Citation
Nutrition education and scratch cooking changes in schools

Material Information

Title:
Nutrition education and scratch cooking changes in schools a mixed methods study of interventions in Aurora Public Schools
Creator:
Guenther, Debra Carol
Place of Publication:
Denver, CO
Publisher:
University of Colorado Denver
Publication Date:
Language:
English
Physical Description:
1 electronic file : ;

Subjects

Subjects / Keywords:
Children -- Nutrition -- Standards -- Colorado ( lcsh )
School children -- Food -- Standards -- Colorado ( lcsh )
Schools -- Health promotion services -- Colorado ( lcsh )
Genre:
bibliography ( marcgt )
theses ( marcgt )
non-fiction ( marcgt )

Notes

Review:
This paper examines the impact of a nutrition education intervention on food choice and consumption in elementary cafeterias. After a school district changed 42 lunch and breakfast entree recipes to include more whole grains, fresh produce and fresh meats, this study examined student food choice and food consumption in five schools with a nutrition education intervention and five schools without. The nutrition education intervention, focusing on a 3 level rating system to teach children about healthful food (Go, Slow, and Whoa), included an assembly, classroom instruction, point of sale cafeteria labeling, parent breakfast, and nutrition related newsletter articles. Intervention schools were matched with comparison schools using propensity score matching. STudent food choice and consumption was measured with digital plate waste methodology with a final sample size of 2223 lunch trays. Students enrolled in intervention schools showed significantly increased odds of choosing white milk (OR=2.17, p=0.05) over chocolate milk or no m ilk. None of the other healthy eating outcome models produced significant differences between intervention and comparison schools when controlling for gender, grade, and entree items. Qualitative interviews with physical education teachers responsible for program implementation revealed low does and fidelity of implementation for a variety of reasons. Additional teacher support, clarification of goals and program components, and technical assistance is advised to improve program implementation.
Thesis:
Thesis (Ph. D.)--University of Colorado Denver. Health and behavioral sciences
Bibliography:
Includes bibliographical references.
General Note:
Department of Health and Behavioral Sciences
Statement of Responsibility:
by Debra Carol Guenther.

Record Information

Source Institution:
|University of Colorado Denver
Holding Location:
|Auraria Library
Rights Management:
All applicable rights reserved by the source institution and holding location.
Resource Identifier:
862975025 ( OCLC )
ocn862975025

Downloads

This item has the following downloads:


Full Text
NUTRITION EDUCATION AND SCRATCH COOKING CHANGES IN SCHOOLS:
A MIXED METHODS STUDY OF INTERVENTIONS
IN AURORA PUBLIC SCHOOLS
by
DEBRA CAROL GUENTHER
B.A., Wake Forest University, 1996
M.S., University of Maryland, 1999
A thesis submitted to the
Faculty of the Graduate School of the
University of Colorado in partial fulfillment
of the requirements for the degree of
Doctor of Philosophy
Health and Behavioral Sciences
2013


This thesis for the Doctor of Philosophy degree by
Debra Carol Guenther
has been approved for the
Health and Behavioral Sciences Program
by
Patrick Krueger, Chair
Deborah Main, Advisor
John Brett
Bryan Wee
April 17, 2013
11


Guenther, Debra Carol (Ph.D., Health and Behavioral Sciences)
Nutrition Education and Scratch Cooking Changes in Schools: A Mixed Methods
Study of Interventions in Aurora Public Schools
Thesis directed by Professor Deborah S. Main.
ABSTRACT
This paper examines the impact of a nutrition education intervention on food
choice and consumption in elementary cafeterias. After a school district changed 42
lunch and breakfast entree recipes to include more whole grains, fresh produce and
fresh meats, this study examined student food choice and food consumption in five
schools with a nutrition education intervention and five schools without. The
nutrition education intervention, focusing on a 3-level rating system to teach
children about healthful food (Go, Slow, and Whoa), included an assembly,
classroom instruction, point of sale cafeteria labeling, parent breakfast, and
nutrition-related newsletter articles. Intervention schools were matched with
comparison schools using propensity score matching. Student food choice and
consumption was measured with digital plate waste methodology with a final
sample size of 2223 lunch trays. Students enrolled in intervention schools showed
significantly increased odds of choosing white milk (OR=2.17, p=0.05) over chocolate
milk or no milk. None of the other healthy eating outcome models produced significant
differences between intervention and comparison schools when controlling for gender,
grade, and entree items. Qualitative interviews with physical education teachers
responsible for program implementation revealed low dose and fidelity of
implementation for a variety of reasons. Additional teacher support, clarification of goals


and program components, and technical assistance is advised to improve program
implementation.
The form and content of this abstract are approved. I recommend its
publication.
Approved: Deborah S. Main
IV


TABLE OF CONTENTS
CHAPTER
I. BACKGROUND AM) SIGNIFICANCE...................................1
II. REVIEW 01 THE LITERATURE....................................22
III. METHODS....................................................49
IV. QUANTITATIVE RESULTS........................................95
V. QUALITATIVE RESULTS........................................108
VI. DISCUSSION.................................................120
REFERENCES.......................................................132
APPENDIX
A. Aurora Public School Sample Menu-December 2011.................153
B. Go Slow Whoa Rating System.....................................154
C. Matching Data..................................................160
D. Qualitative Semi-Structured Interview Guide....................162
E. Field Notes....................................................163
F. Interrater Reliability Kappa Coefficients......................164
G. AIC and BIC for Random Intercept and Random Coefficient Models.167
H. Residual Graphs for Final Models...............................168
v


CHAPTERI
BACKGROUND AND SIGNIFICANCE
Background
The Health Problem
Childhood obesity is a major public health concern in the United States as rates of
overweight and obese children continue to rise. Since 1970, the number of obese
children ages 6-11 has quadrupled and the number of obese adolescents ages 12-19 has
tripled (Hedley et al., 2004; Levi et al., 2010). This exponential increase in childhood
obesity has implications for current and future health of children as they become adults.
The rise in childhood obesity rates may also lead to a reduction in life expectancy
(Fontaine, Redden, Wang, Westfall, & Allison, 2003). Public and private schools
nationwide are targeting physical activity within school hours to increase energy
expenditure among students. This research specifically targeted dietary intake as many
schools are also changing the nutrition environments to target knowledge, attitudes, and
dietary choices among students. Given that disability and death are linked to poor dietary
habits and obesity, improving nutritional intake of children as well as adults could have a
significant impact on health and mortality in the United States.
Childhood obesity has immediate effects on child health. An estimated 61 percent
of overweight young people (ages 5-10) have at least one additional risk factor for heart
disease (i.e. high blood pressure, high cholesterol) (Freedman, Dietz, Srinivasan, &
Berenson, 1999). Over the past several decades, type II diabetes has risen significantly in
children (Vivian, 2006). Obese children and adolescents are at higher risk for a plethora
of health issues including bone and joint problems, sleep apnea, menstrual abnormalities,
1


and social/psychological problems such as stigmatization and poor self-esteem (Daniels,
2006; Daniels et al., 2005; Koplan, Liverman, & Kraak, 2005). There is also evidence
that poor nutrition and physical inactivity leading to increased obesity and health
problems are linked to poor test scores, concentration, and academic achievement
(Kristjansson, Sigfusdottir, & Allegrante, 2010; Taras & Potts-Datema, 2005).
Obese children are at increased risk of becoming obese adults (Serdula et al.,
1993; A. S. Singh, Mulder, Twisk, van Mechelen, & Chinapaw, 2008). According to a
meta-analysis conducted by Serdula and colleagues (1993), 42% to 63% of obese adults
were obese as children and higher levels of obesity (BMI) in childhood were linked to
greater risk of adult obesity. The risk of overweight children becoming overweight adults
is at least twice as high compared with normal-weight children and the persistence of
overweight is greater with increasing levels of overweight (A. S. Singh et al., 2008).
Given the relationship between childhood obesity and adult obesity, targeting childrens
dietary behavior could lead to improved dietary habits throughout life, decreased child
and adult obesity rates, and avoidance of health risk in the long run.
Body Mass Index (BMI) is a commonly used metric to classify obesity, expressed
as weight in kilograms divided by height in meters squared (kg/m ). Child obesity and
overweight criteria are based on the 2000 CDC BMI-for-age-growth charts for the U.S.,
and classify any child with a BMI at or above the 95th percentile of the sex-specific BMI
growth charts as obese, in the 85-95th percentile as overweight. This is a change from
previous years, where children at the 95th percentile and above were overweight, and kids
in the 85th-95th percentiles were at risk for overweight (Barlow & Expert, 2007; Krebs
et al., 2007). In 2007-2008, 16.9% of children and adolescents in the United States were
2


at or above the 95th percentile and 31.7% were at or above the 85th percentile of BMI for
age. Narrowing the category down to 6- through 11-year olds reveals that 19.6% are
above the 95th percentile of BMI for age (Ogden, Carroll, Curtin, Lamb, & Flegal, 2010).
Childhood overweight and obesity rates in Colorado are problematic. Colorado is the
23 leanest state for children, but has the second-fastest rate of childhood obesity
increase in the nation, according to the 2010 National Survey of Childrens Health (Levi
et al., 2010). In 2009, the prevalence of overweight (85th to 94th percentile for BMI)
among children ages 2-14 years was 13.2 percent and obesity prevalence (95th percentile
and above) was 13.3 percent (COPAN, 2009), higher than the Healthy People 2010
objective of five percent for obesity.
Well-documented research highlights significant racial/ethnic and
socioeconomic disparities in childhood obesity in the United States (Ogden et al., 2010;
G. K. Singh, Kogan, Van Dyck, & Siahpush, 2008; G. K. Singh, Siahpush, & Kogan,
2010; Wang & Beydoun, 2007). In 2007-2008, Hispanic boys had significantly higher
odds of having high BMI at three BMI cut-points (97th percentile, 95th percentile, and 85th
percentile) when compared to non-Hispanic White boys. Among girls, non-Hispanic
Black girls were significantly more likely than non-Hispanic White girls to have high
BMI at the three BMI cut points. Singh and colleagues (2010b) highlighted higher
prevalence of obesity in 2007 for Black, Hispanic, and American Indian children (over
23%) compared to White children (12%) and higher prevalence of overweight for
Hispanic and Black children (over 41%) compared to White children (27%). Similar
disparities exist in the state of Colorado. The obesity prevalence among non-Hispanic
White children (10.9%) was significantly lower than for Hispanic children (21.2%) and
3


non-Hispanic Black children (23.4%) between 2006-2008 (COPAN, 2009). The
racial/ethnic breakdown in the current study population reveals more non-white students
than white students. Averaged across 33 elementary schools participating in this
research, 58% of the students were Hispanic, 17% were Black, 17% were White, 1%
were Native American, 4% were Asian, and 3% identified as two or more
races/ethnicities. The nutrition intervention evaluated in this study targeted a district with
high racial/ethnic diversity, and was therefore poised to make a significant contribution to
the dietary habits and obesity rates of a racially diverse population at high risk for
childhood obesity.
The medical costs of obesity rise dramatically each year in the United States. The
estimated annual medical costs of obesity increased from $74 billion in 1998 to $147
billion in 2008 (Finkelstein, Trogdon, Cohen, & Dietz, 2009). After combining costs for
increased medical care, loss of worker productivity due to death, loss of productivity due
to illness and disability of active workers, and loss of productivity due to total disability,
the overall annual costs associated with obesity are likely much higher than those
estimated costs. According to analyses by Finkelstein et al. (2008), across all payers
(normal payer, Medicare, and Medicaid), obese patients had per capita medical spending
that was $1,429 (42%) greater than spending for normal weight people in 2006.
Medicaid and Medicare paid an estimated 42 percent of obesity-related medical
expenditures in 2006. Childhood obesity, in 1998, was related to excess medical
expenditures totaling $124 million (Johnson, Mclnnes, & Shinogle, 2006). These
estimates of childhood and adult obesity expenditures are likely underestimates of the
actual economic costs if other downstream costs are considered.
4


Social determinants of health are important predictors of childhood obesity. The
social determinants of health are the circumstances in which people are born, grow up,
live, work and age. These conditions are influenced by the distribution of money,
resources, and power at the global, national and local levels (World Health Organization,
2012). The inequitable distribution of resources at the local level can be further divided
into opportunities and resources available within neighborhoods and schools. Social
determinants affect childhood obesity risk through access to fresh fruit and vegetables,
safe environments, and income levels that afford opportunities for physical activity,
healthcare and education. Children participating in the National School Lunch Program
(NSLP) or School Breakfast Program (SBP), children in lower SES households, and
children attending public schools all have a higher risk of being overweight or obese (Li
& Hooker, 2010), thus linking social determinants of health, and specifically poverty, to
childhood obesity. Many of the students in our nation and in the specific population
targeted in this study fit the profile for social determinants of health that make healthy
eating and active living difficult rather than easy.
Physical inactivity, dietary intake, genetic predisposition, illness, social
determinants, and other factors contribute to childhood obesity. Given the myriad of
negative social, economic, and health outcomes associated with childhood obesity, public
health officials are searching for new, innovative and effective interventions. The NSLP
and SBP are two national programs aimed at providing healthy food to children from
low-income families during the school day. Social determinants of health that put
children at higher risk of obesity could be offset by positive environmental and nutrition
interventions offered within schools. Given that children spend seven to eight hours each
5


day in school, school-based interventions and policies could positively affect the nutrition
environment for half of a childs waking hours. Even though dietary intake is determined
in part by child and family factors such as food preferences, monetary resources, choices,
and eating behaviors, it is also a simple function of what foods are accessible to children
and what knowledge they have about those foods; therefore school-based childhood
nutrition and nutrition education are two logical intervention targets for improving
dietary intake and preventing obesity. This study will examine two specific childhood
nutrition interventions delivered in elementary schools: School Lunches and the Go,
Slow, Whoa nutrition education program.
The National School Lunch Program (NSLP)
In 1932, the federal government began providing aid for school lunch programs
from agencies such as the Federal Emergency Relief Administration, the Reconstruction
Finance Corporation, and the Civil Works Administration. Soon thereafter, funding
increased for labor in school lunchrooms and donations of surplus farm commodities to
school lunch programs. The National School Lunch Program (NSLP) became permanent
with the passage of the National School Lunch Act in 1946 (Congress, 1946). The act
stipulated a formula for giving cash from the federal government to the states based on
per capita income and population. This cash was, and still is, doled out across the state
for school lunch programs, as long as requirements for school lunch contents (fat,
saturated fat, total calories, iron, calcium, etc.) were followed (Hinrichs, 2010). The
NSLP, subsidized by the federal government, helps to provide healthy food and adequate
calories to students during the school day. Section two of the 1946 act (Congress, 1946)
reads:
6


It is hereby declared to be the policy of Congress, as a measure of national
security, to safeguard the health and well-being of the Nations children and to
encourage the domestic consumption of nutritious agricultural commodities and
other food, by assisting the States, through grants-in-aid and other means; in
providing an adequate supply of foods and other facilities for the establishment,
maintenance, and expansion of non-profit school-lunch programs.
In the 1990s, policymakers discovered that many school lunches failed to meet
nutrition requirements. In 1995, they passed the School Meals Initiative for Healthy
Children which required school meals to meet one-third of the child recommended daily
allowance (RDA) of calories, protein, calcium, iron, and vitamins A and C and limit fat
and saturated fat content (to 30% and 10% respectively) of total calories in order to
receive reimbursement (Schanzenbach, 2009).
According to the USDA Food and Nutrition Service (U.S. Department of Agriculture,
October, 2011):
1. The NSLP is a federally assisted meal program operating in over 101,000 public
and non-profit private schools and residential childcare institutions.
2. The Food and Nutrition Service administers the program at the Federal level. At
the State level, the National School Lunch Program is usually administered by
State education agencies.
3. School lunches must meet Federal nutrition requirements but decisions about
specific ingredients/foods to serve, recipes, and food preparation methods will be
made by local school food authorities.
4. As of July 1, 2011, families are eligible for reduced price school meals if their
income is between 130% and 185% of the federal poverty level, for which
students can be charged no more than 40 cents. Families are eligible for free
meals at 130% of the federal poverty level. To put these guidelines into
7


perspective, a family of four in the 48 contiguous states or the District of
Columbia would be eligible for reduced price school meals with an annual family
income of $41,348 and free meals with an annual family income of $29,055.
5. Even full-priced meals are subsidized to some extent. Local school food
authorities set prices for full-price meals, but must provide meals without profit.
6. Schools with less than 60% student eligibility for free and reduced price lunch
receive cash reimbursement at the following rates: $2.77 for free lunches, $2.37
for reduced-price lunches, and $0.26 for paid lunches. Schools with more than
60% student eligibility for free and reduced price lunch receive the following
reimbursements: $2.78 for free lunches, $2.39 for reduced-price lunches, and
$0.28 for paid lunches.
7. By the end of the first year of the NSLP, 1946-1947, about 7.1 million children
were participating in the program. By 1970, 22 million children were
participating. In FY 2010, more than 31 million children participated in the
NSLP. Since the program officially began in 1946, over 219 billion lunches have
been served.
8. The cost of the NSLP totaled $10.8 billion in FY 2010.
Food consumed by children during the school day has a significant impact on
both health and educational outcomes. Previous research shows that the NSLP improves
educational attainment of students eating school lunch (Hinrichs, 2010), and increases
vitamin and mineral consumption as compared to non-participants in NSLP (Gleason &
Suitor, 2001; Gleason & Suitor, 2003). These positive effects of NSLP are promising
given that the NSLP serves approximately 60 percent of the total U.S. student population
8


daily (Schanzenbach, 2009). In 2010, 31.5 million kids participated in the NSLP: 17.6
million free lunches (56 percent of participants), 3.0 million reduced-price lunches (10
percent of participants), and 11.1 million full-price lunches (34 percent of participants)
(U.S. Department of Agriculture, 2011).
Funds supporting the NSLP have a strong impact on what is offered in schools.
Schools must follow the nutritional guidelines of the NSLP in order to receive continued
funds for school lunch programs. However, decisions about the specific recipes, food
items, and menu choices available in each school district are made at the local level. The
appeal and likability of school lunch items are functions of ingredients and taste, not just
of the nutritional breakdown. A spinach and egg quiche with low-fat cheese and whole
wheat crust could have the same nutritional breakdown as a bean burrito with fresh
tomatoes, brown rice, low-fat cheese, and a whole wheat tortilla. However, differences in
culture, race/ethnicity, past experience with similar foods, and taste/texture might make
one recipe much more acceptable by the student population. Therefore, the nutritional
guidelines of the NSLP combined with the knowledge and creativity of the local nutrition
services personnel will determine the taste, likability, and consumption of school lunches,
and their impact on child health and obesity.
According to the School Nutrition Dietary Assessment Study (SNDA-III), over
85% of schools in the United States prepared lunches that met the standards set forth by
the NSLP for protein, Vitamin A, vitamin C, calcium, and iron. However, fewer than
one-third of public schools served school lunches with less than 30 percent of calories
from fat or less than 10 percent of calories from saturated fat (Crepinsek, Gordon,
McKinney, Condon, & Wilson, 2009). Even though schools might meet national
9


standards for foods served, the actual foods and quantities children consume from what is
served is most important. However, this is seldom documented due to required time and
costs of plate waste studies to determine consumption. The amount of food consumed at
lunch could be influenced by the scheduling of recess before or after lunch (Bergman,
Buergel, Englund, Clem et al., 2004), the length of the lunch period, peer food
consumption and role modeling, and food preferences (Bergman, Buergel, Englund, &
Femrite, 2004a). The estimated amount of calories wasted from school lunches in the
NSLP varies widely from 12% (Devaney, Gordon, & Burghardt, 1995) to 40%
(Bergman, Buergel, Englund, & Femrite, 2004b). This study examined lunches in a local
school district that meet all criteria set by the NSLP by using plate waste methods to
determine foods and food quantities consumed during school lunch. By utilizing
quantitative plate waste methods to identify the percentage of food thrown away from
school lunches, the current study fills an important gap in understanding eating habits by
using quantitative methods to uncover childrens eating habits and preferences.
School Nutrition and Nutrition Education Advocacy and Intervention
In addition to the NSLP, other local and national efforts are tackling the obesity
problem. The U.S. Child Nutrition and WIC Reauthorization Act of 2004 ("Child
Nutrition and WIC Reauthorization Act," 2004) required that school districts in the U.S.
implement a Local Wellness Policy by July 1, 2006. The goals of these wellness policies
were to improve students nutritious food consumption and physical activity expenditure
in order to reduce the U.S. obesity epidemic. This reauthorization act is renewed every 5
years to reflect changing demands for all school meal programs, Child and Adult Care
Food Programs, and WIC.
10


On December 13, 2010, President Obama signed the Healthy, Hunger-Free Kids
Act reauthorizing the Child Nutrition Act governing the National School Lunch Program
("Healthy, Hunger-Free Kids Act of 2010," 2010). This bill includes health-promoting
school food policies such as:
1. Increasing school meal reimbursement for schools by six cents per meal;
2. Setting improved nutrition standards for school meals;
3. Setting policies for vending machines, school stores, etc.;
4. Simplifying the process for gaining access to free meals;
5. Piloting expansion of Farm to School programs as well as organic foods.
Based on their potential to impact childhood obesity and health, nutrition and
nutrition education policies and interventions are an important part of many non-profit
organizations and federal programs. The American Dietetic Association, the Society for
Nutrition Education, and the American School Food Service Association are heavily
involved in nutrition programs and interventions in schools and recommend that
comprehensive nutrition services must be provided to all of the nations preschool
through grade twelve students. These nutrition services shall be integrated with a
coordinated, comprehensive school health program and implemented through a school
nutrition policy (Briggs et al., 2003). The School Nutrition Association established the
School Nutrition Specialist (SNS) Credentialing Program in 1997 to help school nutrition
professionals enhance their skills and elevate professional standards. This national
organization also developed a tool, Keys to Excellence, to help dietetics professionals use
a best practices framework for continuous program review, evaluation, and improvement
(School Nutrition Association, n.d.). The Centers for Disease Control and Prevention and
11


the National Association of State Boards of Education published a document titled The
Role of Schools in Preventing Childhood Obesity (Wechsler, McKenna, Lee, & Dietz,
2004). The Farm to School national program helps schools procure fresh produce and
use farm products in school lunches {Benefits of farm-to-schoolprojects healthy eating
and physical activity for school children: field hearing before the Committee on
Agriculture, Nutrition, and Forestry, United States Senate., 2009). Other national
organizations taking part in school nutrition and school policy interventions include: the
American Association of Family Physicians (AAFP), National Alliance for Nutrition and
Activity (NANA), School Nutrition Association (SNA), Michelle Obamas Lets Move
campaign, The Food Trust, the National School Boards Association (NSBA), the
American Heart Association (AHA), the Coordinated School Health Program, and the
Food Research and Action Center (FRAC). Within Colorado, numerous organizations
have child nutrition, nutrition education and obesity prevention as part of their policy
agendas, funding priorities, and advocacy issues. These organizations include: the
Colorado School Nutrition Association, Colorado Department of Education, Colorado
Childrens Campaign, Colorado Connections for Healthy Schools, Childrens Health
Foundation, Colorado Legacy Foundation, Colorado Association of School Executives,
Anschutz Health and Wellness Center, CU Denver, Colorado Health Foundation, Rocky
Mountain Center for Health Promotion and Education, LiveWell Colorado and more.
Multiple agencies and organizations are collaborating to strengthen policy and advocacy
initiatives aimed at school nutrition and nutrition education. This research evaluated two
such initiatives involving scratch/healthy cooking and the Go, Slow, Whoa program.
12


LiveWell Colorado Origins
LiveWell Colorado, a nonprofit organization committed to reducing obesity in
Colorado, was established as a grant-making collaborative in 2007 in partnership with the
Colorado Department of Public Health and Environment and with initial funding from
The Colorado Health Foundation, Kaiser Permanente, and the Kresge Foundation.
LiveWell Colorado became a 501(c)(3) in 2009 and focuses on policy, environmental and
personal lifestyle changes that remove barriers to and increase healthy behaviors.
LiveWell Colorado is implementing a five-year strategic plan that focuses on the
following areas:
1. Funding community coalitions throughout the state focused on healthy eating and
active living strategies.
2. Informing and advancing multi-sector policy efforts with key stakeholders at the
local, state and national levels.
3. Leading social marketing initiatives that inspire a culture shift and motivate
sustainable healthy behavior change.
This research focused on two specific programs of LiveWell Colorado: Culinary
Boot Camps and the Go, Slow, Whoa Nutrition Education Program. Both programs
are part of LiveWell Colorados LiveWell @ School initiative. Nearly 400,000 Colorado
children participate in the National School Lunch Program, with about 40 percent
qualifying for free or reduced lunch. Therefore, working towards improving the
healthfulness of school food and delivering nutrition education in schools will have a
significant impact on a large number of Colorado children. Although the overall
LiveWell Colorado program is entitled Culinary Boot Camps, this research examined
13


how only one district, Aurora Public Schools, implemented scratch cooking cafeteria
changes following participation in the Culinary Boot Camp program. Scratch cooking
changes include cooking with more whole grains, more fresh fruits and vegetables, and
less processed, frozen, and canned ingredients. Therefore, this research did not
necessarily evaluate the impact of the Culinary Boot Camp program, but instead
evaluated the impact of specific scratch cooking cafeteria changes inspired in this one
district by the Culinary Boot Camp program. Thus, this research refers to these district-
level changes as Scratch Cooking Cafeteria Changes. Go, Slow, Whoa is a nutrition
education program adopted and modified from national health education curriculum for
use in Aurora Public Schools.
Culinary Boot Camp Inspired Scratch Cooking Cafeteria Changes:
As of the beginning of this research (August 2011), 63 Colorado school districts
and 280 school food service managers and workers had participated in 5-day culinary
boot camps. In Aurora Public Schools, 53 school cafeteria managers and central office
leadership staff attended the LiveWell Colorado and Colorado Health Foundation
sponsored Cook for America Culinary School Food Boot Camp, which represents 16%
of the total Nutrition Services Staff (kitchen managers, kitchen staff, and central office
leadership and staff). This Boot Camp is taught by experienced chefs and promotes the
preparation of fresh food and scratch cooking in school cafeterias at the same or lower
costs than schools currently pay for processed, ready to heat and serve foods. The goal of
Scratch Cooking Cafeteria Changes in Aurora Public Schools is to increase healthful
nutrients (i.e. fiber, calcium, vitamins A and C) and decrease unhealthful nutrients (i.e.
total fat, saturated fat, sodium, and cholesterol) in school lunch items while maintaining
14


or increasing attractiveness, taste, and desirability. The Scratch Cooking Cafeteria
Changes reflect multiple contributing strategies: using more healthful ingredients (i.e.
whole grains, fresh fruits and vegetables, non-processed ingredients), creating and using
new recipes with those healthful ingredients, and utilizing new techniques and kitchen
equipment to prepare the recipes.
Go, Slow, Whoa
Go, Slow, Whoa (GSW) is a nutrition education program promoted by the
LiveWell Colorado LiveWell @ School initiative. GSW aims to improve student and
school community awareness of healthful foods, increase students healthful food choices
and ultimately impact childhood obesity rates. GSW teaches children and their families
about healthful foods using a green light (Go), yellow light (Slow), red light (Whoa)
visual aid to symbolize foods that should be eaten frequently, sometimes, and seldom.
LiveWell Colorado has developed a business plan to scale up this intervention and
offer it to additional school districts throughout Colorado. Given the paucity of data on
program implementation and effectiveness, this research will make a significant
contribution to program justification and program improvement.
Why Aurora?
Data collected in this school district demonstrates the need for addressing the
obesity and BMI issues of elementary students. Aurora Public Schools (APS), the target
school district for this study, includes two different counties: Arapahoe and Adams.
Research shows that across the two counties, an average of 13.6% of children are obese,
and an additional 12.4% are overweight (COPAN, 2009). These rates are significantly
higher than the Healthy People 2010 goal of five percent. In APS specifically, childhood
15


rates are much higher than the county averages. APS contains a very ethnically diverse
and socioeconomically challenged student population, increasing the risk for childhood
obesity. Student data collected in APS physical education classes show increasing BMIs
as students advance by grade level. This points to the critical need to prevent further
weight gain at the elementary school level. Given that overweight and obesity rates in
APS are generally higher than the national and Colorado rates, this school district is an
excellent target for healthy eating education, promotion, and programming (C. Fenton,
personal communication, February 7, 2012). Many chronic diseases are related to
individual and social patterns of behavior (food consumption included), thus
interventions, policies, and programming targeting individual and community behavior
could result in risk reduction for obesity and related diseases. This district could benefit
from successful interventions to change the nutrition environment and improve student
nutrition knowledge.
Significance
Scratch Cooking Cafeteria Changes and the Go, Slow, Whoa Program have
potential to improve the nutritional intake of children in Aurora Public Schools.
Changing childrens food options to make them more healthful and improving childrens
knowledge and attitudes regarding healthy food, together, may impact nutrient
consumption. Determining whether the joint effects of changing the food environment
and providing nutrition education are greater than changing the food environment alone is
an important question. Answering this question is key for educators and policy makers
alike to implement the most effective and efficient interventions.
16


Objective, Rationale, and Purpose
This research, Scratch Cooking Initiatives and Nutrition Education in Public
Elementary Schools, evaluated Scratch Cooking Cafeteria Changes and nutrition
education implementation in Aurora Public Schools to determine their effect on cafeteria
consumption of healthy foods. Students participating in the GSW program have
opportunities to gain knowledge and skills to improve their choices and consumption of
healthy cafeteria food. Aurora Public Schools implemented many scratch cooking
changes beginning in August 2011, and this research assessed the consumption of menu
items and ascertained differences between students in schools with GSW and schools
without GSW. Results of this research inform program justification and program
improvement for both the GSW program and Scratch Cooking Cafeteria Changes.
Overview of Theoretical and Conceptual Framework
The rationale for and potential impact of school interventions addressed in this
research are informed by Social Cognitive Theory (SCT) and Social Ecological Models
(SEM) (Bandura, 1986; McLeroy, Bibeau, Steckler, & Glanz, 1988). Historically, many
nutrition education interventions have utilized Social Cognitive Theories to guide
programming targeting individual, behavioral, and environmental factors (Bandura,
2004). Although changing knowledge is an important part of nutrition education, SCT
further utilizes skill building, role modeling, self-efficacy, and environmental change to
reach behavior change goals. The Go, Slow, Whoa program utilizes these SCT
constructs to impact student nutrition knowledge and behavior.
Social Ecological Models offer an important theoretical framework for this
research by acknowledging factors outside an individual that may influence behavior.
17


Ecological models recognize that behavior and health outcomes are the result of more
than just individual motives, and explore the interaction between an individual and the
environment (Contento, 2011). The Scratch Cooking Cafeteria Changes seek to improve
the food environment that students encounter on a daily basis. The Go, Slow, Whoa
program works within the school nutrition environment to discuss, promote, and
reinforce healthy selections from the cafeteria menu choices.
Gaps Filled
This research addressed several gaps in the current academic literature related to
scratch cooking initiatives and nutrition education evaluation. Scratch Cooking school
food initiatives have not been systematically evaluated in a K-12 public school context,
although there is a related body of literature that examines taste preferences and
consumption of fruits, vegetables, and whole grains. This research examined student
responses to large-scale menu changes offering more fresh, whole grain, unrefined menu
items to students in the NSLP. Plate waste studies following healthful food additions and
changes are very uncommon due to the expense and time required for plate waste studies.
Nutrition education programs are oftentimes evaluated with pre-post knowledge
questionnaires, 24-hour recall surveys, food journals, or food preference surveys. This
research used a robust and reliable method of determining what kids actually eat by
collecting data on their food consumption in the cafeteria. Digital photography plate
waste protocols are in their infancy, yet show promise in collecting important and valid
data about food consumption. This research further examined this method as a tool for
effectively measuring school food consumption.
18


Lastly, unlike many studies of food consumption using quantitative data, this
research also included a qualitative component to explore teacher perceptions of the
intervention and program implementation and fidelity issues. This study utilized
qualitative methods to understand the impact and significance of various components of
the Go, Slow, Whoa program and to help explain the quantitative plate waste findings.
Research Question
How influential is the Go, Slow, Whoa nutrition education program combined
with Scratch Cooking Cafeteria Changes over and above the impact of Scratch Cooking
Cafeteria Changes alone on elementary students choices and dietary intake related to
healthy foods?
Specific Aims and Hypotheses
Aim 1,
Determine the relative impact of GSW and Scratch Cooking Cafeteria Changes
over and above the impact of Scratch Cooking Cafeteria Changes alone on students
choices and consumption of school lunch.
Hypothesis la.
Students receiving the GSW program will choose more healthy foods at school
than non-GSW students.
Hypothesis lb.
Students receiving the GSW program will consume more of the healthy foods on
their school lunch tray than non-GSW students.
19


Hypothesis lc.
Students receiving the GSW program will choose and consume more overall
fruits and vegetables, white milk, and fiber, and less fat and saturated fat than non-GSW
students.
Aim 2,
Determine the relative dose, quality and adherence/fidelity of GSW
implementation in each GSW school and relate implementation to observed outcomes.
Hypothesis 2a.
Students in schools with higher dose and better quality and adherence/fidelity of
GSW programming will show more positive results of GSW programming (Hypotheses
la, lb, lc).
In order to address these Aims and Hypotheses, quantitative digital plate waste
photography and qualitative teacher interviews were conducted in 10 Aurora public
elementary schools. Five schools were the GSW schools for the 2011-2012 school year,
and five schools without the GSW program were matched as control schools.
To address Aim 1, Quantitative data were collected from 1st, 3rd, and 5th grade students in
each school. Plate waste data were collected on the same three consecutive days in each
school, therefore allowing comparison of plate waste for the same menu in each school
on each day.
To address Aim 2, Qualitative physical education teacher interviews were conducted in 7
schools (three schools with GSW implementation in 2010-2011, 1 school with
implementation in 2011-2012, and three schools with implementation in 2012-2013).
20


Informal observations and conversations within schools also contributed important
qualitative data to this study aim.
21


CHAPTER II
REVIEW OF THE LITERATURE
Children in the United States are eating excess calories from unhealthy food. The
predisposition for children to like high-fat and high-sugar foods, wide availability of
unhealthy food, media persuasion, and social environments all encourage poor dietary
intake (Birch, 1999; Roblin, 2007). Nutrition education in schools is one way to
influence a large percentage of children in the United States. Nutrition education, as
discussed in this research and in most academic literature, encompasses much more than
simply lecturing about nutrients in foods and their relation to health. According to
Contento (2008) nutrition education needs to be a much more comprehensive enterprise
than information dissemination in order to be effective. Nutrition education needs to
address food preferences and sensory-affective factors; person-related factors such as
perceptions, beliefs, attitudes, meanings, and social norms; and environmental factors
(p. 176). This chapter addresses research and theory related to different nutrition
education interventions targeting food preferences and food habits, knowledge and
attitudes, social norms, parental involvement, classroom instruction, and school-level
environmental changes.
Theoretical Basis for Nutrition Education
Social Cognitive Theory
Social Cognitive Theory (SCT) is the most widely used theory for designing
nutrition education and health promotion programming (Contento, 2011; Reynolds &
Spruijt-Metz, 2006). SCT describes the interaction of dynamic and reciprocal factors
(personal, behavioral, and environmental) that influence health behavior (Bandura, 1986).
22


Personal factors involve feelings and thoughts (outcome expectations). Behavioral
factors include food-, nutrition-, and health-related knowledge and skills (behavioral
capability) and skills in handling personal behaviors (self-regulation skills and self-
efficacy). Environmental factors include those external to individuals, including physical
and social environments (Contento, 2011). The interaction of outcome expectations,
individual agency (including the skills to complete behaviors and the self-efficacy to
initiate behaviors), and environmental support for behavior change are key factors of
SCT that facilitate behavior change in the GSW program.
Personal factors influencing behavior include our individual thoughts and beliefs
about potential actions and our personal capabilities. Human beings of all ages are self-
reflective beings; self-reflection and evaluation influence personal behavior and future
decisions. Among these self-reflective personal behaviors are outcome expectations.
Outcome expectations can be related to physical outcomes, social outcomes, and self-
evaluative outcomes (Bandura, 2000). Physical outcome expectations, in the health
domain, include perceived risk of disease from not engaging in healthy behaviors or
engaging in unhealthy behaviors. Positive physical outcomes related to food also include
pleasant sensory experiences such as the taste or smell of something sweet (Bandura,
2000, 2004). Taste, as an outcome expectation, is a very powerful predictor of nutrition
behavior in adults (Anderson, Winett, Wojcik, Winett, & Bowden, 2001), but has not
been well-studied in children. Exploring and influencing the physical outcome
expectations of children eating school lunch is an important aspect of nutrition education
in APS.
23


Social outcomes include the social consequences of a behavior; for example
positive peer reactions to a behavior, such as soda drinking among teenagers or
consumption of Hot Cheetos by elementary school peers. The inclusion of curriculum
related to positive social outcome expectations of eating healthful food will be explored
with proposed qualitative methods in the current study. Self-evaluative outcomes are the
beliefs one has about his or her personal actions or behaviors that result in a sense of self-
worth and avoidance of behaviors that lead to dissatisfaction. Self-evaluative outcomes
might include satisfaction from reducing saturated fat or soda consumption. In the EatFit
program, goal setting, working towards goals, and satisfaction with achieving goals
helped improve childrens eating behavior (Horowitz, Shilts, & Townsend, 2004). In
adults, self-satisfaction for personal accomplishments and health behaviors is one of the
most powerful regulators of behavior (Bandura, 2000), however, self-satisfaction
research with children and nutrition is scarce. The GSW program includes self-
satisfaction learning objectives as well as learning objectives related to pleasant physical
outcomes and social outcomes.
Behavioral capability, a personal factor construct in SCT, is related to health
knowledge, and the cognitive, affective, and behavioral skills needed to carry out
healthful behavior (Bandura, 2000, 2004). Nutrition knowledge, a target of many
nutrition interventions, can pertain to facts, procedures, or behavioral skills. The GSW
program specifically targets facts about nutrition to help students understand how to
choose and eat more healthful food. Factual knowledge targeted in the GSW lessons
includes information about food items, specific nutrients, the food guide pyramid and
recommended daily allowances of nutrients. Although some research reveals that
24


increasing knowledge about fruits and vegetables did not have a significant influence on
fruit and vegetable consumption in children (Reynolds, Hinton, Shewchuk, & Hickey,
1999), other researchers show knowledge acquisition to be a precursor to behavioral
capability and behavior change related to healthful eating (DeVault et al., 2009; Sauberli,
Lee, Contento, Koch, & Calabrese Barton, 2008; Stevens et al., 1995; Van Horn,
Obarzanek, Friedman, Gernhofer, & Barton, 2005). The GSW program is firmly rooted
in knowledge acquisition, and the proposed plate waste methods will help determine
whether the program (and knowledge acquisition) positively impacts dietary intake.
Observational learning and modeling, also referred to as vicarious experiences,
are additional tools for increasing behavioral skills, confidence, and capabilities in
children (Bandura, 1998). In elementary schools, peers, teachers, cafeteria workers, and
administrators are constantly modeling behaviors related to food and nutrition. Peer,
adult, and parent modeling have been strategic components of many successful school-
based nutrition education interventions (Edmundson et al., 1996; Hoelscher et al., 2004;
Horowitz et al., 2004; Luepker et al., 1996). The GSW program encourages school
personnel to foster positive role modeling, an important component related to adoption
and continuation of healthful eating and behavior. Qualitative interviews with physical
education teachers explored the presence and magnitude of this modeling in selected
schools.
Self-efficacy is another major motivator of action, because people must believe
they can do something to actually want to initiate new behavior. Self-efficacy is crucial
to overcome impediments or barriers to adopting and maintaining healthful behaviors
(Bandura, 1989, 2000). The higher the level of perceived self-efficacy, the harder and
25


longer people will be willing to persist in a new, healthy behavior or lifestyle change.
Self-efficacy is especially important in beginning, modifying, and maintaining complex
behaviors such as healthful eating (Contento, 2011). The GSW program and specific
lessons target self-efficacy by helping students feel confident that they have the
knowledge to select healthy foods, the skills to prepare healthy snacks, and the ability to
convince parents to supply healthy food at home. All of these self-efficacy areas have the
potential to affect student dietary intake. One nutrition education intervention for
second- and third-grade students in Alabama included components to increase nutrition
knowledge and student self-efficacy skills to select healthy foods both at school and at
home, and reported significant knowledge gains and improved dietary behavior (Powers,
Struempler, Guarino, & Parmer, 2005). The GSW program has not been evaluated for
its impact on self-efficacy. However, GSW indirectly targets self-efficacy through role
modeling from teachers eating school lunch in the cafeteria and healthy food messages
posted throughout the school and cafeteria. Most nutrition interventions include multiple
components; thus it is hard to determine the most powerful components influencing
nutrition knowledge and dietary behavior. However, self-efficacy has been a component
of many promising and successful nutrition education interventions with children
(Edmundson et al., 1996; Gortmaker et al., 1999; Hoelscher et al., 2004; Horowitz et al.,
2004; Sauberli et al., 2008). Self-efficacy is rooted in the concept of agency, or an
individuals influence over his or her own behaviors and surroundings. A strong sense of
personal agency is not just confidence in knowledge and skills to perform some action,
but is also the ability to regulate thoughts, motivations, feelings, and behaviors or to
change environmental conditions to achieve desired results (Bandura, 1989, 1998, 2000).
26


Qualitative interviews with physical education teachers examined their perspectives on
students personal agency and self-efficacy related to acquiring, choosing, and eating
healthy foods and their inclusion of learning objectives and curriculum to target these
concepts.
The final SCT construct associated with this research is environmental influence
and how the environment interacts with personal and behavioral factors to impact
behavior change (Bandura, 1998). The nutrition environment is the specific environment
related to this research; students participating in the NSLP are surrounded by food
choices and a nutrition environment chosen in part by the federal government and in part
by the local school district. Scratch Cooking Cafeteria Changes to decrease fat, sugar,
and sodium have altered the nutrition environment in all Aurora Public Schools, but
student reactions to these changes have not been assessed. Students in GSW schools,
who essentially receive two interventions (Scratch Cooking Cafeteria Changes and a
nutrition education intervention), may have significantly more improved dietary intake as
a result of the GSW program than students in comparison schools receiving only Scratch
Cooking Cafeteria Changes. The concepts surrounding environmental changes will be
discussed more fully in the next section: Social Ecological Models.
Interventions based on SCT constructs have proven effective for improving
dietary behavior and reducing obesity levels in children. Successful interventions are
often comprised of several components, each of which might be based on different SCT
constructs and involve different levels of intervention intensity and contact hours, and
thus it is difficult to outline a magic formula for necessary components. The GSW
program and Scratch Cooking Cafeteria Changes utilize SCT personal and behavioral
27


factors targeting thoughts, knowledge, behavioral capability and self-efficacy, in addition
to environmental factors to influence behavior change and healthful eating. This research
addresses important gaps by utilizing qualitative methods to uncover the presence of
learning objectives and curriculum related to these SCT constructs. Comparing GSW
school children to non-GSW school children helped uncover the impact of GSW
participation.
Social Ecological Models
Altering a persons social and physical environment to improve health and health
behavior fits within the framework of Social Ecological Models. Social Ecological
strategies underlie the school districts Scratch Cooking Cafeteria Changes, nutrition
environment modifications, and desire to strengthen the nutrition environment in schools.
Factors external to an individual can have a huge impact on health and health behavior.
Environmental intervention strategies target physical surroundings, social climates,
information accessibility, organizational systems, and policy to provide support for
improving health, and in this case, healthy eating and active living (Contento, 2011).
Social ecological models of health promotion and behavior change address several
ecologies or contexts in which people live at once. The main levels of influence are
intrapersonal factors, interpersonal processes and primary groups,
institutional/organizational factors, community factors, and public policy and legislation
(DiClemente, Crosby, & Kegler, 2002; Green & Kreuter, 2005; McLeroy et al., 1988). In
this research, children are influenced by many of these levels including their own
intrapersonal factors (likes and dislikes, self-efficacy, knowledge), peer perceptions of
healthy eating, and school culture and policies. Bronfenbrenners Ecological Systems
28


