes, no)?’’ The potential changes in perceived
each comparison, wherein one of the choices was either reinforcement from the different social groups (par-
lower in sodium, lower in fat, or lower in both than the ents, teachers, friends) were salient to investigate: we
other. The standardized coefficient a for this scale was expected to see an increase in the influence of peers as
acceptable (third grade a Å 0.78; fifth grade a Å 0.76). the students grew from childhood to preadolescence.
Therefore, coefficient a’s (third and fifth grade, respec-
Usual food choices. This 14-item scale specifically tively) for each source of social reinforcement are as
asked about usual food choices (behavior) in a forced- follows: social reinforcement from family (a Å 0.58;
choice format, again focusing on lower fat or lower so- 0.73) teachers (a Å 0.74; 0.87), and friends (a Å
dium food options (third grade a Å 0.76; fifth grade a 0.67; 0.83).
Å 0.76). In each item students were asked which of the
two foods presented they eat more often, e.g., ‘‘What Dietary self-efficacy. The 16 items that constituted
foods do you eat most of the time, cookies or apples?’’ this scale measured the children’s self-efficacy to
Comparable foods were presented across a variety of choose foods lower in fat and sodium. The scale used a
binary choices, with one of the two foods being more 3-point ordinal response set (i.e., not sure, a little sure,
heart healthy than the other. This scale was designed very sure) for each item. As with the other dietary
to measure usual food selections, rather than serve as scales, the children were asked to choose between two
a daily record of consumption. Therefore, this scale foods or preparation techniques wherein one of the
measures what a child eats most of the time. foods was more heart healthful, e.g., ‘‘How sure are you
that you can eat cereal instead of a donut?’’ (third grade
Dietary knowledge. The third scale was knowledge- a Å 0.83; fifth grade a Å 0.87)
based and targeted foods considered heart healthful,
i.e., those foods with lower fat and lower sodium. Stu- Physical activity self-efficacy. The last scale com-
dents were asked to identify which food was ‘‘better for prised five items (third grade a Å 0.67; fifth grade a
446 EDMUNDSON ET AL.

Å 0.69) and followed the 3-point ordinal response set terminants was taken at each time point. With respect
applied to the dietary self-efficacy scale. These items to children, the analysis is cross-sectional, and the fol-
measured children’s confidence in their ability to par- low-up rate for individual children is not of concern.
ticipate in various age appropriate aerobic activities. With respect to schools, the randomized units, all 96
A sample item from this scale is ‘‘How sure are you schools participated at every time point.
that you can be physically active 3–5 times a week?’’ Questionnaires were also distributed during the next
Several strategies were undertaken to investigate 2 years to each student in attendance during a regular
the validity of these scales. Factorial validity was esti- class period, in Spring 1993 and Spring 1994 (after the
mated for the physical activity measures. The food classroom curriculum had been completed). The ques-
choices presented in the dietary measures were compa- tionnaire was administered as an approved school dis-
rable across all four of the scales, in order to ensure trict activity so all students present on that class day
that any findings observed could be attributed to real were encouraged and expected to participate.
differences between the constructs rather than scaling Survey administrators for all 3 years were members
artifacts. Additionally, a multiple regression analysis of the CATCH evaluation staff at each site, who were
was employed to explore the concurrent form of crite- trained to implement the standardized protocols devel-
rion-related validity17 for the dietary measures, in oped by the Coordinating Center. For the purpose of
which the usual food choices scale was the dependent this study, a standardized protocol refers to strict ad-
variable. Usual food choice is a measure of behavior herence to one specific protocol, applicable to all four
and therefore the logical choice for examining the valid- field sites, which was implemented during every ad-
ity of the other dietary scales. We conducted this pre- ministration of the survey. This protocol included in-
liminary analysis to examine whether the dietary psy- structions for survey administration to ensure the in-
chosocial determinants were performing according to tegrity of the data collection process and protect
theory. This type of analysis is commonly used to inves- against a positive response bias. Many strategies were
tigate latent constructs for evidence of validity (as fac- employed to maximize the integrity of the data; several
tor analysis is used to provide evidence of one type of will be discussed here.
construct validity). First, the sequence of the scales within the survey
The outcome of the multiple regression analysis pro- was such that intention and behavior were measured
vided additional evidence of validity for the dietary prior to knowledge, social reinforcement, and self-effi-
scales. Dietary intention was highly correlated with cacy. This arrangement minimized the potential for
the dependent variable; to avoid potential suppressor students to base their behavioral responses on what
variable effects, it was not included in the analysis. they thought was healthy for them, or what they
CATCH field site, school type, baseline values for the thought their parents (or teachers or friends) wanted
psychosocial determinants, and determinants by gen- them to eat.
der interaction terms were included in the model as Second, evaluation staff, not intervention staff or
fixed effects. Statistically significant (P õ 0.0001) vari- teachers, administered the surveys. None of the inter-
ables in the model were knowledge, self-efficacy, and vention staff were allowed to participate in the data
perceived social reinforcement for healthful food collection for these assessments.
choices from parents and friends. Self-efficacy and per- The third strategy pertained to the protocol for the
ceived social reinforcement from parents for healthy survey administration. The Coordinating Center pro-
food choices accounted for approximately 41% of the vided an annual training which included at least one
variance in usual food choices (partial r Å 0.26 and representative from each site. These CATCH evalua-
0.15, respectively). Perceived social reinforcement from tion staff, in turn, coordinated and supervised the
friends had a partial r Å 0.10. Results indicated the training of additional evaluation staff back at their site.
interactions by gender were not significant for any of To ensure staff compliance with the survey protocol,
the variables in the model. Overall, the result of this each administrator was randomly observed by a data
analysis supported the validity of the dietary psychoso- collection supervisor during the assessment period. Ad-
cial constructs. ditionally, site visits were conducted by the Coordinat-
ing Center during the data collection periods.
Procedures The HBQ was read aloud; therefore, survey adminis-
trators were trained to avoid voice inflections which
During the 1991–1992 academic year, the HBQ was could be construed to prefer one choice over another.
administered twice, at the beginning and at the end of Each potential administrator was required to master
the third-grade year, to 6,956 students enrolled in 96 this skill before being allowed to read the survey to the
schools at four sites across the United States. The anal- children. Additionally, at least two CATCH staff were
ysis in this paper employs a cohort of 96 schools, in present in the room for each administration. One per-
which a cross-sectional ‘‘snapshot’’ of psychosocial de- son read the questions aloud, while the students fol-
 DETERMINANTS OF DIET AND PHYSICAL ACTIVITY 447

