CHAPTER ONE

INTRODUCTION

Background

Most Americans, especially low-income individuals, do not consume adequate quantities

of fruits and vegetables for disease prevention and health promotion. Use of psychosocial and
fruit and vegetable consumption survey data derived from consumers, may be helpful in
understanding the barriers and promoters to fruit and vegetable consumption in the low-income
population. This information may be beneficial in planning nutrition education interventions for
the low-income population and informing policy makers.

According to nationwide nutrition surveys, Americans chronically lack fruits and

vegetables in their diet (Krebs-Smith & Kantor, 2001). Fruits and vegetables are vital for good
health because they are good sources of essential vitamins and minerals, and fiber. In addition to
essential vitamins and minerals, fruits and vegetables provide other phytonutrients with powerful
chronic disease risk reduction potential (Steinmetz & Potter, 1991a, 1996).

Lack of good nutrition, more specifically, low intake of fruits and vegetables, is clearly
implicated in a number of serious diseases. Based on growing scientific evidence, the incidence
of chronic diseases, such as cancer, coronary heart disease, atherosclerosis, and stroke can be
reduced through increased fruit and vegetable consumption. Fruit and vegetable consumption
may also play a preventive role in birth defects, cataract formation, hypertension, asthma,
diverticulosis, obesity, and diabetes (Van Duyn & Pivonka, 2000).

Due to the bulk of scientific evidence linking fruit and vegetable consumption with
disease prevention and health promotion, it is important to convey to consumers and policy
makers the health benefits of eating more fruits and vegetables, and to make individual and
systems level changes to enhance the rate at which these healthy dietary changes occur. People
who live in poverty experience worse health than people who are more affluent (Klerman, 1992).
The poor also obtain their health services in a less coordinated and comprehensive manner than
the affluent. This tends to create a greater financial burden on the government for health care.
For these reasons, good nutrition for disease prevention and health promotion is exceedingly
important for low-income individuals.

Fruit and Vegetable Consumption data

The 2000 Dietary Guidelines for Americans and the USDA Food Guide Pyramid
establish a minimum of 5 servings of fruits and vegetables as a dietary recommendation for good
health. The Year 2010 Health Objectives include a nutrition objective to increase fruit and
vegetable consumption to 75% of all Americans consuming at least 2 servings of fruits and 3
servings of vegetables by the year 2010 (DHHS, 2000).

In the 1980’s, dietary consumption survey data indicated a major deficit in fruit and
vegetable intake and underscored a need for a nutrition education program to promote
consumption of fruits and vegetables. In 1991, the 5 A Day for Better Health Program was
launched as a research initiative of the National Cancer Institute to conduct research on dietary

1
behavior change in fruit and vegetable consumption. A nationwide nutrition education/media
campaign grew out of the initiative (Heimendinger, et al., 1996).

In the fall of 1991, a baseline nationwide survey of fruit and vegetable consumption was
completed for the 5 A Day for Better Health Program. Data on a nationally representative
sample of 2,837 adults, with an oversampling of African-Americans and Hispanics, were
collected by telephone using a food frequency questionnaire. The results indicated that the
overall median intake was 3.4 fruit and vegetable servings per day and the mean intake was 3.8
daily servings. Overall, only 23 percent of the population was consuming five or more servings
per day. Disparities in fruit and vegetable consumption based upon income were noted. Those
respondents with less than 130% of the poverty index ratio, consumed a median intake of 3.1
servings daily. In contrast, those in the higher income category of greater than 300% of the
poverty index ratio, consumed a median intake of 3.7 servings daily. Therefore, those in the
low-income category ate, an average of 0.6 servings per day less than those in the high-income
category (Subar, et al., 1996).

Six years later in the fall of 1997, a follow-up survey was conducted using the same
methodology to assess the progress of the 5 A Day for Better Health Program. The results
indicated a modest positive, but statistically significant overall mean increase in fruit and
vegetable consumption of 0.23 daily servings (Stables, et al., in press). In addition, over the
same six-year time frame, there was a small and statistically significant increase in the percent of
those who consume at least five or more servings per day to 28% (Stables, et al., in press). As
was true with the baseline survey, low-income adults lagged behind in servings of fruits and
vegetables consumed.