Theory outlines the impact of environment on children and adolescents (Bronfenbrenner,
1989). Ecological Systems Theory posits that children are affected by multiple levels of
influence. The microsystem includes direct influences such as peers, family, school, and
neighborhood. The mesosystem connects structures of the microsystem. The exosystem
includes the larger, indirect social system and the macrosystem includes values, customs,
and laws. The chronosystem incorporates the relationship between external events, the
timing of those events, and the physical maturation process. The bi-directionality of each
of these levels of influence is important in Ecological Systems Theory. Effective
interventions are often designed for multiple levels of influence.
The following diagram outlines the use of Social Ecological Models for changing dietary
behavior (McLeroy et al., 1988):
Social Structure/Public Policy
Community
Institutional/Organizational
Interpersonal
Individual
Figure 1: Theoretical Framework: Social Ecological Models
29


Table 1: Theoretical Framework: Social Ecological Models
Social Structure/Public Policy Local State and Federal policies and laws that regulate or support healthy actions.
Community Social networks, norms, or standards (e.g. public agenda, media agenda, partnerships).
Institutional/Organizational Rules, regulations, policies and informal structures (worksites, schools, religious groups).
Interpersonal Interpersonal processes and primary groups (family, peers, social networks, associations) that provide social identity and role definition.
Individual Individual characteristics that influence behaviors such as knowledge, attitudes, beliefs, and personality traits.
30


A similar depiction of Social Ecological Models related to nutrition education,
food choice and dietary practices is offered below (Contento, 2011).
Adapted from Contento. 2011. g>. 53
Figure 2: Social Ecological Model for Nutrition Education
At the basic level of influence, the individual level, social psychological theories
of behavior change can guide interventions directed at individuals. Leading theories for
nutrition education include SCT, the Health Belief Model, the Transtheoretical Model of
Change, and the Theory of Planned Behavior. As already discussed, several
intrapersonal constructs of SCT (beliefs, knowledge, attitudes, behavioral capability, and
self-efficacy) are the basis and focus of GSW programming and were evaluated with this
research. GSW programming also targets other levels of social ecological models. At
the interpersonal level, peer and adult role modeling and parental involvement to
influence dietary behavior are emphasized with PE teacher lessons, involvement of
31


teachers, and parent information sessions and newsletters. Institutional, organizational,
and community settings, and specifically the elementary schools in Aurora, offer
opportunities to involve decision makers and policy makers in obesity prevention
strategies to change relevant food and activity environments. In APS, the food service
directors, cafeteria managers, administrators, and school board members are influential in
improving the nutrition environment and supporting the GSW program. Developing,
supporting, and introducing 41 new recipes and new scratch cooking techniques prior to
the 2011-2012 school year involved collaboration from multiple organizational
constituents. Beyond the local school and community level, LiveWell Colorado has
offered the Culinary Boot Camp program to districts throughout the state. LiveWell
Colorado also hopes to expand the GSW program to over 40 districts in Colorado and
develop the program as a national model for successful nutrition education programming.
These efforts will undoubtedly educate new audiences at the local, state, and national
level regarding the importance of obesity prevention strategies and helpful environmental
change efforts. Each of the levels of influence within social ecological models holds
important promise for improving dietary habits in children (Contento, 2011), and this
research was designed to evaluate two programs that address multiple levels of influence.
The school environment plays an important role in what foods kids have access to
during the school day. Children could eat 35-40% of their calories at school based on the
school meals program, a la carte offerings, vending machines, classroom snacks, and
school stores (Briggs et al., 2003). Changing that environment to ensure the most
healthful food offerings to students, and reducing availability of unhealthy food should
affect the food and nutrients children consume at school. Merely increasing availability
32


and accessibility of healthy foods in schools significantly increases consumption, as
evidenced by one study targeting fruits and vegetables (Hearn et al., 1998). Simply
changing what is offered shows promise in improving student dietary intake. More
healthful school lunches and lunch practices are successfully reducing the consumption
of energy-dense, low-nutrient foods among elementary school children. Successful
strategies include offering French fries less than once a week and offering fresh
vegetables and fruit daily (Briefel, Crepinsek, Cabili, Wilson, & Gleason, 2009). Schools
that offer French fries in school lunches more than once a week and schools that offer
desserts more than once a week are more likely to have higher rates of obesity among
students (Fox, Dodd, Wilson, & Gleason, 2009). Exposure to and availability of fruits
and vegetables encourage increased consumption. Students with increased access to FV
and more exposure to a variety of fruits and vegetables are more likely to consume larger
amounts of fruits and vegetables (Briggs et al., 2003; Eriksen, HaraldsdUttir, Pederson, &
Flyger, 2003). Recognizing the benefits of access to fruits and vegetables, APS has
included a salad bar at all elementary schools this year, and a goal of offering 3 fresh
fruits or vegetables each day. Quantitative methods in this study assessed the amount of
fruits and vegetables students are taking and the amount they are consuming each day.
The High Five project and the CATCH project changed the school environment to
improve intake of fruit and vegetables with fourth- and fifth-graders, respectively (Perry,
Bishop et al., 1998; Reynolds et al., 2000). Specifically, they improved variety and
attractiveness of fruits and vegetables, served extra fruit choices whenever a dessert was
offered, and utilized point-of-purchase signs to advocate for healthy eating. It is
important to note that other nutrition education components were utilized in the High
33


Five project and the CATCH project, but inclusion of fruit and vegetable attractiveness,
advertising, and availability components played some role in influencing fruit and
vegetable intake. The High Five Project reported increases in fruit and vegetable
consumption from 2.6 servings per day to 3.96 servings one year later to 3.20 servings
two years later (Reynolds et al., 2000). However, the CATCH project did not report
significant differences in FV intake between treatment and control groups after the
nutrition intervention at the end of the school year (Perry, Bishop et al., 1998). Of
particular note here is that the two programs, the High Five Project and the CATCH
project, targeted similar age groups (fourth and fifth graders), utilized food environment
changes and parental homework activities, included weekly lessons, and employed hands
on learning, modeling, reinforcement, and skills building. Surprisingly, the successful
program, the High Five Project, had far fewer contact hours with students (14, 30 minute
nutrition lessons = 7 total contact hours) than the CATCH program (47, 40 minute
nutrition lessons = 31 total contact hours) (Perry, Bishop et al., 1998; Reynolds et al.,
2000). Another study (Cafeteria Power Plus) involving younger students (first- and third-
grade students) specifically targeted environmental strategies to: (1) increase
opportunities during school lunch to eat a variety of fruits and vegetables, (2) provide
new healthful role models (i.e. cafeteria workers) who eat fruits and vegetables, and (3)
institute social support for kids to eat fruits and vegetables at lunch (Perry et al., 2004).
Students in the intervention schools (2 years of environmental cafeteria interventions)
had significantly higher fruit intake (0.79 servings per lunch for intervention schools vs.
0.63 servings for control schools), but not vegetable intake (0.52 servings per lunch for
intervention schools vs. 0.58 servings for control schools). These variable research
34


outcomes highlight that it is very difficult to compare different nutrition interventions
when the methods, curriculum, materials, contact hours, populations, and other strategies
are quite varied. Perry et al. (2004) is one of the few research teams employing process
evaluation and analysis to determine the impact of various program components. They
discovered significant impact of verbal encouragement by food service staff and the
number of fruits and vegetables students can choose, but non-significant effects of
increasing the number of fruits and vegetables on a snack cart and increasing the appeal
of fruits and vegetables. The GSW program and APS cafeteria changes employ
environmental strategies to increase availability and attractiveness of fruits and
vegetables, but they also use additional and varied methods, materials, contact hours and
strategies. The dose and fidelity of implementation at each school is an important
consideration of these environmental strategies and other GSW components and was
addressed in this research.
Although some research shows that improving attractiveness and availability of
healthy food increases consumption, other researchers report that if students are offered
healthy and unhealthy choices side by side, they oftentimes make less healthy choices.
For example, one study found that when elementary students moved from their
elementary school offering only NSLP to middle school offering a la carte and snack bar
meals, their intake of fruits, vegetables, and milk decreased, and consumption of
sweetened beverages and high-fat, high-sugar foods increased (Cullen & Zakeri, 2004).
The food environment and available foods shift as students move from elementary to
middle to high schools, thus influencing their environmental supports for healthful eating.
Because of the increased availability of unhealthy foods in middle school, it is important
35


to bolster attitudes, personal motivation, and self-efficacy for choosing healthy foods in
elementary school. Elementary children with greater ability to choose healthy foods
should be more likely to navigate the transition to a more varied nutrition environment in
middle school with more success and healthier food consumption. Although the Scratch
Cooking Cafeteria Changes in Aurora elementary schools improve the nutrition
environment, the GSW program aims to specifically strengthen intrapersonal factors for
healthy eating in addition to other interpersonal and ecological factors related to healthy
eating. These intrapersonal factors may help protect kids from unhealthy food choices
when they move to middle school or when they are offered a variety of foods in
environments outside of schools.
Availability and familiarity with fruits and vegetables may not be the only ways
of encouraging increased consumption using the school food environment. In hopes of
enticing students to eat healthful cafeteria foods, many school nutrition programs are
using marketing strategies to help compete with fast-food restaurants. For example,
schools are using food courts, portable food and salad bars, more a la carte offerings, and
local chefs to improve the visual and taste elements of food (Briggs et al., 2003).
Researchers are also examining the potential of creative signage to increase attractiveness
and consumption of fruits and vegetables (S. Smith, personal communication, September
18, 2011). APS uses two such strategies by offering salad bars in all schools and using
Go, Slow, Whoa signage to identify healthy food options.
Meals and snacks during the school day play an important role in developing
childrens eating patterns. Even though NSLP dictates nutrient requirements for school
lunches, the preparation, quality, and taste of foods offered by individual districts and
36


schools is not monitored at the national level. If schools are offering healthful foods that
do not taste good, students are likely throwing away a good portion of the food on their
lunch tray. Alternatively, if students are offered lunches that meet the NSLP guidelines
but do not include new, healthful recipes with whole and fresh food ingredients, children
are unlikely to develop new taste preferences for healthful food. These issues might
continue the problems of low-nutrient intake or neophobia (fear of new foods). Using the
cafeteria to increase childrens consumption of healthful foods is a promising nutrition
environment strategy, but one that requires constant evaluation. APS pioneered new
healthy recipes in 2011, and this study will help determine if students are reacting
positively and developing taste preferences for these healthy foods.
Research on the school food environment shows it to be a positive influence on
both low and middle socioeconomic status (SES) student food consumption (Cullen,
Watson, & Fithian, 2009). Most low SES students selected the healthful NSLP meal and
did not buy many a la carte items. These decisions, most likely due to economic
necessity, keep low-income students from purchasing additional a la carte items and
enable them to be positively affected by the nutrition environment. When snack bar
foods were limited and controlled, middle SES kids also selected the NSLP, thus pointing
to the potential influence of the NSLP on nutrient intake when competitive foods are not
available for financial or policy reasons (Cullen et al., 2009). Elementary schools in APS
do not offer competitive foods, making the nutrition environment equal for all students,
regardless of SES. Given the higher risk of obesity for lower SES students, it is
important to pay particular attention to the nutrition intervention effects for different SES
levels. The national prevalence of childhood obesity for children below the poverty
37


threshold was 27.4% in 2007, or 2.7 times higher than the prevalence for children with
family income over 400% of the poverty level (G. K. Singh et al., 2010). Given that 75%
of APS elementary school students participate in Free and Reduced Lunch programming,
it is important to be sensitive to the impact of SES on nutrition interventions. The High 5
project found that high SES students in intervention schools had the largest FV
consumption when compared to control schools, but that middle SES and low SES
students in intervention schools still consumed significantly more fruits and vegetables
than control schools (Reynolds et al., 2000). The CATCH study, upon which the GSW
program is based, compares data by study sites, ethnic groups, and gender, but does not
consistently report outcome data for SES groups (Lytle et al., 1996; Perry, Lytle et al.,
1998). Although specific SES data will not be obtained for each tray collected in
proposed plate waste methods in this study, aggregate free and reduced lunch data (a
proxy for SES) is available for each school and will be used to match GSW schools to
comparison non-GSW schools.
Social ecological models suggest multiple levels of influence on nutrition
behavior. The GSW program and Scratch Cooking Cafeteria Changes in APS offer
various intervention strategies at each level of social ecological modeling and influence.
Quantitative data collection methods will determine the effectiveness of overall
programming (all levels of influence) at improving dietary intake. Qualitative data
collection techniques will help qualify quantitative results by determining the inclusion
and dose of specific strategies at different social ecological levels.
38


Other Research on Eating Behaviors, Food Choice, and Nutrition Education
Taste Preference and Food Habits of Children
Prevailing theories of taste preference and food habits support the notion that
infants and children prefer specific tastes. Those tastes lead to food patterns and habits in
childhood that influence growth and development and future health and chronic disease
protection or risk (Nicklas, Webber, Srinivasan, & Berenson, 1993). Food habits that
persist throughout childhood and adolescence are more likely to continue into adulthood
(Kelder, Perry, Klepp, & Lytle, 1994). Therefore, encouraging the development of
preferences for healthful foods in childhood is helpful for creating healthful food habits
in adulthood. APS could make a significant impact on lifelong health by encouraging
students to develop healthful habits in elementary schools.
Neophobia (fear of new foods) is present in many children, but can be reduced by
repeated opportunities to sample new foods. Reducing neophobia requires five to fifteen
exposures to new foods (Birch, 1995). If children are continually exposed to high-sugar,
high-fat, high-salt foods, those foods will become familiar and children will continue to
crave them over more healthy, but less familiar options. Changing nutrition
environments in APS to offer more healthful food options for children of young ages will
help to make those foods more familiar and possibly more valued by students. When
presented with new, whole grain foods, elementary students had favorable responses to
the look and taste of whole grain cereal and cheese bread, but were not sure they would
want to eat it in a cafeteria lunch (Burgess-Champoux, Marquart, Vickers, & Reicks,
2006). Kids reported that new foods in the cafeteria would be accepted if they looked
good, tasted good, and were familiar. Children reported the following tips for
39


introducing new, healthful foods: If it was healthy, Id probably hide it, Id try to make
it look like what the other food was. I would just switch it and not tell everybody and let
them eat it (Burgess-Champoux et al., 2006). The challenge to school cafeteria and
nutrition services planners in Aurora Public Schools is to make new foods appear familiar
enough in appearance and taste or utilize creative marketing or incentive strategies to
encourage students to try new foods. Qualitative methods in this research assessed
teacher perceptions of whether students realize the foods they are eating are new and
healthy and whether students accept these new foods.
Teaching and Best Practices for Nutrition Education
A recent survey by Action for Healthy Kids reveals that parents think schools are
providing nutrition education to all students and they would like such education to be part
of the core curriculum two days a week (Action For Healthy Kids, 2005). In reality, kids
in the U.S. get an average of 13 hours of nutrition education per year (Celebuski & Farris,
2000; Lytle et al., 1994). Teacher preparation and allotted time for nutrition education
are important factors in the quality of information and skills students receive. APS has
made nutrition education a priority by allotting increased instruction time for nutrition
education in physical education classes.
Nutrition education varies by district, and even more so by schools and
classrooms. Classroom teachers, those most often delivering nutrition education, have
little training in the topic area because of lack of time and funding and focus on academic
core subject instruction. Only about half of elementary school teachers have formal
training in nutrition education (Celebuski & Farris, 2000). However, 88% of elementary
school teachers reported teaching lessons about nutrition to their students in the 1996-
40


1997 school year. The mean number of hours spent in a school year on nutrition
education by elementary school teachers (K-5) who taught nutrition was 13, below the
minimum of 50 hours thought to be necessary for impact on behavior (Celebuski &
Farris, 2000; Lytle, 1994). Qualitative methods explored the training and confidence
levels of P.E. teachers in APS elementary schools to deliver nutrition education and the
GSW program. P.E. teacher interviews included questions regarding amount of time
spent preparing for and delivering GSW lessons to document different doses of the
GSW program offered at different schools. Training for and confidence in teaching, as
well as dose/contact hours of GSW programming could have an influential effect on
school-level outcomes.
Incorporating nutrition education into classroom instruction with successful
learning strategies is an important area of research. One meta-analysis of nutrition
education strategies recommends the following for effective nutrition education (Lytle,
1994):
Instruction with a behavioral focus, or a focus on changing specific behaviors
rather than on learning general facts about nutrition;
Employment of active learning strategies instead of relying exclusively on
information dissemination and didactic teaching methods;
Devotion of adequate time and intensity to nutrition education (at least 50 hours
per year to impact attitudes and behavior);
A family involvement component;
A meals program and food-related policies that reinforce classroom nutrition
education;
41


Teachers with adequate training in nutrition education
Other research contributes to best practices for school nutrition education. The
CATCH program demonstrated that nutrition education messages for children are
successful when they are focused on behavior and supported by the school nutrition
programs (Perry, Lytle et al., 1998). Parental and community involvement are important
components of successful nutrition education interventions (Nader et al., 1996; Perez -
Rodrigo et al., 2001). Staff dealing with school meals should be properly trained,
supported and integrated with teaching staff (Fulkerson, French, Story, Snyder, &
Paddock, 2002). Based on a meta-analysis of 29 behavioral interventions studying ages
two to 18, strategies to reduce unhealthy behaviors (decreasing sedentary behaviors and
dietary fat) are more effective than those promoting positive behaviors (increasing
physical activity and consumption of fruits and vegetables) (Kamath et al., 2008).
Longer duration and more contact hours involved with interventions produce greater
benefit to participants. The School Health Education Evaluation, a large evaluation of the
effects of health education programming in schools, found that 8 hours of health
education could produce large effect sizes in program-specific knowledge and 20 hours
could produce large effect sizes in general health education knowledge. However, even
after 35 to 50 contact hours, only moderate effect sizes could be achieved in attitudes and
behaviors (Connell, Turner, & Mason, 1985). The CATCH intervention involved 15 to 20
contact hours per year over three years for third through fifth graders and resulted in
positive eating behavior and physical activity behavior changes but not physiological
changes (Luepker et al., 1996). These changes persisted when re-measured in the eighth
grade (Nader et al., 1999). The Know Your Body program included 30-50 contact hours
42


to manipulate diet and create positive effects on serum cholesterol and blood pressure
(Resnicow et al., 1992; Walter, 1989). Given the high number of content hours necessary
for knowledge, attitude, and behavior change, new interventions should plan
appropriately to achieve desired results. An important component of this research was a
careful analysis of GSW implementation in each school. P.E. teachers were interviewed
and asked specific questions about contact hours, learning strategies, school and
administrator support, and other aspects of GSW programming. Analysis of GSW
implementation also provides important process evaluation information to improve
consistency across schools in future programming.
Implementation Issues in Prevention and Education Programs
Continuous delivery of effective interventions is an important component in program
success; few interventions are sustained over time regardless of their success during pilot
periods (Durlak & DuPre, 2008; E. M. Rogers, 2003). Measuring the implementation of
a program is paramount to drawing conclusions about program success. Negative or null
results from a program could signify a failure of that program, or they could simply mean
that the program was not implemented as designed. Conversely, positive program impact
could result from intended implementation or from implementation quite different than
what was intended. It is impossible to make judgments about a program without
assessing implementation. Implementation data also help to test theory behind an
innovation by ascertaining which components and which related theories were effectively
administered (Durlak & DuPre, 2008). A meta-analysis by Durlak (2008) defined key
terms related to implementation: (1) Fidelity, or the extent to which a program
corresponds to the originally intended program, (2) Dosage, or how much of a program
43


has been delivered, and (3) Quality, or how well different program components have
been conducted. Regression analyses in one review of 221 school-based prevention
programs targeting aggressive behaviors reported that implementation was the most
important variable that influenced outcomes (Wilson, Lipsey, & Derzon, 2003). Other
studies of school-based interventions and community-based interventions targeting a
variety of outcomes found significantly higher effect sizes in programs that monitored
implementation and adjusted analyses for factors related to implementation (Ananiadou,
Schneider, Smith, & Smith, 2004; Derzon, Sale, Springer, & Brounstein, 2005; D.
DuBois, Holloway, Valentine, & Cooper, 2002; Tobler, 2000).
Researchers have historically analyzed implementation in two major ways: (1)
categorically, with groups for different levels of implementation (i.e. high vs. low) or (2)
continuously, with percentages assessing level of dosage or fidelity (Durlak & DuPre,
2008). The latter method reports a broader range of implementation and might give more
statistical power to conclusions about the effects of implementation. The two most
common methods for arriving at implementation values are provider self-reports and
independent behavioral observations (Durlak & DuPre, 2008). Few studies have directly
compared the two strategies, but objective observations would likely return more
unbiased, accurate, and comparable assessments of implementation. Including measures
of implementation in qualitative and quantitative analyses help predict program
outcomes. This research utilized qualitative strategies to assess implementation.
A social ecological framework can be used to identify influences on program
implementation. Contextual factors are important in understanding motives, support, and
barriers for implementation. Durlak and colleagues identified the following main
44


categories in an ecological framework for program implementation: innovations,
individuals and communities, and features associated with the prevention delivery and
support systems (2008). The innovation can be defined as the program itself. Program
characteristics that affect implementation include adaptability (program flexibility to
meet the needs of providers/teachers) and compatibility (contextual appropriateness and
fit with the organization/school) (Everett M. Rogers, 2003). Although there must be a
balance between adaptability and fidelity of proposed implementation, programs that are
adaptable for teachers and that fit within the culture of a school and school district show
higher fidelity and quality of implementation.
Factors related to individual providers and communities will affect program
implementation. For example, school staff forced to launch a program and school staff
who volunteer to create or deliver a program could vary in the effectiveness of their
program implementation. Provider/teacher characteristics of implementation include: (1)
providers who recognize program need, (2) providers with high self-efficacy for program
delivery, (3) providers who have the skills to implement the program, and (4) providers
who believe in the program and the potential benefits (Barr, Tubman, Montgomery, &
Soza-Vento, 2002; Cooke, 2000; Ringwalt et al., 2003). Supportive principals are also an
integral part of high fidelity for program implementation (Kam, Greenberg, & Walls,
2003).
The prevention delivery and support system for providers are key components of
implementation. Training and technical assistance for people implementing a new
prevention program are key elements of program success. The goals of training should
include development of mastery in specific intervention teaching and skills, as well as
45


attention to provider expectations, motivation and sense of self-efficacy (Durlak &
DuPre, 2008). After a program begins, technical assistance should help maintain
providers motivation and commitment, improve their skills where needed, and support
problem solving efforts. Two studies show that early monitoring of implementation
followed promptly by retraining for providers with initial difficulties doubled the fidelity
of implementation to over 85% (Dufrene, Duhon, Gilbertson, & Noell, 2005; Greenwood,
Tapia, Abbott, & WAlton, 2003). Dialogue and communication with providers is crucial
immediately after the beginning of a new program, and remains important throughout the
duration of a program.
Summary
According to the American Dietetic Association, about half of all Americans
believe they are doing all they can to achieve balanced nutrition and a healthful diet, but
given that such a significant portion of our population is overweight or obese, it appears
that knowledge alone does not lead to behavior change (American Dietetic Association,
2008). Given that information dissemination is necessary but not sufficient to change
behaviors, Contento (2011) recommends that nutrition education focus on personal
motivations and competence, interpersonal interactions, and environmental factors that
influence individual and community patterns of behavior. Aurora Public Schools are
following that recommendation with the Go, Slow, Whoa program to educate students,
increase knowledge, and change attitudes and behaviors surrounding nutrition, and the
Scratch Cooking initiative to improve the nutrition environment and healthful food
offerings in schools. The theoretical basis for this research included SCT and SEM. The
constructs of SCT and SEM provide support for intervention components and facilitate
46


knowledge and action by individuals and make the environment more conducive and
supportive of desired changes. This research assessed the effectiveness of this
educational and ecological combination of Aurora Public School programming for
promoting dietary intake and health.
The most similar intervention to the current GSW intervention is the Dietary
Intervention Study in Children (DISC): a 3-year intervention aimed at decreasing intake
of fat, saturated fat, and cholesterol among pre-adolescent children by increasing
knowledge and skills for identifying healthy food (Stevens et al., 1995; Van Horn et al.,
2005). The DISC study utilized intervention sessions (number unreported) to teach
children how to identify healthy foods (Go foods) to choose and eat all the time and less-
healthy foods (Whoa foods) to choose and eat only occasionally. The intervention
lessons utilized a Go-Guide wheel with eight wedges pertaining to eight categories of a
childs diet. Within each wedge, Go foods were highlighted in green and Whoa foods
were highlighted in red. Dietary intake was measured with 24-hour dietary recall at both
pre-intervention and post-intervention. Although the intervention sessions in the DISC
research targeted a visual nutrition guide very similar to GSW, assessing dietary intake
with subjective 24-hour recall is much different than objective visual plate waste analysis
used in this research. However, utilizing a 24-hour recall methodology, the intervention
group showed increased consumption of Go Foods in all measured food categories except
for fruits and those increases were significant for dairy products, desserts, and fats/oils.
The intervention group showed decreased Whoa Dairy, Whoa fats/oils, and Whoa
vegetables (specifically French fries) when compared with the control group. This
nutrition education intervention utilized very similar visual aids and teaching tools as the
47


GSW program in Aurora Public Schools. Teaching healthy food identification skills with
a Go-Whoa visual aid improved diet quality and leads to hypotheses about the current
GSW programming in APS. Given the success of the DISC program, I hypothesize that
students in GSW schools with high fidelity, dose, and quality of implementation who
received Scratch Cooking Cafeteria Changes will have significantly higher choice and
consumption of Go foods than non-GSW schools who only receive Scratch Cooking
Cafeteria Changes. Overall, I hypothesize that students in GSW schools will report
increased choice and consumption of healthy foods and new healthy recipes in APS
cafeterias.
48


CHAPTER HI
METHODS
Research Question and Hypotheses
How influential is the Go, Slow, Whoa nutrition education program combined
with Scratch Cooking Cafeteria Changes over and above the impact of Scratch Cooking
Cafeteria Changes alone on elementary students choices and dietary intake related to
healthy foods?
Aim 1: Determine the relative impact of
Figure 3: Aims and Hypotheses
49


Research Design
This study employed a quasi-experimental, posttest only, nonequivalent control
group design using mixed-methods and focused on outcome evaluation. In this study, the
Go, Slow, Whoa program and Scratch Cooking Cafeteria Changes, existing pieces of
LiveWell Colorado programming, were in need of outcome evaluation. Outcome, or
summative, evaluation focuses on two issues: effect assessment (production of desired
effects) and efficiency assessment (benefits in relation to costs) (Singleton & Straits,
2010). The focus of this research was effect assessment of both GSW and Scratch
Cooking Cafeteria Changes, which helps in drawing conclusions about efficiency of the
programs. Quasi-experimental designs are utilized when legal, ethical, or practical
considerations make true experimental design impossible (Singleton & Straits, 2010). In
Aurora Public Schools, all schools were exposed to Scratch Cooking Cafeteria Changes
in the 2010-2011 school year. Only five elementary schools were exposed to GSW, and
these five schools were selected without randomization. Thus, a nonequivalent control
group design was chosen to compare schools with GSW to control schools without GSW.
Although similarity of schools and students cannot be ensured without randomization,
this quasi-experimental design matched experimental and control schools with propensity
score and Mahalanobis matching techniques in an attempt to minimize differences
between schools on observable characteristics. In this instance, pretest data were not
collected by the school district, therefore a posttest-only design offered the best
information for determining program effects.
This quasi-experimental, posttest only, nonequivalent control group outcome
evaluation research was selected over competing study designs because of existing
50


constraints of the school district and LiveWell programming. Combining quantitative
and qualitative data in a methodologically rigorous design returned rich data regarding
student behaviors and teacher impressions resulting from school cafeteria changes and
nutrition education. This study design was best suited to answer the research questions
related to Scratch Cooking Cafeteria Changes and GSW programming.
Background and Target Population
Aurora Public Schools and LiveWell Colorado Programming
Aurora Public Schools
Aurora Public Schools span two counties: Adams and Arapahoe. The average
obesity and overweight percentages for all three counties is significantly higher than the
Healthy People 2010 goal of 5%. According to 2009 data (COPAN, 2009):
Table 2: Child Obesity Rates in Aurora
County % obese children (ages 2-14) (95th percentile) % overweight children (85th-94th percentile) % children who eat 5 or more servings of fruits and vegetables daily
Arapahoe 11.9% 11.8% 28%
Adams 18.6% 14.3% 26.9%
Data collected from three Aurora elementary schools show even higher rates of
obesity and overweight than the general county rates. Body Mass Indices (BMIs) showed
47 percent of students were overweight or obese and 19 percent were obese.
Furthermore, BMI analysis in Aurora identified a trend of increasing BMIs as students
advance by grade in elementary school (L. Scott, 2010). This trend points to the critical
need to prevent weight gain at the elementary school level. This school district is an
51


excellent target for healthful eating education, promotion, and programming.
Scratch Cooking Cafeteria Changes in Aurora
As of April 2012 in Aurora Public Schools, 53 school food service managers and
central office leadership staff had attended the LWC and the Colorado Health Foundation
sponsored Cook for America Culinary School Food Boot Camp. This represents 16%
of the total Nutrition Services Staff (kitchen managers, kitchen staff, and central office
leadership and staff). The culinary boot camps teach necessary scratch cooking skills
to increase the use of whole grains, utilize more fresh fruits and vegetables, prepare raw
meat instead of reheating frozen and processed meats, and use knives with proper knife
skills to increase control over the healthfulness of ingredients. Following the boot camps
and prior to the 2011-2012 school year, Aurora Public Schools Nutrition Services
Department made significant changes to their menu selection and recipes. Forty-one new
recipes utilized by all Aurora Public Schools were developed or modified minimizing
added fat, sodium and sugar. APS has incorporated stealth health in many recipes by
adding healthful foods and ingredients that the students might not even detect (i.e. fresh
carrots, zucchini, onions, celery, and garlic in spaghetti/red sauce). APS began offering a
salad bar at most schools in Fall 2011. In September, 2011, Aurora Public Schools
served over 220,000 lunches in 30 elementary schools, averaging 459 lunches per school.
Although foods consumed and likability of cafeteria offerings were not assessed
before APS implemented broad menu changes, post-test only assessment of what kids eat
and what kids do not eat will help with APS Nutrition Services planning, program
evaluation, and food ordering.
52


The Go. Slow. Whoa Program
Aurora Public Schools piloted the Go, Slow, Whoa program in one school in
Spring 2010. After analyzing pre-post surveys from students in the program as well as
interviewing program managers, the program was expanded to five elementary schools
for the 2010-2011 school year. All elementary schools are slated to receive the program
by the 2014-2015 school year.
The Go, Slow, Whoa Program was developed as a component of the Coordinated
Approach to Child Health (CATCH program) at the University of Texas School of Public
Health at Houston and adopted as a part of the National Heart, Lung and Blood Institute
WECAN! program to help kids identify healthy foods in school meals and at home
(NHLBI, 2009; Perry, Bishop et al., 1998; Perry et al., 1997). Using a green light, yellow
light, red light visual aid, students are taught how to incorporate more healthy foods into
their diets. Go foods, or more nutrient dense foods, should be consumed more often than
Slow or Whoa foods that have added fat and sugar. Slow foods should be eaten less often
and in smaller quantities. Whoa foods should be eaten only rarely.
GSW, in Aurora Public School District, utilizes a kick-off assembly, physical
education class time, parent meetings, classroom teacher reinforcement, and cafeteria
food labeling to educate about and reinforce the consumption of Go foods. In late
September, 2011, each of the five GSW elementary schools hosted a kick-off GSW
assembly featuring the P.E. teacher, a TV personality from Channel 7, and several
mascots in costume (Power Panther and the Colorado Avalanche mascot). The assembly
reinforced the proper identification of Go, Slow, and Whoa foods, introduced kids to the
cafeteria labeling system, and served as a cheerleading session for the program. In
53


October, parents were invited to attend either a parent breakfast or a parent afternoon
snack meeting to describe the program and encourage family participation at home.
Parents receive monthly newsletters and menus with Go, Slow, and Whoa foods
highlighted. The newsletters are created and managed by each individual GSW school,
and, therefore, differ across schools.
At the beginning of the 2011-2012 school year the APS Healthy Schools
Coordinator held trainings for all principals, gym teachers, nurses, and cafeteria managers
from the GSW schools to explain program components and how to implement them
according to LiveWell and APS standards. Program components are listed in Table 3.
The nutrition education components of the program are taught and reinforced in P.E.
classes as well as in homeroom classrooms. Students are encouraged by teachers and
parents to look at the menu and cafeteria labels and choose Go foods on a daily basis.
The Nutrition Services Department in the school district label all foods on the school
menus (received by all APS families) and daily in the GSW school cafeterias themselves
as either Go (green apple), Slow (yellow circle), or Whoa (red square). See Appendix A
for a copy of the December 2011 menu and Appendix B for a copy of the rating system
for GSW foods (Coordinated Approach to Child Health (CATCH), 2012). It is important
to note that the CATCH rating system is for individual food items such as whole wheat
bread, fresh tomatoes, 2% cheese, and butter, which would each be labeled as Go, Go,
Slow, and Whoa foods, respectively. However, APS Nutrition Services staff label
complete menu items in the cafeterias and on print menus. Therefore, taking into account
each of those ingredients, a grilled cheese sandwich made with whole wheat bread, 2%
cheese, tomato slices, and a small amount of butter would likely be labeled as a Slow
54


food. An important element of the GSW program is point-of-sale labeling of GSW
foods. Kitchen managers set out labels in acrylic frames each day above or beside food
options identifying food items as Go, Slow, or Whoa.
Incentive days allow nutrition services and school personnel to promote healthy
foods and the GSW program. Three incentive days are offered during the school year to
each GSW school during which a special fruit, vegetable, or skim milk is offered. If
students take AND eat the item they receive a Channel 7 wristband (similar to Lance
Armstrongs Live Strong bands). The bands come in three different colors, thus a new
color is offered on each incentive day.
Table 3: GSW Program Components
Program Component Program Component Description Responsible Party (note: there is no mandate as to who completes different components. This column is based on qualitative interviews and informal observations.)
Nutrition Education/GSW Curriculum Lessons to teach and reinforce the Go, Slow, Whoa labeling system P.E. Teachers, Cafeteria Managers
Kick Off Assembly Assembly to introduce GSW, increase energy, awareness, and excitement. District GSW Administrators, Principals, P.E. Teachers, Cafeteria Managers, Channel 7 News Personalities
GSW Labels on Lunch Menus Encouragement and assistance for students to look for GSW labels on printed lunch menus Classroom Teachers, P.E. Teachers, Parents, Cafeteria Staff
Labeling of Cafeteria Food with GSW Labels GSW labels affixed to plastic sneeze guard or in front of lunch items each day. Cafeteria Staff
Three Incentive Days Prizes (bracelets) awarded if students take/consume a certain product (white milk, specific fruit, specific vegetable) on three days during the school year Cafeteria Staff, Lunch Room Aides, Administrators
Parent Breakfast Informational meeting about GSW Principal, Assistant Principal, Cafeteria Manager
GSW Articles in School Newsletters Newsletters to entire school community should contain GSW information or nutrition-related articles P.E. Teachers, Front Office
Extra Programming Guest Speakers, More Incentive Days P.E. Teachers
55


Methods for Matching GSW Schools with Comparison Schools
Five schools were selected for participation in the GSW program in the 2011-
2012 school year. All schools will implement the GSW program by 2014-2015. The
procedure for placing schools on the rotating start list was not random, nor was the
selection process documented. Seven schools participated in GSW programming in
2010-2011 and five elementary schools started in fall 2011. Given that schools were not
selected randomly for treatment in the 2011-2012 school year, propensity score matching
was an appropriate statistical technique to identify five schools most similar to the GSW
schools on observed school characteristics to serve as comparison schools.
Propensity score matching allows researchers to determine the extent to
which a treatment group and comparison group is similar based on some observed
covariates, even when the groups are assigned without randomization. The goal of
propensity score matching is to match experimental and control subjects on
observed covariates so that the main difference between them is whether or not they
received the treatment (Khandker, Koolwal, & Samad, 2010; Rosenbaum & Rubin,
1983). Propensity scores reduce all of the observed covariates into a numerical,
scalar summary for selecting matched samples: the probability of receiving the
treatment (ranging from 0 to 1), conditional on the covariates. Matching can then
be done on this scalar summary rather than on all of the covariates directly
(Rosenbaum & Rubin, 1983). The premise behind this technique is that if two units
have the same propensity score but are in different treatment groups, the
determination of which unit received treatment and which did not was random
based on observables. Therefore, in the school-level setting, treatment and
56


comparison schools with similar propensity scores should have similar joint
distributions of the covariates used for propensity score matching. Treatment effects
can then be estimated based on the mean differences in outcomes between
participants and matched non-participants. In this particular research, schools were
treated as matching units because entire schools were selected to receive treatment.
Matching methods such as propensity score matching are becoming more
and more popular as ways to estimate causal effects by using observational
data. These methods, which select subsets of the original treatment and
control units that are the most similar on the observed covariates, can be
conceived as a way to replicate a randomized experiment by selecting
treatment and control units (schools) that look only randomly different from
one another on all of the observed covariates (Stuart, 2007).
Oftentimes observational studies simply utilize models with a treatment
indicator and a set of covariates as predictors of some outcome variable. Instead, the
design of observational studies can be structured with the rigor of randomized
experiments: without using the outcome data and with a good understanding of the
treatment and control conditions (Rubin, 2001). This approach involves outlining
the extent to which the treatment and comparison groups are similar on background
covariates and then using methods such as propensity score matching to ensure that
the treatment effects are estimated by using schools or subjects that look similar to
each other. The strongly ignorable treatment assumption (SITA), a key assumption
of propensity score matching, implies that treatment assignment is independent of
the potential outcomes given the observed covariates (Rosenbaum & Rubin, 1983).
SITA is met if treatment assignment (Z;) and the potential outcomes Y, = (Y0i, Y] j)
are conditionally independent given the observed covariates X; (or, alternatively,
the propensity score). The formula for this probability of treatment assignment is:
57


Pr(Zi|Xi,Yi) = Pr(Z;|Xi), with 0 < Pr(Z, = 1 |X;) < 1,
where Yo; is the potential comparison outcome (for Z, = 0) and Yu the potential
treatment outcome (for Z; = 1) for all subjects i = 1,..N. These outcomes are
called potential outcomes because they refer to the outcome one would observe if
subject i gets treated (Yu) or not treated (Yoi) (Shadish & Steiner, 2010).
There are several advantages of Propensity Score Matching (PSM)(Khandker et
al., 2010). PSM provides good comparisons if observed characteristics drive
program participation and unobserved characteristics are negligible. PSM does not
require a baseline survey or data. PSM is a semi-parametric method, imposing
fewer constraints on the functional form of the treatment model, as well as fewer
assumptions about the distribution of the error term. This means that PSM
increases the likelihood of sensible comparisons across treated and matched control
units.
The information that was available for all public elementary schools in
Aurora is summarized in Table 4. This study used various propensity score models
with some and all of these covariates to determine the best fitting model.
58


Table 4: Covariates for Matching
Variable Definition
Race/Ethnicity % of students in each school who are: White omitted from model as comparison group Black Asian Hispanic Native American Native Hawaiian Two or More Race/Ethnicities
Free/ Reduced Lunch (FRL) % of students in each school who are eligible for free or reduced lunch. Governmental program that provides low-income students with free or reduced price school meals. Percentages for schools are determined by taking the number of students eligible for free or reduced price lunch divided by the total number of students enrolled in the school.
Gifted and Talented (GT) % of students in each school who are gifted and talented in either math or language or both. They are identified by teachers as they search for students who perform at exceptional levels. Nominations for GT students are made with supporting evidence from cognitive and culture free measures, academic aptitude and performance, language acquisition skills, etc.)
Special Education (SPED) % of students in each school who are classified as special education and have an individualized education plan (IEP).
English Language Learners (ELL) % of students in each school who are classified as English Language Learners (either no proficiency or low proficiency in English)
Nutrition Programs (NP) Total Number of additional Nutrition Programs offered in a school (Range 0-5) The five possible programs offered at various Aurora schools are: Breakfast in the Classroom Fresh Fruit and Vegetable Program INEP Nutrition Education Healthier U.S. Schools Challenge Award for implemented changes and activity time (not necessarily a program, but a commendation for changes) Coordinated School Health.
The Nutrition Programs variable was created as a sum of additional nutrition
programs offered in a school. This variable was created by combining 5 different
dichotomous variables (for each of five programs) to save degrees of freedom in
modeling. The logic of including the Nutrition Program variable relates to the ability and
motivation of each school to bring in additional nutrition programs. If a school
59


(administrators, teachers, and/or parents) is able to bring in 3 additional programs, that
school is likely different in various ways (motivation of school community, SES of
children, previous grant funding) compared to a school that does not bring in any
additional program.
Propensity scores were calculated in STATA using the PSMATCH2 program
which runs multiple methods of matching using propensity scores and full Mahalanobis
matching (Leuven & Sianesi, 2003). Propensity scores were estimated with a logit model
using two different matching methods: nearest neighbor matching and Mahalanobis
matching. These two matching methods were chosen because of the small sample size
and based on previous research with Mahalanobis matching and smaller sample sizes
(Zhao, 2004). Nearest-neighbor matching randomly orders the treatment schools, and
then finds a control school with the closest propensity score (D'Agostino, 1998). Because
of the small sample size of untreated schools, those two schools were then taken out of
the pool (not replaced) and the second treatment school was matched with the nearest
neighbor in terms of propensity score.
The second method used for matching was Mahalanobis metric matching
combined with propensity score matching. This method randomly ordered subjects and
then calculated the distance between the first treated subject and all controls using
matching variables (including the propensity score) and the sample covariance matrix of
matching variables from the full set of control subjects (Baser, 2006; D'Agostino, 1998).
The first randomly ordered treatment subject was matched based on the smallest
Mahalanobis distance. The process was then repeated until each treatment subject
received a match.
60


In this research, propensity score matching was based on data available for all
elementary schools included in Table 4: (See Appendix C for complete Data Table.).
Running 12 different logit models helped determine the best fitting models for the
available data. As indicated in the literature, different variables were logged and squared
in different models to determine improvements in the Akaike Information Criterion
(AIC) and Bayesian Information Criterion (BIC). Different variables (12 total) were
included and excluded in successive models to determine improvements in model fit.
Models with fewer covariates improved AIC and BIC values. Even though none of the
individual independent variable coefficients were significant in any of the models
(perhaps due to small sample size), three models emerged with very similar AIC and BIC
values. Using fewer covariates in these three models with such a small pool of available
matches helped to avoid overspecification of matching models. Post-hoc analyses of
variable means show the differences between covariates (mean values) for treatment
schools and match schools. Model 1 used propensity score matching with four covariates
and nearest-neighbor matching. All variables with large mean differences between
matches were logged and run in different models to determine the best fitting model.
Logging the Hispanic variable resulted in lower AIC and BIC values, and thus this
variable is logged in Model 2. Model 2 used propensity score matching with four
covariates (1 logged) and nearest-neighbor matching. Model 3 used full Mahalanobis
matching with five covariates (the covariates in Model 2 plus propensity scores).
61