lowed and recorded responses onto their copies of the tion, in addition to the main effects of time (3 df),
questionnaire. The second staff person proctored the CATCH intervention group (2 df), and time 1 interven-
survey by walking around answering questions from tion interaction (6 df).
the children. Children were not allowed to ask ques- The interaction term was divided into three orthogo-
tions out loud—they were instructed to raise their nal contrasts for the following principal research ques-
hand if they had a question, and the proctor would tions or tests: (a) Would the separation between inter-
come to their desk to answer it. Children were also vention and control HBQ scores be different at the end
instructed to use a cover sheet while they completed of CATCH (semester 6), i.e., would CATCH produce a
the survey so that their answers would be private. lasting effect on the HBQ score? (b) Did the separation
At the beginning of every section of the survey, chil- between intervention and control HBQ scores change
dren were informed that there were no right or wrong between baseline (semester 1) and the interim time
answers and that this was not a test. Teachers stayed points (semesters 2 and 4)? That is, did CATCH pro-
in the classroom during the administration to maintain duce at least a transient effect on the HBQ score? (c)
class discipline, but were requested not to walk around Did the separation, if any, between schools receiving
and were prohibited from viewing children’s responses. school-based intervention only and schools receiving
The questionnaire required approximately 30–40 min additional home-and-family intervention change over
to complete. When the administration was complete, the four measurement periods, i.e., did the family com-
the questionnaires were collected and immediately re- ponent of CATCH have some effect on the HBQ score?
moved from the school. Numerator degrees of freedom for these three tests
Students whose primary language was Spanish re- were 1, 2, and 3, respectively. Denominator degrees of
ceived a Spanish version of the questionnaire, while a freedom were 279 in each case. To supplement each
bilingual CATCH staff person read the questionnaire significance test, an effect size was calculated as fol-
aloud in Spanish. The initial Spanish translation of the lows: h Å [F/(F / dfn/dfd,)]1/2, where F indicates the
HBQ was supervised by one site and cross-translated Fisher F statistic from MANOVA and dfn and dfd indi-
for accuracy by another. The Texas and California sites cate numerator and denominator degrees of freedom,
alternated these two responsibilities during the 3-year respectively.18
period. Professional Spanish translators, with experi- The SAS/STAT MIXED procedure was used for re-
ence in translating elementary school educational ma- peated-measures analysis.19,20 Additional SAS/STAT
terials, were hired at both sites to ensure the accuracy procedures were used for Cronbach a and correla-
and readability level. Less than 2% of the total third- tions.21 Pearson product moment correlations were cal-
grade sample, and less than 1% of the fifth-grade sam- culated among all of the psychosocial determinants to
ple, took the Spanish version of the HBQ. investigate several theoretical relationships. First,
SCT5 emphasizes the interactive nature of the relation-
Analysis ships among these determinants; thus a sense of the
general strength of each relationship would be ob-
Because schools were the randomized units in served among reliable and valid measures of the deter-
CATCH (rather than individual children), the present minants. Second, Bandura and others have reported
analysis was conducted on school-level average mea- that, as the skill acquisition process unfolds, constructs
sures of determinants of health behavior. All respon- other than self-efficacy within SCT tend to exert more
dents to the HBQ were included in the school average, influence on behavior, particularly after individuals
whether or not they were members of the CATCH co- have practiced and reached a moderate level of mas-
hort as defined by venipuncture data. The items within tery. Therefore, these correlations were calculated for
each scale of the HBQ were summed to produce an the intervention and control groups separately, for each
individual scale score. The individual scores were ag- of the four measurement periods, in order to test the
gregated to produce school average scale scores at each relative strength of these relationships over time.
of the four measurement periods (semesters 1, 2, 4, and The effect of gender upon the psychosocial variables
6). The number of data points available for each scale was examined with the multivariate procedure, MAN-
was thus 384 (96 schools 1 4 periods). COVA. MANCOVA was used to examine the intermit-
Repeated-measures multivariate analysis of vari- tent and the final effects of each psychosocial determi-
ance (MANOVA) was conducted to determine whether nant by gender, and an h coefficient was calculated for
the CATCH intervention schools diverged from the con- each variable.
trol schools over time. The analysis was performed sep-
arately for each HBQ scale as dependent variable. In- RESULTS
dependent variables in the MANOVA included a main
effect for CATCH field site (3 df), to adjust for any Figures 1, 2, and 3 represent mean differences and
demographic, geographic, or other site-related varia- standard errors of the two intervention conditions from
448 EDMUNDSON ET AL.