These modest positive changes are consistent with other nationally representative surveys
of fruit and vegetables consumption. The 1989-1991 USDA Continuing Survey of Food Intakes
by Individuals (CSFII) shows an average adult intake of 4.1 servings of fruits and vegetables.
The 1994-1996 CSFII survey data showed that the average adult intake of fruits and vegetables
increased to an estimate of 4.6 servings, when fried potatoes are not included (Krebs-Smith,
1998, Krebs-Smith & Kantor, 2001). While this increase was statistically significant, it still
remains shy of the minimum of 5 servings per day. In the 1994-1996 CSFII survey, the low-
income population consumed 0.8 servings of fruits and vegetables less than the high-income
group (Krebs-Smith & Kantor, 2001).

Even though the data suggest that average intakes of fruits and vegetables are increasing,
and on average are approaching 5 servings daily, it should be noted that 5 daily servings
represent the minimum dietary guidance recommendation. Mean intakes should be closer to the
middle of the 5-9 serving range. Furthermore, there is a wide variation in what is considered a
fruit or vegetable for disease reduction purposes. A recent report by the World Cancer Research
Fund and the American Institute for Cancer Research (1997) suggested that potatoes prepared in
any way should not be considered a vegetable, and bananas and plantains should not be
considered fruits, in helping to reduce chronic disease risk, as they do not contribute to disease
reduction in epidemiology studies. If potatoes are excluded, the fruit and vegetable intakes
averaged 3.2 servings per person per day in 1989-1991; and comparable figures for 1994 would
be 3.6 servings (Krebs-Smith, 1998). This would make reported average fruit and vegetable

2
daily intake much lower, and the picture for decreasing risk much bleaker than currently
considered.

Even though there has been a major national initiative in place to increase fruit and
vegetable consumption over the past 10 years, the gains in consumption are not adequate for
decreasing chronic disease risk in this country. Certainly, further improvements could be made,
not only in quantity, but also in quality of fruits and vegetables consumed. Vegetable
consumption is dominated by less nutrient dense potato intake. Greater than 50% of the potatoes
marketed in this country eventually are eaten as fat laden French fries (Putnam, 1996). Nutrient
dense vegetables such as broccoli, other dark green vegetables, and carrots represent only a small
fraction of vegetable consumption (Putnam, 1996). No matter how the data is interpreted, the
current average fruit and vegetable intake in this country is poor to barely reaching the minimum
recommendation, at best, and consumption is considerably lower in the low-income population.

Economics of Fruits and Vegetables

The American food supply provided 5.3 servings, on average, of fruits and vegetables per
person per day in 1998 (Krebs-Smith & Kantor, 2001). This data reveals that total produce
supplies are adequate in this country to meet the minimum requirements. To meet the dietary
guidance requirements of 5-9 servings daily, fruits and vegetables must be uniformly available
and affordable to all demographic groups in this country. A recent USDA report reveals that
between 1982 and 1997, fruits and vegetables increased in price more than all other food groups.
The price increases for fruits and vegetables increased more than double the amount of increases
for processed foods (Putman, 1999). Some of this large increase is due to new trade practices
between produce shippers and retailers, that adds new fees, such as volume discounts and
slotting fees, the cost of which is passed on to the consumer (Calvin, et al., 2001). By and large,
the farmers are not experiencing a large increase in dollars received for the fresh product.
Rather, retailers tend to gain from the emerging trade practices. In fact, the fresh produce section
is currently the most profitable section of grocery stores. This translates into a higher profit
margin for retail grocery stores and higher than necessary prices for all consumers.

Health Policy
These issues of cost are particularly relevant for low-income consumers. Accessibility is
also a concern. A few studies suggest that low income households in poor central cities and low
populated rural areas often have less access to supermarkets and tend to pay more for food
(Kaufman, et al., 1997; Kaufman, 1999; Koralek, 1996). Even if costs are the same and fruits
and vegetables are accessible, low-income people simply have less disposable income. Data
from USDA Consumer Expenditure Surveys show that the low-income apportion their food
budget differently from higher income groups. For example, in 1998 the poorest households
(annual incomes of $12,367 or less) spent $295 per person on fruits and vegetables compared to
$739 spent per person by households in the highest income category (U.S. Department of Labor
Statistics, 2000).

Understandably the government is interested in the pursuit of health and the public
policies that relate to health. A healthy population is crucial from many perspectives, and the
search for health plays an important role in our nation’s economy. American’s spent more than
$1 trillion in the pursuit of health in 1997, and it is estimated that amount could increase to $1.5

3
trillion by 2002 (Thorpe, 1997). Public policies are authoritative decisions that are made in the
legislative, executive, or judicial branches of government. These decisions are intended to direct
or influence the actions, behaviors, or decisions of others. Generally, health policies affect or
influence groups or classes of individuals, such as the poor, the elderly, or children.