Formulae for Propensity Score and Mahalanobis Metric Logit Models:
Model 1: Propensity Score Matching
Y (GSW treatment inclusion) = a (intercept) + biXi (percentage Hispanic students) +
b2X2 (percentage English Language Learner Students) + b3X3 (percentage Free and
Reduced Lunch Students) + b4X4 (Number of Nutrition Programs) + 8
Model 2: Propensity Score Matching
Y (GSW treatment inclusion) = a (intercept) + biXi (logged percentage Hispanic
students) + b2X2 (percentage English Language Learner Students) + b3X3 (percentage
Free and Reduced Lunch Students) + b4X4 (Number of Nutrition Programs) + 8
Model 3: Mahalanobis Matching
Y (GSW treatment inclusion) = a (intercept) + biXi (logged percentage Hispanic
students) + b2X2 (percentage English Language Learner Students) + b3X3 (percentage
Free and Reduced Lunch Students) + b4X4 (Number of Nutrition Programs) + s +
b5X5(Propensity Score) + s
Results of all three matching models using the Stata PSMATCH2 program
(Leuven & Sianesi, 2003), are shown in Figure 4.
Model Log Likelihood (null) Log Likelihood (model) AIC BIC
Model 1 Propensity Scores -12.51 -12.20 34.40 40.49
Model 2 Propensity Scores -12.51 -12.05 34.09 40.20
Model 3 Matching Mahalanobis -12.51 -12.05 34.09 40.20
Figure 4: Matching Model Results
Model 3 (Mahalanobis matching) produces similar model statistics to Model 2
(Propensity Score matching), because Model 3 was calculated with the same predictor
variables as Model 2. Therefore, the propensity scores are identical for both models.
However, the comparison schools chosen by Model 3 are different because Mahalanobis
matching uses the propensity scores as an additional matching variable.
62


The STATA PSMATCH 2 program determined the following matches for each
GSW school for each of the three models.
Table 5: Matched Schools Based on Three Matching Models
GSW School Model 1 Match Model 2 Match Model 3 Match
Altura Wheeling Peoria Wheeling
Lansing Yale Fulton Jewell
Vistapeak Sixth Avenue Yale Arkansas
Crawford Murphy Creek Sable Fletcher
Vassar Dartmouth Dartmouth Dartmouth
The following figures (Figures 5-9) show the absolute values of covariate
differences between GSW schools and their matches for each model. These figures
provide a visual of how similar the GSW schools were to their matches on each matching
variable using 3 different Models. If the y-axis values in any of the following graphs are
zero, then there is zero difference between the GSW school and the match school on that
covariate.
Altura
Altura-Model 1 match
(Wheelig]
Altura-Model 2 match
(Peoria]
Altura-Model 3 match
(Wheelig]
Figure 5: Altura Matching Results
63


Lansing
c
m
O rH
O
-C 0)
a §
£ £
C/5 s
= i
g *
0)
S o
2*
u
aj t/i
T3
c a;
iS
5 S
Lansing-Model 1 match
(Yale)
Lansing-Model 2 match
(Fulton)
Lansing-Model 3 match
(Jewell)
Figure 6: Lansing Matching Results
E
o
re n
O rH
O _
JS V
u -o
o
£ s
t/5
u
a
m
OJ t/5
S O
u
aj t/5
-a
S 0)
tu Ja
3= re
S S
Vista Peak
Vista Peak-Model 1
match (Sixth Avenue)
Vista Peak-Model 2
match (Yale)
- Vista Peak-Model 3
match (Arkansas)
Matching Variables
Figure 7: Vista Peak Matching Results
64


Crawford
Crawford-Model 1 match
(Murphy Creek)
Crawford-Model 2 match
(Sable)
- Crawford-Model 3 match
(Fletcher)
Figure 8: Crawford Matching Results
Vassar
Vassar-Model 1 match
(Dartmouth)
Vassar-Model 2 match
(Dartmouth)
- Vassar-Model 3 match
(Dartmouth)
Matching Variables
Figure 9: Vassar Matching Results
65


Visual inspection of the differences in means between GSW schools and their
matches using Models 1, 2, and 3 indicated that Model 3 (Mahalanobis Matching with
Propensity Scores) had the smallest differences on the most number of covariates.
Smallest differences between schools on covariates simply meant those schools were
most similar on that covariate. For example, in looking at the covariate White
(percentage of White students in the school) in Figure 5, the difference between Altura
and the Model 2 match (Peoria) is the smallest (shaded red), and therefore, Model 2 is the
best match for that particular covariate. There were 5 covariates for each school.
Multiplying 5 covariates by 5 schools resulted in 25 different variable combinations to be
considered in the best fitting model across all schools. Totaling which model (Model 1,
Model 2, or Model 3) produced the smallest difference between the GSW School and
Match School on each of the variables is summarized in Table 6.
Table 6: Comparison of Matching Models for Best Fit
Model 1 Model 2 Model 3
Number of variables in which the model produced the best match 9 12 16
*note that the total number of variables in this table does not total 25. If a variable
resulted in the same lowest differences between GSW school and Match school for more
than one Model, then both models were counted in this table.
Because it produced the greatest number of smallest differences between match
school means and GSW school means, Model 3 (Mahalanobis Matching with Propensity
Scores) was used to identify untreated comparison schools for the GSW treatment
schools. Final school matches are listed in Table 7.
66


Table 7: Final Comparison Schools
GSW School Model 3 Match
Altura Wheeling
Lansing Jewell
Vista Peak Arkansas
Crawford Fletcher
Vassar Dartmouth
Data Collection Methods
Aim 1: Quantitative Plate Waste Data
Aim 1: Determine the relative impact of GSW and Scratch Cooking Cafeteria
Changes over and above the impact of Scratch Cooking Cafeteria Changes alone on
students choices and consumption of school lunch.
This research utilized both quantitative and qualitative data collection methods.
Identifying what children choose and what they eat at school helped address the research
questions. Children, especially young children, are oftentimes omitted from data
collection in schools because of their underdeveloped reading skills and variation in
cognitive skills and development (J. Scott, 2008). This study used quantitative data
collection that did not require reading on the part of the students in grades 1,3, and 5.
Grades 1,3, and 5 were chosen for data collection to represent the range of ages and
grades in an elementary school without collecting data from all grades.
Dietary intake using quantitative survey methodology, while providing some
useful information based on personal food recall, does not always produce reliable results
(Beaton, Burema, & Ritenbaugh, 1997; Tran, Johnson, Soultanakis, & Matthews, 2000).
Children and adolescents, specifically, do not accurately recall food intake, which
becomes a problem when such assessments are used to monitor nutritional status
67


(Livingstone, Robson, & Wallace, 2004). Children under-report, over-report, and
incorrectly identify foods and portion sizes. A meta-analysis of 11 studies using food
recall surveys with children found a wide range in mean energy intake and validity of
food recall techniques (McPherson, Hoelscher, Alexander, Scanlon, & Serdula, 2000).
Food diaries/records are another subset of food recall techniques commonly used to
eliminate issues with memory, but they do not eliminate issues related to estimating
portion size and properly identifying foods consumed. Validity and reliability of food
diaries/records is not consistent in multiple studies using this technique (McPherson et
al., 2000). Therefore, the quantitative method selected for this study assessed plate
waste, an accurate assessment of food intake for a specified time period that did not
involve individual subjective reporting. In order to assess the effect of GSW and Scratch
Cooking Cafeteria Changes on students in grades 1, 3, and 5, plate waste methodology
was selected considering the varying cognitive abilities of children ages 5-11.
The most accurate way to measure food intake is weighing foods before and after
consumption (Wolper, Heshka, & Heymsfield, 1995). However, this method is time
consuming and costly. Using plate waste or the quantity/weight of food left after a
food is consumed has been utilized to assess the National School Lunch and School
Breakfast programs. Given the importance of good data concerning dietary intake in
children, researchers developed other reliable and valid methods for collecting data that
are less costly and less intrusive. Plate waste protocols based on direct visual observation
of foods before and after a meal is well suited to cafeteria settings or other public eating
situations (Comstock, St Pierre, & Mackiernan, 1981; Wolper et al., 1995). The portion
of food that remains (plate waste) is estimated by visual observation when a tray is
68


returned at the end of a meal. Visual estimation of plate waste is a promising and
inexpensive technique for assessing dietary intake.
Original visual estimation plate waste studies utilized trained observers directly
observing food trays before and after eating. The observers estimated portion sizes for
each food in reference to standard portion sizes using some sort of percentage rating
scale. By weighing several samples of the reference food (to get accurate baseline
serving sizes and nutrition information) and comparing them to the percentage of food
left over, the percentage of food eaten (along with weight, energy, nutrients) can be
estimated from the average reference food weight and nutrition information. This visual
method has been used extensively in studies of nutrient intake and eating behaviors in
schools and other institutional settings (Auld, Romaniello, Heimendinger, Hambidge, &
Hambidge, 1999; Davidson, Hayek, & Altschul, 1986; Friedman & Hurd-Crixell, 1999).
Direct visual estimation has been proven comparable to weighed protocols in several
studies (Comstock et al., 1981; S. Dubois, 1990; Kirks & Wolff, 1985).
One of the barriers to using direct visual estimation in a school cafeteria
environment is the required time for food tray assessment. Observing food trays at the
end of a lunch line before a student sits down to eat takes precious minutes away from
food consumption time. At the end of the lunch period, collecting a large sample of trays
necessitates on-site estimation to avoid transportation, sanitation, time, space, and
manpower considerations of transporting food trays. On-site post-lunch estimation in a
timely manner is difficult due to utilization of cafeteria space for other activities
following lunch. Without a large number of data collectors, visual estimation would
likely take a long period of time, even for a random sample of trays. It is also difficult to
69


conduct inter-rater reliability assessments if trays are subsequently discarded after post-
lunch waste estimation.
A more recently developed method avoids some of these problems by taking a
digital photograph of cafeteria trays both before and after food is consumed. The digital
photograph is stored for future visual estimation, and can be rated by several trained
researchers to ensure data quality. The digital plate waste photography method, utilized
successfully with both adults and children, enables reliable and valid data collection for a
fraction of the cost of weighed plate waste estimates (Swanson, 2008; Williamson et al.,
2003). This study utilized a slightly different method where data collectors recorded on
an index card what was on student trays at the end of a lunch line. That record was then
used as the pre-consumption data, and a digital photograph of each tray at the end of the
lunch period was compared to the pre-consumption data to determine plate waste and
food consumption. This modified digital plate waste protocol was developed to avoid
delays in lunch lines and reduced eating time resulting from the time required for pre-
consumption photographs (S. Smith, personal communication, September 18, 2011).
Sampling Frame for Quantitative Data:
In Aurora Public Elementary Schools, there are 12,800 students. Five schools
implemented GSW programming in 2010-2011, and 20 schools were eligible to be match
schools because they had not yet received any GSW training or materials. Five GSW
schools and five match schools were included in the quantitative portion of this study.
Research Assistants aimed to collect twenty cafeteria trays from each of three grades (1,
3, and 5) in each school for three consecutive days. Therefore, in total, 1800 trays were
targeted for collection and photographing (20 trays per grade x 3 grades x 10 schools x 3
70


days). This sampling strategy would return 900 trays from GSW schools and 900 trays
from non-GSW comparison schools. We targeted 50% males and 50% females (900
males and 900 females) and 33% from each grade (600 1st Graders, 600 3rd graders, 600
5 th graders).
Quantitative Data Collection Process
Because data were collected in 10 different schools on the same day for three
consecutive days, over 100 Research Assistants (RAs) (4 volunteers per school/per day)
were utilized. These RAs were Kaiser Permanente employees participating in a work-
related volunteer program, University of Colorado Denver graduate and undergraduate
students, University of Colorado Colorado Springs undergraduate students, and LiveWell
Colorado Volunteers. One lead RA for each school was trained in research protocols
during a 2-hour training. These lead RAs were present in the same school for all three
days and trained all other RAs at that school.
Supplies Used
From School
Spider Cart with removable large sheet trays
3x4 or larger rectangular table
From Data Collection Team
Digital cameras with extra batteries (each school had a different memory card for all data
collection days)
1 tripod
Base Board (a board made out of plywood shows placement of tripod and trays for
photographing)
71


Ziplocs (to bag random samples for weighing)
Wet wipes
Protractor
Latex gloves
Paper towels
Masking tape
Sharpie markers (for marking tray cards)
Cooler to transport 5 samples of each food
3 clipboards note: for ease and efficiency, pre-tear 2-inch pieces of masking tape and
affix to the clipboard around the perimeter. These tape pieces were used to tape
fluorescent cards to student trays.
Fluorescent Tray Cards -75 for each day with correct lunch offerings printed on card.
Used at CU Denver
Digital Scale Ohaus Scout Pro SP401 Portable Digital Gram Scale
APS menu planning and nutritional analysis software
Plate Waste Protocols
Each elementary school separates lunch periods by grade. For statistical power,
we needed 20 trays per lunch period per grade (1,3, and 5). Given the potential loss of
tray cards (students hiding tray cards or throwing away cards before RAs could catch
them) and malfunction with photography equipment, we aimed to collect 28 trays per
lunch period per grade (1, 3, and 5). RAs attempted to alternate between male and
female students, although this was not always possible as clusters of same-gender
students often came through the line together.
72


Four research assistants arrived 30 minutes prior to the first lunch period to set up.
Each of the cafeterias had a different flow, but the main data collection protocol was the
same. The research assistants arranged trash cans in a central spot and positioned a
spider cart with removable trays near those trash cans to collect trays/plate waste. RAs
set up a table for photography in a less-traveled area of the cafeteria. Upon this table
RAs placed the BaseBoard (elevated 2 inches using a 2-inch binder at the end where food
trays will be placed), tripod, and camera. Two RAs positioned themselves near the end of
the lunch line(s) with clipboards and fluorescent lunch tagging cards. One RA randomly
selected the 4th boy to exit the lunch line. The other RA randomly selected the 4th girl to
exit the lunch line. The RAs explained the purpose of the study and gave students the
opportunity to refuse participation.
Hi, we are from the University of Colorado Denver and we are doing a study on what
kids like to eat in the cafeteria. Would it be OK if we collect your tray at the end of the
lunch period? Wait for answer. If yes, Great! Thank you very much. Dont throw
your tray out at the end of lunch, just put it on that cart over there or give it to one of
those two people standing near the cart. If no, Thats OK, enjoy your lunch.
If the student agreed to participate, the RA quickly checked off the pre-portioned (main
entree and side items) or selected (salad bar fruits and veggies) items on the tray using a
florescent card (see example below).
73


Date: April 18 School:
Grade: 13 5
Gender: M F
Beef Stew
Biscuits quantity?
Green Apples quantity? Milk oz left skim white skim chocolate 1% white
Salad Bar none lg svg (>1 cup) sm serving (< 1 cup) Other
Other F/V quantity? Other F/V quantity?
Figure 10: Tray Cards for Quantitative Plate Waste Data Collection
The RAs then taped the completed card to the students tray in order to identify
participant trays at the end of lunch. This interaction with the student took less than 30
seconds. After an RA affixed the florescent card to a tray, he or she took the next
available student. RAs attempted to alternate between male and female students. The
other two RAs positioned themselves near the garbage cans and began scanning trays for
florescent cards as students finished lunch and walked towards the trashcans. They took
participant trays and racked them on Spider Trays (tall rolling carts with room for
stacking trays). When the RAs at the end of the lunch line finished flagging 28 students
(approximately 14 males and 14 females) during a lunch period (usually 20-30 minute
periods), they joined the RAs near the trashcans to receive finished trays. Depending on
74


the flow of the cafeteria and the number of students, some RAs were able to begin the
arranging and tray photographing process at this point.
Photographing Trays
Research assistants set up a photography area set apart from the main cafeteria
activity and trashcan chaos. That area contained a table, baseboard, tripod, and digital
camera (Nikon Coolpix L24). The tripod was set 26 inches above the table and angled
down on the tray at 45-degrees (protractor included in supply list). Prior to
photographing the tray, RAs placed two letters (corresponding to the first two letters of
the school) and a number (numbers began with 1 on the first data collection day, 200 on
the second data collection day, and 300 on the third data collection day) on each
fluorescent card so that trays/cards were photographed in a numerical order. RAs then
poured any remaining milk into a measuring cup and recorded the ounces remaining to
the nearest V2 ounce on the florescent card. RAs rearranged food on the tray to make it
more clear how much of each lunch item was consumed. For example, if a student
placed his or her leftover hamburger over the carrot sticks, RAs moved the hamburger to
the side and arranged the carrot sticks so the photo clearly showed the leftover hamburger
and how many carrot sticks remain. Finally, the RA took one photo of the tray with the
florescent card still attached to it. The florescent card was then collected and all waste
thrown in the trash.
In addition to photographing all participant trays, RAs also photographed and
collected three sample lunch trays for lab weighing. The weights of these three lunch
trays (and all items on that tray) were averaged and used as the comparison serving size
for all post-lunch plate waste. Thus, if the average cheeseburger (from those 3 samples)
75


was 5.5 ounces, then all plate waste from that particular school was compared to a pre-
lunch tray containing a 5.5-ounce cheeseburger (and nutrients associated with a 5.5-
ounce cheeseburger).
Aim 2: Qualitative Interview Data Collection
Aim 2: Determine the relative dose, quality and adherence/fidelity of GSW
implementation in each GSW school and relate implementation to observed outcomes.
Qualitative data were collected from semi-structured interviews with physical
education teachers in GSW schools to gather data on program implementation, impact,
and effectiveness. P.E. teachers are responsible for delivering and reinforcing much of
the GSW curriculum. Individual semi-structured interviews were chosen to encourage
honest communication about the quality of GSW program materials and curriculum,
support and assistance from school and district administrators, and individual teacher
enthusiasm for and cooperation with the GSW program goals. Interviews were chosen to
elicit complete responses, encourage depth of responses, and determine relative emphasis
on issues. Interviews were chosen over focus groups because focus groups are less
appropriate for these outcomes given that members might not share the same emphasis on
topics and because the group dynamics may imply ideas or emphases that are misleading
(Harrell & Bradley, 2009).
Qualitative Data Collection Process
Seventeen eligible physical education teachers were contacted by email in
October 2012. Follow-up emails were sent each week until December 1 to all teachers
who did not reply to emails. Seven physical education teachers agreed to participate in
30-45 minute individual interviews. Interviews were conducted in November and
76


December 2012 with teachers in year 1 (3 teachers), year 2 (1 teacher), or year 3 of GSW
implementation (3 teachers). There were 17 teachers eligible to be interviewed for their
GSW participation: three had left the district and 7 declined participation or did not
answer emails. Interviews were conducted at convenient times for teachers at their
individual schools. Semi-structured interviews were conducted using the interview guide
shown in Appendix D. Interviews were recorded and transcribed.
Informal observations were also recorded when researchers were present in the
schools. The lead data collector at each school filled out the field notes sheet each day of
data collection. These observations were recorded in field notes and included
descriptions of observed interactions in cafeterias, visual aids posted on school walls,
comments from conversations with school staff, available foods in salad bars, etc.
Unstructured field notes are another qualitative technique used to collect important
data(Mulhall, 2003). Field notes were recorded on the field notes sheet shown in
Appendix E.
Analyses
Quantitative Plate Waste Analysis: Aim 1
The average weight of each food was determined by weighing the three samples from
each cafeteria with a digital scale. Nutrition information for each food was assessed with
Vboss software (http://www.horizon-boss.eom/l utilized by Aurora Public Schools
Nutrition Services. This software package encompasses menu planning, ordering,
recipes, nutrition, accounting, point-of-sale, online payments, free and reduced lunch
management, and more. The menu planning, recipe, and nutrition analysis features of the
software provided information for the nutrition content of lunch items and meals on data
77


collection days. The software analyzes nutrition data for all items in a lunch (recipe for
an entree, any additional side items, and milk).
The amount of each food item remaining was estimated in 10% increments.
Various estimation methods have been used in previous visual plate waste studies
including a 6-increment scale (all remaining, almost all remaining, 3/4 remaining, V2
remaining, Vi remaining, none remaining) (Comstock et al., 1981; S. Dubois, 1990;
Friedman & Hurd-Crixell, 1999), and 10%-increment scale (all, 90%, 80%, 70% etc.
remaining) (Swanson, 2008; Williamson et al., 2003). A 10% increment scale was used
in this research because that method was successfully used in two recent plate waste
studies employing digital photography (Swanson, 2008; Williamson et al., 2003). These
estimates were used to determine the nutrient intake of each student. Specifically, if a
student left 20% of a chicken breast on her plate (consuming 80%), and the chicken
breast has 300 calories, 6 grams of fat, and 20 grams of protein, data for that student will
reflect consumption of 240 calories (300*.80), 4.8 grams of fat (6*.80), and 16 grams of
protein (20*.80). This method of nutrient estimation was repeated for total energy and 13
total macro and micronutrients, similar to data collected in previous plate waste research
(Templeton, Marlette, & Panemangalore, 2005). Totals for each macronutrient and
micronutrient were summed from all food on a students tray. Nutrient totals were
recorded for both amounts taken (using information from the tray card) and amounts
consumed (deduced from visual analysis of waste in digital photographs). Salad bar
items were self-selected in each school; therefore there was no way to quantify a before
eating serving size. Instead, the presence of specific salad bar items was recorded on the
78


tray cards. The variable Total FV is a sum total of the number of unique fruits and
vegetables selected from both the hot lunch line and the salad bar.
The type of milk selected by students was recorded on each tray card (skim white,
1% white, and 1% chocolate). Two dichotomous variables were created for milk choice:
(1) whether students chose white milk or whether they chose chocolate milk/No milk, and
(2) whether students who chose milk chose white milk over chocolate milk.
Plate waste photographs were analyzed by two different research assistants and
compared for inter-rater reliability. Similar to previous digital plate waste research, the
two raters estimated plate waste percentages were averaged for data analysis. If the
discrepancy between the two raters was greater than 50%, the principal investigator
assessed the photograph for a third rating, and the closest two of the three ratings was
used to determine average plate waste (Swanson, 2008). Interrater reliability was
measured for each individual category of collected data (grade, gender, and individual
food/salad bar items): 34 categories of data on Day 1, 32 on Day 2, and 44 on Day 3.
Kappa coefficients for all 110 rated categories are presented in Appendix F. Kappas less
than 0.20 are poor in strength of agreement, Kappas of 0.21-0.40 are fair, Kappas of 0.41-
0.60 are moderate, Kappas of 0.61-0.80 are good, and Kappas of 0.81-1.0 are considered
very good in strength of agreement (Cohen, 1960). The number of Kappa coefficients in
each of these categories for raw data and for data reconciled by a third researcher are
summarized in Table 8.
79


Table 8: Interrater Reliability Kappa Coefficients
Number of Variables with this Kappa
Kappa Statistic Raw Data Reconciled Data
<0.20 2 2
0.21 -0.40 10 6
0.41 -0.60 26 16
0.61 -0.80 6 2
0.81 1.00 66 84
Note: total number of variables rated by two raters and reconciled by a third rater=l 10
Statistical Analyses were performed with Stata 11 on amounts taken and amounts
consumed for both GSW schools and non-GSW schools. Nutrient data for each
food/menu for each of the three data collection days are presented in Tables 9-12.
80


Table 9: Menus for Data Collection Days
April 18-Weds April 19-Thurs April 20-Fri April 24-Tues (Weather Back Up)
Beef Stew new Biscuits Green apples Salad bar Milk BBQ Chicken -new WH grain dinner roll Light Peaches Baked beans Salad bar Milk Big Daddy Cheese Pizza with whole grain crust-new Can Green Beans Red grapes Salad bar Milk Calzone-new Broccoli Red Grapes Salad bar Milk
Table 10: Data for Menu Served on April 18
Nutrients Units Beef Stew/ 100g Beef Stew/ 1 cup serving Angel Biscuits /100g Angel Biscuits/ loz biscuit Apple Fresh/ 100g Apple Fresh/ 1/2 cup 1% White Milk/ 100g 1% Wh Milk/ 8oz or 227g Entire School Lunch Units
Adjusted weight gm 100 340.04 100 33.79 100 138 100 226 737.83 gm
Energy kcals 105.98 360.37 270.38 91.36 52 71.76 44.09 100 623.49 kcals
Protein O 6.21 21.11 6.56 2.22 0.26 0.36 3.53 8 31.69 O &
Vitamin A (RE) RE 132.19 449.49 0 0 5.0 6.9 44.09 100 556.39 RE
Vitamin A (IU) IU 659.62 2242.96 0 0 54.0 74.52 220.5 500 2817.4 8 IU
Vitamin C mg 2.72 9.25 0 0 4.6 6.35 1.06 2.4 18 mg
Iron mg 1.24 4.23 1.1 0.373 0.12 0.17 0 0 4.773 mg
Calcium mg 28.35 96.39 10.24 3.46 6.0 8.28 132.28 300 408.13 mg
Total Fat g 2.45 8.33 7.16 2.41 0.17 0.23 1.1 2.5 13.47 O C5
Saturated Fat g 0.72 2.45 1.93 0.65 0.03 0.04 0.66 1.5 4.64 o &
Trans Fat g 0.00 0 2.0 0.68 0.00 0 0 0 0.68 g
Carbohydrate g 14.52 49.38 44.96 15.19 13.81 19.06 5.29 12 95.63 g
Total Fiber g 1.82 6.17 0.78 0.27 2.4 3.31 0 0 9.75 §
Cholesterol mg 8.94 30.38 0.90 0.03 0.00 0 4.41 10 40.41 mg
Sodium mg 197.32 670.95 393.22 132.87 1.0 1.38 55.12 125 930.2 mg
81


Table 11: Data for Menu Served on April 19
Nutrients Units BBQ Chicken/ 100g BBQ Chicken/ 1 piece Whole Grain Dinner Roll/ 100g Whole Grain Dinner Roll/ 1 oz roll Light Peach/ 100g Light Peach/ V2 cup Baked Beans/ 100g Baked Beans/ 1 cup 1% White Milk/ 100g 1% White Milk/ 8oz carton or 227g Entire School Lunch Units
Adjusted weight gm 100 192.85 100 42.47 100 120.20 100 257.06 100 226 838.58 gm
Energy kcals 214.18 413.04 249.72 106.06 55.56 66.78 107.69 276.83 44.09 100 962.71 kcals
Protein g 15.86 30.58 6.3 2.68 0.00 0 4.62 11.86 3.53 8 53.12 O O
Vitamin A (RE) RE 9.30 17.94 0.0 0 47.62 57.24 0.00 0 44.09 100 175.18 RE
Vitamin A (IU) IU 46.51 89.69 0.0 0 238.10 286.19 0.00 0 220.5 500 875.88 IU
Vitamin C mg 0.74 1.41 0.01 0.00 0.95 1.14 0.00 0 1.06 2.4 4.95 mg
Iron mg 0.34 0.66 1.84 0.78 0.00 0 1.39 3.56 0 0 5 mg
Calcium mg 1.51 2.92 10.28 4.37 0 0 30.77 79.09 132.2 8 300 386.38 mg
Total Fat g 15.82 30.50 5.03 2.14 0 0 0.77 1.98 1.1 2.5 37.12 g
Sat. Fat g 4.99 9.62 0.98 0.41 0 0 0.00 0 0.66 1.5 11.53 g
Trans Fat g 0.00 0 0.00 0 0 0 0.00 0 0 0 0 O O
Carbo- hydrate g 1.60 3.08 44.85 19.05 13.49 16.22 22.31 57.34 5.29 12 107.69 g
Total Fiber g 0.07 0.14 0.51 0.22 0 0 3.85 9.89 0 0 10.25 g
Cholesterol mg 53.93 104.00 0.05 0.02 0 0 423.08 0 4.41 10 114.02 mg
Sodium mg 142.82 275.42 432.48 183.69 7.94 9.54 0.00 1087.6 55.12 125 1681.2 5 mg
82


Table 12: Data for Menu Served on April 20
Nutrients Units Big Daddy Cheese Pizza/ 100g Big Daddy Cheese Pizza/1 slice Canned Green Beans/ 100g Canned Green Beans/ Vi cup Red Grapes/ 100g Red Grapes/ Vi cup 1% Wli Milk/ 100g 1% White Milk/ 8oz carton or 227g Entire school lunch Units
Adjusted weight gm 100 135 100.00 128.85 100.00 90.72 100 226 580.57 gm
Energy kcals 237.04 320 16.67 21.48 67.00 60.78 44.09 100 502.26 kcals
Protein g 14.07 19 0.83 1.06 0.63 0.57 3.53 8 28.63 g
Vitamin A (RE) RE 55.90 41.67 53.69 10.00 9.07 44.09 100 162.76 RE
Vitamin A (IU) IU 279.51 208.33 268.43 100.00 90.72 220.5 500 859.15 IU
Vitamin C mg 0.00 0.83 1.08 4.00 3.63 1.06 2.4 7.11 mg
Iron mg 1.79 0.79 1.01 0.29 0.26 0 0 1.27 mg
Calcium mg 223.61 36.66 47.24 14.00 12.70 132.2 8 300 359.94 mg
Total Fat g 6.67 9 0.00 0 0.35 0.32 1.1 2.5 11.82 g
Saturated Fat g 2.59 3.5 0.00 0 0.11 0.10 0.66 1.5 5.1 g
Trans Fat g 0.00 0.00 0 0.00 0 0 0 0 g
Carbohydrate g 28.89 39 3.31 4.26 17.15 15.56 5.29 12 70.82 O
Total Fiber g 2.96 4 1.65 2.13 0.90 0.82 0 0 6.95 g
Cholesterol mg 11.11 15 0.00 0 0.00 0 4.41 10 25 mg
Sodium mg 362.96 490 322.31 415.30 2.00 1.81 55.12 125 1032.1 1 mg
Although 13 different nutrients are reported for every food served in the Aurora
Public School Cafeterias, not all 13 nutrients would be expected to vary greatly based on
GSW program participation. As a whole, students have few choices in the hot lunch line
and most of their choices are Go and Slow foods resulting from Auroras adoption of new
scratch cooking changes across the entire district. Students do, however, have the ability
to choose as many fruits and vegetables from the salad bar as they as they want. Given
83


that the GSW curriculum (incentive days, menu review, activities in class) focuses on
choosing Go foods with less processing, more whole grain, more whole foods (i.e.
more fruits and vegetables), and less added sugar, it is most realistic that students in
GSW schools would choose more fruits and vegetables, would choose milk (and white
milk instead of chocolate milk), and would consume less fat/saturated fat and more fiber
from eating more Go foods on their tray. For example, Chocolate Cake was served on
Day 2 of this study. GSW administrators would hypothesize that fewer students from
GSW schools would take cake (and therefore less total fat), and would take more fruits
and vegetables, thus increasing their intake of fiber. If they chose not to take or to take
but not eat all of the cake (average piece of cake across all schools=155 calories), those
students might have room for more white milk (average of 90 calories per carton for
1% or skim white milk) or more fruits and vegetables (varied caloric content). They
might have also chosen to take or drink more white milk regardless of cake consumption,
simply because white milk is recommended as a GO food and because an incentive day is
held each year where students earn a prize for drinking white milk. Therefore,
descriptive statistics were included for all total nutrient, milk, and fruit/vegetable
variables, but statistical models were run on only the following most relevant dependent
variables that GSW is expected to influence:
Did students choose white milk (skim and 1%) over chocolate milk or no milk?
For students choosing milk, did they choose white milk (skim and 1%) over
chocolate milk?
Percentage of white milk consumed.
Total number of fruits/vegetables taken.
84


Total fiber taken.

Total fiber consumed.
Total fat taken.
Total fat consumed.
Total saturated fat taken.
Total saturated fat consumed.
There is an important distinction between milk, fiber, fat, and saturated fat taken
and consumed. Students food choices in the lunch line may be influenced by GSW
curriculum but other factors may come into play once they begin to eat that might make
their actual food consumption lower on these key variables (white milk, high fiber, FV,
lower fat options). These factors, detailed in the literature review, include: length of the
lunch period, recess before lunch, peer role modeling, taste, smell, and teacher/staff role
modeling.
The two dependent variables for choosing white milk are dichotomous dependent
variables, whereas the remaining eight dependent variables are linear variables. Separate
statistical models were calculated for each of the 10 dependent variables listed above.
Because of the hierarchical nature of collecting data from students nested in schools,
multilevel modeling was appropriate for our analyses. Student responses to an
intervention are likely to be more correlated for students in the same school than for
students in different schools, thus violating the assumption of independent responses and
necessitating a statistical test other than traditional ANOVA and ordinary least-squares
regression. A 2-level multilevel regression model can generate more accurate standard
errors for program effect estimates and other important parameters in clustered designs
85


than other statistical techniques like linear regression that do not account for data
clustering. (Rabe-Hesketh & Skrondal, 2008; Raudenbush & Bryk, 2002) Multilevel
models have more error terms and therefore more flexibility in defining the covariance
structure. This flexibility leads to two distinct advantages: (1) more flexibility in the
questions researchers ask about the covariance structure and because of the complexity of
covariance structures able to be modeled, (2) better estimates of standard errors of the
regression coefficients and resulting accuracy of confidence intervals (Berkhof &
Kampen, 2004; Dedrick et al., 2009). Multilevel models also ensure more accurate effect
sizes for an intervention and help avoid Type 1 errors and biased parameter estimates
(Peugh, 2010; Wampold & Serlin, 2000). However, not all nested datasets require
multilevel modeling. Multilevel modeling is necessary only if response variables vary
across level-2 units (e.g. schools) (Peugh, 2010). In this study, response variable data
was likely to differ across schools (mostly due to school characteristics such as
demographics, achievement scores, and characteristics of PE teachers and other school
personnel), therefore multilevel modeling was selected for analyses. Variance in the
level-2 data was assessed by measuring intraclass correlation (ICC).
In this study, the ICC is both the proportion of nutrient/food values variation that
occurs across schools (level-2 units) and the expected correlation between the
nutrient/food values of two students (level-1 units) from the same school (Peugh, 2010).
The ICC is similar to the R effect size from regression and represents the proportion of
student nutrient score variance that can be explained by mean nutrient score differences
across schools. The ICC provides help in determining the necessity of multilevel
modeling. An ICC value of zero indicates: (a) no mean nutrient score variation across
86


schools (level-2), (b) all nutrient score variation occurs across students (i.e. level-1), and
(c) traditional analysis techniques such as ANOVA and regression can be used to analyze
the student data (Peugh, 2010). As the ICC value increases, the proportion of dependent
variable score variation that occurs across schools increases, resulting in violations of the
independence assumption and requiring multilevel modeling to analyze data.
Given that the ICC>0 in all models used in this study, a multilevel model that
separates variation between students and variation between schools was utilized to
determine if mean nutrient values differed significantly across schools. The most basic
multilevel model to explain the potential differences in nutrient/food scores is shown with
the following equations (Hox, 2010; Raudenbush & Bryk, 2002):
Equation 1: Level 1 : Yy = P Equation 2: Level 2 : p0j = yoo + uoj
Equation 1 shows that the differences in nutrient score of student / in school j (Yy)
can be modeled as a function of the mean nutrient score for school j (Poj) plus a residual
term that reflects individual student differences around the mean nutrient score of school
j (ry). Equation 2 shows that the mean nutrient score for school j (Poj) can be modeled as
a function of the grand-mean nutrient score (yoo) plus a school-specific deviation from the
grand mean (uoj). Substituting Equation 2 into Equation 1 yields the combined
unconditional multilevel modeling equation:
Combined multilevel modeling equation: Yy = y0o + u This equation combines nutrient score variability into within-group (i.e., level-1, ry) and
between-group (i.e., level-2, uoj) components.
87


After estimating a multilevel model, the ICC is calculated with the following:
ICC= t0o/(too + o2)
where Too is the variance in nutrient score means across schools and o is the
nutrient score variation across students within a school. An ICC predicts the proportion
of variance that occurs across schools, and typically an ICC value between .05 and .20 is
common in cross-sectional multilevel modeling applications in social research studies
(Peugh, 2010). In this research, ICC values ranged from 0.04-0.47.
It should be noted that a random effects multilevel linear regression was not
appropriate for one outcome variable. The percent of white milk consumed represented a
proportion variable bound on the upper and lower ends by zero and one. This outcome
variable was analyzed with a generalized linear model with logit link and a binomial
outcome ranging from zero to one and clustering on school ID. Although it did not
estimate the intraclass correlation, this model allowed for correct estimation of standard
errors and analysis of the overall effect of the intervention and student-level covariates.
To determine the extent to which GSW is associated with the outcome
independent of individual student characteristics, this study used multilevel modeling to
partition the variance between schools and student level characteristics for nine outcome
variables. Models were run with restricted maximum likelihood (REML) estimation
because of the ability of REML to attain more accurate variance estimates in smaller
sample sizes (Peugh, 2010; Raudenbush & Bryk, 2002). REML treats regression
coefficients as unknown quantities and estimates coefficients and variance estimates from
sample data, subtracting the appropriate amount of degrees of freedom. Conversely,
maximum likelihood estimation treats regression coefficients as known population
88


parameters and does not allocate any degrees of freedom to the estimation (Peugh, 2010).
The resulting variance estimates from maximum likelihood estimation are
underestimated, especially when the sample size is small. Brown and Draper (2000)
show that restricted maximum likelihood estimation provides reasonable variance
estimates with as few as 6-12 groups (Browne & Draper, 2000).
Because the effect of individual schools on overall results was unknown, schools
were treated as a random effect in our model. Student level characteristics (grade and
gender) were added as fixed effects to examine their independent contributions to food
and nutrient consumption. Treatment condition was added as both a random effect to
some models and a fixed effect to others and models were compared to determine the
best fitting model. Analyzing treatment condition (GSW or non-GSW) as a random
effect utilized a random-coefficient model where the effect of GSW was allowed to vary
across schools, producing a different coefficient for each school. Analyzing treatment
condition as a fixed effect utilized a random-intercept model where the effect of GSW
was reported as an intercept, or a mean effect for all schools. Testing whether the
random-coefficient model fit better than the nested random-intercept model was
accomplished in Stata 11 with a likelihood-ratio test of the two models (Rabe-Hesketh &
Skrondal, 2008). Of the nine dependent variables, none of the likelihood-ratio tests were
significant, indicating that the random-intercept model should not be rejected in favor of
the random-coefficient model. Therefore, treatment condition was treated as a fixed
effect in all models.
Model selections were confirmed by examining Akaikes Information Criterion
(AIC) and Schwartzs Bayesian Information Criterion (BIC) for each model (Akaike,
89


1974); Schwarz, 1978).
AIC = -2LL+2p
where LL is the log likelihood andp is the number of ML-estimated
parameters.
BIC=-2LL+p ln(N)
where N is the sample size at Level 1.
Both the AIC and BIC start with the log-likelihood value and subsequently penalize for
the number of covariance parameters estimated. The BIC creates a bigger penalty for the
number of parameters estimated (Dedrick et al., 2009). AIC and BIC for all 10 random
coefficient models and all 10 random intercept models are shown in Appendix G.
Results from likelihood ratio tests and AIC/BIC comparisons favor the random-
intercept models over random coefficient models, treating schools as a random effect and
treatment condition, grade, gender, and day of the study/entree as fixed effects. The final
models in this study contain both fixed and random effects and are considered mixed-
effects models.
Other post-estimation commands were used in this study to show the superiority
of multilevel modeling over standard regression techniques in the final models. Using
Stata 11, likelihood-ratio (LR) tests for all ten final multilevel models compared
multilevel modeling (Linear and Logistic Mixed Effects Models) to standard linear and
logistic regression. The LR test is testing whether an estimated variance component is
different from zero, and whether or not multilevel modeling is a better fitting model than
standard linear or logistic regression. All ten likelihood-ratio tests were significant at
p=.001, indicating that multilevel modeling was superior to standard linear and logistic
90


regression for all models in this study. ICC and likelihood ratio tests were utilized to
determine that multilevel modeling was the most appropriate modeling technique for
collected data.
Several assumptions should be met for multilevel modeling. The normality
assumption was assessed in all models by examining the distribution of Level 1 residuals.
Deviance residuals were calculated for Level 1 residuals in the two white milk logit
models as recommended by previous studies (McCullagh & Nelder, 1989). Deviance
residuals have the best properties for examining goodness of fit. Residual graphs for all
multilevel models are found in Appendix H.
Data were screened for outliers to identify data-entry errors, inaccurate coding of
a student, missing values coded incorrectly, or simply individuals who are different from
others students in the sample.
Formulae for Multi-Level Model.
*Note: Outcome variable is listed as Total Fat Grams Taken in the equations below.
Separate multilevel models were created for all outcome variables (9 in total).
Model 1: Y (Total Fat Grams TAKEN) = a (intercept) + b|X, (GSW participation-Level2 Main
Effect) + b2X2 (Gender-Level 1) + b3X3 (Grade-Level 1) + b4X4 (Day of Data Coll.-Level 1) +e
Model 2: Y (Total Fat Grams CONSUMED) = a (intercept) + biXj (GSW participation-Level 2 Main
Effect) + b2X2 (Gender-Level 1) + b3X3 (Grade-Level 1) b4X4 (Day of Data Coll.-Level 1) +e
Table 13: Multi-level Model for Control and Main Effect Predictor Variables
Predictor Variables (Main Effects)
GSW Participation
Level 1-Control Variables
Gender
Grade Level
Day 1, 2, or 3
Level 2-Control Variables
None school sample size too small
91


Adding in additional predictor variables at the school level (i.e. mean school SES,
ELL, etc.) was not possible due to the very small number of schools (10) included in this
study.
Qualitative Data Analysis: Aim 2
Using a combination of deductive and inductive methods, salient themes were
identified related to teacher opinions regarding Scratch Cooking Cafeteria Changes and
GSW implementation and effectiveness. Teacher interviews were analyzed using codes
emerging from literature/hypotheses and in vivo codes emerging from collected data, and
subsequently determining themes within and across schools.
A qualitative analysis was used with a five stage iterative process to analyze the
transcripts: (1) development of a coding schedule; (2) coding of the data; (3) description
of the main categories; (4) linking of categories into major themes; and (5) the
development of explanations for the relationships between themes (Hsieh & Shannon,
2005; Pope, Ziebland, & Mays, 2000). An ecological framework for program
implementation was utilized as a lens through which to analyze the implementation and
impact of GSW (Durlak & DuPre, 2008). Interview transcripts were analyzed within the
context of three main categories of this ecological framework: innovations/program
characteristics, individuals and communities, and features associated with program
delivery and program support.
For all interviews, codes were generated and recorded in a table. This table also
includes excerpts from teacher comments relating to each code. The resulting codes and
comments were grouped, as appropriate, into overall themes and findings from
92


qualitative data collection. These qualitative data and analyses assisted in several ways
including: (1) helping to make sense of quantitative data; and (2) documenting dose,
implementation fidelity, likability, and impact of the GSW program in APS.
GSW implementation (dose, quality, adherence/fidelity) was also assessed using
informal conversations with cafeteria staff and school personnel and observations by
researchers while in schools. The implementation of GSW programming was assessed
by coding transcribed interviews, organizing them into themes, and drawing conclusions
from these themes. Implementation of Scratch Cooking Cafeteria Changes was also
assessed using observations by researchers during data collection days. The
implementation of Scratch Cooking Cafeteria Changes should be similar across schools,
as all schools follow the same district-wide menus and food ordering is done in a central
location for all schools. However, given the variability that could occur due to shipping
of food, leftover food, and cafeteria worker buy-in to Scratch Cooking Cafeteria
Changes, there could be variability in implementation. Data from collected field notes
assessed menu similarity across schools.
Qualitative data helped to inform the interpretation of quantitative analyses.
Linking qualitative data to quantitative data has several benefits: (a) to enable
confirmation or corroboration via comparison; (b) to elaborate or develop analysis,
providing richer detail; and (c) to initiate new lines of thinking through attention to
surprises or paradoxes (Rossman & Wilson, 1994). Analyzing qualitative interviews and
observations and quantitative plate waste data together provided far greater information
than either method of data collection alone.
93


Human Subjects Review
Because this study used original data collected with children in Aurora Public
schools, proper safeguards were taken to ensure the safety of all study participants.
Approval from the Colorado Multiple Institutional Review Board (COMIRB), LiveWell
Colorado, and Aurora Public Schools was obtained before data collection began.
Students were asked for verbal informed consent in the lunch line. Student names were
not collected, and faces were not photographed.
94


CHAPTER IV
QUANTITATIVE RESULTS
Descriptive Statistics
Of 2,223 collected and photographed lunch trays, 1123 (51%) came from GSW
schools and 1100 (49%) came from non-GSW comparison schools. The sample was
split fairly evenly among 1st graders, 3rd graders, and 5th graders and among male and
female students. Among the entire sample of all schools over three days, the average
number of calories taken in the hot lunch line was 553 (s.d. 134), whereas the average
number of calories consumed by the end of lunch was 292 (s.d. 154). The characteristics
of lunch trays collected from intervention and comparison groups are shown in Table 14.
95