analysis of each of the subscales of the social reinforce-

ment scale (i.e., reinforcement from friends, parents,
and teachers) revealed a significant effect for gender
during the fifth grade: girls perceived greater reinforce-
ment for healthful eating than boys. Self-efficacy for
dietary choices and physical activity increased moder-
ately during the third grade, but subsequently declined
during the last 2 years of the intervention.
No statistically detectable effect for the perceived
negative or positive support for physical activity was
found at the end of the intervention. Perceived positive
support for physical activity did improve immediately
following the third-grade program, but this effect was
not sustained in the subsequent years. The CATCH
program did not affect a reduction in perceived nega-
tive support for physical activity. (See Fig. 3.) Signifi-
cant improvements were observed in self-efficacy for
physical activity after the third-and fourth-grade inter-
ventions, but dropped to the same levels as controls
during fifth grade. This overall decline in the effects
provides additional support that these findings were
not due to a positive response bias.
The results presented in these figures display base-
line values as well as the outcomes at the completion
of each year of the intervention. The values for the
effect sizes, in this case the h coefficients, provided evi-

FIG. 1. Effects of CATCH intervention on diet-related behavior

and knowledge, according to self-report in elementary-school chil-
dren. All three scales rose significantly in intervention schools com-
pared with control schools in the interim period (spring semester,
grades 3 and 4) and remained significantly elevated at follow-up
(grade 5). Range of each scale is indicated in parentheses. Symbols
indicate estimated mean { standard error, from repeated-measures
analysis of variance. For statistical details see Table 1.

the control condition (which is represented by the bro-

ken line in each figure) at each of the four measurement
periods. Scores for the control condition served as the
contrast values for the repeated measures analysis,
which adjusts for lack of independence among the re-
peated measures.
Dietary intention, usual food choice, and knowledge
improved significantly within the intervention schools,
and that effect was sustained across the 3 years. (See
Fig. 1.) The largest improvement occurred at the begin-
ning of the program. Declines in the effects were ob-
served as the children were exposed to 3 years of the
FIG. 2. Effects of CATCH intervention on diet-related social rein-
intervention; this pattern of a plateau effect or of a forcement and self-efficacy, according to self-report in elementary-
decrease in the size of the effect was not unexpected school children. Both scales rose significantly in intervention schools
for a longitudinal program. compared with control schools at the first interim observation (spring
Figure 2 presents the results for dietary self-efficacy semester, grade 3). Social reinforcement scale remained significantly
and perceived social reinforcement for healthy food elevated at subsequent follow-up periods (grades 4 and 5), while
self-efficacy scale returned to control levels. Range of each scale is
choices. The intervention effect was strongest for the indicated in parentheses. Symbols indicate estimated mean { stan-
reinforcement determinant, and these effects were sus- dard error, from repeated-measures analysis of variance. For statisti-
tained at relatively high levels during all 3 years. An cal details see Table 1.
 DETERMINANTS OF DIET AND PHYSICAL ACTIVITY 449

Only one significant effect was noted between the

school-only condition and the school-plus-family condi-
tion. Students enrolled in the school-plus-family inter-
vention scored significantly higher on the construct di-
etary knowledge (P õ 0.02). None of the other psychoso-
cial determinants were meaningfully different between
the two intervention conditions.
Recall that a MANCOVA was conducted to deter-
mine the effects of gender upon the variable outcomes.
A gender effect was detected only for perceived social
reinforcement from friends for healthful eating, which
was significant for girls. The influence of the family
intervention for girls was observed in the effect size for
dietary knowledge (P õ 0.006). No other significant
gender differences were observed from this analysis.
Pearson correlations between the psychosocial deter-
minants were calculated for all four measurement peri-
ods (i.e., preintervention third-grade baseline and ev-
ery spring after completion of the curriculum for grades
3, 4, and 5). The patterns over time are illustrated here
by the data collected at the end of the third grade and
the end of the fifth grade curriculum: these correlations
are presented in Tables 2a, 2b, 3a, and 3b. Table 2a
illustrates the interrelationships among the variables
when the subjects were in the third grade, while the
second part of the table, 2b, presents the correlations
calculated during the fifth grade for the intervention
FIG. 3. Effects of CATCH intervention on social support and
self-efficacy related to physical activity, according to self-report in
elementary-school children. Positive support and self-efficacy scales
diverged modestly from control at the first interim period (spring TABLE 1
semester, grade 3) but returned to control levels by follow-up (grade
5). Range of each scale is indicated in parentheses. Symbols indicate Effects of CATCH Intervention on Psychosocial
estimated mean { standard error, from repeated-measures analysis Determinants of Cardiovascular Risk Behavior
of variance. For statistical details see Table 1.
Interima Finalb