Full Text

PAGE 1

NUTRITION EDUCATION AND SCRATCH COOKING CHANGES IN SCHOOLS: A MIXED METHODS STUDY OF INTERVENTIONS IN AURORA PUBLIC SCHOOLS by DEBRA CAROL GUENTHER B.A., Wake Forest University, 1996 M.S., University of Maryland, 1999 A thesis submitted to the Faculty of the Graduate School of the University of Colorado i n partial fulfillment of the requirements for the degree of Doctor of Philosophy Health and Behavioral Sciences 2013

PAGE 2

ii This thesis for the Doctor of Philosophy degree by Debra Carol Guenther has been approved for the Health and Behavioral Sciences Program b y Patrick Krueger, Chair Deborah Main, Advisor John Brett Bryan Wee April 17, 2013

PAGE 3

iii Guenther, Debra Carol (Ph.D., Health and Behavioral Sciences) Nutrition Education and Scratch Cooking Changes in Schools : A Mixed Methods Study of Interventions in Aurora Public Schools Thesis directed by Professor Deborah S. Main. ABSTRACT This paper examines the impact of a nutrition education intervention on food choice and consumption in elementary cafeterias. After a school district changed 42 lunch and breakfast entre recipes t o include more whole grains, fresh produce and fresh meats, this study examined st udent food choice and food consumption in five schools with a nutrition education intervention and five schools without. The nutrition education intervention, focusin g on a 3 level rating system to teach children about healthful food (Go, Slow, and Whoa) included an assembly, classroom instruction, point of sale cafeteria labeling, parent breakfast, and nutrition related newsletter articles. Intervention schools were matched with comparison schools using propensity score matching. Student food choice an d consumption was measured with digit al plate waste methodology with a final sample si ze of 22 23 lunch trays Students enrolled in intervention schools showed significantly increased odds of choosing white milk (OR=2.17, p=0.05) over chocolate milk or no milk. None of the other healthy eating outcome models produced significant differences between intervention and comparison schools when controlling for gender, grade, and entre items Qualitative interviews with physical education teachers responsible f or program implementation revealed low dose and fidelity of implementation for a variety of reasons. Additional teacher support clarification of goals

PAGE 4

iv and program components, and technical assistance is advised to improve program implementation. The fo rm and content of this abstract are approved. I recommend its publication. Approved: Deborah S. Main

PAGE 5

v TABLE OF CONTENTS I. BACKGROUND AND SIGNIFICANCE ................................ ............................... 1 II. REVIEW OF THE LITERATURE ................................ ................................ ....... 2 2 III. METHODS ................................ ................................ ................................ ............ 4 9 IV. QUANTITATIVE RESULTS ................................ ................................ ............... 9 5 V. QUALITATIVE RESULTS ................................ ................................ ................ 10 8 VI. DISCUSSION ................................ ................................ ................................ ...... 1 2 0 REFERENCES ................................ ................................ ................................ ................ 132 APPENDIX A. Aurora Public School Sample Menu December 2011 ................................ ............... 153 B. Go Slow Whoa Rating System ................................ ................................ .................. 154 C. Matching Data ................................ ................................ ................................ ............ 160 D. Qualitative Semi Structured Interview Guide ................................ ........................... 162 E. Field Notes ................................ ................................ ................................ ................. 163 F. Interrater Reliability Kappa Coefficients ................................ ................................ ... 164 G. AIC and BIC for Random Intercept and Random Coefficient Models ..................... 167 H. Residual Graphs for Final Models ................................ ................................ ............. 168

PAGE 6

1 CHAPTER I BACKGROUND AND SIGNIFICANCE Background The Health Problem Childhood obesity is a major public health concern in the United States as rates of overweight and obese children continue to rise. Since 1970, the number of obese children ages 6 11 has quadrupled and the number of obese adolescents ages 12 19 has triple d (Hedley et al., 2004; Levi et al., 2010) This exponential increase in childhood obesity has implications for current and future health of children as they become adults. The rise in childhood obesity rates may also lead to a reduction in life expectancy (Fontaine, Redden, Wang, Wes tfall, & Allison, 2003) Public and private schools nationwide are targeting physical activity within school hours to increase energy expenditure among students. This research specifically targeted dietary intake as many schools are also changing the n utrition environments to target knowledge, attitudes, and dietary choices among students. Given that disability and death are linked to poor dietary habits and obesity, improving nutritional intake of children as well as adults could have a significant im pact on health and mortality in the United States. Childhood obesity has immediate effects on child health. An estimated 61 percent of overweight young people (ages 5 10) have at least one additional risk factor for heart disease (i.e. high blood pressur e, high cholesterol) (Freedman, Dietz, Srinivasan, & Berenson, 1999) Over the past several decades, type II diabetes has risen significantly in children (Vivian, 2006) Obese children and adolescents are at higher risk for a plethora of health issues including bone and joint problems, sleep apnea, menstrual abnormalities,

PAGE 7

2 and social/psychological problems such as stig matization and poor self esteem (Daniels, 2006; Daniels et al., 2005; Koplan, Liverman, & Kr aak, 2005) There is also evidence that poor nutrition and physical inactivity leading to increased obesity and health problems are linked to poor test scores, concentr ation, and academic achievement (Kristjansson, Sigfusdottir, & Allegrante, 2010; Taras & Potts Datema, 2005) Obese children are at increased risk of becoming obese adults (Serdula et al., 1993; A. S. Singh, Mulder, Twisk, van Mechelen, & Chinapaw, 2008) According to a meta analysis conducted by Serdula and colleagues (1993) 42% to 63% of obese adults were obese as children and higher levels of obesity (BMI) in childhood were linked to greater risk of adult obesity. The risk of overweight children becoming overweight adults is at least twice as high compared with normal weight children and the persistence of overweight is gr eater with increasing levels of overweight (A. S. Singh et al., 2008) dietary behavior could lead to improved dietary habits throu ghout life, decreased child and adult obesity rates, and avoidance of health risk in the long run. Body Mass Index (BMI) is a commonly used metric to classify obesity, expressed as weight in kilograms divided by height in meters squared (kg/m 2 ). Child o besity and overweight criteria are based on the 2000 CDC BMI for age growth charts for the U.S., and classify any child with a BMI at or above the 95 th percentile of the sex specific BMI growth charts as obese, in the 85 95 th percentile as overweight. Th is is a change from previous years, where children at the 95 th percentile and above were overweight, and kids in the 85 th 95 th percentile s (Barlow & Expert, 2007; Krebs et al., 2007) In 2007 2008, 16.9% of children and adolescents in the United States were

PAGE 8

3 at or above the 95 th percentile and 31.7% were at or above the 85 th percentile of BMI for age. Narrowing the c ategory down to 6 through 11 year olds reveals that 19.6% are above the 95 th percentile of BMI for age (Ogden, Carroll, Curtin, Lamb, & Flegal, 2010) Childhood overweight and obesity rates in Colorado are problematic. Colorado is the 23 rd leanest state for children, but has the second fastest rate of childhood obesity (Levi et al., 2010) In 2009, the prevalence of overweight (85 th to 94 th percentile for BMI) among children ages 2 14 years was 13.2 percent and obesity prevalence (95 th percentile and above) was 13.3 percent (COPAN, 2009) higher than the Healthy People 2010 objective of five percent for obesity Well documented research highlights significant racial/ethnic and socioeconomic disparities in childhood obesity in the United States (Ogden et al., 2010; G. K. Singh, Kogan, Van Dyck, & Siahpush, 2008; G. K. Singh, Siahpush, & Kogan, 2010; Wang & Beydoun, 2007) In 2007 2008, Hispanic boys had significantly higher odds of having high BMI at three BMI cut points (97 th perce ntile, 95 th percentile, and 85 th percentile) when compared to non Hispanic W hite bo ys. Among girls, non Hispanic B lack girls were significantly more likely than non Hispanic W hite girls to have high BMI at the three BMI cut points. Singh and colleagues ( 2010 b ) highlighted higher prevalence of obesity in 2007 for Black, Hispanic, and American Indian children (over 23%) compared to White children (12%) and higher prevalence of overweight for Hispanic and Black children (over 41%) compared to White children (27%). Similar disparities exist in the state of Colorado. The obesity prevalence among non Hispanic White children (10.9%) was significantly lower than for Hispanic children (21.2%) and

PAGE 9

4 non Hispanic Black children (23.4%) between 2006 2008 (COPAN, 2009) The racial/ethnic breakdown in the current study population reveals more non white students than white students. Averaged across 33 elementary schools participating in this research 58% of the students were Hispanic, 17% wer e Black, 17% were White, 1% were Native American, 4% were Asian, and 3% identified as two or more races/ethnicities. The nutrition intervention evaluated in this study targeted a district with high racial/ethnic diversity, and was therefore poised to make a significant contribution to the dietary habits and obesity rat es of a racially diverse population at high risk for childhood obesity. The medical costs of obesity rise dramatically each year in the United States. The estimated annual medical costs of obesity increased from $74 billion in 1998 to $147 billion in 2008 (Finkelstein, T rogdon, Cohen, & Dietz, 2009) After combining costs for increased medical care, loss of worker productivity due to death, loss of productivity due to illness and disability of active workers, and loss of productivity due to total disability, the overa ll annual costs associated with obesity are likely much higher than those estimated costs According to analyses by Finkelstein et al. (2008), ac ross all payers (normal payer, Medicare, and M edicaid), obese patients had per capita medical spending that wa s $1,429 (42%) greater than spending for normal weight people in 2006. Medicaid and Medicare paid a n estimated 42 percent of obesity related medical expenditures in 2006. Childhood obesity, in 1998, was related to excess medical expenditures totaling $12 4 million (Johnson, McInnes, & Shinogle, 2006) These estimates of childhood and adult obesity expenditures are likely underestimates of the actual economic costs if other downstream costs are considered.

PAGE 10

5 Social determinants of health are important predictors of childhood obesity. The social determinants of health are the circumstances in which people are born, grow up, live, work and age. These conditions are influenced by the distribution of money, r esources, and power at the global, national and local levels (World Health Organization, 2012) The inequitable distribution of resources at the local level can be further divided into opp ortunities and resources available within neighborhoods and schools. Social determinants affect childhood obesity risk through access to fresh fruit and vegetables, safe environments, and income levels that afford opportunities for physical activity, heal thcare and education. Children participating in the National School Lunch Program (NSLP) or School Breakfast Program (SBP), children in lower SES households, and children attending public schools all have a higher risk of being overweight or obese (Li & Hooker, 2010) thus linking social determinants of health, and specifically poverty, to childhood obesity. Many of the students in our nation and in the specific population targeted in this study fit the profile for social determinants of health that make healthy eating and active living difficult rather than easy. Physical inactivity, dietary intake, genetic predisposition, illness, social determinants, and other factors contribute to childhood obesity. Given the myriad of negative social, economic, and health outcomes associated with childhood obesity, public hea lth officials are searching for new, innovative and effective interventions. The NSLP and SBP are two national programs aimed at providing healthy food to children from low income families during the school day. Social determinants of health that put chil dren at higher risk of obesity could be offset by positive environmental and nutrition interventions offered within schoo ls. Given that children spend seven to eight hours each

PAGE 11

6 day in school, school based interventions and policies could positively affect the nutrition in part by child and family factors such as food preferences, monetary resources, choices, and eating behaviors, it is also a simple function of what f oods are accessible to children and what knowledge they have about those foods; therefore school based childhood nutrition and nutrition education are two logical intervention targets for improving dietary intake and preventing obesity. This study will ex amine two specific childhood nutrition interventions delivered in elementary schools: School Lunches and the Go, Slow, Whoa nutrition education program. The National School Lunch Program (NSLP) In 1932, the federal government began providing aid for school lunch programs from agencies such as the Federal Emergency Relief Administration, the Reconstruction Finance Corporation, and the Civil Works Administration. Soon thereafter, funding increased for labor in school lunchrooms and donations of surplus farm commodities to school lunch programs. The National School Lunch Program (NSLP) became permanent with the passage of the National School Lunch Act in 1946 (Congress, 1946) The act stipula ted a formula for giving cash from the federal government to the states based on per capita income and population. This cash was, and still is, doled out across the state for school lunch programs, as long as requirements for school lunch contents (fat, s aturated fat, total calories, iron, calcium, etc.) were followed (Hinrichs, 2010) The NSLP, subsidized by the federal government, helps to provide healthy food and adequate calories to students during the school day. Section two of the 1946 act (Congress, 1946) reads:

PAGE 12

7 It is hereby declared to be the policy of Congress, as a measure of national security, to safeguard the health and well encourage the domestic consumption of nutritious agricultural commodities and other food, by assisting the States, through grants in aid and other means; in provi ding an adequate supply of foods and other facilities for the establishment, maintenance, and expansion of non profit school lunch programs. In the 1990s, policymakers discovered that many school lunches failed to meet nutrition requirements. In 1995, th third of the child recommended daily allowance (RDA) of calories, protein, calcium, iron, and vitamins A and C and limit fat and saturated fat content (to 30% and 10% respectively) of total calories in order to receive reimbursement (Schanzenbach, 2009) According to the USDA Food and Nutrition Service (U.S. Department of Agri culture, October, 2011) : 1. The NSLP is a federally assisted meal program operating in over 101,000 public and non profit private schools and residential childcare institutions. 2. The Food and Nutrition Service administers the program at the Federal level. At the State level, the National School Lunch Program is usually administered by State education agencies. 3. School lunches must meet Federal nutrition requirements but decisions about specific ingredients/foods to serve, recipes, and food preparation method s will be made by local school food authorities. 4. As of July 1, 2011, families are eligible for reduced price school meals if their income is between 130% and 185% of the federal poverty level, for which students can be charged no more than 40 cents. Famil ies are eligible for free meals at 130% of the federal poverty level. To put these guidelines into

PAGE 13

8 perspective, a family of four in the 48 contiguous states or the District of Columbia would be eligible for reduced price school meals with an an nual family income of $41,348 and free meals with an annual family income of $29,055. 5. Even full priced meals are subsidized to some extent. Local school food authorities set prices for full price meals, but must provide meals without profit. 6. Schools with less than 60% student eligibility for free and reduced price lunch receive cash reimbursement at the following rates: $2.77 for free lunches, $2.37 for reduced price lunches, and $0.26 for paid lunches. Schools with more than 60% student eligibility for free and r educed price lunch receive the following reimbursements: $2.78 for free lunches, $2.39 for reduced price lunches, and $0.28 for paid lunches. 7. By the end of the first year of the NSLP, 1946 1947, about 7.1 million children were participating in the program By 1970, 22 million children were participating. In FY 2010, more than 31 million children participated in the NSLP. Since the program officially began in 1946, over 219 billion lunches have been served. 8. The cost of the NSLP totaled $10.8 billion in F Y 2010. Food consumed by children during the school day has a significant impact on both health and educational outcomes. Previous research shows that the NSLP improves educational attainment of students eating school lunch (Hinrichs, 2010) and increases vitamin and mineral consumption as compared to non participants in NSLP (Gleason & Suitor, 2001; Gleason & Suitor, 2003) These positive e ffects of NSLP are promising given that the NSLP serves approximately 60 percent of the total U.S. student population

PAGE 14

9 daily (Schanzenbach, 2009) In 2010, 31.5 million kids participated in the NSLP: 17.6 million free lunches (56 percent of participants), 3.0 million reduced price lunches (10 percent of participants), and 11.1 million full price lunches (3 4 percent of participants) (U.S. Department of Agriculture, 2011) Funds s upporting the NSLP have a strong impact on what is offered in schools. Schools must follow the nutritional guidelines of the NSLP in order to receive continued funds for school lunch programs. However, decisions about the specific recipes, food items, an d menu choices available in each school district are made at the local level. The appeal and likability of school lunch items are functions of ingredients and taste, not just of the nutritional breakdown. A spinach and egg quiche with low fat cheese and whole wheat crust could have the same nutritional breakdown as a bean burrito with fresh tomatoes, brown rice, low fat cheese, and a whole wheat tortilla. However, differences in culture, race/ethnicity, past experience with similar foods, and taste/textu re might make one recipe much more acceptable by the student population. Therefore, the nutritional guidelines of the NSLP combined with the knowledge and creativity of the local nutrition services personnel will determine the taste, likability, and consu mption of school lunches, and their impact on child health and obesity. According to the School Nutrition Dietary Assessment Study (SNDA III), over 85% of schools in the United States prepared lunches that met the standards set forth by the NSLP for prot ein, Vitamin A, vitamin C, calcium, and iron. However, fewer than one third of public schools served school lunches with less than 30 percent of calories from fat or less than 10 percent of calories from saturated fat (Crepinsek, Gordon, McKinney, Condon, & Wilson, 2009) Even though schools might meet national

PAGE 15

10 standards for foods served the actual foods and quantities children consume from what is served is most important. However, this is seldom documented due to required time and costs of plate waste studies to determine consumption. The amount of food consumed at lunch could be inf luenced by the scheduling of recess before or after lunch (Bergman, Bue rgel, Englund, Clem et al., 2004) the length of the lunch period, peer food consumption and role modeling, and food preferences (Bergman, Buergel, E nglund, & Femrite, 2004a) The estimated amount of calories wasted from school lunches in the NSLP varies widely from 12% (Devaney, Gordon, & Burghardt, 1995) to 40% (Bergman, Buergel, Englund, & Femrite, 2004b) This study examine d lunches in a local school district that m eet all criteria set by the NSLP by using plate waste methods to determine foods and food quantities consu med during school lunch. By utilizing quantitative plate waste methods to identify the percentage of food thrown away from school lunche s, the curren t study fill s an important gap in understanding eating habits by using qua ntitative methods to u School Nutrition and Nutrition Education Advocacy and Intervention In addition to the NSLP, other local and nat ional efforts are tackling the obesity problem The U.S. Child Nutrition and WIC Reauthorization Act of 2004 ("Child Nutrition and WIC Reauthorization Act," 2004) required that school districts in the U.S. implement a Loca l Wellness Policy by July 1, 2006. The goals of these wellness po licies nutritious food consumption and physical activity expenditure in order to reduce the U.S. obesity epidemic. This reauthorization act is renewed every 5 year s to reflect changing demands for all school meal programs, Child and Adult Care Food Programs, and WIC.

PAGE 16

11 On December 13, 2010, President Obama signed the Healthy, Hunger Free Kids Act reauthorizing the Child Nutrition Act governing the National School Lunc h Program ("Healthy, Hunger Free Kids Act of 2010," 2010) This bill includes health promoting school food policies such as: 1. Increasing school meal reimbursement for schools by six cents per meal; 2. Setting improved nutrition stan dards for school meals; 3. Setting policies for vending machines, school stores, etc.; 4. Simplifying the process for gaining access to free meals; 5. Piloting expansion of Farm to S chool programs as well as organic foods Based on their potential to impact childhood ob esity and health, nutrition and nutrition education policies and interventions are an important part of many non profit organizations and federal programs. The American Dietetic Association, the Society for Nutrition Education, and the America n School Food Service Association are heavily involved in nutrition programs and interventions in schools and reco m mend that through grade twelve students. These nutrition services shall be integrated with a coordinated, comprehensive school health program and implemented through a school (Briggs et al., 2003) The School Nutrition Association established the School Nutrition Specialist (SNS) Credentialing Program in 1997 to help school nutrition professionals enhance their skills and elevate professional standards. This national organization also developed a tool, Keys to Excellence, to help dietetics professionals use a best practices framework for conti nuous program review, evaluation, and improvement (School Nutrition Association, n.d.) The Centers for Disease Control and Prevention and

PAGE 17

12 the National Association of State Boards of Education (Wechsler, McKenna, Lee, & Dietz, 2004) The Farm to School national program helps schools procure fresh produce and use farm products in school lunches ( Benefits of farm to school projects healthy eating and physical activity for school children: field hearing before the Committee on Agriculture, Nutrition, and Forestry, United States Senate. 2009) Other national organizations taking part in s chool nutrition and school policy interventions include: the American Association of Family Physicians (AAFP), National Alliance for Nutrition and campaign, The Food Trust, th e National School Boards Association (NSBA), the American Heart Association (AHA), the Coordinated School Health Program, and the Food Research and Action Center (FRAC). Within Colorado, numerous organizations have child nutrition, nutrition education and obesity prevention as part of their policy agendas, funding priorities, and advocacy issues. These organizations include: the Colorado School Nutrition Association, Colorado Department of Education, Colorado Foundation, Colorado Legacy Foundation, Colorado Association of School Executives, Anschutz Health and Wellness Center, CU Denver, Colorado Health Foundation, Rocky Mountain Center for Health Promotion and Education, LiveW ell Colorado and more. Multiple agencies and organizations are collaborating to strengthen policy and advocacy initiatives aimed at school nutrition and nutrition education. This research evaluated two such initiatives involv ing scratch/healthy cooking and the Go, Slow, Whoa program.

PAGE 18

13 LiveWell Colorado Origins LiveWell Colorado, a nonprofit organization committed to reducing obesity in Colorado, was established as a grant making collaborative in 2007 in partnership with the Colorado Department of Public Health and Environment and with initial funding from The Colorado H ealth Foundation, Kaiser Permanente, and the Kresge Foundation. LiveWell Colorado became a 501(c)(3) in 2009 and focuses on policy, environmental and personal lifestyle changes that remove barriers to and increase healthy behaviors. LiveWell Colorado is i mplementing a five year strategic plan that focuses on the following areas: 1. Funding community coalitions throughout the state focused on healthy eating and active living strategies. 2. Informing and advancing multi sector policy efforts with key stakeholders at the local, state and national levels. 3. Leading social marketing initiatives that inspire a culture shift and motivate sustainable healthy behavior change. T his research focused on two specific programs of LiveWell Colorado: Culinary Boot Camps and the Go Slow, Whoa Nutrition Education Program Both programs LiveWell @ School initiative Nearly 400,000 Colorado children participate in the National School Lunch Program, with about 40 percent qualifying for free or reduced lunch. Therefore, working towards improving the healthfu lness of school food and delivering nutrition education in schools will have a significant impact on a large number of Colorado children. Although the overall LiveWell Colorado program is entitled Culinary Boot Camps this research examined

PAGE 19

14 how only one distr ict, Aurora Public Schools, implemented scratch cooking cafeteria changes following participation in the Culinary Boot Camp program. Scratch cooking changes include cooking with more whole grai ns, more fresh fruits and vegetables, and less processed, frozen, and canned ingredients. Therefore, this research did not necessarily evaluate the impact of the Culinary Boot Cam p program, but instead evaluated the impact of specific scratch cooking cafe teria changes inspired in this one district by the Culinary Boot Camp pr ogram. Thus, this research refer s to these district Go, Slow, Whoa is a nutrition education program adopted and modified from na tional health education curriculum for use in Aurora Public Schools. Culinary Boot Camp Inspired Scratch Cooking Cafeteria Changes: As of the beginning of this research (August 2011) 63 Colorado school districts and 280 school food s ervice managers and wo rkers had participated in 5 day culinary boot camp s. In Aurora Public Schools, 53 school cafeteria managers and central office leadership staff attended the LiveWell Colorado and Colorado Health Foundation sponsored Cook for America Culinary School Food Boot Camp which represents 16% of the total Nutrition Services Staff (kitchen managers, kitchen staff, and central office leadership and staff). This Boot Camp is taught by experienced chefs and promotes the preparation of fresh food and scratch cooking in school cafeterias at the same or lower costs than schools currently pay for processed, ready to heat and serve foods. The goal of Scratch Cooking Cafeteria Changes in Auro ra Public Schools is to increase healthful nutrients (i.e. fiber, calcium, vitamins A and C) and decrease un healthful nutrients (i.e. total fat, saturated fat, sodium, and cholesterol) in school lunch items while maintaining

PAGE 20

15 or increasing attractiveness, t aste, and desirability. The Scratch Cooking Cafeteria Changes reflect multiple contributing strategies: using more healthful ingredients (i.e. whole grains, fresh fruits and vegetables, non processed ingredients), creating and using new recipes with those healthful ingredients, and utilizing new techniques and kitchen equipment to prepare the recipes. Go, Slow, Whoa Go, Slow, Whoa (GSW) is a nutrition education program promoted by the LiveWell Colorado LiveWell @ School initiative GSW aims to improve stu dent and school community awareness of healthful healthful food choices and ultimately impact childhood obesity rates. GSW teaches children and their families about healthful foods using a green light (Go), yellow light (Slow), r ed light (Whoa) visual aid to symbolize foods that should be eaten frequently, sometimes, and seldom. LiveWell Colorado has developed a business plan to scale up this intervention and offer it to additional school districts throughout Colorado. Given the paucity of data on program implementation and effectiveness, this research will make a significant contributio n to program justification and program improvement. Why Aurora? Data collected in this school district demonstrates the need for addressing the obesity and BMI issues of elementary students. Aurora Public Schools (APS), the target school distric t for this study, includes two different counti es: Arapahoe and Adams Rese arch shows that across the two counties, an average of 13.6% of children are obese, and an additional 12.4% are overweight (COPAN, 2009) These rates are significantly higher than t he Healthy People 2010 goal of five percent. I n APS specifically, childhood

PAGE 21

16 rates are much higher than the county average s APS contains a very ethnically diverse and socioeconomically challenged student population, increasing the risk for childhood obesity. Student data collected in APS physical education classes show increasing BMIs as stud ents advance by grade level This points to the critical need to prevent further weight gain at the elementary school level. Given that overweight and obesity rates in APS are generally higher than the nation al and Colorado rates, this school district is an excellent target for healthy eating education, promotion, and programming ( C. Fenton, personal communication, February 7, 2012) Many chronic diseases are related to individual and social patterns of behav ior (food consumption included), thus interventions, policies, and programming targeting individual and comm unity behavior could result in risk reduction for obesity and related diseases. This district could benefit from successful interventions to change the nutrition environment and improve student nutrition knowledge. Significance Scratch Cooking Cafeteria Changes and the Go, Slow, Whoa Program have potential to improve the nutritional intake of children in Aurora Public Schools. knowledge and attitudes regarding healthy food, together, may impact nutrient consumption. Determining whether the joint effects of changing the food environment and providing nutrition educa tion are greater than changing the food environment alone is an important question. Answering this question is key for educators and policy makers alike to implement the most effective and efficient interventions.

PAGE 22

17 Objective, Rationale, and Purpose This research Scratch Cooking Initiatives and Nutrition Education in Public Elementary Schools, evaluate d Scratch Cooking Cafeteria Changes and nutrition education implementation in Aurora Public Schools to determine their effect on cafeteria consumption of healthy foods. Students partic ipating in the GSW program have opportunities to gain knowledge and skills to improve their choices and consumption of healthy cafeteria food. Aurora Public Schools implemented many scratch cooking changes beginning in Au gust 2011 and this research assessed the consumption of menu items and ascertain ed differences between students in schools with GSW and schools without GSW. Results of this research inform program justification and program improvement for both the GSW p rogram and Scratch Cooking Cafeteria Changes. Overview of Theoretical and Conceptual Framework The rationale for and potential impact of school interventions addressed in this research are informed by Social Cognitive Theory (SCT) and Social Ecological Mod els (SEM) (Bandura, 1986; McLeroy, Bibeau, Steckler, & Glanz, 1988) Historically, many nutrition e ducation interventions have utilized Social Cognitive Theories to guide programming targeting individual, behavioral, and environmental factors (Bandura, 2004) Alth ough changing knowledge is an important part of nutrition education, SCT further utilizes skill building, role modeling, self efficacy, and environmental change to reach behavior change goals. The Go, Slow, Whoa program utilizes these SCT constructs to impact student nutrition knowled ge and behavior. Social Ecological Models offer an important theoretical framework for this research by acknowledging factors outside an individual that may influence behavior.

PAGE 23

18 Ecological models recognize that behavior and health outcomes are the result of more than just individual motives, and explore the interaction between an individual and the environment (Contento, 2011) The Scratch Cooking Cafeteria Changes seek to improve the food environment that students encounter on a daily basis. The Go, Slow, Whoa program works within the school nutrition environment to discuss, promote, and reinforce healthy selections from the cafeteria menu choices. Gaps Filled This research addressed several gaps in the cu rrent academic literature related to scratch cooking initiatives and nutrition education evaluation Scratch Cooking school food initiatives have not been systematical ly evaluated in a K 12 public school context, although there is a related body of litera ture that examines taste preferences and consumption of fruits, vegetables, and whole grains. This research examine d student responses to large scale menu changes offering more fresh, whole grain, unrefined menu items to students in the NSLP. Plate waste studies following healthful food additions and changes are very uncommon due to the expense and time required for plate waste studies. Nutrition education programs are oftentimes evaluated with pre post knowledge questionnaires, 24 hour recall surveys, food journals, or food preference surveys. This research used a robust and reliable method of determining what kids actually eat by collecting data on their food consumption in the cafeteria. Digital photography plate waste protocols are in their infancy yet show promise in collecting important and valid data about food consumption. This research further examine d this method as a tool for effectively measuring school food consumption.

PAGE 24

19 Lastly, unlike many studies of food consumption using quant itative da ta, this research also include d a qualitative component to explore teacher perceptions of the intervention and program implementation and fidelity issues This study utilized qualitative methods to understand the impact and significance of various compone nts of the Go, Slow, Whoa program and to help explain the quantitative plate waste findings. Research Question How influential is the Go, Slow, Whoa nutrition education program combined with Scratch Cooking Cafeteria Changes over and above the impact of Scratch Cooking choices and dietary intake related to healthy foo ds? Specific Aims and Hypotheses Aim 1 Determine the relative impact of GSW and Scratch Cooking Cafeteria Changes choices and consumption of school lunch. Hypothesis 1 a Students receiving the GSW program will choose more healthy foods at school than non GSW students Hypothesis 1 b Students receiving the GSW program will consume more of the healthy foods on their school lunch tray than non GSW students.

PAGE 25

20 Hypothesis 1 c Students receiving the GSW program will choose and consume more overall fruits and vegetables, white milk, and fiber, and less fat and saturated fat than non GSW students. Aim 2 Determine the relative dose, quality and adherence/fidelity of GSW impleme ntation in each GSW school and relate implementation to observed outcomes Hypothesis 2 a. Students in schools with higher dose and better quality and adhere nce/fidelity of GSW programming will show more positive results of GSW prog ramming (Hypotheses 1a, 1b, 1c ). In order to address these Aims and Hypothese s, quantitative digital plate waste photography and qualitative teacher interviews were conducted in 10 Aurora public elementary schools. Five schools were the GSW schools for the 2011 2012 school yea r, and five schools without the GSW program were matched as control schools. To address Aim 1 Quantitative data were collected from 1 st 3 rd and 5 th grade students in each school. Plate waste data were collected on the same three consecutive days in ea ch school, therefore allowing comparison of plate waste for the same menu in each school on each day. To address Aim 2 Qualitative physical education teacher interviews were conducted in 7 schools (three schools with GSW implemen tation in 2010 2011, 1 school with impl ementation in 2011 2012, and three schools with implementation in 2012 2013 )

PAGE 26

21 Informal observations and conversations with in schools also contribute d important qualitative data to this study aim.

PAGE 27

22 CHAPTER II REVIEW OF THE LITERATURE Children in the United States are eating excess calories from unhealthy food. The predisposition for children to like high fat and high sugar foods, wide availability of unhealthy food, media persuasion, and social environments all encourage poor dietary i ntake (Birch, 1999; Roblin, 2007) Nutrition education in s chools is one way to influence a large percentage of children in the United States. Nutrition education, as discussed in this research and in most academic literature, encompasses much more than simply lecturing about nutrients in foods and their relation to health. According to than information dissemination in order to be effective. Nutrition education needs to address food preferences and sensory affective factors; pe rson related factors such as perceptions, beliefs, attitudes, meanings, and social norms; and environmental fa (p. 176). This chapter address es research and theory related to different nutrition education interventions targeting food preferences and food habits, knowledge and attitudes, social norms, parental involvement, classroom instruction, and school level environmental changes. Theoretical Basis for Nutrition Education Social Cognitive Theory Social Cognitive Theory (SCT) is the most widely use d theory for designing nutrition education and health promotion programming (Contento, 2011; Reynolds & Spruijt Metz, 2006) SCT describes the interaction of dynamic and reciprocal factors (personal, behavioral, and environmental) that influence health behavior (Bandura, 1986)

PAGE 28

23 Personal factors involve feelings and thoughts (outcome expectations). Behavioral factors include food nutrition and health related knowledge and skills (behavioral c apability) and skills in handling personal behaviors (self regulation skills and self efficacy). Environmental factors include those external to individuals, including physical and social environments (Contento 2011) The interaction of outcome expectations, individual agency (including the skills to complete behaviors and the self efficacy to initiate behaviors), and environmental support for behavior change are key factors of SCT that facilitate behavior change in the GSW program. Personal factors influencing behavior include our individual thoughts and beliefs about potential actions and our personal capabilities. Human beings of all ages are self reflective beings; self reflection and evaluation influen ce personal behavior and future decisions. Among these self reflective personal behaviors are outcome expectations. Outcome expectations can be related to physical outcomes, social outcomes, and self evaluative outcomes (Bandura, 2000) Physical outcome expectations, in the health domain, include perceived risk of disease from not engaging in healthy behaviors or engaging in unhealthy behaviors. Positive physical outcomes related to food also include pleasant sensory experiences such as the taste or smell of something sweet (Bandura, 2000, 2004) Taste, as an outcome expectation, is a very powerful predictor of nutrition behavior in adults (Anderson, Winett, Wojcik, Winett, & Bowden, 2001) but has not been well studied in children. Exploring and influencing the physical outcome expectations of children eating school lunch is an important aspect of nutrition education in APS.

PAGE 29

24 Social outcomes include the social consequences of a behavior; for example positive peer reactions to a behavior, such as soda drinking among teenagers or consumption of Hot Cheetos by elementary school peers. The inclusion of curriculum related to p ositive social outcome expectations of eating healthful food will be explored with proposed qualitative methods in the current study. Self evaluative outcomes are the beliefs one has about his or her personal actions or behaviors that result in a sense of self worth and avoidance of behaviors that lead to dissatisfaction. Self evaluative outcomes might include satisfaction from reducing saturated fat or soda consumption. In the EatFit program, goal setting, working towards goals, and satisfaction with achieving goals (Horowitz, Shilts, & Townsend, 2004) In adults, self satisfaction for personal accomplishments and health behaviors is one of the most powerful regulators of behavi or (Bandura, 2000) however, self satisfaction research with chil dren and nutrition is scarce. The GSW program includes self satisfaction learning objectives as well as learning objectives related to pleasant physic al outcomes and social outcomes Behavioral capability, a personal factor construct in SCT, is related to health knowledge, and the cognitive, affective, and behavioral skills needed to carry out healthful beha vior (Bandura, 2000, 2004) Nutrition knowledge, a target of many nutrition interventions, can pertain to facts, procedures, or behavioral skills. The GSW program specifically targets facts about nutrition to help students understand how to choose and eat more healthful food. Factual knowledge targeted in the GSW lessons includes information a bout food items, specific nutrients, the food guide pyramid and recommended daily allowances of nutrients. Although some research reveals that

PAGE 30

25 increasing knowledge about fruits and vegetables did not have a significant influence on fruit and vegetable con sumption in children (Reynolds, Hinton, Shewchuk, & Hickey, 1999) other researchers show knowledge acquisition to be a precursor to behavioral capability and behavior change related to healthful eating (DeVault et al., 2009; Sauberli, Lee, Contento, Koch, & Calabrese Barton, 2008; Stevens et al., 1995; Van Horn, Obarzanek, Friedman, Gernhofer, & Barton, 2005) The GSW program is firmly rooted in knowledge acquisition, and the proposed plate waste methods will help determine whether the program (and knowledge acqui sition) positively impacts dietary intake. Observational learning and modeling, also referred to as vicarious experiences, are additional tools for increasing behavioral skills, confidence, and capabilities in children (Bandura, 1998) In elementary schools, peers, teachers, cafeteria workers, and administrators are constantly modeling behaviors related to food and nutrition. Peer, adult, and p arent modeling have been strategic components of many successful school based nutrition education interventions (Edmundson et al., 1996; Hoelscher et al., 2004; Horowitz et al., 2004; Luepker et al., 1996) The GSW program encourages school personnel to foster positive role modeling, an im portant component related to adoption and continuation of healthful eating and behavior. Qualitative interviews with physical education teachers explored the presence and magnitude of this modeling in selected schools. Self efficacy is another major motivator of action, because people must believe they can do something t o actually want to initiate new behavior. Self efficacy is crucial to overcome impediments or barriers to adopting and maintaining healthful behaviors (Bandura, 1989, 2000) The higher the level of perceived self efficacy, the harder and

PAGE 31

26 longer people will be willing to persist in a new, healthy behavior or lifestyle change. Self efficacy is especially important in beginning, modifying, and maintaining complex behaviors such as healthful eating (Contento, 2011) The GSW program and sp ecific lessons target self efficacy by helping students feel confident that they have the knowledge to select healthy foods, the skills to prepare healthy snacks, and the ability to convince parents to supply healthy food at home. All of these self effica cy areas have the potential to affect student dietary intake. One nutrition education intervention for second and third grade students in Alabama included components to increase nutrition knowledge and student self efficacy skills to select healthy foods both at school and at home, and reported significant knowledge gains and improved dietary behavior (Powers, Struempler, Guarino, & Parmer, 2005) The GSW program has not been evaluated for its impact on self efficacy. However, GSW indirectly targets self efficacy through role modeling from teachers eating school lunch in the cafeteria and healthy food messages posted throughout the school and cafeteria Most nutrition interventions include multiple components; thus it is hard to determine the most powerful components influencing nutrition knowledge and dietary behavior. However, self efficacy has been a component of many promising and successful nutri tion education interventions with children (Edmundson et al., 1996; Gortmaker et al., 1999; Hoelscher et al., 2004; Horowitz et al., 2004; Sauberli et al., 2008) Self efficacy is rooted in the concept of agency, or an personal agency is not just confidence in knowledge and skills to perform some action, but is also the ability to regulate thoughts, motivations, feelings, and behaviors or to change environmental conditions to achieve desired results (Bandura, 1989, 1998, 2000)

PAGE 32

27 Qualitative interviews with physical education teachers examine d their perspectives on personal agency and self efficacy related to acquiring, choosing, and eating health y foods and their inclusion of learning objectives and curriculum to target these concepts. The final SCT construct associated with this research is environmental influence and how the environment interacts with personal and behavioral factors to impact behavior change (Bandura, 199 8) The nutrition environment is the specific environment related to this research; students participating in the NSLP are surrounded by food choices and a nutrition environment chosen in part by the federal government and in part by the local school d istrict. Scratch Cooking Cafeteria Changes to decrease fat, sugar, and sodium have altered the nutrition environment in all Aurora Public Schools, but student reactions to these c hanges have not been assessed. S tudents in GSW schools, who essentially re ceive two interventions (Scratch Cooking Cafeteria Changes and a nutrition education intervention), may have significantly more improved dietary intake as a result of the GSW program than students in comparison schools receiving only Scratch Cooking Cafete ria Changes. The concepts surrounding environmental changes will be discussed more fully in the next section: Social Ecological Models. Interventions based on SCT constructs have proven effective for improving dietary behavior and reducing obesity levels i n children. Successful interventions are often comprised of several components, each of which might be based on different SCT constructs and involve different levels of intervention intensity and contact hours, and thus it is difficult to outline a magic formula for necessary components. The GSW program and Scratch Cooking Cafeteria Changes utilize SCT personal and behavioral

PAGE 33

28 factors targeting thoughts, knowledge, behavioral capability and self efficacy, in addition to environmental factors to influence b ehavio r change and healthful eating. This research addresses important gaps by utilizing qualitative methods to uncover the presence of learning objectives and curriculum related to these SCT constructs Comparing GSW school children to non GSW school chi ldren help ed uncover the impact of GSW participation. Social Ecological Models behavior fits within the framework of Social Ecological Models. Social Ecological envir onment modifications, and desire to strengthen the nutrition environment in schools. Factors external to an individual can have a huge impact on health and health behavior. Environmental intervention strategies target physical surroundings, social climate s, information accessibility, organizational systems, and policy to provide support for improving health, and in this case, healthy eating and active living (Contento, 2011) Social ecological models of heal th promotion and behavior change address several intrapersonal factors, interpersonal processes and primary groups, institutional/organizational factors, community fact ors, and public policy and legislation (DiClemente, Crosby, & Ke gler, 2002; Green & Kreuter, 2005; McLeroy et al., 1988) In this research, children are influenced by many of these levels including their own intrapersonal factors (likes and dislikes, self efficacy, knowledge), peer perceptions of healthy eating, and school culture and policies. Bronfe n Systems

PAGE 34

29 Theory outlines the impact of environment on children and adolescents (Bronfenbrenner, 1989) Ecological Systems Theory posits that children are affected by multiple levels of influence. The microsystem includes direct influencers such as peers, family, school, and neighborhood. The mesosystem connects structures of the microsystem. The exosystem includes the larger, ind irect social system and the macrosystem includes values, customs, and laws. The chronosystem incorporates the relationship between external events, the timing of those events, and the physical maturation process. The bi directionality of each of these lev els of influence is important in Ecological Systems Theory. Effective interventions are often designed for multiple levels of influence. The following diagram outlines the use of Social Ecological Models for changing dietary behavior (McLer oy et al., 1988) : Figure 1: Theoretical Framework: Social Ecological Models

PAGE 35

30 Table 1: Theoretical Framework: Social Ecological Models Social Structure/Public Policy Local State and Federal policies and laws that regulate or support healthy actions. Community Social networks, norms, or standards (e.g. public agenda, media agenda, partnerships). Institutional/Organizational Rules, regulations, policies and informal structures (worksites, schools, religious groups). Interpersonal Interpersonal proces ses and primary groups (family, peers, social networks, associations) that provide social identity and role definition. Individual Individual characteristics that influence behaviors such as knowledge, attitudes, beliefs, and personality traits.

PAGE 36

31 A similar depiction of Social Ecological Models related to nutrition education, food choice and dietary practices is offered below (Contento, 2011) Figure 2: Social Ecologica l Model for Nutrition Education At the basic level of influence, the individual level, social psychological theories of behavior change can guide interventions directed at individuals. Leading theories for nutrition education include SCT, the Health Belief Model, the Transtheoretical Model of Change, and th e Theory of Planned Behavior. As already discussed, several intrapersonal constructs of SCT (beliefs, knowledge, attitudes, behavioral capability, and self efficacy) are the basis and focus of GSW programming and were evaluated with this research. GSW programming also targets other levels of social ecological models. At the interpersonal level, peer and adult role modeling and parental involvement to influence dietary behavior are emph asized with PE teacher lessons, involvement of

PAGE 37

32 teachers, and parent information sessions an d newsletters. Institutional, organizational, and community s ettings, and specifically the elementary schools in Aurora, offer opportunities to involve decision makers and policy makers in obesity prevention str ategies to change relevant food and activity environments. In APS, the food service directors, cafeteria managers, administrators, and school board members are influential in improving the nutrition environment and supporting the GSW program. Developing, supporting, and introducing 41 new recipes and new scratch cooking techniques prior to the 2011 2012 school year involved collaboration from multiple organizational constituents. Beyond the local school and community level, LiveWell Colorado has offered the Culinary Boot Camp program to districts throughout the state. LiveWell Colorado also hopes to expand the GSW program to over 40 districts in Colorado and develop the program as a national model for successful nutrition education programming. These ef forts will undoubtedly educate new audiences at the local, state, and national level regarding the importance of obesity prevention strategies and helpful environmental change efforts. Each of the levels of influence within social ecological models holds important promise for improving dietary habits in children (Contento, 2011) and this research was designed to evaluate two programs that address multiple levels of influence. The school environment plays an important role in what foods kids have access to during the school day. Children could eat 35 40% of their calories at school based on the school meals program, a la carte offerings, vending machines, classroom snacks, and school stores (Briggs et al., 2003) Changing that environment to ensure the most healthful food offerings to students, and reducing availability of unhealthy food should affect the food and nutrients children consume at school. Merely increasing availability

PAGE 38

33 and accessibility of healthy foods in schools significantly increases consumption, as evidenced by one study targeting fruits and vegetables (Hearn et al., 1998) Simply changing what is offered shows promise in improving student dietary intake. More healthful school lunches and lunch practices are successfully reducing the consumption of energy dense, low nutrient foods amo ng elementary school children. Successful strategies include offering French fries less than once a week and offering fresh vegetables and fruit daily (Briefel, Cre pinsek, Cabili, Wilson, & Gleason, 2009) Schools that offer French fries in school lunches more than once a week and schools that offer desserts more than once a week are more likely to have higher rates of obesity among students (Fox, Dodd, Wilson, & Gleason, 2009) Exposure to and availability of fruits and vegetables encourage increased consumption. Students with increased access to FV and more exposure to a variety of fruits and vegetables are more likely to consume larger amounts of fruits and vegetables (Briggs et al., 2003; Eriksen, Haraldsdttir, Pederson, & Flyger, 2003) Recognizing the benefits of access to fruits and vegetables, APS has included a salad bar at all elementary schools this year, and a goal of offering 3 fresh fruits or vegetables each day. Quantitative methods i n this study assess ed the amount of fruit s and vegetables students are taking and the amount they are consuming each day. The High Five project and the CATCH project changed the school environment to improve intake of fruit and vegetables with fourth and fifth graders, respectively (Perry, Bishop et al., 1998 ; Reynolds et al., 2000) Specifically, they improved variety and attractiveness of fruits and vegetables, served extra fruit choices whenever a dessert was offered, and utilized point of purchase signs to advocate for healthy eating. It is important to note that other nutritio n education components were utilized in the High

PAGE 39

34 Five project and the CATCH project but inclusion of fruit and vegetable attractiveness, advertising, and availability components played some role in influencing fruit and vegetable intake. The High Five Pr oject reported increases in fruit and vegetable consumption from 2.6 servings per day to 3.96 servings one year later to 3.20 servings two years later (Reynolds et al., 2000) However the CATCH project did not report significant differences in FV intake between treatment and control groups after the nutrition intervention at the end of the school year ( Perry, Bishop et al., 1998 ) Of particular note here is that the two programs, the High Five Project and the CATCH project, targeted similar age groups (fourth and fifth graders), utilized food environment c hanges and parental homework activities, included weekly lessons, and employed hands on learning, modeling, reinforcement, and skills building. Surprisingly, the successful program, the High Five Project, had far fewer contact hours with students (14, 30 minute nutrition lessons = 7 total contact hours) than the CATCH program (47, 40 minute nutrition lessons = 31 total contact hours) (Perry, Bishop et al., 1998 ; Reynolds et al., 2000) Another study (Cafeteria Power Plus) involving younger students (first and third grade students) specifically targeted environmental strategies to: (1) increase opportunities during school lunch to eat a variety of fruits and vegetables, (2) p rovide new healthful role models (i.e. cafeteria workers) who eat fruits and vegetables, and (3) institute social support for kids to eat fruits and vegetables at lunch (Perry et al., 2004) Students in the intervention schools (2 years of environmental caf eteria interventions) had significantly higher fruit intake (0.79 servings per lunch for intervention schools vs. 0.63 servings for control schools), but not vegetable intake (0.52 servings per lunch for intervention schools vs. 0.58 servings for control s chools). These variable research

PAGE 40

35 outcomes highlight that it is very difficult to compare different nutrition interventions when the methods, curriculum, materials, contact hours, populati ons, and other strategies are quite varied. Perry et al. (2004) is one of the few research teams employing process evaluation and analysis to determine the impact of various program components. They discovered significant impact of verbal encouragement b y food service staff and the number of fruits and vegetables students can choose, but non significant effects of increasing the number of fruits and vegetables on a snack cart and increasing the appeal of fruits and v egetables. The GSW program and APS caf eteria changes employ environmental strategies to increase availability and attractiveness of fruits and vegetables, but they also use additional and varied methods, materials, contact hours a nd strategies. The dose and fidelity of implementation at each school is an important consideration of these environmental strategies and other GSW components and was addressed in this research. Although some research shows that improving attractiveness and availability of healthy food increases consumption, other res earchers report that if students are offered healthy and unhealthy choices side by side, they oftentimes make less healthy choices. For example, one study found that when elementary students moved from their elementary school offering only NSLP to middle school offering a la carte and snack bar meals, their intake of fruits, vegetables, and milk decreased, and consumption of sweetened beverages and high fat, high sugar foods increased (Cullen & Zakeri, 2004) The food environment and available foods shift as students move from elementary to middle to high schools, thus influencing their environmental supports for healthful eating. Because of the increased availability of unhealthy foods in middle school, it is important

PAGE 41

36 to bolster attitudes, personal motivation, and self efficacy for choosing healthy foods in elementary school. Elementary children with greater ability to choose healthy foods should be more likely to navigate the transition to a more varied nutrition environment in middle school with more success and healthier food consumption. Altho ugh the Scratch Cooking Cafeteria Changes in Aurora elementary schools improve the nutrition environment, the GSW program aims to specifically strengthen intrapersonal factors for healthy eating in addition to other interpersonal and ecological factors rel ated to healthy eating when they move to middle school or when they are offered a variety of foods in environments outside of schools. Availability and familiarity with fruit s and vegetables may not be the only ways of encouraging increased consumption using the school food environment. In hopes of enticing students to eat healthful cafeteria foods, many school nutrition programs are using marketing strategies to help compete with fast food restaurants. For example, schools are using food courts, portable food and salad bars, more a la carte offerings, and local chefs to improve the visual and taste elements of food (Briggs et al., 2003) Researchers are also examining the potential of creative signage to increase attractiveness and consumption of fruits and vegetables (S. Smith, personal communication, September 18, 2011). APS uses two such strategies by offering salad bars in all schools and using Go, Slow, Whoa signage to identify healthy food options. Meals and snacks during the school day play an important role in developing nutrient requirements for school lunches, the preparation, quality, and taste of foods offered by individual districts and

PAGE 42

3 7 schools is not monitored at the national level. If schools are offering healthful foods that do not taste good, students are likely throwing away a good portion of the food on their lunch tray. Alternatively, if students are offered lunches that meet the NSLP guidelines but do not include new, healthful recipes with whole and fresh food ingredients, children are unlikely to develop n ew taste preferences for healthful food. These issues might continue the problems of low nutrient intake or neophobia (fear of new foods). Using the environment stra tegy, but one that requir es constant evaluation. APS pioneer ed new healthy recipes in 2011 and this study will help determine if students are reacting positively and developing taste preferences for these healthy foods. Research on the school food enviro nment shows it to be a positive influence on both low and middle socioeconomic status (SES) student food consumption (Cullen, Watson, & Fithian, 2009) Most low SES students selected the healthful NSLP meal and did not buy many a la carte items. T hese decisions, most likely due to economic necessity, keep low income students from purchasing additional a la carte items and enable them to be positively affected by the nutrition environment. When snack bar foods were limited and controlled, middle SE S kids also selected the NSLP, thus pointing to the potential influence of the NSLP on nutrient intake when competitive foods are not available for financial or policy reasons (Cullen et al., 2009) Elementary schools in APS do not offer competitive foods, making the nutrition environment equal for all students, regardless of SES. Given the higher risk of obesity for lower SES students, it is important to pay particular attention to the nutrition intervention effects for different SES levels. The national prevalence of childhood obesity for children below the poverty

PAGE 43

38 threshold was 27.4% in 2007, or 2.7 times higher than the prevalence for children with family income over 400% of the poverty level (G. K. Singh et al., 2010) Given that 75% of APS elementary school students participate in Free and Reduced Lunch programming, it is important to be sensitive to the impact of SES on nutrition interventions. The High 5 project found that high SES students in intervention schools had the largest FV consumption when compared to control school s, but that middle SES and low SES students in intervention schools still consumed significantly more fruits and vegetables than control schools (Reynolds et al., 2000) The CATCH stu dy, upon which the GSW program is based, compares data by stu dy sites, ethnic groups, and gender but does not consistently report outcome data for SES groups (Lytle et al., 1996; Perry, Lytle et al., 1998) Although specific SES data will not be obtained for each tray collected in proposed plate waste methods in this study, aggregate free and reduced lunch data (a proxy for SE S) is available for each school and will be used to match GSW schools to comparison non GSW schools. Social ecological models suggest multiple levels of influence on nutrition behavior. The GSW program and Scratch Cooking Cafeteria Changes in APS offer various intervention strategies at each level of social ecological modeling and influence. Quantitativ e data collection methods will determine the effectiveness of overall programming (all levels of influence) at improving dietary intake. Qualitative data collection techniques will help qualify quantitative results by determining the inclusion and dose of specific strategies at different social ecological levels.

PAGE 44

39 Other Research on Eating Behaviors, Food Choic e, and Nutrition Education Taste Preference and Food Habits of Children Prevailing theories of taste preference and food habits support the notion t hat infants and children prefer specific tastes. Those tastes lead to food patterns and habits in childhood that influence growth and development and future health and chronic disease protection or risk (Nicklas, Webber, Srinivasan, & Berenson, 1993) Foo d habits that persist throughou t childhood and adolescence are more likely to continue into adulthood (Kelder, Perry, Klepp, & Lytle, 1994) Therefore, encouraging the development of preferences for h ealthful foods in childhood is helpful for creating healthful food habits in adulthood. APS could make a significant impact on lifelong health by encouraging students to develop healthful habits in elementary schools. Neophobia (fear of new foods) is pres ent in many children, but can be reduced by repeated opportunities to sample new foods. Reducing neophobia requires five to fifteen exposures to new foods (Birch, 1995) If children are continually exposed to high sugar, high fat, high salt foods, those foods will become familiar and children will continue to crave them over more healthy, but less familiar options. Changing nutrition env ironments in APS to offer more healthful food options for children of young ages will help to make those foods more familiar and possibly more valued by students. When presented with new, whole grain foods, elementary students had favorable responses to t he look and taste of whole grain cereal and cheese bread, but were not sure they would want to eat it in a cafeteria lunch (Burgess Champoux, Marquart, Vickers, & Reicks, 2006) Kids reporte d that new foods in the cafeteria would be accepted if they looked good, tasted good, and were familiar. Children reported the following tips for

PAGE 45

40 it look like wha t the other food was. I would just switch it and not tell everybody and let (Burgess Champoux et al., 2006) The challenge to school cafeteria and nutrition services planners in Aurora Public Schools is to make new foods appear familiar enough in appearance and taste or utilize creative marketing or incentive strategies to encourage students to try new foods. Qualitative methods in this research assessed teacher perceptions of whether students realize the foods they are new foods. Teaching and Best Practices for Nutrition Education A recent survey by Action for Healthy Kids reveals that parents think schools are p roviding nutrition education to all students and they would like such education to be part of the core curriculum two days a week (Action For Healthy Kids, 2005) In reality, kids in the U.S. get an average of 13 hours of nutrition education pe r year (Celebuski & Farris, 2000; Lytle et al., 1994) Teacher preparation and allotted time for nutrition education are important factors in the quality of information and skills students receive. APS has made nutrition education a priority by allotting increased instruction time for nutrition education in physical education classes. Nutrition education varies by district, and even more so by schools and classrooms. Classroom teachers, those most often delivering nutrition education, have little training in the topic area because of lack of time and funding and focus on academic core subject instruction. Only about half of elementary school teachers have formal training in nutrition education (Celebuski & Farris, 2000) However, 88% of elementary school teachers reported teaching lessons about nutrition to their students in the 1996

PAGE 46

41 1997 school year. The mean number of hours spent in a school year on nutrition education by elemen tary school teachers (K 5) who taught nutrition was 13, below the minimum of 50 hours thought to be necessary for impact on behavior (Celebuski & Farris, 2000; Lytle, 1994) Q ualitative methods explored the training and confidence levels of P.E. teachers in APS elementary schools to deliver nutrition education and the GSW program. P.E. teacher interviews include d questions regarding amount of time spent preparing for and delivering GSW lessons to docume GSW program offered at different schools. Training for and confidence in teaching, as well as dose/contact hours of GSW programming could have an influential effect on school level outcomes. Incorporating nutrition education in to classroom instruction with successful learning strategies is an important area of research. One meta analysis of nutrition education strategies recommends the following for effective nutrition education (Lytle, 1994) : Instruction with a behavioral focus, or a focus on changing specific behaviors rather than on learning general facts about nutrition; Employment of active learning strategies instead of relying exc lusively on information dissemination and didactic teaching methods; Devotion of adequate time and intensity to nutrition education (at least 50 hours per year to impact attitudes and behavior); A family involvement component; A meals program and food r elated policies that reinforce classroom nutrition education;

PAGE 47

42 Teachers with adequate training in nutrition education Other research contributes to best practices for school nutrition education. The CATCH program demonstrated that nutrition education messag es for children are successful when they are focused on behavior and supported by the school nutrition programs (Perry, Lytle et al., 1998) Parental and community involvement are important components of successful nutrition education interventions (Nader et al., 1996; Perez Rodrigo et al., 2001) Staff dealing with school meals should be properly trained, supported and integrated with teaching staff (Fulkerson, French, Story, Snyder, & Paddock, 2002) Based on a meta analysis of 29 behavioral interventions studying ages two to 18, strategies to reduce unhealthy behaviors (decreasing sedentary behaviors and dietary fat) are more e ffective than those promoting positive behaviors (increasing physica l activity and consumption of fruits and vegetables ) (Kamath et al., 2008) Longer duration and more contact hours involved with interventions produce greater benefit to participants. The School Health Education Evaluation, a large evaluation of the effects of health education programming in schools, found that 8 hours of health e ducation could produce large effect sizes in program specific knowledge and 20 hours could produce large effect sizes in general health education knowledge. However, even after 35 to 50 contact hours, only moderate effect sizes could be achieved in attitu des and behaviors (Connell, Turner, & Mason, 1985) The CATCH intervention involved 15 to 20 contact hours per year over three years for third through fifth graders and resulted in positive eating behavior and physical activity behavior changes but not physiologica l changes (Luepker et al., 1996) These changes persisted when re measured in the eighth grade (Nader et al., 1999) The Know Your Body program included 30 50 contact hours

PAGE 48

43 to manipulate diet and create positive effects on serum cholesterol and blood pressure (Resnicow et al., 1992; Walter, 1989) Given the high number of content hours necessary for knowledge, attit ude, and behavior change, new interventions should plan appropriately to achieve desired results An important component of this research wa s a careful analysis of GSW implementation in each school. P.E. teachers were interviewed and asked specific questions about cont act hours, learning strategies school and administrator support, and other aspects of GSW programming. Analysis of GSW implementation also provides important process evaluation information to improve co nsistency across schools in future programming. Implementation Issues in Prevention and Education Programs Continuous delivery of effective interventions i s a n important component in program success ; few interventions are sustained over time regardless of their success during pilot period s (Durlak & DuPre, 2008; E. M. Rogers, 2003) Measuring the implementation of a program is paramount to drawing conclusions about program success. Negativ e or null results from a program could signify a failure of that program, or they could simply mean that the program was not implemented as designed. Conve rsely, positive program impact could result from intended implementation or from implementation quite different than what was intended. It is impossible to make judgments about a program without assessing implementation. Implementation data also help to test theory behind an innovation by ascertaining which components and which related theories were effectively administered (Durlak & DuPre, 2008) A meta analysis by Durlak (2008) def ined key terms related to implementation: (1) Fidelity, or the extent to which a program corresponds to the originally intended program, (2) Dosage, or how much of a program

PAGE 49

44 has been delivered, and (3) Quality, or how well different program components hav e been conducted. Regression analyses in one review of 221 school based prevention programs targeting aggressive behaviors reported that implementation was the most important variable that influenced outcomes (Wilson, Lipsey, & Derzon, 2003) Other studies of school based interventions and community based interventions targeting a variety of outcom es fou nd significantly higher effect sizes in programs that monitor ed implementation and adjust ed analyses for factors related to implementation (Ananiadou, Schneider, Smith, & Smith, 2004; Derzon, Sale, Springer, & Brounstein, 2005; D. DuBois, Holloway, Valentine, & Cooper, 2002; Tobler, 2000) Researchers have historically analyzed implementation in two m ajor ways: (1) categorically, with groups for different levels of implementation (i.e. high vs. low) or (2) continuously, with percentages assessing level of dosage or fidelity (Durlak & DuPre, 2008) The latter method reports a broader range of implementation and might give more statistical power to conclusions about the effects of implementation. The two most common methods for arriving at implementation values are provider self rep orts and independent behavioral observations (Durlak & DuPre, 2008) Few studies have directly compared the two strategies, but objective observations would likely return more unbiase d, accurate and comparable assessments of implementation. Including measures of implementation in qualitative and quantitative analyses help predict program outcomes. This research utilized qualitative strategies to assess implementation. A social ecol ogical framework can be used to identify influences on program implementation. Contextual factors are important in understanding motives, support, and barriers for implementation. Durlak and colleagues identified the following main

PAGE 50

45 categories in an ecolo gical framework for program implementation: innovations, individuals and communities, and features associated with the prevention delivery and support systems (2008). The innovation can be defined as the program itself. Program characteristics that affe ct implementation include adaptability (program flexibility to meet the needs of providers/teachers) and compatibility (contextual appropriateness and fit with the organization/school) (Everett M. Rogers, 2003) Although there must be a balance between adaptability and fidelity of proposed implementation, programs that are a daptable for teachers and that fit within the culture of a school and school district show higher fidelity and quality of implementation. Factors related to individual providers and communities will affect program implementation. For example, school sta ff forced to launch a program and school staff who volunteer to create or deliver a program could vary in the effectiveness of their program implementation. Provider/teacher characteristics of implementation include: (1) providers who recognize program ne ed, (2) providers with high self efficacy for program delivery, (3) providers who have the skills to implement the program, and (4) providers who believe in the program and the potential benefits (Barr, Tubman, Montgomery, & Soza Vento, 2002; Cooke, 2000; Ringwalt et al., 2003) Supportive principals are also an integral part of high fide lity for program implementation (Kam, Greenberg, & Walls, 2003) The prevention delivery and support system for providers are key components of i mplementation. Training and technical assistance for people implementing a new prevention program are key element s of program success. The goals of training should include development of mastery in specific intervention teaching and skills, as well as

PAGE 51

46 at tention to provider expectations, motivation and sense of self efficacy (Durlak & DuPre, 2008) After a program begins, technical assistance s hould help maintain where needed, and support problem solving efforts. Two studies show that early monitoring of implementation followed promptly by retraining for providers with initial difficult ies doubled the fidelity of implementation to over 85% (Dufrene, Duhon, Gilbertson, & Noell, 2005; Greenwood, Tapia, Abbott & WAlton, 2003) Dialogue and communication with providers is crucial immediately after the beginning of a new program, and remains important throughout t he duration of a program. Summary According to the American Dietetic Association, about half of all Americans believe they are doing all they can to achieve balanced nutrition and a healthful diet, but given that such a significant portion of our population is overweight or obese, it appears that knowledge alone does not lead to behavior change (American Dietetic Association 2008) Given that information dissemination is necessary but not sufficient to change behaviors, Contento (2011) recommends that nutrition education focus on personal motivations and competence, interpersonal interactions, and environmental factors that influence individual and community patterns of behavior. Aurora Public Schools are following that recommen dation with the Go, Slow, Who a program to educate students, increase knowledge, and change attitudes and behaviors surrounding nutrition, and the Scratch Cooking initiative to improve the nutrition environment and healthful food offerings in schools. The t heoretical basis for this researc h included SCT and SEM. Th e constructs of SCT and SEM provide support for intervention components and facilitate

PAGE 52

47 knowledge and action by individuals and make the environment more conducive and supportive of desired changes This research assess ed the effectiveness of this educational and ecological combination of Aurora Public School programming for promoting dietary intake and health. The most similar intervention to the current GSW intervention is t he Dietary Interventio n Study in Children (DISC): a 3 year intervention aimed at decreasing intake of fat, saturated fat, and cholesterol among pre adolescent children by increasing knowledge and skills for identifying healthy food (Stevens et al., 1995; Van Horn et al., 2005) The DISC study utilized intervention sessions (number unreported) to teach children how to identify healthy foods (Go foods) to choose and eat all the time and less healthy foods (Whoa foods) to choose and eat only occasionally. The interventio n lesson wheel with eight wedges pertaining to eight categories of a were highlighted in red. Dietary intake was measured with 24 hour dietary recall at bot h pre intervention and post intervention. Although the intervention sessions in the DISC research targeted a visual nutrition guide very similar to GSW, assessing dietary intake with subjective 24 hour recall is much different than objective visual plate waste analysis used in this research. However, utilizing a 24 hour recall methodology the intervention group showed increased consumption of Go Foods in all measured food categories except for fruits and those increases were significant for dairy product s, desserts, and fats/oils. The intervention group showed decreased Whoa Dairy, Whoa fats/oils, and Whoa vegetables (specifically French fries) when compared with the control group. This nutrition education intervention utilized very similar visual aids a nd teaching tools as the

PAGE 53

48 GSW program in Aurora Public Schools. Teaching healthy food identification skills with a Go Whoa visual aid improved diet quality and leads to hypotheses about the current GSW programming in APS. Given the success of the DISC pro gram, I hypothesize that students in GSW schools with high fidelity, dose, and quality of implementation who received Scratch Cooking Cafeteria Changes will have significantly higher choice and consumption of Go foods than non GSW schools who only receive Scratch Cooking Cafeteria Changes. Overall, I hypothesize that students in GSW schools will report increased choice and consumption of healthy foods and new healthy recipes in APS cafeterias.

PAGE 54

49 CHAPTER III METHODS Research Question and Hypotheses How influential is the Go, Slow, Whoa nutrition education program combined with Scratch Cooking Cafeteria Changes over and above the impact of Scratch Cooking healthy foods? Figure 3 : Aims and Hypotheses Aim 1: Determine the relative impact of GSW and Scratch Cooking Cafeteria Changes over and above the impact of Scratch Cooking Cafeteria Changes alone on lunch. Hypothesis 1a: Students receiving the GSW program will choose more healthy foods than non GSW students at school. Hypothesis 1b: Students receiving the GSW program will consume more of the healthy foods on their school lunch tray than non GSW students. Hypothesis 1c: Students receiving the GSW program will choose and consume more overall fiber and white milk and less total fat and saturated fat during school lunch than non GSW students. Aim 2: Determine the relative dose, quality and adherence/fidelity of GSW implementation in each GSW school and relate implementation to observed outcomes. Hypothesis 2a: Students in schools with higher dose and better quality and adherence/fidelity of GSW programming will show more positive results of GSW programming (Hypotheses 1a, 1b, 1c).

PAGE 55

50 Research Design This study employ ed a quasi experimental, posttest only, nonequivalent control group design using mixed methods and focused on outcome evaluation. In this study, the Go, Slow, Whoa program and Scrat ch Cooking Cafeteria Changes, existing pieces of LiveWell Colorado programming were in need of outcome evaluation. Outcome, or summative, evaluation focuses on two issues: effect assessment (production of desired effects) and efficiency assessment (bene fits in relation to costs) (Singleton & Straits, 2010) T he focus of this research was effect assessment of both GSW and Scratch Cooking Cafeteria Changes, which help s in draw ing conclusions about efficiency of the programs. Quasi expe rimental designs are utilized when legal, ethical, or practical considerations make true experimental design impossible (Singleton & Straits, 2010) In Aurora Public Schools, all schools were exposed to Scratch Cooking Cafeteria Changes in the 2010 2011 school year. Only five elementary schools were exposed to GSW, an d these five schools were selected without randomization. Thus, a nonequi valent control group design was chosen to compare schools with GSW to control schools without GSW. Although similarity of schools and students cannot be ensured without randomization, this quasi experimental design match ed experimental and control schools with propensity score and Mahalanobis matching techniques in an attempt to minimize differences b etween schools on observable characteristics In this instance, pretest data were not collected by the school district, therefore a posttest only design offered the best information for determining program effects. This quasi experimental, posttest only nonequivalent control group outcome evaluation research was selected over competing study designs because of existing

PAGE 56

51 constraints of the school district and LiveWell programming. Combining quantitative and qualitative data in a methodologically rigorous design returned rich data regarding student behaviors and teacher impressions re sulting from school cafeteria changes and nutrition education. This study design was best suited to answer the research questions related to Scratch Cooking Cafeteria Changes and GSW programming. Background and Target Population Aurora Public Schools and LiveWell Colorado Programming Aurora Public Schools Aurora Public Schools span two counties: Adams and Arapahoe. The average obesity and overweight percentages for all three counties is significantly higher than the Healthy People 2010 goal of 5%. According to 2009 data (COPAN, 2009) : Table 2: Child Obesity Rates in Aurora County % obese children (ages 2 14) (95 th percentile) % overweight children (85 th 94 th percentile) % children who eat 5 or more servings of fruits and vegetables daily Arapahoe 11.9% 11.8% 28% Adams 18.6% 14.3% 26.9% Data collected from three Aurora elementary schools show even higher rates of obesity and overweight than the general county rates. Body Mass Indices (BMIs) showed 47 percent of students were overweight or obese and 19 percent were obese. Furthermore, BM I analysis in Aurora identified a trend of increasing BMIs as students advance by grade in elementary school (L. Scott, 2010) This trend points to the critical need to prevent weight gain at the elementary school level. This school district is an

PAGE 57

52 excellent target for healthful e ating education, promotion, and programming. Scratch Cooking Cafeteria Changes in Aurora As of April 2012 in Aurora Public Schools, 53 school food service managers and central office leadership staff had attended the LWC and the Colorado Health Foundation sponsored Cook for America Culinary School Food Boot Camp. This represents 16% of the total Nutrition Services Staff (kitchen managers, kitchen staff, and central office kills to increase the use of whole grains, utilize more fresh fruits and vegetables, prepare raw meat instead of reheating frozen and processed meats, and use knives with proper knife skills to increase control over the healthfulness of ingredients. Follo wing the boot camps and prior to the 2011 2012 school year, Aurora Public Schools Nutrition Services Department made significant changes to their menu selection and recipes. Forty one new recipes utilized by all Aurora Public Schools were developed or mod ified minimizing adding healthful foods and ingredients that the students might not even detect (i.e. fresh carrots, zucchini, onions, celery, an d garlic in spaghetti/red sauce ). APS began offering a salad bar at most schools in Fall 2011 In September, 2011, Aurora Public Schools served over 220,000 lunches in 30 elementary schools, averaging 459 lunches per school. Although foods consumed and likability o f cafeteria offerings were not assessed before APS implemented broad menu changes, post test only assessment of what kids eat and what kids do not eat will help with APS Nutrition Services planning, program evaluation, and food ordering.

PAGE 58

53 The Go, Slow, W hoa Program Aurora Public Schools piloted the Go, Slow, Whoa program in one school in Spring 2010. After analyzing pre post surveys from students in the program as well as interviewing program managers, the program was expanded to five elementary schools for the 2010 2011 school year. All elementary schools are slated to receive the program by the 2014 2015 school year. The Go, Slow, Whoa Program was developed as a component of the Coordinated Approach to Child Health (CATCH program) at the University of Texas School of Public Health at Houston and adopted as a part of the National Heart, Lung a nd Blood Institute WECAN! p rogram to help kids identify healthy foods in school meals and at home (NHLBI, 2009; Perry, Bishop et al., 1998 ; Perry et al., 1997) Using a green light, yellow light, red light visual aid, students are taught how to incorporate more healthy foods into their diets. Go foods, or more nutrient dense food s, should be consumed more often than Slow or Whoa foods that have added fat and sugar. Slow foods should be eaten less often and in smaller quantities. Whoa foods should be eaten only rarely. GSW, in Aurora Public School District, utilizes a kick off as sembly, physical education class time, parent meetings, classroom teacher reinforcement, and cafeteria food labeling to educate about and reinforce the consumption of Go foods. In late September, 2011, each of the five GSW elementary schools hosted a kick off GSW assembly featuring the P.E. teacher, a TV personality from Channel 7, and several mascots in costume (Power Panther and the Colorado Avalanche mascot). The assembly reinforced the proper identification of Go, Slow, and Whoa foods, introduced kids to the cafeteria labeling system, and served as a cheerleading session for the program. In

PAGE 59

54 October, parents were invited to attend either a parent breakfast or a parent afternoon snack meeting to describe the program and encourage family participation at home. Parents receive monthly newsletters and menus with Go, Slow, and Whoa foods highlighted The newsletters are created and managed by each individual GSW school, and, therefore, differ across schools. At the beginning of the 2011 2012 school year the APS Healthy Schools Coordinator held trainings for all principals, gym teachers, nurses, and cafeteria managers from the GSW schools to explain program components and how to implement them according to LiveWell and APS standards. Program components ar e listed in Table 3 The nutrition education components of the program are taught and reinforced in P.E. classes as well as in homeroom classrooms. Students are encouraged by teachers and parents to look at the menu and cafeteria labels and choose Go foo ds on a daily basis. The Nutrition Services Department in the school district label all foods on the school menus (received by all APS families) and daily in the GSW school cafeterias themselves as either Go (green apple), Slow (yellow circle), or Whoa (r ed square). See Appendix A for a copy of the December 2011 menu and Appendix B for a copy of the rating system for GSW foods (Coordinated Approach to Child Health (CATCH), 2012) It is important to note that the CATCH rating system is for individual food items such as whole wheat bread, fresh tomatoes, 2% cheese, and butter, which would each be labeled as Go Go, Slow, and Whoa foods, respectively. However, APS Nutrition Services staff label complete menu items in the cafeterias and on print menus. Therefore, taking into account each of those ingredients, a grilled cheese sandwic h made with whole wheat bread, 2% cheese, tomato slices, and a small amount of butter would likely be labeled as a Slow

PAGE 60

55 food. An important element of the GSW program is point of sale labeling of GSW foods. Kitchen managers set out labels in acrylic frame s each day above or beside food options identifying food items as Go, Slow, or Whoa. Incentive days allow nutrition services and school personnel to promote healthy foods and the GSW program. Three incentive days are offered during the school year to ea ch GSW school during which a special fruit, vegetable, or skim milk is offered. If students take AND eat the item they receive a Channel 7 wristband (similar to Lance co lor is offered on each incentive day. Table 3: GSW Program Components Program Component Program Component Description Responsible Party (note: there is no mandate as to who completes different components. This column is based on qualitative interviews an d informal observations.) Nutrition Education/GSW Curriculum Lessons to teach and reinforce the Go, Slow, Whoa labeling system P.E. Teachers, Cafeteria Managers Kick Off Assembly Assembly to introduce GSW, increase energy, awareness, and excitement. District GSW Administrators, Principals, P.E. Teachers, Cafeteria Managers, Channel 7 News Personalities GSW Labels on Lunch Menus Encouragement and assistance for students to look for GSW labels on printed lunch menus Classroom Teachers, P.E. Teachers, P arents, Cafeteria Staff Labeling of Cafeteria Food with GSW L abels GSW labels affixed to plastic sneeze guard or in front of lunch items each day. Cafeteria Staff Three Incentive Days Prizes (bracelets) awarded if students take/consume a certain product (white milk, specific fruit, specific vegetable) on three days during the school year Ca feteria Staff, Lunch Room Aides, Administrators Parent Breakfast Informational meeting about GSW Principal, Assistant Principal, Cafeteria Manager GSW Articles in School Newsletters Newsletters to entire school community should contain GSW information or nutrition related articles P.E. Teachers, Front Office Extra Programming Gues t Speakers, More Incentive Days P.E. Teachers

PAGE 61

56 Methods for Matching GSW S chools with Comparison Schools Five schools were selected for participation in the GSW program in the 2011 2012 school year. All schools will implement the GSW program by 2014 2015. The procedure for placing schools on the rotating start list was not random, nor was the selection process documente d. Seven schools participated in GSW programming in 2010 2011 and f ive elementary schools started in fall 201 1 Given that schools were not selected randomly for treatment in the 2011 2012 school y ear, propensity score matching wa s an appropriate statistical technique to identify five schools most similar to the GSW schools on observed school characteristics to serve as comparison schools Propensity score matching allows researchers to determine the extent to which a treatment group and comparison group is similar based on some observed covariates, even when the groups are assigned without randomization. The goal of propensity score matchi ng is to match experimental and control subjects on observed covariates so that the main difference between them is whether or not they received the treatment (Khandker, Koolwal, & Samad, 2010; Rosenbaum & Rubin, 1983) Propensity scores reduce all of t he observed covariates into a numerical, scalar summary for selecting matched samples: the probability of receiving the treatment (ranging from 0 to 1), conditional on the covariates. Matching can then be done on this scalar summary rather than on all of the covariates directly (Rosenbaum & Rubin, 1983) The premise behind this technique is that if two units have the same propensity score but are in different treatment groups, the determination of which unit received treatment and which did not was random based on observables. Therefore, in the school level setting, treatment and

PAGE 62

57 comparison schools with simi lar propensity scores should have similar joint distributions of the covariates used for propensity score matching. Treatment effects can then be estimated based on the mean differences in outcomes between participants and matched non participants. In thi s particular research, schools we re treated as matching units because entire schools were selected to receive treatment. Matching methods such as propensity score matching are becoming more and more popular as ways to estimate causal effects by using obser vational data. These methods, which select subsets of the original treatment and control units that are the most similar on the observed covariates, can be conceived as a way to replicate a randomized experiment by selecting treatment and control units (sc hools) that look only randomly different from one another on all of the observed covariates (Stuart, 2007) Oftentimes observational studies simply utilize models with a treatment indicator and a set of covariates as predictors of some outcome variable. Instead, the design of observational studies can be structured with the rigor of randomized experiments : without using the outcome data and with a good understanding of the treatment and control conditions (Rubin, 2001) This approach involves outlining the extent to which the treatment and comparison groups are similar on background covariates and then using methods such a s propensity score matching to ensure that the treatment effects are estimated by using schools or subjects that look similar to each other. The strongly ignorable treatment assumption (SITA), a key assumption of propensity score matching, implies that tr eatment assignment is independent of the potential outcomes given the observed covariates (Rosenbaum & Rubin 1983) SITA is met if treatment assignment (Z i ) and the potential outcomes Y i = (Y 0i ,Y 1i ) are conditionally independent given the observed covariates X i (or, alternatively, the propensity score). The formula for this probability of treatment assignment is:

PAGE 63

58 Pr(Z i |X i ,Y i ) = Pr(Z i |X i ), with 0 < Pr(Z i = 1|X i ) < 1, where Y 0i is the potential comparison outcome (for Z i = 0) and Y 1i the potential treatment outcome (for Z i called potential outcom es because they refer to the outcome one would observe if subject i gets treated (Y 1i ) or not treated (Y 0i ) (Shadis h & Steiner, 2010) There are several advantages of Propensity Score Matching (PSM) (Kha ndker et al., 2010) PSM provides good comparisons if observed characteristics drive program participation and unobserved characteristics are negligible. PSM does not require a baseline survey or data. PSM is a semi parametric method, imposing fewer constraints on the functional form of the treatment model, as well as fewer assumptions about the distribution of the error term. This means that PSM increases the likelihood of sensible comparisons across treated and matched control units. The informatio n that was available for all public elementary schools in Aurora is summarized in Table 4 This study used various propensity score models with some and all of these covariates to determine the best fitting model.

PAGE 64

59 Table 4: Covariates for Matching Varia ble Definition Race/Ethnicity % of students in each school who are: White omitted from model as comparison group Black Asian Hispanic Native American Native Hawaiian Two or More Race/Ethnicities Free/ Reduced Lunch (FRL) % of students in each school who are eligible for free or reduced lunch. Governmental program that provides low income students with free or reduced price school meals Percentages for schools are determined by taking the number of students eligible for free or reduced price lunch divided by the total number of students enrolled in the school. Gifted and Talented (GT) % of students in each school who are gifted and talented in either math or language or both. They are identified by teachers as they search for students who perform at exceptional levels. Nominations for GT students are made with supporting evidence from cognitive and culture free measures, academic aptitude and performance, languag e acquisition skills, etc.) Special Education (SPED) % of students in each school who are classified as special education and have an individualized education plan (IEP). English Language Learners (ELL) % of students in each school who are classified as English Language Learners (either no proficiency or low proficiency in English) Nutrition Programs (NP) Total Number of additional Nutrition Programs offered in a school (Range 0 5) The five possible programs offered at various Aurora schools are: Breakfast in the Classroom Fresh Fruit and Vegetable Program INEP Nutrition Education Healthier U.S. Schools Challenge Award for implemented changes and activity time (not necessarily a program, but a commendation for changes) Coordinated School Health The Nutrition Programs variable was created as a sum of additional nutrition programs offered in a school. This variable was created by combining 5 different d ichotomous variables (for each of five programs) to save degrees of freedom in modeling. Th e logic of including the Nutrition Program variable relates to the ability and motivation of each school to bring in additional nutrition programs. If a school

PAGE 65

60 (administrators, teachers, and/or parents) is able to bring in 3 additional programs, that scho ol is likely different in various ways (motivation of school community, SES of children, previous grant funding) compared to a school that does not bring in any additional program. Propensity scores were calculated in STATA using the PSMATCH2 program whi ch runs multiple methods of matching using propensity scores and full Mahalanobis matching (Leuven & Sianesi, 2003) Propensity scores were estimated with a logit model using two different matching methods: nearest neighbor ma tching and Mahalanobis matching. These two matching methods were chosen because of the small sample size and based on previous research with Mahalanobis matching and smaller sample sizes (Zhao, 2004) Nearest neighbor matching randomly orders the treatment schools, and then finds a control s chool with the closest propensity score (D'Agostino, 1998) Because o f the small sample size of untreated schools, those two schools were then taken out of the pool (not replaced) and the second treatment school was matched with the nearest neighbor in terms of propensity score. The second method used for matching was Mah alanobis metric matching combined with propensity score matching. This method randomly ordered subjects and then calculated the distance between the first treated subject and all controls using matching variables (including the propensity score) and the s ample covariance matrix of matching variables from the full set of control subjects (Baser, 2006; D'Agostino, 1998) The first randomly ordered treatment subject was matched based on the smallest Mahalanobis distance. The process was then repeated until each treatment subject received a match.

PAGE 66

61 In this research, propensity score matching was based on data available for all elementary schools included in Table 4 : (See Appendix C for complete Data T able.). Running 12 different logit models helped determine the best fitting models for the available data. As indicated in the literature, different variables were logged and squared in different models to determine improvements in the Akaike Information Criterion (AIC) and Bayesian Information Criterion (BIC). Different variables (12 total) were included and excluded in successive models to determine improvements in model fit. Models with fewer covariates improved AIC and BIC values. Even though none of the individual independent variable coefficients were significant in any of the models (perhaps due to small sample size), three models emerged with very similar AIC and BIC values. Using fewer covariates in these three models with such a small pool of available matches helped to avoid overspecification of matching models. Post hoc analyses of variable means show the differences between covariates (mean values) for treatment schools and match schools. Model 1 used propensity score matching with four c ovariates and nearest neighbor matching. All variables with large mean differences between matches were logged and run in different models to determine the best fitting model. Logging the Hispanic variable resulted in lower AIC and BIC values, and thus t his variable is logged in Model 2. Model 2 used propensity score matching with four covariates (1 logged) and nearest neighbor matching. Model 3 used full Mahalanobis matching with five covariates (the covariates in Model 2 plus propensity scores).

PAGE 67

62 Formulae for Propensity Score and Mahalanobis Metric Logit Models: Mod el 1: Propensity Score Matching 1 X 1 (percentage Hispanic students) + b 2 X 2 (percentage English Language Learner Students) + b 3 X 3 (percenta ge Free and Reduced Lunch Students) + b 4 X 4 Model 2: Propensity S core Matching 1 X 1 (logged percentage Hispanic students) + b 2 X 2 (percentage English Language Learner Students) + b 3 X 3 (percentage Free and Reduced Lunch Students) + b 4 X 4 Model 3: Mahalanobis Matching 1 X 1 (logged percentage Hispanic students) + b 2 X 2 (percentage English La nguage Learner Students) + b 3 X 3 (percentage Free and Reduced Lunch Students) + b 4 X 4 b 5 X 5 Results of all three matching models using the Stata PSMATCH2 program (Leuven & Sianesi, 2003) are shown in Figure 4 Model Log Likelihood (null) Log Likelihood (model) AIC BIC Model 1 Propensity Scores 12.51 12.20 34.40 40.49 Model 2 Propensity Scores 12.51 12.05 34.09 40.20 Model 3 Mahalanobis Matching 12.51 12.05 34.09 40.20 Figure 4: Matching Model Results Model 3 (Mahalanobis matching) produces similar model statistics to Model 2 (Propensity Score matching), because Model 3 was calculated with the same predictor variables as Model 2. Therefore, the propensity scores are identical for both models. However, the comparison schools chosen by Model 3 are different because Mahalanobis matching uses the propensity scores as an additional matching va riable.

PAGE 68

63 The STATA PSMATCH 2 program determined the following matches for each GSW school for each of the three models. Table 5: Matched Schools Based on Three Matching Models GSW School Model 1 Match Model 2 Match Model 3 Match Altura Lansing Vistapeak Crawford Vassar Wheeli n g Yale Sixth Avenue Murphy Creek Dartmouth Peoria Fulton Yale Sable Dartmouth Wheeli n g Jewell Arkansas Fletcher Dartmouth The following figures (Figures 5 9) show the absolute values of covariate differences between GSW schools and their matches for each model. These figures provide a visual of how similar the GSW schools we re to their matches on each matching variable using 3 different Models. If the y axis value s i n any of the following graphs are zero, then there is zero difference between the GSW school and the match school on that covariate. Figure 5: Altura Matching Results 0 2 4 6 8 10 12 14 16 Difference between GSW school and matched schools from Model 1, 2, 3 Matching Variables Altura Altura-Model 1 match (Wheelig) Altura-Model 2 match (Peoria) Altura-Model 3 match (Wheelig)

PAGE 69

64 Figure 6: Lansing Matching Results Figure 7: Vista Peak Matching Results 0 2 4 6 8 10 12 14 16 18 20 Difference between GSW school and matched schools from Model 1, 2, 3 Matching Variables Lansing Lansing-Model 1 match (Yale) Lansing-Model 2 match (Fulton) Lansing-Model 3 match (Jewell) 0 5 10 15 20 25 30 35 40 45 Difference between GSW school and matched schools from Model 1, 2, 3 Matching Variables Vista Peak Vista Peak-Model 1 match (Sixth Avenue) Vista Peak-Model 2 match (Yale) Vista Peak-Model 3 match (Arkansas)

PAGE 70

65 Figure 12 Figure 8: Crawford Matching Results Figure 9: Vassar Matching Results 0 10 20 30 40 50 60 70 80 Difference between GSW school and matched schools from Model 1, 2, 3 Matching Variables Crawford Crawford-Model 1 match (Murphy Creek) Crawford-Model 2 match (Sable) Crawford-Model 3 match (Fletcher) 0 1 2 3 4 5 6 7 8 Difference between GSW school and matched schools from Model 1, 2, 3 Matching Variables Vassar Vassar-Model 1 match (Dartmouth) Vassar-Model 2 match (Dartmouth) Vassar-Model 3 match (Dartmouth)

PAGE 71

66 Visual inspection of the differences in means between GSW schools and their match es using Models 1, 2, and 3 indicated that Model 3 ( Mahalanobis Matc hing with Propensity Scores) had the smallest differences on the most number of covariates. iates simply meant those schools we re most similar on that covariate. For example, in looking at th e covariate White (percentage of White students in the school) in Figure 5 the difference between Altura and the Model 2 match (Peoria) is the smallest (shaded red), and therefore, Model 2 is the best match for that particular covariate. There we re 5 co variates for each school. Multiplying 5 covariates by 5 schools resulted in 25 different variable combinations to be considered in the best fitting model across all schools. Totaling which model (Model 1, Model 2, or Model 3) produced the smallest differ ence between the GSW School and Match School on each of the variables is summarized in Table 6. Table 6: Comparison of Matching Models for Best Fit Model 1 Model 2 Model 3 Number of variables in which the model produced the best match 9 12 16 *note that the total number of variables in this table does not total 25. If a variable resulted in the same lowest differences between GSW school and Match school for more than one Model, then both models were counted in this table. Because it produce d the greatest number of smallest differences between match sc hool means and GSW school means Model 3 (Mahalanobis Matching with Propensity Scores) was used to identify untreated comparison schools for the GSW treatment schools. Final school matches are listed in Table 7.