Effect Effect
dence to determine whether there were longitudinal size Significance size Significance
sustained effects at the end of the 3 years: these results Psychosocial scale (h) (P) (h ) (P)
are presented in Table 1. Intermittent effects that may
Dietary intention 0.581 õ0.001 0.265 õ0.001
have occurred during the third and/or fourth grade, but Usual food choice 0.553 õ0.001 0.236 õ0.001
perhaps did not prevail into the fifth grade, are also Dietary knowledge 0.661 õ0.001 0.406 õ0.001
presented. Scores from the intervention schools were Physical activity
significantly higher for each dietary related scale, with Positive support 0.202 0.003 0.021 0.70
Negative support 0.048 0.72 0.047 0.43
the most substantial differences observed for the di- Food choice
etary social reinforcement, intention to eat healthy, Social
and knowledge scales. Table 1 illustrates the effect reinforcement,
sizes and P values for the mean HBQ scale scores, com- total 0.591 õ0.001 0.452 õ0.001
paring the control group with the combined interven- From friends 0.395 õ0.001 0.123 0.040
From parents 0.501 õ0.001 0.342 õ0.001
tion groups, adjusted for preintervention values as well From teachers 0.645 õ0.001 0.600 õ0.001
as site and school variation. There was no significant Self-efficacy
interaction between intervention effect and gender. Dietary 0.235 õ0.001 0.102 0.088
Ethnic composition of the school at each time point, Physical activity 0.171 0.016 0.107 0.074
when added to the MANOVA as a school-level covariate a
From F test of H0 : (intervention 0 control)semesters2,4 Å (interven-
(in the form of the percentages of Caucasian, African- tion 0 control)semester1; df Å 2, 279.
American, and Hispanic respondents), had negligible b
From F test of H0 : (intervention 0 control) semester 6 Å (intervention
impact on the statistical inferences and effect sizes. 0 control) semester1 ; df Å 1, 279.
450 EDMUNDSON ET AL.

TABLE 2a
Descriptive Statistics and Pearson Correlations among Psychosocial Determinants for Third-Grade Students
after 1 Year of CATCH Intervention—Intervention Condition (n Å Approx 4,400)

Measure 1 2 3 4 5 6 7 Mean SD

1. Dietary intention — 6.29 6.39

2. Usual food choices 0.72* — 4.27 7.28
3. Dietary knowledge 0.45* 0.38* — 9.46 4.76
4. Diet. soc. reinforcement 0.45* 0.44* 0.37* — 6.92 9.78
5. Dietary self-efficacy 0.51* 0.51* 0.36* 0.40* — 7.69 6.57
6. Positive support for PA 0.26* 0.29* 0.20* 0.38* 0.31* — 6.40 4.43
7. Negative support for PA 0.14* 0.12* 0.28* 0.20* 0.19* 0.24* — 4.34 2.89
8. Phys. act. self-efficacy 0.29* 0.31* 0.26* 0.30* 0.53* 0.38* 0.18* 2.97 2.24

* All correlations are significant at P õ 0.0001.

group. Tables 3a and 3b present the correlations for structs are not the same, and the scales used to mea-
the control group. sure the constructs were valid and adequately sensitive
The correlation matrices revealed weak to moderate to detect the dissimilarity as well as the similarity
interrelationships among most of the determinants, among the constructs.
with one exception. The relationship between intention Correlations among the physical activity determi-
and usual food choices remained high (r Å 0.73 to 0.76) nants remained about the same across the 3 years of
across all 3 years. According to Bandura,7,8 intention measures for control and intervention groups. Physical
is a proximal goal; thus the students were setting short- activity self-efficacy was moderately correlated with
term goals which were compatible with their reported perceived positive support for physical activity, while
usual behavior. Although the correlation between in- the relationship between self-efficacy and perceived
tention and usual food choice behavior suggested simi- negative support was weak.
larity between the constructs, the relationship demon-
strated a theoretically consistent amount of unshared
DISCUSSION
variance, in this case about 66% (the coefficient of
alienation value, or k).21 Therefore, the empirical evi-
dence reflected very well the hypothesized theoretical The combined effects of the classroom curriculum
relationship for both the intervention and the control strategies, environmental changes in the school food
group. service, and the physical education program did lead
Excluding the correlation between intention and to changes in what children themselves expect to do,
usual food choices, the correlations among the dietary their confidence in being able to perform the behavior,
constructs (knowledge, usual behavior, self-efficacy, and their perceived support for their behavior to con-
and intention) were moderate, which provided empiri- sume lower fat, lower sodium food, and to engage in
cal evidence for the unique variance expressed by those additional moderate-to-vigorous physical activity. Sig-
constructs. This evidence demonstrates that the con- nificant effects were found early in the intervention for

TABLE 2b
Descriptive Statistics and Pearson Correlations among Psychosocial Determinants for Fifth-Grade Students
after 3 Years of CATCH Intervention—Intervention Condition (n Å Approx 4,250)

Measure 1 2 3 4 5 6 7 Mean SD

1. Dietary intention — 1.19 6.58

2. Usual food choices 0.76* — 0.33 6.87
3. Dietary knowledge 0.32* 0.32* — 10.09 4.88
4. Diet. soc. reinforcement 0.45* 0.44* 0.30* — 3.54 10.20
5. Dietary self-efficacy 0.54* 0.54* 0.32* 0.37* — 5.77 7.15
6. Positive support for PA 0.25* 0.26* 0.23* 0.31* 0.28* — 5.76 4.38
7. Negative support for PA 0.16* 0.17* 0.34* 0.18* 0.23* 0.32* — 4.93 2.61
8. Phys. act. self-efficacy 0.24* 0.27* 0.24* 0.22* 0.46* 0.39* 0.22* 2.79 2.24

* All correlations are significant at P õ 0.0001.