PAGE 72

67 Table 7 : Final Comparison Schools GSW School Model 3 Match Altura Lansing Vista P eak Crawford Vassar Wheeli n g Jewell Arkansas Fletcher Dartmouth Data Collection Methods Aim 1 : Quantitative Plate Waste Data Aim 1: Determine the relative impact of GSW and Scratch Cooking Cafeteria Changes over and above the impact of Scratch Cooking Cafeteria Changes alone on This research utilize d both quantitative and qualitative d ata collection methods. Identify ing what children choose and what they eat at school helped address the research questions. Children, especially young children, are oftentimes omitted from data collection in schools because of their underdeveloped readin g skills and variation in cognitive skills and development (J. Scott, 2008) This study used quanti tative data collection that did not require reading on the part of the students in grades 1, 3, and 5. Grades 1, 3, and 5 were chosen for data collection to represent the range of ages and grades in an elementary school without collecting data from all grades. Dietary intake us ing quantitative survey methodology, while providing some useful information based on personal food recall, does not always produce reliable results (Beaton, Burema, & Ritenbaugh, 1997; Tran, Johnson, Soultanakis, & Matthews, 2000) Children and adolescents, specifically, do not accurately recall food intake, whi ch becomes a problem when such assessments are used to monitor nutritional status

PAGE 73

68 (Livingstone, Robson, & Wallace, 2004) Children under report, over report, and incorrectly identify foods and portion sizes. A meta analysis of 11 studies using food recall surveys with children found a wide range in mean energy intake and validity of food recall techniques (McPherson, Hoelscher, Alexander, Scanlon, & Serdula, 2000) Food diaries/records are another subset of food recall techniques commonly used to eliminate issues with memory, but they do not eliminate issues related to estimating portion size and properly identifying foods consumed. Validity and reliability of food diaries/records is not consistent in multiple studies using this technique (McPherson et al., 2000) Therefore, the quantitative method selected for this study assessed plate waste, an accurate assessment of food intake for a specified time period that did not involve individual subjective reporting. In order to assess the effect of GSW and Scratch Cooking Cafeteria Changes on stude nts in grades 1, 3, and 5, plate waste methodology was selected considering the varying cognitive abilities of children ages 5 11. The most accurate way to measure food intake is weighing foods before and aft er consumption (Wolper, Heshka, & Heymsfield, 1995) However, this method is time food is consumed ha s been utilized to assess the National School Lunch and School Breakfast programs. Given the importance of good data concerning dietary intake in children, researchers developed other reliable and valid methods for collecting data that are less costly and less intrusive. Plate waste protocols based on direct visual observation of foods before and after a meal is well suited to cafeteria settings or other public eating situations (Comstock, St Pierre, & Mackiernan, 1981; Wolper et al., 1995) The portion of food that remains (plate waste) is estimated by visual observation when a tray is

PAGE 74

69 return ed at the end of a meal. Visual estimation of plate waste is a promising and inexpensive technique for assessing dietary intake. Original visual estimation plate waste studies utilized trained observers directly observing food trays before and after eatin g. The observers estimated portion sizes for each food in reference to standard portion sizes using some sort of percentage rating scale. By weighing several samples of the reference food (to get accurate baseline serving sizes and nutriti on information) and comparing them to the percentage of food left over, the percentage of food eaten (along with weight, energy, nutrients ) can be estimated from the average reference food weight and nutrition information. This visual method has been used extensively in studies of nutrient intake and eating behaviors in schools and other institutional settings (Auld, Romaniello, Heimendinger, Hambidge & Hambidge, 1999; Davidson, Hayek, & Altschul, 1986; Friedman & Hurd Crixell, 1999) Direct visual estimation has been proven comparable to weighed protocols in several studies (Comstock et al., 1981; S. Dubois, 1990; Kirks & Wolff, 1985) One of the barriers to using direct visual estimation in a school cafeteria environment is the required time for food tray assessment Observing food trays at the end of a lunch li ne before a student sits down to eat takes precious minutes away from food consumption time. At the end of the lunch period, c ollecting a large sample of trays necessitates on site estimation to a void transportation, sanitation, time, space, and manpower considerations of transporting food trays. On site post lunch estimation in a timely manner is difficult due to utilization of cafeteria space for other activities following lunch. Without a large number of data collectors, visual estimation would likely take a long period of time even for a random sample of trays It is also difficult to

PAGE 75

70 conduct inter rater reliability assessments if trays are subsequently discarded after post lunch waste estimation. A more recently developed method avoids some of th ese problems by taking a digital photograph of cafeteria trays both before and after food is consumed. T he digital photograph is stored for future visual estimation, and can be rated by several trained researche rs to ensure data quality. The digital plat e waste photography method, utilized successfully with both adults and children, enables reliable and valid data collection for a fraction of the cost of weighed plate waste estimates (Swanson, 2008; Williamson et al., 2003) This study utilize d a slightly different m ethod where data collectors record ed on an index card what wa s on student trays at the end of a lunch line. That record w as then used as the pre consumption data, and a digital photograph of each tray at the end of the lunch period was compared to the pre consumption data to determine plate waste and fo od consumption. This modified digital plate waste protocol was developed to avoid delays in lunch lines and reduced eating time resulting from the time required for pre consumption photographs (S. Smith, personal communication, September 18, 2011). Sampl ing Frame for Quantitative Data : In Aurora Public Elementary Schools, there are 12,800 students. Five schools implemented GSW programming in 2010 2011, and 20 schools were eligible to be match schools because they had not yet received any GS W training or materials. Five GSW schools and five match schools were included in the quantitative portion of this study. Research Assistants a im ed to collect twenty cafeteria trays from each of three grades (1, 3, and 5) in each school for three consecut ive days. Therefore, in total, 1800 trays were targeted for collection and photographing (20 trays per grade x 3 grades x 10 schools x 3

PAGE 76

71 days). This sampling strategy would return 900 trays from GSW schools and 900 trays from non GSW comparison schools. We targeted 50% males and 50% females (900 males and 900 females) and 33% from each grade (600 1 st Graders, 600 3 rd graders, 600 5 th graders). Quantitative Data Collection Process Because data were collected in 10 different schools on the same d ay for three consecutive days, over 100 Research Assistants (RAs) (4 volunteers per school /per day) were utilized These RAs were Kaiser Permanente employees participating in a work related volunteer program, University of Colorado Denver graduate and undergraduate students, University of Colorado Colorado Springs undergraduate students, and LiveWell Colorado Volunteers. One lead RA for each school was trained in research proto cols during a 2 hour training. These lead RAs were present in the same school for all three days and trained all other RAs at that school. Supplies Used From School Spider Cart with removable large sheet trays From Data C ollection Team D igital cameras with extra batteries (each school had a different memory card for all data collection days) 1 tripod Base Board (a board made out of plywood shows placement of tripod and trays for photographing)

PAGE 77

72 Ziplocs (to bag random sample s for weighing) Wet wipes Protractor Latex gloves Paper towels Masking tape Sharpie markers (for marking tray cards) Cooler to transport 5 samples of each food 3 clipboards note: for ease and efficiency, pre tear 2 inch pieces of masking tape and affi x to the clipboard around the perimeter. These tape pieces were used to tape fluorescent cards to student trays. Fluorescent Tray Cards 75 for each day with correct lunch offerings printed on card. Us ed at CU Denver Digital Scale Ohaus Scout Pro SP401 Portable Digital Gram Scale APS menu planning and nutritional analysis software Plate Waste Protocols Each elementary school separates lunch periods by grade. For statistical power, we needed 20 trays per lunch period per grade (1, 3, and 5 ). Given the p otential loss of tray cards (students hiding tray cards or throwing away cards before RAs could catch them) and malfunction with photography equipment, we aimed to collect 28 trays per lunch period per grade (1, 3, and 5 ) RAs attempted to alternate betw een male and female students, although this was not always possible as clusters of same gender students often came through the line together.

PAGE 78

73 Four research assistants arrived 30 minutes prior to the first lunch period to set up. Each of the cafeterias h ad a different flow, but the main data collection protocol was the same The research assistants arrange d trash cans in a central spot and position ed a spider cart with removable trays near those trash cans to collect trays/plate waste. RAs set up a table for photography in a less traveled area of the cafeteria Upon this table RAs placed the BaseBoard (elevated 2 inches using a 2 inch binder at the end where food trays will be placed), tripod, and c amera. Two RAs position ed themselves near the end of the lunch line(s) with clipboards and fluorescent lu nch tagging cards. One RA randomly select ed the 4 th boy to exit the lunch line. The other RA randomly select ed the 4 th girl to exi t the lunch line. The RAs explain ed the purpose of the study and ga v e students the opportunity to refuse participation. kids like to eat in the cafeteria. Would it be OK if we collect your tray at the end of the your tray out at the end of lunch, just put it on that cart over there or give it to one of If the student agreed t o participate, the RA quickly check ed off the pre portioned (main entre and side items) or selected ( salad bar fruits and veggies) items on the tray using a florescent card (see example below).

PAGE 79

74 Date: April 18 School: Grade: 1 3 5 Gender: M F Beef Stew Biscuits quantity? _____ Green Apples quantity?____ Milk oz left ______ ____ skim white ____ skim chocolate ____ 1% white Salad Bar _____none _____lg svg (>1 cup) _____sm serving (< 1 cup) Other ________ Other F/V ________ quantity? _____ Other F/V _______ quantity? ______ Figure 10: Tray Cards for Quantitative Plate Waste Data Collection Th e RAs then taped the participant trays at the end of lunch. This interaction with the student took less than 30 seconds. After an RA affixed the florescent car d to a t ray, he or she took the next available student. RAs attempted to alternate between male and female students. The other two RAs position ed themselve s near the garbage cans and bega n scanning trays for florescent cards as students finish ed lunch and walk ed towards the trashcans. They took participant trays and rack ed them on Spider Trays (tall rolling carts with room for stacking trays). When the RAs a t the end of the lunch line finished flagging 28 students (approximately 14 males and 14 females) during a lunch period (usually 20 30 minute periods), they joined the RAs near the trashcans to receive finished trays. Depending on

PAGE 80

75 the flow of the cafeteria and the number of students, some RAs were able to begin the arranging and tray photographing process at this point. Photographing Trays Research assistants set up a photography area set apart from the main cafeteria activity and trashcan chaos. That area contained a table, baseboard, tripod, and digital camera ( Nikon Coolpix L24). The tripod was set 26 inches above the table and angled down on the tray at 45 degrees (protractor included in supply list). Prior to p hotographing the tray, RAs placed two letters (corresponding to the first two letters of the school) and a number (numbers began with 1 on the first data collection day, 2 0 0 on the se cond data collection day, and 3 0 0 on the third data collection day) on each fluores cent card so that trays/cards were photographed in a numerical order. RAs then pour ed any remaining milk into a measuring c up and record ed the ounces remaining to the nearest ounce on the florescent card. RAs rearrange d food on the tray to make it more clear how much of each lunch item wa s consumed. For example, if a student placed his or her leftover hamburger over the ca rrot sticks, RAs moved the hamburger to the side and arrange d the carrot sticks so the photo clearly show ed the leftover hamburger and how many carrot sticks remain. Finally, the RA took one photo of the tray with the florescent card still attached to it. The florescent card was then collected and all waste thrown in the trash. In addition to photographing all participant trays, RAs also photographed and collected three sample lunch trays for lab weig hing. The weights of these three lunch trays ( and all items on that tray) were averaged and used as the comparison serving size for all post lunch plate waste. Thus, if the av erage cheeseburger (from those 3 samples)

PAGE 81

76 was 5.5 ounces, then all plate waste from that particular school was compared to a pre lunch tray containing a 5.5 ounce cheeseburger (and nutrients associated with a 5.5 ounce cheeseburger). Aim 2: Qualitative Interview Data Collection Aim 2: Determine the relative dose, quality and adherence/fidelity of GSW implementation in each GSW school and relate implementation to observed outcomes. Qualitative data were collected from semi structured interviews with physical education teachers i n GSW schools to gather data on program implementation, impact, and effectiveness P.E. teachers a re respons ible for delivering and reinforcing much of the GSW curriculum. Individual semi structured interviews were chosen to encourage honest communication about the quality of GSW program materials and curriculum, support and assistance from school and district administrators, and individual teacher enthusiasm for and cooperation with the GSW program goals. Interviews were chosen to elicit complete responses, encourage depth of responses, and determine relative emphasis on issues. Interviews were chosen over f o cus groups because focus groups are less appropriate for these outcomes given that members might not share the same emphasis on topics and because the group dyn amics may imply ideas or emphase s that are misleading (Harrell & Bradley, 2009) Qualitative Data Collection Process Seventeen eligible physical education teachers were contacted by email in October 2012. Follow up emails were sent each week until December 1 to all teachers who did not reply to emails. Seven physical education teachers agreed to participate in 30 45 mi nute individual interviews. Interviews were conducted in November and

PAGE 82

77 December 2012 with teachers in year 1 (3 teachers), year 2 (1 teacher), or year 3 of GSW implementation (3 teachers). There were 17 teachers eligible to be interviewed for their GSW pa rticipation: three had left the district and 7 declined participation or did not answer emails Interviews were conducted a t convenient times for teachers at their individual schools. Semi structured interviews were conducted using the interview guide sh own in Appendix D Interviews were recorded and transcribed. Informal observations were als o recorded when researchers we re present in the sc hools. The lead data collector at each school filled out the field notes sheet each day of data collection. Th ese observations were recorded in field notes and include d descriptions of observed interactions in cafeterias, visual aids posted on school walls, comments from conversations with school staff, available foods in salad bars, etc. Unstructured field notes are another qualitative technique used to collect important data (Mulhall, 2003) Field notes were recorded on the field not es sheet shown in Appen dix E Analyses Quantitat ive Plate Waste Analysis: Aim 1 The average weight of each food was det ermined by weighing the three samples from each cafeteria with a digital scale. Nutrition information for each food was assessed with Vboss software ( http://www.horizon boss.com/ ) utilized by Aurora Public Schools Nutrition Services This software pack age encompass es menu planning, ordering, recipes, nutrition, accounting, point of sale, online payments, free and reduced lunch management, and more. The menu planning, recipe, and nutrition analysis features of the software provided information for the nutrition content of lunch items and meals on data

PAGE 83

78 collection days. The software analyzes nutrition data for a ll items in a lunch (recipe for an entre, any additional side items, and milk). The amount of each food item remaining was estimated in 10% increments. Various estimation methods have been used in previous visual plate waste studies including a 6 increm ent scale (all remaining, almost all remaining, remaining, remaining, remaining, none remaining) (Comstock et al., 1981; S. Dubois, 1990; Friedman & Hurd Crixell, 1999) and 10% increment scale (all, 90%, 80%, 70% etc. remaining) (Swanson, 2008; Williamson et al., 2003) A 10% increment scale was used in this research because that method was successfully used in two recent plate waste studies employing digital photogra phy (Swanson, 2008; Williamson et al., 2003) These estimates we re used to determine the nutrient intake of each student. S pecifically, if a student left 20% of a chicken breast on her plate (consuming 80%), and the chicken breast has 300 calories, 6 grams of fat, and 20 grams of protein, data for that student will re flect consumption of 240 calories (300*.80), 4.8 grams of fat (6*.80), and 16 grams of protein (20*.80). This method of nutrient estimation was repeated for total energy and 13 total macro and micronutrients, similar to data collected in previous plate wa ste research (Templeton, Marlette, & Panemangalore, 2005) Totals for each macronutrient and micronutrient were summed Nutrient t otals were recorded for both amounts taken (using information from the tray card) and amounts consumed (deduced from visual analysis of waste in digital photographs). Salad bar items were self selected in each school; therefore there was e

PAGE 84

79 tray cards. The is a sum total of the number of unique fruits and vegetables selected from both the hot lunch line and the salad bar. The ty pe of milk selected by students was recorded on each tray card (skim white, 1% white, and 1% chocolate). Two dichotomous variables were created for milk choice: ( 1) whether students chose white milk or whether they chose chocolate milk/No milk, and ( 2) wh ether students who chose milk chose white milk over chocolate milk. Plate waste photographs were analyzed by two different research assistants and compared for inter rater reliability. Similar to previous digital plate waste research, the two esti mated plate waste percentages were averaged for data analysis. If the discr epancy between the two raters was greater than 50%, the principal investigator assess ed the photograph for a third rating, and the closest two of the three ratings was used to dete rmine average plate waste (Swanson, 2008) Interr ater reliability was measured for each individual category of collected data (grade, gender, and individual food/salad bar items): 34 categories of data on Day 1, 32 on Day 2, and 44 on Day 3. Kappa coefficients for all 110 rated categories are presented in Appendix F Kappas less than 0 .20 are poor in strength of agreement, Kappas of 0 .21 0 .40 are fair, Kappas of 0.41 0.60 are moderate, Kappas of 0 .61 0 .80 are good, and Kappas of 0 .81 1.0 are considered very good in strength of agreement (Cohen, 1960) The number of Kappa coefficients in each of these categories for raw data and for data reconciled by a thir d researcher are summarized in Table 8

PAGE 85

80 Table 8: Interrater Reliability Kappa Coefficients Number of Variables with this Kappa Kappa Statistic Raw Data Reconciled Data < 0.20 2 2 0.21 0.40 10 6 0.41 0.60 26 16 0.61 0.80 6 2 0.81 1.00 66 84 Note: total number of variables rated by two raters and reconciled by a third rater=110 Statistical Analyses were performed with Stata 11 on amounts taken and amounts consumed for both GSW schools and non GSW schools. Nutrient data for each food/menu for each of the three data collection days are presented in Tables 9 12

PAGE 86

81 Table 9 : Menus for Data Collection Days April 18 Weds April 19 Thurs April 20 Fri April 24 Tues (Weather Back Up) Beef Stew new Biscuits Green apples Salad bar Milk BBQ Chicken new WH grain dinner roll Light Peaches Baked beans Salad bar Milk Big Daddy Cheese Pizza with whole grain crust new Can Green Beans Red grapes Salad bar Milk Calzone new Broccoli Red Grapes Salad bar Milk Table 10: Dat a for Menu Served on April 18 Nutrients Units Beef Stew/ 100g Beef Stew/ 1 cup serving Angel Biscuits /100g Angel Biscuits/ 1oz biscuit Apple Fresh/ 100g Apple Fresh/ 1/2 cup 1% White Milk/ 100g 1% Wh Milk/ 8oz or 227g Entire School Lunch Units Adjusted weight gm 100 340.04 100 33.79 100 138 100 226 737.83 gm Energy kcals 105.98 360.37 270.38 91.36 52 71.76 44.09 100 623.49 kcals Protein g 6.21 21.11 6.56 2.22 0.26 0.36 3.53 8 31.69 g Vitamin A (RE) RE 132.19 449.49 0 0 5.0 6.9 44.09 100 556.39 RE Vitamin A (IU) IU 659.62 2242.96 0 0 54.0 74.52 220.5 500 2817.4 8 IU Vitamin C mg 2.72 9.25 0 0 4.6 6.35 1.06 2.4 18 mg Iron mg 1.24 4.23 1.1 0.373 0.12 0.17 0 0 4.773 mg Calcium mg 28.35 96.39 10.24 3.46 6.0 8.28 132.28 300 408.13 mg Total Fat g 2.45 8.33 7.16 2.41 0.17 0.23 1.1 2.5 13.47 g Saturated Fat g 0.72 2.45 1.93 0.65 0.03 0.04 0.66 1.5 4.64 g Trans Fat g 0.00 0 2.0 0.68 0.00 0 0 0 0.68 g Carbohydrate g 14.52 49.38 44.96 15.19 13.81 19.06 5.29 12 95.63 g Total Fiber g 1.82 6.17 0.78 0.27 2.4 3.31 0 0 9.75 g Cholesterol mg 8.94 30.38 0.90 0.03 0.00 0 4.41 10 40.41 mg Sodium mg 197.32 670.95 393.22 132.87 1.0 1.38 55.12 125 930.2 mg

PAGE 87

82 Table 11: Data for Menu Served on April 19 Nutrients Units BBQ Chicken/ 100g BBQ Chicken/ 1 piece Whole Grain Dinner Roll/ 100g Whole Grain Dinner Roll/ 1 oz roll Light Peach/ 100g Light Peach/ cup Baked Beans/ 100g Baked Beans/ 1 cup 1% White Milk/ 100g 1% White Milk/ 8oz carton or 227g Entire School Lunch Units Adjusted weight gm 100 192.85 100 42.47 100 120.20 100 257.06 100 226 838.58 gm Energy kcals 214.18 413.04 249.72 106.06 55.56 66.78 107.69 276.83 44.09 100 962.71 kcals Protein g 15.86 30.58 6.3 2.68 0.00 0 4.62 11.86 3.53 8 53.12 g Vitamin A (RE) RE 9.30 17.94 0.0 0 47.62 57.24 0.00 0 44.09 100 175.18 RE Vitamin A (IU) IU 46.51 89.69 0.0 0 238.10 286.19 0.00 0 220.5 500 875.88 IU Vitamin C mg 0.74 1.41 0.01 0.00 0.95 1.14 0.00 0 1.06 2.4 4.95 mg Iron mg 0.34 0.66 1.84 0.78 0.00 0 1.39 3.56 0 0 5 mg Calcium mg 1.51 2.92 10.28 4.37 0 0 30.77 79.09 132.2 8 300 386.38 mg Total Fat g 15.82 30.50 5.03 2.14 0 0 0.77 1.98 1.1 2.5 37.12 g Sat. Fat g 4.99 9.62 0.98 0.41 0 0 0.00 0 0.66 1.5 11.53 g Trans Fat g 0.00 0 0.00 0 0 0 0.00 0 0 0 0 g Carbo hydrate g 1.60 3.08 44.85 19.05 13.49 16.22 22.31 57.34 5.29 12 107.69 g Total Fiber g 0.07 0.14 0.51 0.22 0 0 3.85 9.89 0 0 10.25 g Cholesterol mg 53.93 104.00 0.05 0.02 0 0 423.08 0 4.41 10 114.02 mg Sodium mg 142.82 275.42 432.48 183.69 7.94 9.54 0.00 1087.6 55.12 125 1681.2 5 mg

PAGE 88

83 Table 12: Data for Menu Served on April 20 Nutrients Units Big Daddy Cheese Pizza/ 100g Big Daddy Cheese Pizza/1 slice Canned Green Beans/ 100g Canned Green Beans/ cup Red Grapes/ 100g Red Grapes/ cup 1% Wh Milk/ 100g 1% White Milk/ 8oz carton or 227g Entire school lunch Units Adjusted weight gm 100 135 100.00 128.85 100.00 90.72 100 226 580.57 gm Energy kcals 237.04 320 16.67 21.48 67.00 60.78 44.09 100 502.26 kcals Protein g 14.07 19 0.83 1.06 0.63 0.57 3.53 8 28.63 g Vitamin A (RE) RE 55.90 41.67 53.69 10.00 9.07 44.09 100 162.76 RE Vitamin A (IU) IU 279.51 208.33 268.43 100.00 90.72 220.5 500 859.15 IU Vitamin C mg 0.00 0.83 1.08 4.00 3.63 1.06 2.4 7.11 mg Iron mg 1.79 0.79 1.01 0.29 0.26 0 0 1.27 mg Calcium mg 223.61 36.66 47.24 14.00 12.70 132.2 8 300 359.94 mg Total Fat g 6.67 9 0.00 0 0.35 0.32 1.1 2.5 11.82 g Saturated Fat g 2.59 3.5 0.00 0 0.11 0.10 0.66 1.5 5.1 g Trans Fat g 0.00 0.00 0 0.00 0 0 0 0 g Carbohydrate g 28.89 39 3.31 4.26 17.15 15.56 5.29 12 70.82 g Total Fiber g 2.96 4 1.65 2.13 0.90 0.82 0 0 6.95 g Cholesterol mg 11.11 15 0.00 0 0.00 0 4.41 10 25 mg Sodium mg 362.96 490 322.31 415.30 2.00 1.81 55.12 125 1032.1 1 mg Although 13 different nutrients are reported for every food served in the Aurora Public School Cafeterias, not all 13 nutrients would be expected to vary greatly based on GSW program participation. As a whole, students have few choices in the hot lunch line scra tch cooking changes across the entire district. Students do, however, have the ability to choose as many fruits and vegetables f rom the salad bar as they as they want Given

PAGE 89

84 that the GSW curriculum (incentive days, menu review, activities in class) focuse s on with less processing more whole grain, more whole foods (i.e. more fruits and vegetables), and less added sugar, it is most realistic that students in GSW schools would choose more fruits and vegetables, would choose milk (and whi te milk instead of chocolate milk ) and would consume less fat/saturated fat and more fiber from eating more Go foods on their tray. For example, Chocolate Cake was served on Day 2 of this study. GSW administrators would hypothesize that fewer students f rom GSW schools would take cake (and therefore less total fat), and would take more fruits and vegetables, thus increasing their intake of fiber If they chose not to take or to take but not eat all of the cake (average piece of cake across all schools=15 5 calories), those 1% or skim white milk) or more fruits and vegetables (varied caloric content). They might have also chosen to take or drink more white milk regardless of cake consumption, simply because white milk is recommended as a GO food and because an incentive day is held each year where students earn a prize for drinking white milk. The refore, descriptive statistics we re included for all total nutrient, milk, a nd fruit/vegetable variables, but statistical models we re run on only the following most relevant dependent variables that GSW is expected to influence: Did students choose white milk (skim and 1%) over chocolate milk or no milk? For students choosing milk did they choose white milk (skim and 1%) over chocolate milk? Percentage of white milk consumed. Total number of fruits/vegetables taken.

PAGE 90

85 Total fiber taken. Total fiber consumed. Total fat taken. Total fat consumed. Total saturated fat taken. Total saturated fat consumed. There is an important distinction between milk, fiber, fat, and saturated fat taken and consumed Students food choices in the lunch line may b e influenced by GSW curriculum but other factors may come into play once they beg in to eat that might make their actual food consumption lower on these key variables (white milk, high fiber, FV, lower fat options). These factors, detailed in the literature review, include: length of the lunch period, recess before lunch, pee r role mod eling, taste, smell, and t eacher/staff role modeling. The two dependent variables for choosing white milk are dichotomous dependent variables, whereas the remaining eight dependent variables are linear variables. Separate s tatistical model s were calculated for each of the 10 dependent variables listed above Because of the hierarchical nature of collecting data from students nested in schools, multilevel modeling wa s appropriate for our analyses. S tudent responses to an intervention are likely to be more correlated for students in the same school than for students in different schools, thus violating the assumption of independent responses and necessitating a statistical test other than traditional ANOVA and ordinary least squares regression. A 2 level multilevel regression model can generate more accurate standard errors for program effect estimates and other important parameters in clustered designs

PAGE 91

86 than other statistical techniques like linear regression that do not account for data clusteri ng. (Rabe Hesketh & Skrondal, 20 08; Raudenbush & Bryk, 2002) Multilevel models have more error terms and therefore more flexibility in defining the covariance structure. This flexibility leads to two distinct advantages: (1) more flexibility in the questions researchers ask about t he covariance structure and because of the complexity of covariance structures able to be modeled, (2) better estimates of standard errors of the regression coefficients and resulting accuracy of confidence intervals (Berkhof & Kampen, 2004; Dedrick et al., 2009) Multilevel models also ensure more accurate effect sizes for an intervention and help avoid Type 1 errors and biased parameter estimates (Peugh, 2010; Wampold & Serlin, 2000) However, not all nested datasets require mul tilevel modeling Multilevel modeling is necessary only if response variables vary across level 2 units (e.g. schools) (Peugh, 2010) In this study, r es ponse variable data wa s likely to differ across schools (mostly due to school characteristics such as demographics, achievement scores, and characteristics of PE teachers and other school personnel), therefore multilevel modeling was selected for analyses V ariance in the level 2 data was assessed by measuri ng intraclass correlation (ICC) In this study, the ICC is both the proportion of nutrient /food values variation that occurs across schools (level 2 units) and the expected correlation between the n utrient /food values of two students (level 1 units) from the same school (Peugh, 2010) The ICC is similar to the R 2 effect size from regression and repre sents the proportion of student nutrient score variance that can be explained by mean nutrient score differences across schools. The ICC provides help in determining the necessity of multilevel modeling An ICC value of zero indicates: (a) no mean nutri ent score variation across

PAGE 92

87 schools (level 2), (b) all nutrient score variation occurs across students (i.e. level 1), and (c) traditional analysis techniques such as ANOVA and regression can be used to analyze the student data (Peugh, 2010) As the ICC value increases, the proportion of dependent variable score variation that occurs across schools increases, resulting in violations of the independence assum ption and requiring multilevel modeling to analyze data. Given that the ICC>0 in all models used in this study a multilevel model that utilized to determine if mean nutrient value s differ ed significantly across schools. The most basic multilevel model to explain the potential differences in nutrient /food scores is shown with the following equations (Hox, 2010; Raudenbush & Bryk, 2002) : Equation 1: Level 1 : Y ij 0j + r ij Equation 2: Level 0j 00 + u 0j Equation 1 shows that the differences in nutrient score of student i in school j (Y ij ) can be modeled as 0j ) plus a residual term that reflects individual student differences around the mean nutrient score of school j ( r ij ). Equation 2 shows that the mean nutrient score for school j ( 0j ) can be modeled as a function of the grand mean nutrient score ( 00 ) plus a school specific deviation from the grand mean ( u 0j ). Substituting Equation 2 into Equation 1 yields the combined unconditional multilevel modeling equation: Combined multilevel modeling equation: Y ij 00 + u 0j + r ij This equation combines nutrient score variability into within group (i.e., level 1, r ij ) and between group (i.e., level 2, u 0j ) components.

PAGE 93

88 After estimating a multilevel model, the ICC is calculated with the following: ICC 00 00 2 ) 00 is the variance in nutrient score means across schools and 2 is the nutrient score variation across students within a school. An ICC predicts the proportion of variance that occurs across schools, and typically an I CC value between .05 and .20 is common in cross sectional multilevel modeling applications in social research studies (Peugh, 2010) In this research, ICC values ranged from 0.04 0.47. It should be noted that a random effects multilevel linear regression was not appropriate for one outcome variable. The percent of white milk consumed represented a proportion variable bound on the upper and lower ends by zero and one. This ou tcome variable was analyzed with a generalized linear model with logit link and a binomial outcome ranging from zero to one and clustering on school ID Although it did not estimate the intraclass correlation, this model allowed for correct estimation of standard errors and analysis of the overall effect of the intervention and student level covariates. To determine the extent to which GSW is associated with the outcome independent of indiv idual student characteristics, this study used multilevel modelin g to partition the v ariance between schools and student level characteristics for nine outcome variables Models were run with restricted maximum l ikelihood (REML) estimation because of the ability of REML to attain more accurate variance estimates in smaller sample sizes (Peugh, 2010; Raudenbush & Bryk, 2002) REML treats regression coefficients as unknown quantities and estimates coefficients and variance estimates from sample data, subtracti ng the appropriate amount of de grees of freedom. Conversely, maximum likelihood e stimation treats regression coefficients as known population

PAGE 94

89 parameters and does not allocate any degrees of freedom to the estimation (Peugh, 2010) The resulting variance estimates from maximum likelihood e stimation are underestimated, especially when the sample size is small. Brown and Draper (2000) show that restricted ma ximum likelih ood estimation provide s r easonable variance estimates with as few as 6 12 groups (Browne & Draper, 2000) Because the effect of individual schools on overall results was unknown, schools were treated as a random effect in our model. Student level charact eristics (grade and gender) were added as fixed effects to examine their independent contributions to food and nutrient consumption. Treatment condition was added as both a random effect to some models and a fixed effect to others and models were compared to determine the best fitting model. Analyzing treatment condition (GSW or non GSW) as a random effect utilized a random coefficient model where the effect of GSW was allowed to vary across schools, producing a different coefficient for each school. Ana lyzing treatment condition as a fixed effect utilized a random intercept model where the effect of GSW was reported as an intercept, or a mean effect for all schools. Testing whether the random coefficient model fit better than the nested random intercept model was accomplished in Stata 11 with a likelihood ratio test of the two models (Rabe Hesketh & Skrondal, 2008) Of the nine dependent variables, none of the likelihood ratio tests were significant, indicating that the random intercept model should not be rejected in favor of the random coefficient model. Therefore treatment condition was treated as a fixed effec t in all models ( Akaike,

PAGE 95

90 1974) ; Schwarz 1978 ) AIC = 2 LL +2 p where LL is the log likelihood and p is the number of ML estimated parameters. BIC= 2 LL + p ln(N) where N is the sample size at Level 1. Both the AIC and BIC start with the log likelihood value and subsequently penalize for the number of covariance parameters estimated. The BIC creates a bigger penalty for the number of parameters estimated (Dedrick et al., 2009) AIC and BIC for all 10 random coefficient models and all 10 random intercept models are shown in Appendix G Results from likelihood ratio tests and AIC/BIC comparisons favor the random intercept models over random coefficient models, treating schools as a random effect and treatment condition, grade, gender, and day of the study /entree as fixed effects. The final model s in this study contain both fixed and random effects and are considered mixed effect s model s O ther post estimation commands were used in this study to show the superiority of multilevel modeling over stand ard regression techniques in the final models. Using Stata 11 likelihood ratio (LR) tests for a ll ten final multilevel models compared multilevel modeling (Linear and Logistic Mixed Effects Model s ) to standard linear and logistic regression. The LR test is testing whether an estimated variance component is different from zero, and whether or not multilevel modeling is a better fitting model than standard linear or logistic regression. All ten likelihood ratio tests were significant at p=.001, indicating that multilevel modeling was superior to standard linear and logistic

PAGE 96

91 regression for all models in this study. ICC and likelihood ratio tests were utilized to determine that multilevel modeling was the most appropriate modeling technique for collected data. Several assumptions should be met for multilevel modeling The normality assumption was assessed in all models by examining the dist ribution of Level 1 residuals. Deviance residuals were calculated for Level 1 residuals in the two white milk logit models as recommended by previous studies (McCullagh & Nelder, 1989) Deviance residuals have the be st properties for examining goodness of fit. Residual graphs for all multileve l models are found in Appendix H Data were screened for outliers to identify data entry errors, inaccurate coding of a student, missing values coded incorrectly, or simply individuals who are different from others students in the sample. Formulae for Multi Level Model. Taken Separate multilevel models were created for all outcome variables ( 9 in total ). Model 1: 1 X 1 (GSW particip ation Level2 Main Effect ) + b 2 X 2 (Gender Level 1) + b 3 X 3 (Grade Level 1) + b 4 X 4 (Day of Data Coll. Level 1) + Model 2: 1 X 1 (GSW participation Level 2 Main Effect ) + b 2 X 2 (Gender Level 1) + b 3 X 3 (Grade Level 1) b 4 X 4 (Day of Data Coll. Table 13 : Multi level Model for Control and Main Effect Predictor Variables Predictor Variables (Main Effects) GSW Participation Level 1 Control Variables Gender Grade Level Day 1, 2, or 3 Level 2 Control Variables None school sample size too small

PAGE 97

92 Adding in additional predictor variables at the school level (i.e. mean school SES, ELL, etc.) was not possible due to the very small number of schools (10) included in this study. Qualitative Data Analysis: Aim 2 Using a combination of ded uctive and ind uctive methods, salient themes were identified related to teacher opinions regarding Scratch Cooking Cafeteria Changes and GSW implementation and effectiveness. Teacher interviews were analyzed using codes emerging from literature/hypotheses and in vivo codes emerging from collected data, and subsequently determining themes within and across schools. A qualitative analysis was used with a five stage iterat ive process to analyze the t ranscript s : (1) development of a coding schedule; (2) coding of the data; (3) description of the main categories; (4) linking of categories into major themes; and (5) the development of explanations for the relationships between themes (Hsieh & Shannon, 2005; Pope, Ziebland, & Mays, 2000) An ecological framework for program implementation was utilized as a lens through which to analyze the implementation and impact of GSW (Durlak & DuPre, 2008) Interview transcripts were analyzed within the context of three main categories of this ecological framework: innovations/pr ogram characteristics, individuals and communities, and features associated with program delivery and program support. For all interviews, codes were generated and recorded in a table. This table also include s excerpts from teacher comments relating to ea ch code. The r esulting codes and comments were grouped, as appropriate, into overall themes and findings from

PAGE 98

93 qualitative data collection. These qualitative data and analyses assisted in several ways including: (1) helping to make sense of quantitative data; and (2) documenting dose, implementation fidelity, likability and impact of the GSW program in APS. GSW imple mentation (dose, quality, adhere nce/fidelity) was also assessed using informal conversations with cafete ria staff and school personnel and observations b y researchers while in schools The implementation of GSW programming was assessed by coding transcribed interviews organizing them into themes, and drawing conclusions from these themes. Implementation o f Scratch Cooking Cafeteria Changes was also assessed using observations by rese archers during data collection days The implementation of Scratch Cooking Cafeteria Changes should be similar across schools, as all schools follow the same district wide men us and food ordering is done in a central location for all schools. However, given the variability that could occur due to shipping Changes, there could be variability in i mplementation. Data from collected field notes assessed menu similarity across schools. Qualitative data helped to inform the interpretation of quantitative analyses. Linking qualitative data to quantitative data has several benefits: (a) to enable conf irmation or corroboration via comparison; (b) to elaborate or develop analysis, providing richer detail; and (c) to initiate new lines of thinking through attention to surprises or paradoxes (Rossman & Wilson, 1994) Analyzing qualitative interviews and observations and quantitative plate w aste data together provided far greater information than either method of data collection alone.

PAGE 99

94 Human Subjects Review Because this study used original data collected with children in Aurora Public schools, proper safeguards were taken to ensure the safety of all study participants. Approval from the Colorado Multiple Institutional Review Board (COMIRB), LiveWell Colorado, and Aurora Public Schools was obtained before data collection began. Students were asked for verbal info r med consent in the lunch line. Student names were not collected, and faces were not photographed.

PAGE 100

95 CHAPTER I V QUANTITATIVE RESULTS Descriptive Statistics Of 2,223 collected and photographed lunch trays, 1123 (51%) came from GSW schools and 1100 (49%) came from non GSW comparison schools. The samp le was split fairly evenly among 1 st graders, 3 rd graders, and 5 th graders and among male and female students. Among the entire sample of all schools over three days, the average number of calories taken in the hot lunch line was 553 (s.d. 134), whereas the average number of calories consumed by the end of lunch was 292 (s.d. 154). The characteristics of lunch tr ays collected from intervention and comparison groups are shown in Table 14

PAGE 101

96 Table 14: Characteristics of GSW and Non GSW Schools GSW School s Comparison School s Level: Student Students, n 1123 1100 Gender no./total no. (%) Male 512/1038 (49) 526/1038 (51) Female 528/1048 (50) 520/1048 (50) Missing Data 77/137 (56) 60/137 (44) Grade no./total no. (%) Grade 1 374/755 (50) 381/755(50) Grade 3 356/734 (49) 378/734 (51) Grade 5 369/728 (51) 359/728 (49) Missing Data 1/6 (17) 5/6 (83) Tray Contents Took White Milk no./total no. (%) 254/386 (66)* 132/386 (34)* % White Milk Consumed mean (s.d.) .55 (0.43) .53 (0.40) Total FV Taken mean (s.d.) 1.17 (0.86)* 1.44 (0.84)* Total Fiber Taken mean ( s.d.)** 5.12 (2.66)* 5.85 (2.87)* Total Fiber Consumed mean (s.d.)** 2.00 (1.35)* 2.23 (1.62)* Total Fat Taken mean (s.d.) 19.82 (11.34)* 17.89 (9.04)* Total Fat Consumed mean (s.d.) 10.83 (7.52)* 10.04 (7.12)* Total Saturated Fat Taken mean ( s.d.) 7.07 (3.01)* 6.27 (2.73)* Total Saturated Fat Consumed mean (s.d.) 4.02 (2.64)* 3.62 (2.65)* Total Calories Taken mean (s.d.) 540 (121.00)* 566 (143.65)* Total Calories Consumed mean (s.d.) 294.79 (145.23) 290.13 (161.58) Level: School Schools, n 5 5 Race/Ethnicity White % of students in school 18.3 19.16 Black % of students in school 17.62 17.86 Asian % of students in school 3.18 3.52 Hispanic % of students in school 56.74 54.74 Native American % of students in school 0.96 0.88 Native Hawaiian % of students in school 0.14 0.3 Two or more Race/Ethn % of students in school 3.04 3.54 Free/Reduced Lunch % of students in school 75.74 72.72 Gifted and Talented % of students in school 2.02 2.28 Special Education % of students in school 8.86 9.04 English Language Learners % of students in school 48.34 45.56 Number of Nutrition Programs no. in school 0.8 0.8 Significant mean differences between tray contents of GSW Schools and Comparison Schools (p<.05) ** Total Fiber Taken and Consumed does NOT take into account salad bar items. While the average s of all dependent variables are organized by intervention and comparison groups and reported in T able 14 individual averages for each school are

PAGE 102

97 important when assessing program implementation and program outcomes. Individual averages are also important for school administrators teachers, and parents at each school. The GSW program is expected to influence the choosing and consumption of white milk, total fruits and vegetables, fiber, total fat, and saturated fat. The se particular characteristics of lunch trays collected from each school are shown in T able 15 Graph s of these tray contents at each school are shown in Figures 11 19

PAGE 103

98

PAGE 104

99 Figure 11: Students Taking White Milk Figure 12: White Milk Consumed Figure 13: Total Fruits and Vegetables Taken 0 20 40 60 Altura Arkansas Crawford Fletcher Jewell Lansing Vassar Wheeling Percentage % of students who took white milk over no milk/chocolate milk GSW Schools Non-GSW Schools 0 20 40 60 80 100 Altura Arkansas Crawford Dartmouth Fletcher Jewell Lansing Vassar Vista Peak Wheeling % consumed of 8oz carton Of those taking white milk, % white milk consumed of 8oz carton GSW Schools Non-GSW Schools 0 0.5 1 1.5 2 2.5 Altura Arkansas Crawford Dartmouth Fletcher Jewell Lansing Vassar Vista Peak Wheeling Total FV Taken Total Fruits and Vegetables Taken GSW Schools Non-GSW Schools

PAGE 105

100 Figure 14: Total Fiber Taken Figure 15: Total Fiber Consumed Figure 16: Total Fat Taken 0 5 10 15 20 25 30 Altura Arkansas Crawford Dartmouth Fletcher Jewell Lansing Vassar Vista Peak Wheeling Grams of Fat Total Fat Taken GSW Schools Non-GSW Schools 0 2 4 6 8 10 Altura Arkansas Crawford Dartmouth Fletcher Jewell Lansing Vassar Vista Peak Wheeling Grams of Fiber Total Fiber Taken GSW Schools Non-GSW Schools 0 2 4 6 8 Altura Arkansas Crawford Dartmouth Fletcher Jewell Lansing Vassar Vista Peak Wheeling Grams of Fiber Total Fiber Consumed GSW Schools Non-GSW Schools

PAGE 106

101 Figure 17: Total Fat Consumed Figure 18: Total Saturated Fat Taken Figure 19: Total Saturated Fat Consumed 0 10 20 30 Altura Arkansas Crawford Dartmouth Fletcher Jewell Lansing Vassar Vista Peak Wheeling Grams of Fat Total Fat Consumed GSW Schools Non-GSW Schools 0 2 4 6 8 Altura Arkansas Crawford Dartmouth Fletcher Jewell Lansing Vassar Vista Peak Wheeling Grams of Saturated Fat Total Saturated Fat Consumed GSW Schools Non-GSW Schools 0 2 4 6 8 10 Altura Arkansas Crawford Dartmouth Fletcher Jewell Lansing Vassar Vista Peak Wheeling Grams of Saturated Fat Total Saturated Fat Taken GSW Schools Non-GSW Schools

PAGE 107

102 The overall numbers of male trays and female trays included in this study are similar, with 1048 females (47.1%), 1038 males (46.7%) and 137 missing (6.2%) due to research assistant oversight However, the breakdown of males and females by school, day of data collection, and grade di d not result in such even gender numbers of tray cards. Although researchers aimed to collect 14 trays from female students and 14 trays from male students per grade per day, the dynamics of school cafeterias and clusters of same gender students in line p revented such precision. When looking at the sample size at the lowest level (number of students per grade on one day at one school), the cell sizes for gender (which ideally would be 14 for each gender) range from 2 to 20. Of 180 sample size cells (10 s chools x 3 days of data collection x 3 grades x 2 gend ers), 41 (23%) cells had sample sizes less than 10. The other 139 (77 %) had sample sizes between 10 and 20. Fourtee n of the 180 cells (8%) had fewer than 8 students. Overall, only 6 out of 2223 tray cards were missing grade level information. Table 16 show s the gender breakdown by school, day, and grade.

PAGE 108

103 Table 16: Sample Sizes Number of Trays Collected School Grade Day 1 No. Trays Day 2 No. Trays Day 3 No. Trays reported as: no. females[f], no. males[m], no. missing sex[*] Altura 1 23 (10f,11m, 1*) 25 (16f,9m,0*) 25 (10f,10m,5*) 3 25 (15f,6m,4*) 26 (17f,3m,6*) 25 (15f,7m,5*) 5 22 (11f,9m,2*) 25 (15f,10m,0*) 23 (10f,11m,2*) Arkansas 1 25 (13f,10m,2*) 26 (11f,14m,1*) 25 (15f,10m,0*) 3 23 (13f,9m,1*) 26 (9f,16m,1*) 22 (10f,11m,1*) 5 25 (10f,11m,4*) 24 (8f,12m,4*) 25 (11f,14m,0*) Crawford 1 26 (13f,13m,0*) 25 (13f,11m,1*) 26 (13f,10m,3*) 3 24 (14f,9m,1*) 22 (14f,6m,2*) 26 (12f,12f,2*) 5 18 (9f,9m,0*) 26 (15f,11m,0*) 28 (18f,9m,1*) Dartmouth 1 25 (9f,16m,0*) 25 (8f,15m,2*) 27 (10f,14m,3*) 3 23 (7f,14m,2*) 2 1 (7f,14m,0*) 26 (8f,14m,3*) 5 28 (11f,16m1*) 26 (9f,17f,0*) 21 (7f,11m,3*) Fletcher 1 25 (14f,8m,3*) 26 (14f,10m,2*) 26 (14f,9m,3*) 3 26 (12f,14m,0*) 26 (14f,12m,0) 27 (11f,15m,1*) 5 24 (13f,11m,0*) 24 (13f,11m,0*) 26 (12f,13m,1*) Jewell 1 27 (17f,9m,1*) 25 (19f,2m,4*) 24 (16f,8m,0*) 3 26 (16f,10m,0*) 25 (11f,13f,1*) 25 (13f,11m,1*) 5 17 (8f,9m,0*) 23 (12f,11f,0*) 23 (13f,10m,0*) Lansing 1 24 (11f,13m,0*) 27 (12f,15m,0*) 26 (11f,14m,1*) 3 26 (7f,18m,1*) 27 (10f,17m,0*) 26 (6m,19f,1*) 5 25 (7f,16m,2*) 25 (10f,13m,2*) 26 (10f,16m,0*) Vassar 1 23 (9f,6m,8*) 26 (13f,11m,2*) 24 (9f,15m,0*) 3 0 (field trip this day) 26 (7f,16m,3*) 25 (10f,13m,2*) 5 26 (11f,15m,0*) 26 (12f,13m,1*) 26 (13f,13m,0*) Vista Peak 1 25 (11f,13m,1*) 23 (14f,9m,0*) 26 (13f,13m,0*) 3 26 (15f,11m,0*) 26 (14f,11m,1*) 26 (13f,13,0*) 5 25 (13f,12m,0*) 24 (13f,11m,0*) 24 (14f,10m,0*) Wheeling 1 2 7 (8f,3m,16*) 23 (9f,14m,0*) 25 (11f,12m,2*) 3 31 (11f,20m,0*) 25 (8f,14m,3*) 26 (14f,7m,5*) 5 25 (15f,10m,0*) 25 (12m,12f,1*) 23 (12f,9m,2*) Total: 715 (333f,331m,51*) 755 (360f,354m,41*) 753 (355f,353m,45*) Aim 1: Determine the relative impact of GSW and Scratch Cooking Cafeteria Changes over and above the impact of Scratch Cooking Cafeteria Changes alone on n of school lunch Hypothesis 1 a: Students receiving the GSW program will choose more health y foods at school than non GSW students Hypothesis 1b: Students receiving the GSW program will consume more of the healthy foods on their school lunch tray than non GSW students.

PAGE 109

104 Hypothesis 1c: Students receiving the GSW program will choose and consume mo re overall fruits and vegetables, white milk, and fiber, and less fat and saturated fat than non GSW students. Results R elated to Hypotheses Effects of GSW on Healthy Food Choice and Consumption Student s enrolled in schools with the GSW program showed significantly increased odds of choosing white milk (OR=2.17 p= 0 .05 ) over chocolate milk or no milk when controlling for gender, grade, and day of data collection (with different food offerings each da y). None of the other models produced significant differences between students in GSW schools and students in non GSW schools when controlling for gender, grade, and day of data collection. However, although the differences are not significant, students in GS W schools dra nk more milk (Coeff= 0 .07, p= 0 .52 ) took fewer vegetables (Coeff= 0.26, p= 0 .10 ) took and consume d less grams of fiber (Coeff= 0.63, p= 0 .043; Coeff= 0.25, p=0.42 ) took and consume d more total grams of fat (Coeff=1.45, p=0.40; Coeff= 0 .35, p=0.6 6 ) and took and consume d more grams of saturated fat ( Coeff= 0.64, p=0.21; Coeff= 0.26, p=0.30 ) than students in non GSW schools Additional Results Effects of Grade Significant effects of grade were found for choosing white milk, the amount of white milk consumed, total fruits and vegetables taken, total fiber consumed, total fat consumed and total saturated fat consumed. Students in grade five took more white milk t han students in grade one (Coeff=0.69, p= 0.02 ). Students in grade three and grade five drank more of their white milk than students in grade one (Coeff=0.56, p= 0.00 ;

PAGE 110

105 Coef f=0.78, p= 0.00 ). Students in grade three and five took less fruits and vegetables th an students in grade one (Coeff = 0.12, p=0.00; Coeff= 0.07, p= 0.00 ). Students in grades 3 and 5 consumed more total fiber than students in grade one (Coeff= 0 .36, p= 0.00 ; Coeff= 0 .45, p= 0.00 ). Students in grade s three and five consumed more total fat than students in grade one (Coeff= 1.06, p= 0.00 ; Coeff=0.78, p= 0.00 ) and more saturated fat than students in grade one (Coeff=0.52, p= 0.00 ; Coeff=0.41, p= 0.00 ). Effects of Gender Significant effects of gender were found for choosing white milk, total fiber consumed, total saturated fat consumed, and total fat consumed. Boys took white milk less often than girls (OR=0.78, p= 0.04), consumed less total fruits and vegetables (Coeff= 0.08, p=0.04) and consumed more total fiber than girls (Coeff= 0.14, p= 0 02 ). B oys consumed more total fat and saturated fat than girls (Coeff=0.93, p= 0.00 ; Coeff=0.43, p= 0.00 ). Effects of Meal/Day of Study Significant effects were also found for day of the study (with three different menus for each of the three days). Students cho se significantly more white milk on pizza day compared to beef stew day (OR=2.11, p=0.00) took significantly less fruits and vegetables on chicken day and pizza day compared to beef stew day (Coeff= 0.27, p=0.00; Coeff= 0.13, p=0.00) took significantly l ess fiber on chicken day and pizza day compared to beef stew day (Coeff= 2.47, p=0.00; Coeff= 5.21, p=0.00) and consumed less fiber on chicken day and pizza day compared to beef stew day (Coeff= 0.62, p=0.00; Coeff= 0.29, p=0.00) Students took more tota l fat on chicken day and pizza day compared to beef stew day (Coeff= 21.38, p=0.00; Coeff 6.57, p=0.00) consumed more

PAGE 111

106 total fat on chicken day and pizza day compared to beef stew day (Coeff= 10.72, p=0.00; Coeff=7.79, p=0.00 ), took more saturated fat on chicken day and pizza day compared to beef stew day ( Coeff=5.59, p=0.00; Coeff=4.54, p=0.00 ) and consumed more saturated fat on chicken day and pizza day compared to beef stew day (Coeff=2.55, p=0.00; Coeff=4.25, p=0.00). Table 17 shows the multivariable m odel s for all outcome variables a nd the adjusted odds ratios, coefficients and 95% confidence intervals.

PAGE 112

107

PAGE 113

108 CHAPTER V QUALITATIVE RESULTS Aim 2: Determine the relative dose, quality and adherence/fidelity of GSW implementation in each GSW school and relate implementation to observed outcomes. Hypothesis 2a: Students in schools with higher dose and better quality and adhere nce/fidelity of GSW programming will show more positive results of GSW prog ramming (Hypotheses 1a, 1b, 1c ). Analyses of quali tative interviews revealed several important themes related to the GSW progr am and program implementation. These themes include program leadership, program details and information personal motivation and enthusiasm for the GSW program, lack of time for GSW curriculum, administration support, influence of parents and family on food intake, and student responses to scratch cooking changes and GSW components. An ecological framework for program implementation will be utilized as a lens through whi ch to anal yze the implementation of GSW (Durlak & DuPre, 2008) Three main categories of this ecological framework include: innovations/program characteristics, individuals and commu nities, and features associated with program delivery and program support. Although qualitative interviews revea led important themes related to all three framework groups the third category, features of the program deli very and support system, had the gr eatest impact on a cascade of negative GSW program implementation issues and program outcomes. Although the program itself along with the teachers, administrators, and families in each school have a

PAGE 114

109 role in program effectiveness, this analysis primarily f ocused on program delivery and the program support system Program Delivery and Support System Overall Program Goals and Expectations The most commonly reported theme was insufficient district leadership for the GSW program, and specifically uncertainty regarding the overall goals of the GSW program, the individual components included in ideal implementation, and the components of the GSW program falling under PE teacher responsibility. Insufficient leadership is defined as insufficient materials, commu nication, motivation, follow up, technical assistance, and involvement for optimal program success. Outcomes related to this insufficient leadership include poor shared understanding of program goals and components resulting in varying dose and fidelity o f program implementation. Uncertainty re lated to GSW program goals and ideal imple mentation was reported by five of seven physical education teachers. The program is kicked off in the fall at any new schools beginning the GSW program with a school wide a ssembly. This assembly is scripted with exact lines for PE teachers, 9 News spokespeople, and mascots (the Avalanche Mascot and GSW specific Power Panther). Although teachers appreciate the clarity of assembly materials and ease of assembly implementatio n, most reported uncertainty as to expectations following the assembly. One teacher reported not doing one single thing related to GSW this school year between the date of the assembly (in September) and the date of the interview (in December). Other tea chers reported a lack of understanding of the goals of the GSW pr ogram and of what, exactly, P.E. teachers are responsible for teaching and doing in their classrooms.

PAGE 115

110 W we were already menu. (Teacher F) T his particular teacher began the GSW program with an assembly this year, but continues to be confused by what is expected of the P.E. teacher and the other people involved with GSW at each school. t incentive day. Okay, so am I doing that? Or is my principal doing that? Is it just wit h in the bui l ding? Am I supposed to be doing this? Because the same thing happened with the su r (Teacher F) One teacher who has been implementing GSW curriculum for three years had no idea that the school l unch menus were coded with GSW labels. Other teacher s identified uncertainty in the program and program implementation. et up or how it was being done. (Teacher D) Teachers have lots of important health information from various programs (Flat Fourteeners, GSW, 5 th Gear Kids pacing guide standards) and struggle with how to fit them all in with little guidance. I got p figure out how I can make all three programs run at the same time, yet no one is around to help or to show you. I t would just be nice to k ind of have a meeting. (Teacher D) Another teacher reported: I mean, you guys gave basic ideas but even then you really still need to you do in first gra de, this is what you do with second grade, that works for

PAGE 116

111 now, for a lot of people that are still trying to come in and figure out how they are going to teach every day, to have something in a book, concrete would be helpful. (Teacher C) Unfortunately, when confronted with unclear expectations, vague ideas for implementation, and other health information and programs, many of the teachers are ignoring GSW curriculum and not implementing anything in addition to kick off assemblies and three incentive days per year. These two components are vastly different than the ideal implementation created by GSW administ rators and LiveWell Colorado. One of the administrators of the GSW program acknowledged that mandating specific curriculum would be helpful to teachers, but that Nutrition Services is not authorized to mandate such additions to P.E. teacher curriculum. Wh en asked if she visits the teachers and shares expectations for implementing the GSW program, the administrator responded with: found resources that would align with those standards that they could use (GSW administrator) Therefore, the GSW administrators can only recommend resources and activities. Because of this, whether or not GSW activities and curriculu m are implemented in each school depends less on program expectations and more on teacher enthusiasm motivation and available time. Waning Enthusiasm Five of s even teachers mentioned that the excitement and enthusiasm of the kick off assembly at their sch ools was followed by diminished interest support, and sustainability by GSW, district, and school administrators. Unfortunately, this waning

PAGE 117

112 trying to check boxes for th e GSW roll out without paying particular attention to the long term sustainability and success of the program. the district level. But, what is really going on? And, are you real ly (Teacher A) Another teacher added: You know I hate to say it but right now it feels like they just wanted us to get it kicked off just to get it kicked off. You know what I mean? Just to put a da to have (Teacher D) After the kick off assembly, the enthusiasm diminished and contact with the district GSW coordinator declined. And then attention t o GSW popped up again when it was time to complete an incentive day, but to many teachers it felt like just one more check box for data collection instead of a solid and sustainable program component. vegetables and other healthy foods on more than just three days per year. By encouraging food tasting with only three incentive days, teachers interpret those days as yet another way that the GSW program is only skimming the surface with minimal impact. to form a habit? (Teacher F) Another teacher responded with: T he incentive days, they [GSW district administrators] just want to know [P.E. teachers] done them. Yep, so we can get Channel 7 to come out and pass out the bracelets that they bought. We can make it look fancy and show we are talking a bout eating healthy even though ppease, (Teacher D)

PAGE 118

113 Teachers perceiving diminished district level enthusiasm and sup port for the GSW program feel frustration with the district, reduc e their motivation to implement the program, and d ecrease desire to adopt additional programs in the future. Individuals and Communities Involved with GSW Principal Support Most of the teachers interviewed (five of seven ) commended their principals for being supportive of the GSW program and other nutriti on initiatives in Physical Education. Supportive principals are imperative to teacher motivation and success. Teachers commented on the benefit and effect of principals who dedicate their own time and energy to staying personally healthy. Two teachers w ithout positive principal support felt overworked and under supported in their roles as GSW liaisons. Teacher Enthusiasm and Motivation for GSW Durlak (2008) emphasized the effect of key individuals on program implementation. In the case of GSW, physical education teachers have a large effect on program implementation and success. Out of seven physical education teachers interviewed, it was obvious that some teachers were more personally interested in nutrition than others. All teachers supported nutrition education on a t heoretical basis, but some were more enthusiastic about nutrition education and the GSW program than others Those teachers with higher personal motivation for GSW rep orted more time spent on GSW activities and lessons. makes kids take a second look a t what, you know, (Teacher G)

PAGE 119

114 When another teacher was asked if she had low enthusiasm to implement ye t another mandated program in her class: No, definitely not because I want the kids to have information. I wanted to have something put together that will give kids a purpose to eating and diet hopefully. Not a Hot Chee tos diet. A healthy diet is not a bad thing. A healthy diet is what we eat. (Teacher F) Interestingly, Teacher F was one of the teachers with an unsupportive and unhelpful principal. Students benefit from her personal motivation for nutrition even ami dst the absence of principal support and assistance. This particular teacher made a concerted effort to utilize the daily menus with GSW labels stations with food cutouts and My Plate guidelines, guest speakers related to nutrition topics, and active gam es involving nutrition information in her class. Other P.E. teachers we re less enthusiastic about developing and delivering lesson plans with GSW information. This lower motivation was related to several specific things: the lack of time to develop resou the desire to keep kids moving as much as possible during p hysical education, or the tea Another nutrition education program in Aurora Public Schools pays teachers a fairly sizeable amount of money to be liaisons betw een the program and the school. Teachers commented that this monetary incentive is quite effective at keeping teachers on track with following program outlines and activities. GSW has no such incentive, and this lack of incentive might be a detriment to program implementation given that other programs with incentives might be prioritized.

PAGE 120

115 Student Responses to GSW: Fostering Enthu siasm and Pride It is hard to evaluate overall student response to GSW because of the varied level of implementation and without district permission to ask students directly However, one theme mentioned by several teachers wa s the creation of a sense of enthusiasm and pride among students in what they are eating at school as a result of GSW activities. Having a teacher who is excited about nutrition education and enthusiastic about the GSW program helps to foster enthusiasm in students. I wish you coul me things in their lunch and bring them over and it was just really cool to see. I would have this little crowd around me just to show me that they drank all their milk. They would show me someth ing from their brown why it was healthy. Just seeing how interested they were in showing, (Teacher F) After imp lementing GSW activities, this teacher notes: And now, the kids in the cafeteria, I mean, they call me over to look at their plates and in their lunch boxes. Oh, they are all about it. So, they are definitely empowered. (Teacher F) I had a kid come up to me saying I eat H ot Cheetos but I make sure I exercise a fterwards, so there's a balance. (Teacher E) Influence of Parents and Home Life on Food Intake Four of seven teachers commented on the dissonance between the efforts to create a culture of wellne ss and good nutrition at school and the negative influence of parents and home life on wellness and food intake. through Friday but you go home on the weekends. Kids come back and they have any control over that. (Teacher D)

PAGE 121

116 Teachers commented on two negative family influences on nutrition: finances and nutrition knowledge /motivation Six of the seven teachers inter viewed teach at schools where more than 80% of students receive free or reduced price lunch. And two of those schools have more than 93% of students who receive free or reduced price lunch. Using qualification for free and reduced p rice lunch as a proxy for income, teachers at lower income schools indicate d that income strongly influences what parents buy at grocery stores. Families with lower incomes have less access to healthy foods. choices. And then a lot of them are limited to where they shop, how much how they grew up and what they grew up on. (Teacher F) The other family factors reported by teachers is a lack of parental nutrition knowledge and support for GSW principles along with strong cultural and oftentimes unhealthy influences on diet. Parents cook foods that are cultura lly fami liar and those foods may be quite different than many of the healthy recipes served in Aurora Public Schools. One teacher commented that students are very unlikely to try foods at school that are so vastly different from foods cooked at home. Weekends, h olidays, and summer vacations often derail the goals of the program because of family influences on what kids are eating. Program Charac teristics Lack of Time Six of seven interviewed teachers reported a lack of time to implement GSW activities and other district programs related to health and physical education. Many teachers were frustrated by the ongoing district mandate to implement new programs without all otting more time for physical education classes. Six of seven schools offer physical education one time per week, and one school offers PE every third day. Fifty

PAGE 122

117 minutes of class time per week is such a small increment of time, and teachers struggle to e ven find the time to deliver their curriculum standards. to cover that it would be extra or if it came up and was applicable [we would talk about it]. (Teacher F) Another teache r commented: You know we had another new program that I was kind of given this year for fifth grade and it becomes difficult to fit more and more things into a curriculum that is al ready filled with standards and you know things that I think are important. (Teacher B) Not only is there a paucity of class time for adequate curriculum delivery, there is also diminished planning time for teachers to create and organize GSW activities. e the time myself to create [program activities] when that should be coming, in my opinion, from either LiveWell Colorado or Aurora public schools. Somebody should be responsible for putting that together and getting that reso urce out to that. (Teacher A) The seven teachers interviewed represent schools with varying numbers of extra nutrition pro grams. As outlined in Chapter III elementary schools i n Aurora may have up to five additional nutrition programs depending upon principal interest and average student/family SES. All schools in Aurora are also participating in the 5 th Gear Nutrition Education program for 5 th graders as of 2012 2013. The five optional nutrition education programs are: Breakfast in the Classroom Fresh Fruit and Vegetable Program INEP Nutrition Education Healthier U.S. Schools Challenge Award for implemented changes and activity time (not necessarily a program, but a commendation for changes and initiatives ) Coordinated School Health.

PAGE 123

118 The interviewed teachers whose schools have a larger number of these extra programs (>=3) are also the teachers with very strong comments about not having enough time to fit everything in. Comments from these three teachers with three or four additional nutrition programs related to lack of time included: Too many programs for the available time and money Lack of GSW implementation due to lack of ti me Not enough time to fit in general standards let alone extra curriculum. Keeping Kids Moving During C lass Most of the interviewed teachers are worried about keeping their children moving instead of devoting time to sedentary nutrition lesson s one guy who gets t hem moving once a week one fifty minute period. And so for me to sit and talk about anything I try to make it as have t o teach and show them things building doors. So, the game and combine it. (Teacher B) I think the program is great. Probably if I had more time or I saw the kids two or three times a week I could see myself using a fifteen minute block to talk about nutrition every class. . Even in fifty minutes trying to meet all those PE standards this is between me and you, is very unrealistic. (Teacher B) a P.E. teacher in elementary to hit everything that we need to, to make them proficient. And, you think about reading and writing and how much time you got these kids for forty five minutes a week. Twenty seven and a half now everything going to need to know in fifth grade to be healthy? (Teacher D) Some t eachers have creatively combined nutrition and GSW lessons with active games. This strategy involves a bit of extra planning for the teachers, but is a way to

PAGE 124

119 fo cus on nutrition information without sacrificing movement and physical activity time. Table 18 summarizes all of the identified themes from physical education teacher interviews. Note: The letters of the teachers in Table 18 correspond to the lettered ide ntification of each quote in this section. Table 18: Physical Education Teacher Interview Themes Identified Themes Teacher/School Identifier and Year of GSW Initiation TeacherA 2010 Teacher B 2010 Teacher C 2010 Teacher D 2012 Teacher E 2012 Teacher F 2012 Teacher G 2011 Lack of District GSW Leadership/Not sure of Program Goals and Components X X X X X Passing duties to someone else, or not sure who is responsible for what regarding GSW X X Big fuss for roll out, followed by waning enthusiasm and support at district level. No check ins or communication. X X X X X Lack of Details on Incentive Day X X X X Not enough time to do everything X X X X X X Burdening P.E. Teachers/Teachers feel resentful of new programs and demands X X X X Want to spend time getting kids moving more than sedentary teaching X X X X Kids backslide in the evenings, weekends, and summers X X X X Need more collaboration/all school meetings about GSW/check ins and accountability X X X X X Kids have pride and enthusiasm for what they are learning and eating X X X Support from Principal X X X X X

PAGE 125

120 CHAPTER VI D ISCUSSION This study aimed to identify the influence of the Go, Slow, Whoa nutrition education program combined with Scratch Cooking Cafeteria Changes on elementary investigated student choices and food consumption in schools with GSW and comparison schools without GSW. Qualitative research components helped outline differences in dose and implementation of the GSW program in relation to quantitative results The goal of this research was to systematically evaluate the Go, Slow, Whoa program and cafeteria changes in Aurora Public Schools and to assist GSW administrators and funders with program improvement and funding recommendations. Several key findings emerged from this mixed methods study. First, there were almost no significant differences between GSW schools and non GSW comparison schools in both nutrients taken and nutrients consumed at school lunch. The only significant difference revealed more students in GSW scho ols choosing white milk over chocolate milk or no milk as compared to students in non GSW schools. Second, various implementation issues were identified by physical education teachers that interfere d with program success. And third, students consumed considerably fewer calories than they took in the lunch line (553 vs. 292 calories) and they took and consumed different amounts of nutrients each day. GSW Program Effect on Nutrients Taken and Nutrients Consumed The only significant d ifference between GSW and non GSW schools was a higher odds ratio of choosing white milk over chocolate milk or no milk for students in GSW

PAGE 126

121 schools. The other variables that approached significance for GSW schools and students were drinking more milk, tak ing and consuming less fiber, taking and consuming more total fat, and taking and consuming more saturated fat. If the GSW program were having an effect, GSW school students would be expected to have the opposite choice and consumption of fiber, total fat and saturated fat. These results indicate lack of a program effect on student consumption of more fruits and vegetables and fiber and less fat and saturated fat. Given the qualitative findings that huge implementation issues are influencing the success of the GSW program, one cannot declare a lack of program success due to the program nor discuss specific flaws in the program However, data do show a lack of program success due to a program that was not implemented with fidelity and consistency. Qual itative analyses reveal significant program implementation and success issues related to teachers, students, families and program support. These issues are summarized in the qualitative results and include class contact hours, t eacher motivation and enthusiasm family knowledge and suppor t, and administration support and assistance. In order for health education to be successful, adequate class contact hours are imperative. The Go, Slow, Whoa program is rooted in the well studied CATCH program which involves 47 40 minute nutrition lessons ( Perry, Bishop, et al., 1998 ). Based on a meta analysis of nutrition education studies researchers recommend 50 hours of nutrition education per year to impact attitudes and behavior (Lytle, 1994) Teachers interviewed for this research reported far fewer than 50 hours of GSW and nutrition education programming in their schools. Two teachers reported less than 20 min utes per month of class time devoted to GSW w hich would equate to less than three hours per school year

PAGE 127

122 Physical education teachers are devoting too little time to nutrition education to effect positive change in attitudes and behavior. The meta analysis by Lytle et al. of nu trition education programs also indicated the importance of a family involvement component to influence nutrition attitudes and behaviors (Lytle, 1994) While the GSW program did include a parent breakfast to inform parents of the program and GSW model of fo od categorization, there we re no additional program components or specific lessons involving families Teachers perceived that lac k of healthy nutrition reinforcement at home, due to both income constraints and lack of knowledge, as negatively influencing GSW program success. Broader identified implementation issue s relate to programming load, teacher burnout, and curriculum sta ndards in Aurora Public Schools. P.E. teachers in APS are feeling overburdened and also feel like many health and physical activity programs do not receive the attention from administrators and classroom teaching time that they deserve or need to have an effect. Exploring and working on these time and burden issues at a policy level within the district could have important implications for special program effects (i.e. Go, Slow Whoa) in physical education and other subjects. Targeting and changing food c onsumption through nutrition education programs fits within a Social Ecological Model and comments by APS teachers encourage addressing a lack of program effect at multiple levels (Contento, 2011) bu t specif ically starting at the policy and s ystem level (S.N. Moore, Murphy, & Moore, 2011) Using a social ecological model of nutrition education not only targets policy issues related to allowable f ood in schools, vending machine regulations and snack/party policies, but also relates to policy and structure associated with adequate instruction and appropriate teacher workload.

PAGE 128

123 Comments from physical education teachers d elivering the GSW program revealed less than ideal teacher enthusiasm, teacher knowledge of GSW and program goals, and support from Aurora administration. Physical education teachers are the individuals responsible for much of GSW delivery, and Aurora adm inistration is responsible for reinforcing that delivery with technical assistance, leadership, and support. Teachers or individuals responsible for delivering a program and the delivery and sup port system available to them are crucial to program implementation and success (Durlak & DuPre, 2008) Lack of GSW implementat ion and program success results in part from these issues with teachers delivering the program a nd the administration responsible for supporting teachers. Effect of Scratch Cooking Changes Given that the school district follows very strict federal guidelines for calories and other nutrients in a standard elementary school lunch (550 650 calories dail y ) and given that students are consuming only about half of the calories in their school lunch the reality is that different menu items hold varying appeal for students. Students respond to internal motivation, external motivation and taste in deciding w hat foods and how much of each food to choose and consume at school each day. Students are taking and consuming vastly different amounts of nutrients each day and across days at lunch. Students consumed an average of 292 calories from school lunch out of 553 total tray calories This reveals some important issues with school lunches in Aurora Public Schools. The amount of nutrients that are thrown away is alarming given that the body and brain will function better in school with adequate nourishment. The volume of trash also indicates a waste of district and federal funding if

PAGE 129

124 food is thrown away Observation and interviews by researchers revealed several possible reasons for this waste in Aurora Public Schools : lack of time to finish lunch, upset stomach from recess before lunch, eagerness to get outside to play in schools with recess after lunch, and dislike of foods based on visual or taste cues. Previous research supports the interaction of physical, temporal, and social cha experience (S. N. Moore, Murphy, Tapper, & Moore, 2010) Including qualitative data from students would help reveal the different facto rs affecting eating habits at lunch. Characteristics of Students in Aurora Several teachers mentioned healthy food access issues of families within their school s due to low income and financial issues. Although Aurora has responded with several creative nutrition programs to deliver additional fresh fruits and vegetables at school, many families are not promoting healthy eating at home. The social determinant s of health of Aurora students, the circumstances in which people are born and grow up, are certainly affecting their nutrition and health status. R esearch has linked poverty to obesity ( Li & Hooker, 2010) but the results of this study do not indicate an overconsumption of calories from school lunch The National School Lunch Program dictates that K 5 students should be receiving an average of 550 650 calories at lunch per day. However, based on this research students are taking an average of 553 calori es at lunch, but only consuming 293 calories. This points to a bigger issue given that a large number of Aurora elementary students are obese. If they are eating so few calories at school, then the bulk of their calories are coming from afterschool snack s and meals served at home. The social determinants of health could be related to increased availability of high

PAGE 130

125 fat, high sugar and high calorie food s at home. The scratch cooking changes initiated in APS could be unappealing to students, and the y could opt to eat less of their school lunch in favor of more desirable foods outside of school. Conclusion The quantitative findings of this research were enhanced by the qualitative analysis, which helped explain the lack of observed program effects The quantitative analysis showed almost no GSW program effect, while the qualitative a nalysis uncovered important and related implementation issues. Qualitative analyses revealed markedly more varied class time devoted to Go, Slow, Whoa than projected by GSW administrators. This range of class time devoted to GSW activities helps explain the lack of program effect on food and nutrients taken and consumed at lunch. The Go, Slow, Whoa program is rooted in the well studied CATCH program ( Perry, Bishop, e t al., 1998 ). However, the CATCH program involved 31 total contact hours (47 40 minute lessons) whereas the GSW program in Aurora Public Schools amounted to a fraction of that amount. Some teachers reported as little as three classroom contact hours per year of GSW curriculum. Without fidelity in implementation and a similar and recommend program dose, it is impossible to draw conclusions about the Go, Slow, Whoa program. This research discovered several implementation and programmatic issues, namely u nclear program goals and expectations, lack of time to prepare and deliver curriculum, P.E. teacher desire to keep kids moving during class, low teacher enthusiasm and motivation for new curriculum, and lack of district support and technical assistance. T hese compounded issues contributed to the lack of program implementation and

PAGE 131

126 program effect. Although a program effect was not found, the quantitative analysis revealed the characteristics of student lunch trays on three different days in Aurora Public Schools. Students consu med about half of what they took in the lunch line, took an average of 1.3 fruits and vegetables and consumed about 2 grams of fiber at lunch. Onl y 17% of students chose white milk (over no milk or chocolate milk) at lunch, and students taking all types of milk only consumed 50% of their milk. This information will be helpful as administrators make decisions about recipes, ordering, and planning. Given that the GSW program targets white milk, fruit and vegetable, and high fiber food consumption, these data could be used as baseline data for future GSW research if implementation of the program is improved. Limitations Plate Waste Methodology Digit al Plate Waste methodology has some limitations. The pre meal weight/size of p articular entrees and items w ere estimated from a sample of three collected items, and not actual weights of each student tray. Students might have eaten di fferently knowing th eir tray had been tagged and researchers we re interested in their food patterns and it difficult to photograph and accurately assess individual items. Matching Techniques Even though 11 variables were included in matching models it is still possible that additional important covariates r elated to treatment assignment we re missing from the analysis (i.e. unobserved). If this is the case, propensity scores could be biased Shadish and Steiner (201 0) warn that careful and detailed attention to identifying

PAGE 132

127 background covariates that could be related to treatment selection is imperative in propensity score analysis. One of the limitations of using propensity s core matching for this research wa s the l ack of detailed and similar information across elementary schools in Aurora. Nearest neighbor, or nearest school in this research, propensity score matching is a co mmonly used matching algorithm. However, this method faces the risk of imprecise matches if the nearest neighbor is numerically quite distant (Baser, 2006) This limitation coul d be exacerbated by the small sample size of potential comparison schools in Aurora. Small sample size is related to several other limitations of the matching methods used in this research. Propensity score methods generally work better in larger samples because smaller samples can have substantial imbalances of some covariates even when using propensity score matching (Zhao, 2004) A small sample size (n=25 level 2 units in this study) also increases the risk of overspecification when using a large number of predictors in a propensity scor e logit model. In one study, propensity score methods (one to one matchin g and other matching techniques ) were not superior to covariate matching (Mahalanobis ) for small sample sizes (n=500), but performed better for larger sample sizes (n>1000), whereas Mahalanobis matching appears to be robust with small and large sample sizes n=500 and n>1000) (Zhao, 2004) There is a paucity of research on matching techniques in very small sample sizes such as this sample of 25 schools. There are 32 elementary schools in Aurora, but 7 schools started the GSW program in 2010 o r 2009. Therefore, 25 schools we re included in the current study sample for quantitative data collection Of those 25 schools, 5 schools we re GSW s chools and 20 schools we re possible control schools. Because of this very small samp le size, several

PAGE 133

128 different propensity score models and matching techniques were utilized and resulting matches were checked for balance and mean differences on covariates to determine the best fitting model and matching technique. If the matching methods i ntroduced bias into the results, the direction and magnitude of the biases are unknown. These biases could have masked a true program effect, although this is unlikely given the qualitative findings regarding program implementation. In future studies emp loying matching techniques based on district data, interviewing staff at intervention and comparison schools could help identify additional factors that could be missing in objective and robust matching techniques. Data Collectors Using 10 0 data collectors enabled data to be collected from 10 different schools on the same day. Detailed and comprehensive data collection trainings were presented to one lead data collector at each school to ensure as much consistency and control over data col lection. Lead data collectors utilized a protocol checklist and protocol fidelity questionnaire each day of data collection to increase adherence to data collection protocols. Qualitative Data Collection This research did not employ objective, quantitative measures of implementation. Information regarding dose and fidelity of implementation was gathered fro m qualitative interviews. Teachers could have tailored answers to implementation questions to match W implementation. To minimize this bias, I assured teachers of the anonymity of their answers and utilized a semi structured

PAGE 134

129 interview guide. Follow up questions in the interview were tailored to each teacher to gather as much relevant and honest informa tion regarding implementation as possible. Generalizability Because the study was conducted in a metropolitan area of Colorado, the ability to generalize results to other geographic locations a nd school districts is limited. Significance and Directions f or Future Research As the number of obese and overweight children and the consumption of fat, sugar, sodium and processed foods outside of school continue to grow, effective nutrition policies and programs delivered during school will continue to be a pri ority. Given limited funding and class time, schools and districts must make smart decisions regarding nutrition programs. The results of this research indicate that programs are only as good as their implementation. After tweaking the delivery and support for the Go, Slow, Whoa program, more research in Aurora Public Schools should focus on program dose fidelity, and implementation. On a broader level, many research studies could benefit from quantitatively measurin g implementation. Specifically, future research should include quantitative implementation data as predictors in regression models. Newton and Llosa encourage evaluators to investigate important questions such as how the quality of program implementation relates to different program outcomes and whether program effectiveness varies by setting, conditio ns, or other characteristics (Newton & Llosa, 2010) Using predictor variables related to implementation in a regression model could

PAGE 135

130 This study looked at only three days and found significant differences between days for white milk taken, fiber taken and consumed, fat and saturated fat taken and consumed, and fruits and vegetables taken. Those dif ferences are likely related to the menu choices for each day of the study (beef stew and biscuits; chicken, roll, baked beans and cake; pizza, green beans and grapes). Students consumed less fiber on days 2 and 3, which could be a result of increased desi re to eat the low fiber entre (chicken or pizza) or cake, and therefore lower desire or less stomach room to eat additional high fiber fruits and vegetables and side items. Or, perhaps it takes longer to eat chicken as a result of having to pick it up or cut it, so there is less time to eat other high fiber sides on the tray. Students also took less fruits and vegetables on days 2 and 3, which could signify that they prefer the main entre and side options more on those days and pass by the salad bar and additional fruits and vegetables. Or, perhaps the quality of the fruits and vegetables was lower later on in the week as a weekly produce shipment aged. Our study days were Wednesday, Thursday, and Friday. Students consumed more total fat and saturated fat on days 2 and 3 compared to day 1, which could be a function of them liking the chicken (with skin) and pizza (with cheese) more than they like beef stew. Cake was also served on day 2, which added fat consumption for most students in the cafeteria. Future trays in order to outline the reasons for observed plate waste data in this research. Nutrition Services would benefit from a more thorough assessment of the l ikability and consumption of different menu items in their menu cycle. APS should consider using qualitative interviews or focus groups with students to gather additional information.

PAGE 136

131 Students would provide rich data on perceptions and impact of the prog ram and scratch cooking changes. Aurora Public Schools would also benefit from increased communication with and incentives for teachers implementing the Go, Slow, Whoa program. Physical education teachers mentioned several possible strategies to improve GSW, including monthly meetings of physical education teachers, monetary incentives to complete program components, and increased support from school administrators for teachers to attend nutrition education professional development. APS and LiveWell Colo rado should consider available resources, both monetary and personnel, in the context of these findings and recommendations before deciding how to improve if at all, the GSW program.

PAGE 137

132 REFERENCES Action For Healthy Kids. ( 2005). Parents' views on school wellness p olicies. Akaike, H. (1974). A new look at the statistical model identification. IEEE Transactions on Automatic Control, 19 (6), 716 723. American Dietetic Association. (2008). Summary of findings: American Dietetic Associ ation's public opinion survey, Nutrition and You: Trends 2008 Ananiadou, K., Schneider, B. H., Smith, J. D., & Smith, P. K. (2004). The effectiveness of whole school antibullying programs: a synthesis of evaluation research. School Psychology Review, 33 ( 4), 547 560. Anderson, E. S., Winett, R. A., Wojcik, J. R., Winett, S. G., & Bow den, T. (2001). A Computerized social cognitive intervention for nutrition behavior: Direct and mediated effects on fat, fiber, fruits, and vegetables, self e fficacy, and Outc ome Expectations Among Food Shoppers. Annals of Behavioral Medicine, 23 (2), 88 100. Auld, G. W., Romaniello, C., Heimendinger, J., Hambidge, C., & Hambidge, M. (1999). Outcomes from a school based nutrition education program alternating special resource t eachers and classroom teachers. Journal of School Health, 69 (10), 403 408. Bandura, A. (1986). Social foundations of thought and action : a social cognitive theory Englewood Cliffs, N.J.: Prentice Hall. Bandura, A. (1989). Human agency in social cognitiv e theory. Am erican Psycho logist 44 (9), 1175 1184.

PAGE 138

133 Bandura, A. (1998). Health promotion from the p e rspective of social cognitive t heory. Psychology and Health 13 (4), 623 650. Bandura, A. (2000). Health promotion from the perspective of social cognitive theory. In P. Norman, C. Abraham & M. Conner (Eds.), Understanding and changing health behavior: From health beliefs to self regulation. Amsterdam: Harwood Academic Publishers. B andura, A. (2004). Health promotion by social cognitive means. Health Educ ation and Behav ior 31 (2), 143 164. Barlow, S. E., & Expert, C. (2007). Expert committee recommendations regarding the prevention, assessment, and treatment of child and adolescent overweight and obesity: summary report. Pediatrics, 120 164 192. Barr, J. E., Tubman, J. G., Montgomery, M. J., & Soza Vento, R. M. (2002). Amenability and i mp lementation in secondary school antitobacco p rograms. American Journal of Health Behavior, 26 (1 ), 3 15. Baser, O. (2006). Too much ado about propensity score models? Comparing methods of propensity score m atching. Value in Health, 9 (6), 377 385. Beaton, G. H., Burema, J., & Ritenbaugh, C. (1997). Errors in the interpretation of dietary assessments The American Journal of Clinical N utrition, 65 (4). Benefits of farm to school projects healthy eating and physical activity for school children: field hearing before the Committee on Agriculture, Nutrition, and Forestry, United States Senate. 111th Con gress, 111 245 Sess. (2009). Bergman, E. A., Buergel, N. S., Englund, T. F., Clem, A., Hansen, E. A., Kutrich, L. A. (2004). Elementary students eat more and waste less when recess is

PAGE 139

134 scheduled before lunch. Journal of the American Dietetic Association., 1 04 (2), 45. Bergman, E. A., Buergel, N. S., Englund, T. F., & Femrite, A. (2004a). The relationship between the length of the lunch period and nutrient consumption in the elementary school lunch setting. Journal of Child Nutrition & Management, 28 (2). Ber gman, E. A., Buergel, N. S., Englund, T. F., & Femrite, A. (2004b). The relationship of meal and recess schedules to plate waste in elementary schools. Journal of Child Nutrition & Management, 28 (2). Berkhof, J., & Ka mpen, J. K. (2004). Asymptotic effect of misspecification in the random part of the multilevel m odel. Journal of Educational and Behavioral Statistics, 29 (2), 201 218. Birch, L. L. (1995). Research in review. children's eating: The d evelopment of food acceptance p atterns. Young Children, 50 (2 ), 71 78. Birch, L. L. (1999). Development of food preferences. Annual Review of N utrition, 19 41 62. Briefel, R. R., Crepinsek, M. K., Cabili, C., Wilson, A., & Glea son, P. M. (2009). School Food environments and practices affect dietary behaviors of U S public school c hildren. Journal of the American Dietetic Association, 109 (2), S91 S107. Briggs, M., Safaii, S., Beall, D. L., American Dietetic, A., Society for Nutrition, E., & American School Food Service, A. (2003). Position of the American Dietetic Association, Society for Nutrition Education, and American School Food Service Association -Nutrition services: an essential component of

PAGE 140

135 comprehensive school health programs. Journal of the American Dietetic Association, 103 (4), 505 514. Bronfenbrenner, U. (1989). Ecological s ystems theory Browne, W. J., & Draper, D. (2000). Implementation and performance issues in the Bayesian and likelkihood fitting of multilevel models. Computational Statistics, 15 391 420. Burgess Champoux, T., Marquart, L., Vickers, Z., & Rei cks, M. (2006). Perceptions of children, parents, and teachers regarding whole grain foods, and implications for a school based i ntervention. Journal of Nutrition Education and Behavior 38 (4), 230 237. Celebuski, C., & Farris, E. (2000) Nutrition education in public elementary school classrooms, K 5 Washington, DC: U.S. Dept. of Education, Office of Educational Research and Improvement, National Cen ter for Education Statistics : Supt. of Docs., U.S. G.P.O., distributor. Child Nutrition and WIC Reauthorization Act § 2507 (2004). Cohen, J. (1960). A coefficient of agreement for nominal s cales. Educational and Psychological Measurement, 20 (1), 37 46. Comstock, E. M., St Pierre, R. G., & Mackiernan, Y. D. (1981). Measuring individual plate waste in school lunches. Visual estimation and children's ratings vs. actual weighing of plate waste. Journal of the American Dietetic Association, 79 (3), 290 296. Pubic Law 396 60 Stat. 231 C.F.R. (1 946).

PAGE 141

136 Connell, D. B., Turner, R. R., & Mason, E. F. (1985). Summary of findings of the school health education evaluation: Health promotion effectiveness, i mplementatio n, and c osts. Journal of School Health, 55 (8), 316 321. Contento, I. R. (2011). Nutriti on education : linking research, theory, and practice (2nd ed.). Sudbury, Mass.: Jones and Bartlett Publishers. Cooke, M. (2000). The dissemination of a smoking cessation program: predictors of program awareness, adoption and maintenance. Health Promotion I nternational, 15 (2), 113 124. Coordinated Approach to Child Health ( CATCH). (2012). Go, Slow, Whoa f ood l ists. Retrieved January 25, 2012, from https://sph.uth.tmc.edu/catch/whats_new_food.htm COPAN. (2009). The weight of the s tate: 2009 report on overw eight and o besity in Colorado (Colorado Physical Activity and Nutrition Program, Trans.). Denver: Colorado Department of Public Health and Environment. Crepinsek, M. K., Gordon, A. R., McKinney, P. M., Condon, E. M., & Wilson, A. (2009). Meals offered and served in US public schools: Do they meet nutrient s tandards? Journal of t he American Dietetic Association 109 (2), 3. Cullen, K. W., Watson, K. B. & Fithian, A. R. (2009). The i mpact of school socioeconomic status on student lunch consumption after impl ementation of the Texas p ublic school nutrition p olicy. Journal of School Health, 79 (11), 525 531.