 DETERMINANTS OF DIET AND PHYSICAL ACTIVITY 451

TABLE 3a
Descriptive Statistics and Pearson Correlations among Psychosocial Determinants
for Third-Grade Students—Control Condition (n Å Approx 3,100)

Measure 1 2 3 4 5 6 7 Mean SD

1. Dietary intention — 1.95 6.22

2. Usual food choices 0.70* — 0.54 6.41
3. Dietary knowledge 0.46* 0.39* — 5.43 5.54
4. Diet. soc. reinforcement 0.42* 0.38* 0.32* — 0.43 9.72
5. Dietary self-efficacy 0.44* 0.41* 0.35* 0.30* — 6.08 6.62
6. Positive support for PA 0.23* 0.23* 0.22* 0.30* 0.22* — 5.56 4.60
7. Negative support for PA 0.18* 0.12* 0.32* 0.13* 0.21* 0.17* — 3.86 3.11
8. Phys. act. self-efficacy 0.25* 0.25* 0.27* 0.23* 0.46* 0.33* 0.18* 2.52 2.34

* All correlations are significant at P õ 0.0001.

some of the physical activity determinants; however, for several of the variables. Therefore, we do not believe
those effects were not significant by fifth grade. this potential source of bias greatly affected the effects
The effects of CATCH could have been diluted by reported in this study.
the loss of exposed students and the in-migration of
unexposed students. This could lead to Type II error, Implications for School-Based Health Promotion
i.e., failing to obtain significant findings where effects Programs for Youth
actually existed. With respect to these results, how-
ever, we did not lack for significant findings. Therefore, Although significant effects were noted for most of
Type II error was not a great concern. the psychosocial determinants, the decline in all of the
Alternately, the effects observed from the CATCH fifth-grade measures implies a need for programming
intervention could have been positively biased by the modifications for older preadolescents. At a time when
children, who may have responded to please the inves- youth are preparing for the transition from childhood
tigators, rather than reporting their true beliefs and to adolescence and from elementary school to middle
behaviors. This potential threat to obtaining valid re- school, the influence of the developmental stage and
sults from survey research is called social desirability the perceived cultural demands rapidly increases.
bias. To minimize this potential influence upon the Thus, the preadolescents are in flux, which lays the
data, a rigorous survey administration protocol was groundwork for the competitive influences of other psy-
implemented. The protocol was grounded in state of chosocial factors with the CATCH program. Develop-
the art recommendations for survey research and psy- mentally, children begin to experience a need for auton-
chometric data collection. Strict adherence to this pro- omy as well as peer approval. These seemingly conflict-
tocol was discussed earlier in this paper. While these ing needs tend to result in a shift from parental/adult
procedures may not have completely prevented the in- influence to peer influence upon behavior. The out-
fluence of social desirability bias, the results indicate comes observed in the fifth graders with regard to the
that students did not aim to please the investigators decreased support from friends for nutritional eating

TABLE 3b
Descriptive Statistics and Pearson Correlations among Psychosocial Determinants
for Fifth-Grade Students—Control Condition (n Å Approx 2,900)

Measure 1 2 3 4 5 6 7 Mean SD

1. Dietary intention — 00.83 5.93

2. Usual food choices 0.75* — 01.16 6.30
3. Dietary knowledge 0.34* 0.34* — 7.73 5.44
4. Diet. soc. reinforcement 0.38* 0.37* 0.29* — 02.31 10.97
5. Dietary self-efficacy 0.46* 0.45* 0.33* 0.21* — 5.99 7.16
6. Positive support for PA 0.17* 0.17* 0.18* 0.24* 0.18* — 5.62 4.33
7. Negative support for PA 0.14* 0.15* 0.30* 0.14* 0.17* 0.28* — 4.66 2.68
8. Phys. act. self-efficacy 0.18* 0.19* 0.22* 0.15* 0.40* 0.32* 0.20* 2.86 2.20

* All correlations are significant at P õ 0.0001.

452 EDMUNDSON ET AL.

among males and declining self-efficacy may be a re- ity. The dominant goal of the PE curriculum was to
flection of this shift. increase the amount of moderate to vigorous physical
In particular, the decline in self-efficacy may relate activity during PE class with activities that were fun
to a failure in the curriculum to effectively identify and cooperative. Although the CATCH classroom cur-
salient barriers encountered by the fifth graders to se- riculum included lessons designed to influence the psy-
lecting healthy food items. Bandura8 has emphasized chosocial determinants of physical activity, fewer les-
the successful mastery of activities which are contin- sons were devoted to physical activity compared with
gent upon experiences that are very similar to situa- dietary change. Because the students were receiving a
tions regularly faced in the environment. Mastery and curriculum devoted to physical activity via CATCH PE,
self-efficacy can be undermined by verbal persuasion and because of the practical limitations of how many
(such as high-tech, sophisticated mass media messages classroom health education lessons could be taught,
portrayed in advertisements) and social influences of the classroom curricula included proportionately more
the environment8 (e.g., the cultural popularity of high dietary lessons. Assumptions regarding the indirect ef-
fat, fast food meals and convenience snacks and seden- fects of such physical activity programs on psychosocial
tary social activities). determinants of behavior should be examined.
Given the increasing pressure of social conformity, The CATCH findings on perceived social support may
shifts in social support networks, and increasing inde- also help explain the decline of physical activity with
pendence experienced during early adolescence,23,24 it age that has been observed in school-age children and
is possible that the environmental and psychosocial adolescents. Physical activity options span a broad
barriers experienced by the fifth graders in their daily range of involvement. This includes individual activi-
lives were not adequately addressed by the curriculum ties, team play, and even sedentary activities, all of
skill-building activities or mechanisms of social sup- which the child chooses to initiate, or to join a some-
port. The predictive strength of self-efficacy has been what organized group, in order to participate. Per-
demonstrated in some studies to decline over time.9,25,26 ceived support for these efforts could prove crucial for
This is due in part to an increase in the salience of the child’s participation. Even when physical activity
other influences as a task becomes less new, and the is a positively valued experience, participation often
dynamic changing nature of behavior and its context, requires planning and resources to ensure access to
rendering previous efficacy judgments less predictive. equipment and a relatively safe place to play; many
Therefore, perhaps the fifth-grade curriculum was of activities also require the availability of teammates.
insufficient intensity and duration to achieve mastery Physical activity choices have been previously demon-
or at least balance the environmental and social strated to change over time, particularly for girls, who
changes encountered in the rapidly changing world of become less active and engage in more sedentary social
the preadolescent. On the other hand, there may be a activities as they become older.29,30
limit to the power of a school-based curriculum in the The lower reliability coefficients for the physical ac-
face of the dominant culture. tivity scales and the trends over time suggest a need
As image and social needs become more important for modifications in the measurement approach of the
in decision making, cultural messages become more psychosocial determinants of physical activity used in
valued. The act of eating may shift in meaning from a CATCH, the measurement of additional determinants,
nutritional event to a social event; food choices may and possibly a better match between the goals and ob-
become signs of camaraderie or symbols of indepen- jectives of the PE curriculum with SCT.5 These recom-
dence. As a result, programs for students in fifth grade mendations are timely: sedentary behavior tracks
and higher may need to account for the differences in over adolescence and adulthood and is resistant to
meanings of dietary behaviors, as well as the high prev- change.30 – 32 Additional research with psychometrically
alence of convenience stores and fast food restaurants sound measures that span several SCT determinants,
where students often socialize. which in turn can be used to ground the physical activ-
ity intervention, could help break the pattern for seden-
Implications for Physical Activity and Measurement tary behavior among youth.
of Determinants
Implications for Future Research
The initial effects upon the physical activity determi-
nants observed during the 1st year of the intervention Our findings support the results of other studies
were not sustained. Possible explanations may relate which demonstrated that SCT and social influences
to the curriculum, environmental support, or actual theory methods can be effectively applied to school
activity trends.27 – 29. For example, the CATCH PE cur- health promotion programs.30 – 36 Additional studies
riculum was not designed to directly influence external which evaluated school-based smoking or substance
sources of support or reinforcement for physical activ- abuse prevention programs have reached similar con-
 DETERMINANTS OF DIET AND PHYSICAL ACTIVITY 453