PAGE 142

137 Cullen, K. W., & Zakeri, I. (2004). Fruits, vegetables, milk, and sweetened bever ages consumption and access to a la carte/snack bar meals at school. Ameri can Journal of Public Health, 94 (3), 463 467. D'Agostino, R. B., Jr. (1998). Propensity score methods for bias reduction in the comparison of a treatment to a non randomized control group. Statistics in M edicine, 17 (19), 2265 2281. Daniels, S. R. (2006). The consequences of childhood overweight and o besity. Future of Children, 16 (1), 47 67. Daniels, S. R., Arnett, D. K., Eckel, R. H., Gidding, S. S., Hayman, L. L., Kumanyika, S. (2005). Overweight in children and adolescents: pathophysiology, consequences, prevention, and treatment. Circulation, 111 (15), 1999 2012. Davidson, F., Hayek, L. A. & Altschul, A. (1986). Towards accurate assessment of children's food consumption. Ecology of Food and Nutrition, 18 (4), 309 317. Dedrick, R. F., Ferron, J. M., Hess, M. R., Hogarty, K. Y., Kromrey, J. D., Lang, T. R. (2009). Multilevel modeling: A revi ew of methodological i s sues and a pplications. Review of Educational Research, 79 (1), 69 102. Derzon, J., Sale, E., Springer, J. F., & Bro unstein, P. (2005). Estimating intervention effectiveness: synthetic projection of field evaluation r esults. Journal o f Primary Prevention, 26 (4), 321 343. Devaney, B. L., Gordon, A. R., & Burghardt, J. A. (1995). Dietary intakes of students. The American Journal of Clinical N utrition, 61 (1).

PAGE 143

138 DeVault, N., Kennedy, T., Hermann, J., Mwavita, M., Rask, P., & Jaworsky, A. ( 2009). I t's All About Kids: Preventing overweight in elementary school c hildren in Tulsa, OK. Journal of the American Dietetic Association, 109 (4), 680 687. DiClemente, R. J., Crosby, R. A., & Kegler, M. C. (2002). Emerging theories in health promotion pr actice and research strategies for improving public health. http://search.ebscohost.com/login.aspx?direct=true&scope=site&db=nlebk &db=nlabk&AN=74173 D uBois, D., Holloway, B., Valentine, J., & Coope r, H. (2002). Effectiveness of mentoring programs for youth: A meta analytic r eview. American Journal of Community Psychology, 30 (2), 157 197. Dubois, S. (1990). Accuracy of direct visual estimates of plate w aste in the determination of food consumption. Journal of the American Dietetic Association, 90 382 387. Dufrene, B. A., Duhon, G. J., Gilbertson, D. N., & Noell, G. H. (2005, 2005 Winter). Monitoring implementation of reciprocal peer tutoring: identifyi ng and intervening with students who do not maintain accurate implementation. School Psychology Review, 34, 74+. Durlak, J., & Du Pre, E. (2008). Implementation matters: A review of research on the influence of i mplementation on program outcomes and the factors affecting i mplementation. American Journal of Community Psychology, 41 (3 4), 327 350. Edmundson, E., Parcel, G., Feldman, H., Elder, J., Perr y, C., Johnson, C. (1996). The e ffects of the Child and Adolescent Trial for Cardiovascular Health upon

PAGE 144

139 psychosocial determinants of diet and physical activity b ehavior. Preventive Medicine, 25 (4), 442. Eriksen, K., Haraldsdttir, J., Pederson, R., & Flyger, H. V. (2003). Effect of a fruit and vegetable subscription in Danish schools. Public Health N utritio n, 6 (1), 57 63. Finkelstein, E. A., Trogdon, J. G., Cohen, J. W., & Dietz, W. (2009). Annual medical spending attributable to obesity: Payer and service specific estimates. Health Affairs, 28 (5), w822 w831. Fontaine, K. R., Redden, D. T., Wang, C., Westf all, A. O., & Allison, D. B. (2003). Years of life lost due to obesity. J ournal of the American Medical Association, 289 (2), 187 193. Fox, M. K., Dodd, A. H., Wilson, A., & Gleason, P. M. (2009). Association between school food environment and p ractices a nd body mass i nde x of US public school c hildren. Journal of the American Dietetic Association, 109 (2), S108 S117. Freedman, D. S., Dietz, W. H., Srinivasan, S. R., & Berenson, G. S. (1999). The relation of overweight to cardiovascular risk factors among c hildren and adolescents: the Bogalusa Heart Study. Pediatrics, 103 (6), 1175 1182. Friedman, B. J., & Hurd Crixell, S. L. (1999). Nutrient intake of children eating school breakfast. Journal of the American Dietetic Association, 99 (2), 219 221. Fulkerson, J. A., French, S. A., Story, M., Snyder, P., & Paddock, M. (2002). Foodservice staff perceptions of their influence on student food choices. Journal of the American Dietetic Association, 102 (1), 97 99.

PAGE 145

140 Gleason, P. M., & Suitor, C. W. (2001). Children's d iets in the mid 1990s : dietary intake and its relationship with school meal participation Alexandria, VA: U.S. Dept. of Agriculture, Food and Nutrition Service. Gleason, P. M., & S uitor, C. W. (2003). Eating at s chool: How the National School Lunch Progra m affects children's d iets. American Journal of Agricultural Economics, 85 (4), 1047 1061. Gortmaker, S. L., Peterson, K., Wiecha, J., Sobol, A. M., Dixit, S., Fox, M. K. (1999). Reducing obesity via a school based interdisciplinary intervention among yout h: Planet Health. Archives of Pediatrics & Adolescent M edicine, 153 (4), 409 418. Green, L. W., & Kreuter, M. W. (2005). Health program planning: an educational and ecological approach New York: McGraw Hill. Greenwood, C. R., Tapia, Y., Abbott, M., & WAlton, C. (2003). A building based case study of evidence based literacy practices: implementation, reading behavior, and growth in reading fluency, K 4. Journal of Special Education, 37 (2), 95 110. Harrell, M. C., & Bradley, M. A. (2009). Data collecti on methods: semi structured interviews and focus g roups. Santa Monica, CA: RAND Corporation. Healthy Hunger Free Kids Act of 2010 § 3307 (2010). Hearn, M. D., Baranowski, T., Baranowski, J., Doyle, C., Smith, M., Li n, L. S. (1998). Environmental influences on dietary behavior among children: Availability and accessibility of fruits and vegetables enable c onsumption. Journal of Health Education, 29 (1), 26 32.

PAGE 146

141 Hedley, A. A., Ogden, C. L., Johnson, C. L., Carroll, M. D., Curtin, L. R., & Flegal, K. M. (2004). Prevalence of overweight and obesity among US children, adolescents, and adults, 1999 2002. J ournal of the American Medical Association, 291 (23), 2847 2850. Heron, M. (2011). Deaths: leading causes for 2007. N ational vital statistics reports : from the C enters for Disease Control and Prevention, National Center for Health Statistics, National Vital Statistics System, 59 (8), 1 95. Hinrichs, P. (2010). The effects of the National School Lunch Program on education and health. Journal of Policy Analysis and M anagement, 29 (3), 479 505. Hoelscher, D. M., Feldman, H. A., Johnson, C. C., Lytle, L. A., Osganian, S. K., Parcel, G. S. (2004). School based health education programs can be maintained over time: results from the CATCH Institutionalization study. Prev e ntive Med icine 38 (5), 594 606. Horowitz, M., Shilts, M. K., & Tow nsend, M. S. (2004). EatFit: A goal oriented intervention that challenges adolescents to improve their eating and fitness c hoices. Journal of Nutrition Education and Behavior 36 (1), 43 44. Hox, J. J. (2010). Multilevel analysis : techniques and applications New York: Routledge. Hsieh, H. F., & Shannon, S. E. (2005). Three approaches to qualitative content a nalysis. Qualitative Health Research, 15 (9), 1277 1288. Johnson, E., McInnes, M. M., & Shinogle, J. A. (2006). Symposium: Obesity economics: What is the economic cost of overweight c hildren? Eastern Economic J ournal., 32 (1), 171.

PAGE 147

142 Kam, C. M., Greenberg, M., & W alls, C. (2003). Examining the role of implementation quality in school base d prevention u sing the PATHS c urriculum. Prevention Science, 4 (1), 55 63. Kamath, C. C., McGovern, L., Erwin, P. J., Montori, V. M., Vickers, K. S., Ehrlich, A. (2008). Behavioral interventions to prevent childhood obesity: A systematic review and metaana lyses of randomized trials. J. Clin. Endocrinol. Metab. Journal of Clinical Endocrinology and Metabolism, 93 (12), 4606 4615. Kelder, S. H., Perry, C. L., Klepp, K. I., & Lytle, L. L. (1994). Longitudinal tracking of adolescent smoking, physical activity, and food choice behaviors. Am erican Journal of Public H ealth, 84 (7), 1121 1126. Khandker, S. R., Koolwal, G. B., & Samad, H. A. (2010). Handbook on impact evaluation: qu antitative methods and practices. Washington, D.C.: The World Bank. Kirks, B. A., & Wolff, H. K. (1985). A comparison of methods for plate waste determinations. Journal of the American Dietetic Association, 85 (3), 328 331. Koplan, J. P., Liverman, C. T., & Kraak, V. I. (2005). Preventing childhood obesity: health in the balance: executive summary. Journal of the American Dietetic Association, 105 (1), 131 138. Krebs, N. F., Himes, J. H., Jacobson, D., Nicklas, T. A., Guilday, P., & Styne, D. (2007). Assess ment of child and adolescent overweight and obesity. Pediatrics, 120 193 228. Kristjansson, A. L., Sigfusdottir, I. D., & Allegrante, J. P. (2010). Health behavior and academic achievement among adolescents: The relative contribution of

PAGE 148

14 3 dietary habits, p hysical activity, body mass index, and self esteem. Health Education and Behavior, 37 (1), 51 64. Leuven, E., & Sianesi, B. (2003). PSMATCH2. Stata module to perform full Mahalanobis and propensity score matching, common support graphing, and covariate im balance testing. (Version Statistical Software Components S432001). Chestnut Hill, Massachussetts: Boston College Department of Economics. Levi, J., Segal, L. M., St. Laurent, R., Vinter, S., Trust for America's, H., & Robert Wood Johnson, F. (2010). F as in fat how obesity threatens America's future: 2010. from http://healthyamericans.org/reports/obesity2010/ Li, J., & Hooker, N. H. (2010). Childhood obes ity and schools: Evidence from the national survey of children's health. Journal of School Health, 80 (2), 96 103. Livingstone, M. B., Robson, P. J., & Wallace, J. M. (2004). Issues in dietary intake assessment of children and adolescents. The British Jour nal of N utrition, 92 213 222. Luepker, R. V., Perry, C. L., McKinlay, S. M., Nader, P. R., Parcel, G. S., Stone, E. J. (1996). Outcomes of a field trial to improve children's dietary patterns and physical activity. The Child and Adolescent Trial for Card iovascular Health. CATCH collaborative group. J ournal of the American Medical Association, 275 (10), 768 776. Lytle, L. A. (1994). Nutrition education for school aged children: A review of research Washington, DC: Office of Analysis and Evaluation.

PAGE 149

144 Lytle, L. A., Johnson, C. C., Bachman, K., Wambsgans, K., Perry, C. L., Stone, E. J. (1994). Successful recruitment strategies for school based health promotion: experiences from CATCH. J ournal of Sch ool Health, 64 (10), 405 409. Lytle, L. A., Stone, E. J., Nich aman, M. Z., Perry, C. L., Montgomery, D. H., Nicklas, T. A. (1996). Changes in nutrient intakes of elementary school children following a school based intervention: results from the CATCH Study. Preventive Medicine, 25 (4). McCullagh, P., & Nelder, J. A. (1989). Generalized linear models. (Vol. 2). London: Chapman and Hall. McLeroy, K. R., Bibeau, D., Steckler, A., & Glanz, K. (1988). An Ecological Perspective on Health Promotion Programs. Health Education & Behavior, 15 (4), 351 377. McPherson, R. S., Ho elscher, D., Alexander, M., Scanlon, K., & Serdul a, M. (2000). Dietary assessment methods among school aged children: Validity and r eliability. Preventive Medicine, 31 (2), 11. Moore, S. N., Murphy, S., & Moore, L. (2011). Health improvement, nutrition rel ated behaviour and the role of school meals: the usefulness of a socio ecological perspective to inform policy design, implementation and evaluation. Critical Public Health, 21 (441 454). Moore, S. N., Murphy, S., Tapper, K., & Moore, L. (2010). The social physical and temporal characteristics of pri mary school dining halls and their implication s for child ren's eating behaviours. Health Education 110 (5), 399 411.

PAGE 150

145 Mulhall, A. (2003). In the field: notes on observation in qualitative research. Journal of Advanced Nursing, 41 (3), 306 313. Nader, P. R., Sellers, D. E., Johnson, C. C., Perry, C. L., Stone, E. J., Cook, K. C. (1996). The effect of adult participation in a school based family intervention to improve Children's diet and physical activity: the C hild and Adolescent Trial for Cardiovascular Health. Preventive Medicine, 25 (4). Nader, P. R., Stone, E. J., Lytle, L. A., Perry, C. L., Osganian, S. K., Kelder, S. (1999). Three year maintenance of improved diet and physical activity: the CATCH cohort. C hild and Adolescent Trial for Cardiovascular Health. Arc hives of Pediatrics & Adolescent M edicine, 153 (7), 695 704. Newton, X. A. & Llosa, L. (2010). Toward a more nuanced approach to program effectiveness and a ssessment: Hierarchical linear models in K 12 Program Evaluation. American Journal of Evaluation, 31 (2), 162 179. NHLBI. (2009, 2009/07//). We Can! www.nhlibi.nih.gov/health/public/heart/obesity/wecan/index.htm W ebsite overview, Medicine on the Net, p. 23. Retrieved from http://go.galegroup.com/ps/i.do?id=GALE%7CA203898517&v=2.1&u=aura ria_main&it=r&p=AONE&sw= w Nicklas, T. A., Webber, L. S., Srinivasan, S. R., & Berenson, G. S. (1993). Secular trends in dietary intakes and cardiovascular risk factors of 10 y old children: the Bogalusa Heart Study (1973 1988). American Journal of Clinical N utrition, 57 (6), 930 9 37.

PAGE 151

146 Ogden, C. L., Carroll, M. D., Curtin, L. R., Lamb, M. M., & Flegal, K. M. (2010). Prevalence of high body mass index in US child ren and adolescents, 2007 2008. Journal of the American Medical Association, 303 (3), 242 249. Perez (2001). The school setting: an opportunity for the implementation of dietary guidelines. Public Health N utrition, 4 (2B), 717 724. Perry, C. L., Bishop, D. B., Taylor, G ., Murray, D. M., Mays, R. W., Dudovitz, B. S. (1998 ). Changing fruit and vegetable consumption among children: the 5 a Day Power Plus program in St. Paul, Minnesota. American Journal of Public H ealth, 88 (4), 603 609. Perry, C. L., Bishop, D. B., Taylor, G. L., Davis, M., Story, M., Gray, C. (2004). A Randomized s chool t rial of environmental strategies to encourage fruit and vegetable consumption among c hildren. Health Education & Behavior, 31 (1), 65 76. Perry, C. L., Lytle, L. A., Feldman, H., Nicklas, T ., Stone, E., Zive, M. (1998). Effects of the Child and Adolescent Trial for Cardiovascular Health (CATCH) on Fruit and Vegetable Intake. Journal of Nutrition Education 30 (6), 354 360. Perry, C. L., Sellers, D. E., Johnson, C., Pedersen, S., Bachman, K. J., Parcel, G. S. (1997). The Child and Adolescent Trial for Cardiovascular Health (CATCH): intervention, implementation, and feasibility for elementary schools in the United States. Health Education and Behavior 24 (6), 716 735. Peugh, J. L. (2010). A pr actical guide to multilevel modeling. Journal of School Psychology, 48 (1), 85 112.

PAGE 152

147 Pope, C., Ziebland, S., & Mays, N. (2000). Qualitative Research in Health Care: Analysing Qualitative Data. British Medical Journal, 320 (7227), 114 116. Powers, A. R., Str uempler, B. J., Guarino, A., & Parmer, S. M. (2005). Effects of a nutrition education program on the dietary behavior and nutrition knowledge of second grade and third grade students. Journal of School Health, 75 (4), 129 133. Rabe Hesketh, S., & Skrondal, A. (2008). Multilevel and Longitudinal Modelling Using Stata (2 ed.). College Station: Stata Press. Raudenbush, D. S. W., & Bryk, D. A. S. (2002). Hierarchical linear m odels: Applicat ions and data analysis m ethods. (2nd. ed.). Thousand Oaks, CA: Sage Publ ications. Resnicow, K., Cohn, L., Reinhardt, J., Cross, D., Futterman, R., Kirschner, E. (1992). A Three Year Evaluation of the Know Your Body Program in Inner City Schoolchildren. Hea lth Education and Behavior, 19 (4), 463 480. Reynolds, K. D., Franklin, F. A., Binkley, D., Raczynski, J. M., Harrington, K. F., Kirk, K. A. (2000). Increasing the fruit and vegetable consumption of fourth graders: results from the high 5 project. Preventive Medicine, 30 (4), 309 319. Reynolds, K. D., Hinton, A. W., Shewchuk, R. M., & Hickey, C. A. (1999). Social cognitive model of fruit and vegetable consumption in elementary school c hildren. Journal of Nutrition Education 31 (1), 23 30. Reynolds, K. D., & Spruijt Metz, D. (2006). Translational research in childhood obesity p revention. Evaluation & the Health Professions, 29 (2), 219 245. Ringwalt, C. L., Ennett, S., Johnson, R., Rohrbach, L. A., Simons Rudolph, A., Vincus, A. (20 03). Factors a ssociated with fidelity to substance use p revention

PAGE 153

148 curriculum guides in the nation's middle s chools. Health Education and Behavior, 30 (3), 375 391. Roblin, L. (2007). Childhood obesity: food, nutrient, and eating habit trends and influences. A pplied Physiology and Metabolism 32 (4), 635 645. Rogers, E. M. (2003). Diffusion of in novations. (5 ed.). New York: Free Press. Rosenbaum, P. R., & Rubin, D. B. (1983). The central role of the propensity score in observational studies for causal e ffects. Biometrika, 70 (1), 41 55. Rossman, G. B., & Wilson, B. L. (1994). Numbers and words re visited: being shamelessly eclectic. Quality & quantity., 28 (3), 315. Rubin, D. B. (2001). Using propensity scores to help design observational studies: Application to the tobacco l itigation. Health Services & Outcomes Research Methodology, 2 (3 4), 3 4. Sauberli, W., Lee, H., Contento, I. R., Koch, P., & Calabrese Bar ton, A. (2008). O34: Enhancing personal a gency and competence in eating and moving: An outcome evaluation of choice, control, and c hange (C3), an inquiry based middle school science curriculu m to reduce obesity r isk. Journal of Nutrition Education and B ehavior., 40 (4), 3. Schanzenbach, D. W. (2009). Do school lunches c ontribute to childhood obesity? Journal of Human Resources, 44 (3), 684 709. School Nutrition Association, S. (n.d.). Retrie ved October 12, 2011, from http://www.asfsa.org/ Schw arz, G. (1978). Estimating the dimension of a m odel. The Annals of Statistics, 6 (2), 461 464.

PAGE 154

149 Scott, J. (2008). Children as respondents: The challenge for quantitative methods. In Pia Monrad Christensen & Allison James (E ds.), Research with children: perspectives and practices, 2nd ed. (pp. 87 108). New York: Routledge. Scott, L. (2010). BMI Trends in Aurora Public School Students. Serdula, M. K., Ivery, D., Coates, R. J., Freedman, D. S., Williamson, D. F., & Byers, T. (1993). Do o bese children become obese adults? A review of the l iterature. Preventive Medicine, 22 (2), 167 177. Shadish, W. R., & Steiner, P. M. (2010). A p rimer on propensity score a nalysis. Newborn and Infant Nursing R eviews, 10 (1), 19 26. Singh, A. S., Mulder, C., Twisk, J. W. R., van Mechelen, W., & Chinapaw, M. J. M. (2008). Tracking of childhood overweight into adulthood: a systematic review of the literature. Obesity Reviews, 9 (5), 474 488. Singh, G. K., Kogan, M. D., Van Dyck, P. C., & Siahpush, M. (2008). Racial/ethnic, socioeconomic, and behavioral determinants of c hil dhood and adolescent o besity i n the United States: Analyzing independe nt and joint a ssociations. Annals of Epidemiology, 18 (9), 682 695. Singh, G. K., Siahpush, M., & Kogan, M. D. (2010). Rising social inequalities in US childhood obesity, 2003 2007. Annals of Epidemiology, 20 (1), 40 52. Singleton, R., & Straits, B. C. (20 10). Approaches to social r esearch New York: Oxford University Press. Stevens, V. J., Obarzanek, E., Franklin, F. A., Steinmuller, P., Snetselaar, L., Lavigne, J. (1995). Dietary Intervention Study in Children (DISC): Intervention design and p articipation Journal of Nutrition E ducation., 27 (3), 133.

PAGE 155

150 St uart, E. A. (2007). Estimating causal effects using s chool level data s ets. Educational Researcher, 36 (4), 187 198. Swanson, M. (2008). Digital photography as a tool to measure school cafeteria consumption. J. Sch. Health Journal of School Health, 78 (8), 432 437. Taras, H., & Potts Datema, W. (2005). Obesity and student performance at s chool. The Journal of School Heal th, 75 (8), 291 295. Templeton, S. B., Marlette, M. A., & Panemangalore, M. (2005). Competitive foods increase the intake of energy and decrease the intake of certain nutrients by adolescents consuming school lunch. Journal of the American Dietetic Associa tion, 105 (2), 215 220. Tobler, N. (2000). Lessons Learned. Journal of Primary Prevention, 20 (4), 261 274. Tran, K. M., Johnson, R. K., Soultanakis, R. P., & Matthews, D. E. (2000). In person vs telephone administered multiple pass 24 hour recalls in wome n: validation with doubly labeled water. Journal of the American Dietetic Association, 100 (7), 777 783. U.S. Department of Agriculture., USDA FNS. (2011). School Lunch and Breakfast Program participation and meals s erved. Retrieved September 1, 2011, fr om http://www.fns.usda.gov/pd/ U.S. Department of Agriculture, USDA FNS. (October, 2011). National School Lunch Program Fact Sheet. Retrieved October 14, 2011, from http://www.fns.usda.gov/cnd/Lunch/

PAGE 156

151 Van Horn, L., Obarzanek, E., Friedman, L. A., Gernhofer, N., & Barton, B. (2005). Children's adaptations to a fat reduced diet: the Dietary Intervention Study in Children (DISC). Pediatrics, 115 (6), 1723 1733. Vivi an, E. M. (2006). Type 2 diabetes in children and adolescents the next epidemic? Current Medical Research & Opinion, 22 (2), 297 306. Walter, H. J. (1989). Primary Prevention of Chronic Disease among Children: The School Based "Know Your Body" Interventi on Trials. Health Education and Behavior Health, 16 (2), 201 214. Wampold, B. E., & Serlin, R. C. (2000). The consequence of ignoring a nested factor on measures of effect size in analysis of variance. Psychological Methods, 5 (4), 425 433. Wang, Y. & Bey doun, M. A. (2007). The obesity epidemic in the United States gender, age, socioeconomic, racial/ethnic, and geographic characteristics: A systematic review and meta r egress ion a nalysis. Epidemiologic Reviews, 29 (1), 6 28. Wechsler, H., McKenna, M., Lee, S., & Dietz, W. (2004). The role of schools in preventing childhood obesity. The State Education Standard (December), 4 11. Williamson, D. A., Allen, H. R., Martin, P. D., Alfonso, A. J., Gerald, B., & Hunt, A. (2003). Comp arison of digital photography to weighed and visual estimation of portion sizes. Journal of the American Dietetic Association, 103 (9), 1139 1145.

PAGE 157

152 Wilson, S. J., Lipsey, M. W., & Derzon, J. H. (2003). The effects of school based intervention programs on ag gressive behavior: A meta analysis. Journal of Consulting and Clinical Psychology, 71 (1), 136 149. Wolper, C., Heshka, S., & Heymsfield, S. B. (1995). Measuring food intake: An overview. In David B. Allison & Monica L. Baskin (Eds.), Handbook of assessmen t methods for eating behaviors and weight related problems : measures, theory, and research (pp. 215 240). Los Angeles: Sage Publications. World Health Organization (2012). Social determinants of health. Retrieved January 4, 2011 http://apps.who.int/gb/ebwha/pdf_files/WHA65/A65_16 en.pdf Zhao, Z. (2004). Using matching to estimate t reatment effects: Data requirements, matching metrics, and monte carlo e vidence. The Review of Economics and Statistics, 86 (1), 91 107.

PAGE 158

153 Appendix A : Aurora Public School Sample Menu December 2011

PAGE 159

154 Appendix B: Go Slow Whoa Rating System (Coordinated Appr oach to Child Health (CATCH), 2012) The CATCH Go Slow Whoa List is a tool to guide children and families toward making healthful choices. The overall message i s that all foods can fit into a healthful diet, which consists of more GO foods than SLOW fo ods, and more SLOW foods than WHOA foods. NOTE: Aurora Public Schools used the following tables to guide the labeling of menu items as Go, Slow, or Who a foods. Menu items might have individual components ( i.e., bread, meat, dressing, cheese) from different categories, and thus the overall food is given an average rating by APS Nutrition Services staff and put into a Go, Slow, or Whoa food category. Foods are divided into seven sections, 5 of which are food groups (Vegetables; Fruits; Grains; Milk and Dairy Foods; Meat, Beans, and Eggs) and two of which are Fats and Other. GO foods: Examples include fruits and vegetables, whole grain foods, and unsweet ened 1% milk. GO foods are commonly described as whole foods, meaning that they a re generally the least processed compared to foods in the same food group/sec tion. These foods are also lowest in salt (sodium) and/or added sugars. In addition, GO foods are lo west in unhealthy fats (i.e. solid fats such as butter or lard ) as opposed to healthy fats, wh ich are vegetable oils. WHOA foods: Examples include candy, cookies, chips, fried foods, ice cream, soft drinks, and sugary cereals. WHOA foods are generally the most processed and are highest in unhealthy fats, added sugars, and/or salt. SLOW foo d s are in between GO foods and WHOA foods. Examples include s weetened (including flavored) 1% milk, refined grain foods, and fruit with added sugar. To determine whether a food is GO, SLOW, or WHOA it i s compared to all the other foods in its category (row) of the food group/section. Although foods are categorized in this way, it i s important to note that eating large quantities of foods can be unhealthy, even if they a re GO foods. The GO SLOW WHOA L ist does no t contain combination foods such as sandwiches or pizzas since each ingredient is either a GO, SLOW, or WHOA food. For instance, a pizza is made up of a crust, sauce, cheese, and toppings. These ingredients belong in more than one food group. To determine if the pizza is a GO, SLOW, or WHOA pizza, you should take all the ingredients into consideration. The most healthful type of meal includes mostly GO foods. Here are two examples: GO Breakfast Go Lunch Oatmeal without added sugar (GO) Turkey sandwich Brown sugar (WHOA) Whole wheat bread (GO) Fresh Blueberries (GO) Turkey without skin (GO) 1 % milk (GO) Mustard (GO) American cheese (WHOA) Tomato (GO) Lettuce (GO) Baked potato chips (SLOW) Canned peaches (no added sugar) (GO) Skim Milk (GO)

PAGE 160

155 VEGETABLES Less than 200 mg of sodium (about one pinch of salt) per cup Examples of vegetables: asparagus, avocado, bean sprouts, beets, bok choy, broccoli, Brussels sprouts, cabbage, carrots, cauliflower, celery, chayote, collard greens, corn, cucumbe rs, eggplant, garlic, green beans, jicama, kale, lettuce, mushrooms, mustard greens, nopalitos, okra, onions, parsnip, peas, peppers (such as bell, jalapeno, poblano, etc.), potatoes, pumpkin, sweet potatoes, spinach, squash, taro root, tomatillos, tomatoe s, turnip greens, turnips, yucca (cassava FRUITS Less than 200 mg of sodium (about one pinch of salt) per cup Examples of fruits: apples, apricots, bananas, blackberries, blueberries, cantaloupe, cherries, figs, grapefruit, grapes, honeydew melons, kiwi, kumquats, lemons, limes, mangos, nectarines, oranges, papaya, peaches, pears, persimmons, pineapple, plums, pom egranate, star fruits, strawberries, tangerines, watermelon G O SLOW WHOA Vegetables vegetables with no salt, sugar, or fat added, or with a small amount of salt* added french fries and hash browns canned vegetables made Fried potatoes, fried french fries, fried hash browns Vegetable Juice sodium vegetable juice GO SLOW WHOA Fruits with no sugar or salt added, or with a small amount of salt* added with sugar and/or salt added syrup Fruit Juice fruit juice bars and smoothies bars and smoothies with added sugar Dried Fruit/ Fruit Leather fruit leather leather with added sugar ups

PAGE 161

156 GRAINS GO SLOW WHOA Breads/Muffins/ Sweet Breads grain bread, buns, rolls, Corn tortillas pancakes, and French toast made with vegetable oils pancakes, and French toast made with solid fats Pasta grain flour noodles fat noodle soups Rice and Grains grains (amaranth, barley, buckwheat, corn, whole cornmeal, millet, oats, quinoa, Whole wheat (spelt, durum, farro [emmer], cracked wheat, wheat berries, bulgur) Fried rice Cereals grain, low sugar cereals (such as toasted oats, shredded wheat, oatmeal, muesli) sugar cereals made with sugar cereals made with refined grains solid fats Crackers fat whole grain crackers fat crackers made with refined grains fat crackers Chips Pretzels chips (such as cheese puffs, corn chips) Cookies/Cake Graham crackers bars Popcorn popped popcorn with no salt added and/or salt popcorn (such as corn

PAGE 162

157 MILK and DAIRY FOODS GO SLOW WHOA Milk free (skim/non 1$ (low almond, and rice milk fat dry milk free (skim/non Fortified soy, almond, and rice milk sweetened (reduced Milkshakes Yogurt free or low fat plain or 100$ fruit juice sweetened free or low fat yogurt drinks unsweetened free or low fat yogurt free or low fat yogurt drinks sweetened Whole milk yogurt drinks Cheese Low fat fat soy cheese or reduced Low Powdered cheese sauce cheese (whole Cottage cheese (whole milk) Sour Cream fat sour cream Dairy Desserts fat or low Low with skim or 1$ milk made with 2$ or whole

PAGE 163

158 MEAT, BEANS and EGGS Less than 200 mg of sodium (about one pinch of salt) per cup GO SLOW WHOA Dried Beans and Peas red, garbanzo), peas (such as black eyed, split, purple hull), and lentils with no salt or fat added, or with a small amount of salt* added Beans, peas, and lentils with salt Falafel lentil s made with solid beans, canned Nuts and Seeds with no added salt, sugar, or fat Peanuts, almonds, pecans, walnuts, cashews, and pistachios with no peanut butter and other nut butters pecans, walnuts, and pistachi os with added Peanut butter and other nut butters with added salt, sugar, and/or fat Eggs substitute Fish baked, grilled or broiled (such as salmon, catfish, shrimp, crab, fish sticks Poultry skin (baked, grilled, or broiled) turkey chicken nuggets Beef round roast, round steak, sirloin, lean ground been drained and rinsed fat (such as brisket, T bone, chuc k roast) Ribs Pork chops or tenderloin without fat) Regular cuts of pork (such as pork roast, shoulder, ham) Other Protein Foods based meat substitutes Processed Meat Turkey or chicken sausage

PAGE 164

159 FATS OTHER GO SLOW WHOA Fats stick cooking spray canola, peanut, soybean, corn, cottonseed, safflower, or sunflower) butter, margarine, shortening, lard, salt, pork) Foods Rich in Fats salad dressing made with vegetable oils sauces, and salad dressing made with solid fats GO SLOW WHOA Herbs and Spices (such as basil, rosemary, salt (such as garlic powder or onion powder) glutamate (MSG) Sugars/ sugar syrup Sweeteners/ Candy Agave nectar Beverages Unsweetened decaffeinated tea flavored Spreads/ Condiments

PAGE 165

160 Appendix C: Matching Data School Name # kids Nat Amer Asian Black Hisp White Nat. Haw. Two Eth or > frl ell gt sped np % n % % % % % % % % % % % n Altura ES 462 1.1 0.6 11.5 74.5 9.7 0.0 2.6 96.5 66.0 1.7 6.3 1 Arkansas ES 548 0.5 3.3 23.5 42.7 24.8 0.9 4.2 60.9 33.0 2.6 6.8 1 Aurora Frontier K 8 628 0.6 12.7 17.8 25.6 37.7 0.2 5.3 29.3 14.8 7.0 8.1 0 Century ES 344 0.6 4.4 27.3 34.0 27.9 0.6 5.2 68.3 27.9 2.9 10.8 0 Crawford ES 641 1.4 7.6 7.8 77.8 2.8 0.0 2.5 89.4 78.3 1.2 9.8 1 Dalton ES 521 1.3 6.3 14.8 32.6 40.7 0.2 4.0 52.0 19.6 7.5 12.9 0 Dartmouth ES 385 1.0 3.1 16.9 36.1 37.1 0.0 5.7 57.4 19.7 3.6 13.8 0 Elkhart ES 704 0.7 5.7 14.5 70.0 6.7 1.4 1.0 94.7 67.2 1.6 10.2 3 Fletcher Primary 382 0.5 2.1 5.2 88.5 2.4 0.0 1.3 93.2 83.5 1.0 6.3 2 Fulton ES 487 0.8 5.7 12.5 72.3 6.2 0.0 2.5 91.6 66.9 1.4 6.6 2 Iowa ES 467 0.9 4.3 23.3 42.0 25.9 0.4 3.2 73.0 35.1 3.2 9.9 0 Jewell ES 531 0.9 5.3 26.2 44.6 19.4 0.4 3.2 70.2 40.1 2.4 9.0 1 Kenton ES 553 0.5 2.0 9.0 81.2 5.1 0.4 1.8 87.9 74.3 2.5 11.2 1 Lansing ES 405 1.5 2.0 31.4 55.1 7.4 0.2 2.5 84.7 52.3 1.5 7.9 2 Lyn Knoll ES 28 0.3 1.0 26.6 64.7 5.9 0.0 1.4 88.9 64.0 2.1 5.2 1 Peoria ES 494 0.4 2.2 11.3 76.7 7.9 0.0 1.4 83.8 67.8 1.0 9.7 0 Sable ES 441 0.7 5.2 22.9 59.4 8.8 0.7 2.3 86.4 60.3 1.4 6.3 3

PAGE 166

161 School Name # kids Nat Amer Asian Black Hisp White Nat. Haw. Two Eth or > frl ell gt sped np % n % % % % % % % % % % % n Sixth Avenue ES 610 0.3 2.3 18.2 71.3 5.4 0.5 2.0 89.2 62.8 2.3 8.0 3 Vanguard 508 0.6 2.2 15.4 51.8 27.2 0.0 3.0 55.7 32.9 1.0 12.8 0 Vassar ES 550 0.2 4.5 22.7 38.2 29.6 0.4 4.4 59.3 23.1 2.7 12.7 0 Vaughn ES 498 0.8 1.0 14.1 71.5 9.8 1.0 1.8 85.9 61.6 2.0 5.8 0 Virginia Court 474 0.6 3.6 15.6 71.1 7.4 0.0 1.7 85.2 65.0 0.4 7.8 1 VISTA Peak P 8 Exploratory 803 0.6 1.2 14.7 38.1 42.0 0.1 3.2 48.8 22.0 3.0 7.6 1 Wheeling ES 613 1.5 3.8 17.5 61.8 12.1 0.2 3.3 81.9 51.5 1.8 9.3 0 Yale ES 501 1.0 2.6 18.4 50.7 25.5 0.0 1.8 69.9 43.7 2.6 6.0 0

PAGE 167

162 Appendix D : Qualitat ive Semi Structured Interview Guide 1. What year did you kick off the GSW program? _____________ 2. Tell me about the nutrition knowledge and food intake of the kids in your school. (Prompt for attitudes, motivation, knowledge, peers, role modeling, etc.) 3. Tell me about the Scratch Cooking Cafeteria Changes in your school. (Prompt for what the teachers know, what they have seen, what the kids might tell them about changes.) What do students like and dislike about recent cafeteria/menu changes? 4. Tell me about your experience with the Go, Slow, Whoa program. 5. What do your students like and dislike about the GSW program? 6. What parts of the GSW program are confusing to kids? 7. In your opinion, what grade levels understand and/or like the program most? Why? 8. Do you think you are adequately prepared to deliver the GSW curriculum? Why ? 9. How much education/training do you have in nutrition ed? Please describe your training. 10. 11. How much tim e did you spend GIVING GSW lessons throughout this past school year? (# of hours per week or per month) 12. What methods or strategies have you used to deliver GSW programming? 13. 14. In your opinion, wh at are the strengths and weaknesses of the GSW program? 15. If you were an administrator, would you want GSW at your school? Would you want to continue or make more scratch cooking changes? (prompt with: are these two interventions a good use of time and mo ney?)

PAGE 168

163 Appendix E: Field Notes Field Notes Recording Sheet for lead data collectors: Wednesday, April 18 1. Pre Lunch Question: What types of milk are available before the start of lunch? ( i.e. skim wh ite, skim chocolate, 1% white) 2. Pre Lunch Question: What is available in the salad bar at the start of lunch? 3. Pre Lunch: What, exactly, is offered in the hot lunch line today? (This is redundant with the samples, but I just want to be sure we are aware of everything.) 4. Pre Lunch: Do you see any Go, Slow, Whoa labels on the hot lunch line foods (Red, Yellow, or Green circles/squares/triangles)? 5. Post Lunch: What type of milk is available at the end of your last lunch shift of the day? 6. Notes about cafeteria environment: (i.e. GSW posters? Describe interaction between staff and students at lunch. What are students talking about and doing at lunch (i.e. socializing vs. eating? Hurrying to get outside and play? Trading food?)

PAGE 169

164 Appendix F: Interrater Reliability Kappa Coefficients Day 1 Interrater Reliability Variable Kappa Kappa AFTER 3rd Rater Reconciliation Grade 1 1 Male 0.98 1 Type of Milk Taken 0.94 1 Ounces of Milk Left 0.97 1 Took Beef? 0.9 1 % Beef Left 0.37 0.39 Took Biscuit? 0.96 1 % Biscuit Left 0.56 0.58 Took Apple? 0.89 1 % Apple Left 0.4 0.43 Took Peas? 0.91 1 % Peas Left 0.4 0.41 Took Pears? 0.9 1 % Pears Left 0.47 0.54 Took Broccoli? 0.78 1 % Broccoli Left 0.35 0.49 Took Carrots? 0.96 1 % Carrots Left 0.23 0.2 Took Plum? 1 1 % Plum Left 0.07 0.07 Took Broccoli? 0.91 1 Took Carrots? 0.92 1 Took Celery? 0.86 1 Took Cherry Tomatoes? 0.93 1 Took Chick Peas? 1 1 Took Cucumbers? 0.8 1 Took Spinach or Lettuce? 0.91 0.97 Took Apple? 0.82 0.96 Took Raisins 0.89 1 Took Strawberry? 0.99 1 Took Cantaloupe? 0.93 1 Took Black Beans? 0.67 1 Took Chicken? 0.85 1

PAGE 170

165 Day 2 Interrater Reliability Variable Kappa Kappa AFTER 3rd Rater Reconciliation Grade 0.99 1 Male 0.96 1 Type of Chicken Taken 0.51 0.99 % Chicken Left 0.32 0.32 Cake Taken? 0.98 1 % cake left 0.5 0.51 Baked Beans Taken? 0.97 1 % Baked Beans left 0.53 0.55 Green Beans Taken? 0.99 1 % Green Beans Left 0.42 0.42 Roll Taken? 0.96 0.99 % Roll Left 0.63 0.66 Carrots taken? 0.83 1 % Carrts left 0.46 0.55 Applesauce Taken? 0.95 1 % Applesauce left 0.19 0.22 Orange Taken? 0.99 0.99 % Orange Left 0.46 0.47 Peach Taken? 0.88 1 % Peach Left 0.48 0.86 Type of Milk Taken 0.98 1 Ounces of Milk Left 0.97 1 Broccoli Taken? 0.84 1 Cauliflower Taken? 0.67 1 Celery Taken? 0.97 1 Cherry Tomato Taken? 1 1 Cucumber Taken? 0.92 1 Spinach/Lettuce Taken? 0.93 1 Raisins Taken? 0.8 1 Strawberries Taken? 1 1 Cantaloupe Taken? 0.88 1 Black Beans Taken? 1 1

PAGE 171

166 Day 3 Interrater Reliability Variable Kappa Kappa AFTER 3rd Rater Reconciliation Grade 0.98 1 Male 0.94 1 Pizza Taken? 1 1 % Pizza Left 0.29 0.3 Cucumber Taken? 0.89 1 % Cucumber Left 0.2 0.28 Grapes Taken? 0.97 1 % Grapes Left 0.58 0.61 Green Beans Taken? 0.96 1 % Green Beans Left 0.54 0.56 Broccoli and Carrots Taken? 0.92 1 Broccoli and Carrots Left 0.25 0.25 Pineapple Taken? 0.59 1 Pineapple Left 0.38 0.47 Cake Taken? 0.74 0.99 Cake Left 0.49 0.49 Watermelon Taken? 0.35 1 Watermelon Left 0.09 0.16 Cauliflower Taken? 0.94 1 Cauliflower Left 0.27 0.43 Type of Milk Taken 0.96 1 Ounces of Milk Left 0.93 1 Broccoli Taken? 0.96 1 Cauliflower Taken? 0.41 1 Carrots Taken? 1 1 Celery Taken? 0.97 1 Cherry Tomatoes Taken? 1 1 Chick Peas Taken? 1 1 Cucumber Taken? 0.88 1 Spinach/Lettuce Taken? 0.95 1 Kiwi Taken? 0.99 1 Raisins Taken? 1 1 Strawberries Taken? 1 1 Cantaloupe Taken? 0.99 1 Black Beans Taken? 1 1 Chicken Taken? 0.43 1 Tropical Salad Taken? 0.83 0.96 Orange Taken? 0.84 1 Cake Taken? 0 1 Peach Taken? 0.95 1 Yogurt Taken? 0.96 1 Banana Taken? 0.91 1 Pear Taken? 0.7 1 Pineapple Taken? 0.75 1

PAGE 172

167 Appendix G : AIC and BIC for Random Intercept and Random Coefficient Models AIC Random Intercept Model AIC Random Coefficient Model BIC Random Intercept Model BIC Random Coefficient Model Choosing White Milk over Choc Milk/No milk 1790 1791 1835 1842 Choosing White Milk over Choc Milk 1697 1696* 1741 1746 Total FV Taken 5117 5119 5167 5175 Total Fiber Taken 7270 7272 7321 7329 Total Fiber Consumed 7403 7405 7454 7461 Total Fat Taken 12128 12130 12179 12186 Total Fat Consumed 13140 13143 13192 13199 Total Saturated Fat Taken 7349 7351 7399 7408 Total Saturated Fat Consumed 8687 8689 8738 8746 The only information criteria test of all models where the Random Coefficient Model showed a better fit than the Random Intercept Model.

PAGE 173

168 Appendix H: Residual Graphs for Final Models

PAGE 174

169