clusions regarding the relative effectiveness of health instruments used to evaluate the psychosocial determi-
promotion programs based on social cognitive or social nants of school-based health promotion programs.
influences theories for youth.37,38,39 – 42 Evidence sup- Since CVD risk factors have their origins early in life,
porting the need to move beyond knowledge-only ap- early interventions through school-based programs
proaches into theory-based programming was de- represent the most promising means of reaching an
scribed in the recent Surgeon General’s report on pre- entire population of children in a community to modify
venting tobacco use among young people.43 risk factors that lead to CVD during adulthood.
The body of evidence on this new generation of school Applying the causal model assumed in the CATCH
health promotion implicates several salient features project, the starting place to reduce risk is to address
which appear to account for the effectiveness of school- the psychosocial determinants that lead to lifestyle be-
based programs. These features include interventions haviors that contribute to increased risk for CVD. The
of longer duration, community and environmental sup- evaluation of CATCH suggests that we can be success-
port, comprehensive programming, and changing social ful in this first step.
norms and influences.42,44 – 47 Analyses of smoking pre-
vention program effectiveness, including meta-analy- REFERENCES
ses, emphasize the importance of modifying social in-
fluences for creating sustained effects.40 – 42,48,49 Results 1. Perry CL, Stone EJ, Parcel GS, Ellison RC, Nader PR, Webber
from smoking interventions that have been primarily LS, Luepker RV. School-based cardiovascular health promotion:
school-based, without the support or training of the the Child and Adolescent Trial for Cardiovascular Health
(CATCH). J Sch Health 1990;60:406–13.
family or community (via mass media or family-based
programs), have also been limited.46,50 2. Perry CL, Parcel GS, Stone EJ, Nader P, McKinlay SM, Luepker
RV, Webber LS. The Child and Adolescent Trial for Cardiovascu-
The comparison between the school-based interven- lar Health (CATCH): overview of the intervention program and
tion condition and the school- plus-family intervention evaluation methods. Cardiovas Risk Fact 1992;2:36–44.
condition revealed a greater improvement in dietary 3. McGraw SA, Stone EJ, Osganian SK, Elder JP, Perry CL, John-
knowledge of lower fat and lower sodium foods for those son CC, Parcel G, Webber LS, Luepker RV. Design of process
students who had the additional family intervention. evaluation within the Child and Adolescent Trial for Cardiovas-
However, there were no additive effects as a result of cular Health (CATCH). Health Educ Q 1994;Suppl 2:S5–26.
the family intervention upon the other dietary determi- 4. Edmundson E, Parcel GS, Perry CL, Feldman HA, Smyth M,
Johnson CC, et al. The effects of the Child and Adolescent Trial
nants nor for the determinants of physical activity be- for Cardiovascular Health intervention on psychosocial determi-
havior. The family program implemented in CATCH nants of CVD risk factor behavior among third grade students.
may not have sufficient power to influence additional Am J Health Promo 1996;10:217–25.
effects in the psychosocial variables. These findings de- 5. Bandura A. Social foundations of thought and action. Englewood
part from previous health promotion research which Cliffs (NJ): Prentice Hall, 1986.
targets a single behavior (eating) among young chil- 6. Perry CL, Baranowski T, Parcel G. How individuals, environ-
dren: the multiple behaviors addressed by CATCH may ments, and health behavior interact: Social Learning Theory. In:
Glanz K, Lewis FM, Rimer BK, editors. Health behavior and
require family programs of longer duration and greater
health education: theory research and practice. San Francisco:
involvement.10 Alternatively, the family program may Jossey–Bass, Inc., 1990:161–86.
not have been sufficiently implemented by parents. 7. Bandura A. Moving into forward gear in health promotion and
This question cannot be answered by the CATCH proj- disease prevention. Keynote address presented at the Annual
ect: parental involvement data were not collected. We Meeting of the Society of Behavioral Medicine, San Diego, March
recommend that future research studies investigate 23, 1995.
these questions on family involvement more thor- 8. Bandura A. Exercise of personal and collective efficacy in chang-
oughly. ing societies. In: Bandura A, editor. Self-efficacy in changing
societies. New York: Cambridge Univ. Press, 1995:1–45.
9. Mitchell TR, Hopper H, Daniels D, George-Falvy J, James LR.
CONCLUSION
Predicting self-efficacy and performance during skill acquisition.
J Appl Psycho 1994;79:506–17.
The results presented in this paper provide evidence 10. Luepker RV, Perry CL, Elder J, Feldman H, Johnson C, Kelder
for the effectiveness of the CATCH program to influ- S, et al. Outcomes of a field trial to improve children’s dietary
ence changes in the psychosocial determinants of diet patterns and physical activity: the Child and Adolescent Trial
and physical activity behavior. These results are en- for Cardiovascular Health (CATCH). JAMA 1996;275:768–776.
couraging because the CATCH program represents an 11. Elder J, Perry CL, Stone EJ, Yang M, Parcel G, Johnson C,
et al. Stages of cigarette smoking acquisition in a multistate,
effort to translate theoretical methods of behavior
multiethnic population: the CATCH study. Prev Med 1996;
change into practical strategies that can work in a 25:486–494.
school setting. These findings can help improve our 12. Stone EJ, Osganien SK, McKinley S, Wu MC, Webber LS,
understanding of the application of theory to practical Luepker RV, et al. Operational design and qualitative control in
problems and improve the development and design of the CATCH multicenter trial. Prev Med 1996;25:384–399.
454 EDMUNDSON ET AL.

13. Perry CL, Mullis RM, Maile MC. Modifying the eating behavior cardiovascular health: the Minnesota Heart Health Program
of young children. J Sch Health 1985;55:399–402. youth program. Health Educ Res Theory Pract 1989;4:87–101.
14. Perry CL, Griffin G, Murray DM. Assessing needs for youth 34. Stone EJ, Perry CL, Luepker RV. Synthesis of cardiovascular
health promotion. Prev Med 1985;14:379–93. behavior research for youth health promotion. Health Educ Q
15. Parcel GS, Simons-Morton BG, O’Hara NM, Baranowski T, Wil- 1989;16:155–69.
son B. School promotion of healthful diet and physical activity: 35. Murray DM, Pirie P, Luepker RV, Pallonen U. Five- and six-year
impact on learning outcomes and self-reported behavior. Health follow-up results from four seventh-grade smoking prevention
Educ Q 1989;16:181–99. strategies. J Behav Med 1989;12:207–18.
16. Parcel GS, Emundson E, Perry CL, Feldman HA, O’Hara-Tomp- 36. Davis SM, Lambert LC, Gomez Y, Skipper B. Southwest cardio-
kins N, Nader PR, et al. Measurement of self-efficacy for diet- vascular curriculum project: study findings for American Indian
related behaviors among elementary school children. J Sch elementary students. J Health Educ 1995;Suppl 26:S72–81.
Health 1995;65:23–7. 37. Botvin GJ, Botvin EM. Adolescent tobacco, alcohol, and drug
17. Crocker L, Algina J. Introduction to classical and modern test abuse: prevention strategies, empirical findings, and assessment
theory. New York: CBS College Pub. 1986. issues. J Dev Behav Pediatr 1992;13:290–301.
18. Rosenthal R. Parametric measures of effect size. In: Cooper H, 38. Botvin GJ, Baker E, Dusenbury L, Botvin EM, Diaz T. Long-
Hedges LV, editors. The handbook of research synthesis. New term follow-up results of a randomized drug abuse prevention
York: Sage, 1994:231–44. trial in a white middle-class population. JAMA 1995;273:1106–
19. Murray DM, Wolfinger RD. Analysis issues in the evaluation of 12.
community trials: progress toward solutions in SAS/STAT 39. Walter HJ, Vaughan RD, Wynder EL. Primary prevention of
MIXED. J Community Psychol CSAP special issue 1994;1:140– cancer among children: changes in cigarette smoking and diet
54. after six years of intervention. J Nat Cancer Inst 1989;81:995–
20. SAS Institute, Inc. SAS procedures guide, version 6. 3rd ed. Cary 9.
(NC): SAS Institute, 1990:209–36. 40. Rundall TG, Bruvold WH. A meta-analysis of school-based smok-
21. SAS Institute, Inc. SAS/STAT user’s guide, version 6. 4th ed. ing and alcohol use prevention programs. Health Educ Q
Cary (NC) SAS Institute, Inc., 1992:289–366. MIXED: Technical 1988;15:317–34.
Report No. P-229. 41. Tobler NS. Meta-analysis of 143 adolescent drug prevention pro-
22. Vogt WP. Dictionary of statistics and methodology. New York: grams: quantitative outcome results of program participants
Sage, 1993. compared to a control or comparison group. J Drug Issues
1986;16:537–67.
23. Copeland EP, Hess RS. Differences in young adolescents’ coping
strategies based on gender and ethnicity. J Early Adolesc 42. Tobler NS. Drug prevention programs can work: research find-
1995;15:203–19. ings. J Addict Dis 1992;11:1–28.
24. Levitt MJ, Guacci-Franco N, Levitt JL. Convoys of social support 43. U.S. Department of Health and Human Services, Public Health
in childhood and early adolescence: structure and function. Dev Service, Centers for Disease Control and Prevention, National
Psychol 1993;29:811–8. Center for Chronic Disease Prevention and Health Promotion,
Office on Smoking and Health. Preventing tobacco use among
25. Williams CL, Toomey T, McGovern P, Wagenaar A, Perry CL. young people: a report of the Surgeon General. Washington: U.S.
Development, reliability and validity of self-report alcohol-use Gov. Printing Office, 1994.
measures with young adolescents. J Child Adolesc Subst Abuse.
1995;4:17–40. 44. Flynn BS, Worden JK, Secker-Walker RH, Pirie PL, Badger GJ,
Carpenter JH, et al. Mass media and school interventions for
26. Holden G. Moncher MS, Schinke SP, Barker KM. Self-efficacy cigarette smoking prevention: effects 2 years after completion.
of children and adolescents: a meta-analysis. Psychol Rep Am J Public Health 1994;84:1148–50.
1990;66:1044–46.
45. Flynn BS, Worden JK, Secker-Walker RH, Badger GH, Geller
27. Malina RM. Physical activity and fitness of children and youth: BM. Cigarette smoking prevention effects of mass media and
questions and implications. Med Exerc Nutr Health 1995;4:123– school interventions targeted to gender and age groups. J Health
35. Educ 1995;Suppl 26:S45–51.
28. Sallis JF, Hovell MF, Hofstetter CR, Barrington E. Explanation 46. Flay BR, Koepke D, Thomson SJ, Santi S, Best JA, Brown KS.
of vigorous physical activity during two years using social learn- Six year follow-up of the first Waterloo school smoking preven-
ing variables. Soc Sci Med 1992;34:25–32. tion trial. Am J Public Health 1989;79:1371–6.
29. Sallis JF. Epidemiology of physical activity and fitness in chil- 47. Ellickson PL, Bell RM, McGuigan K. Preventing adolescent drug
dren and adolescents. Crit Rev Food Sci Nutr 1993;33:403–8. use: long term results of a junior high program. Am J Public
30. Kelder SH, Perry CL, Klepp K-I, Lytle LL. Longitudinal tracking Health 1993;83:856–62.
of adolescent smoking, physical activity, and food choice behav- 48. Elder JP, Wildey M, de Moor C, Sallis JF, Eckhardy LE, Edwards
iors. Am J Public Health 1994;84:1121–6. C, et al. The long-term prevention of tobacco use among junior
31. Clarke WR, Lauer RM. Does childhood obesity track into adult- high school students: classroom and telephone interventions. Am
hood? Crit Rev Food Sci Nutr 1993;33:423–30. J Public Health 1993;83:1239–44.
32. Raitakari OT, Porkka KVK, Taimela S, Telama R, Räsänen L, 49. Bruvold, WH. A meta-analysis of adolescent smoking-prevention
Viikari JSA. Effects of persistent physical activity and inactivity programs. Am J Public Health 1993;83:872–80.
on coronary risk factors in children and young adults. Am J 50. Perry CL, Klepp K-I, Shulz JM. Primary prevention of cardiovas-
Epidemiol 1994;140:195–205. cular disease: community wide strategies for youth. J Consult
33. Perry CL, Klepp K-I, Sillers C. Community-wide strategies for Clin Psychol 1988;56:358–64.