Policy mandated nutrition programs to enhance fruit and vegetable consumption

Only minor policy attention has been paid to nutrition education and price supports to
enhance fruit and vegetable consumption within food assistance programs. The Food Stamp
Program is an entitlement program available to almost all low-income households, and acts as a
social safety net. In 1995, a nutrition education component was added for food stamp
participants in some states. It was deemed important to educate low-income participants in the
importance of eating healthy diets and on the importance of using their food dollars to improve
or maintain health. Education regarding the importance of consuming at least 5 servings of
fruits and vegetables is included as part of the healthy diet message in some of the funded
participating states.

The USDA school lunch program also underwent great scrutiny in 1995, when it was
noted that most school lunches did not meet the U.S. Dietary Guidelines for fat content and fruit
and vegetable offerings. The USDA Undersecretary for Food, Nutrition and Consumer Services
made it policy to decrease the fat content of the meals and increase offerings of fruits and
vegetables in all school breakfast and lunches to meet the Dietary Guidelines for Americans.

The Special Supplemental Nutrition Program for Women, Infants, and Children (WIC)
has had a nutrition education component since it’s beginning in 1972 as a pilot program. More
recently, WIC Farmers’ Market Nutrition Program (FMNP) has been initiated in a few states.
With a $12 million appropriation in 1998, the FMNP provided vouchers for fresh fruits and
vegetables to low income women and children and boosted income for 9,600 farmers around the
nation (Weinberg, 1999). While this program is a contribution to good nutrition for many, it is
usually up to $20 per year per participant and only affects those enrolled in WIC in certain states.

The USDA Farm to School Program helps to create a market and provide a distribution
system for farmers to sell local grown produce to local schools, in an effort to improve the
quality and freshness of fruits and vegetables in schools participating in the National School
Lunch Program (USDA, 2000). While this program does not provide additional funding for
school lunch or directly to families, it does increase the selection and availability of fruits and
vegetables for children in agriculture growing areas of the country.

Statement of the Problem

Most Americans, especially low-income individuals, do not eat enough fruits and
vegetables for good health, and this creates an interesting challenge for nutrition educators and
policy makers. Since 1991, there has been an increased emphasis on improving the fruit and
vegetable consumption of Americans. Even though efforts to educate and motivate all
Americans to eat more fruits and vegetables have been put in place, the improvements are very
modest and the chronic problem of under consumption continues. Equally troublesome is the
disproportionately lower consumption of fruits and vegetables in low-income individuals, and

4
current consumption patterns have not significantly improved, despite the fact that a major public
health nutrition education program has been in place for over 7 years. Public health educators
and policy makers need to more fully understand the barriers and promoters to dietary behavior
change in the low-income population. The social consequences of the health sequelae related to
poor diets in this population are great.

The 5 A Day for Better Health Program 1997 Follow-up Survey includes many questions
that may provide insight into the barriers and promoters of fruit and vegetable consumption.
This is one of the few nationally representative surveys that captures consumption information
and data on demographics and psychosocial factors of fruit and vegetable consumption.

There are currently a few policy-mandated programs in place to improve the availability
and accessibility of fruits and vegetables for low-income people. However, these programs
usually target only a specific portion of the low-income population in a few states. Therefore,
only a few of those who would benefit have access to these programs. A comprehensive
nutrition program needs to be enacted whereby all low-income people have affordable healthy
foods available to them.

Purpose of the Study

The purpose of the study was to examine a nationally representative nutrition survey to
identify psychosocial, educational, and behavioral factors related to fruit and vegetable
consumption among adults, particularly low-income adults.
The findings addressed two foci:
1. Issues for behavior change at the individual level.
2. Issues of systems level change
In addition, there was an attempt to identify opportunities for future research to help in
identifying further factors related to fruit and vegetable consumption in the low-income
population.

Research Questions

The following research questions were addressed in this study:

1. What are the relevant constructs in adult’s fruit and vegetable consumption that
are being measured in the 1997 nationally representative survey for the National 5
A Day for Better Health program?
2. What differences in consumption and related factors are explained by income
level?
3. What profile of characteristics differentiates individuals who a) are aware of the 5
A Day program, b) exhibit high level of self-efficacy, c) have high level of
perceived personal benefits of good health-related behaviors, and d) eat high
levels of fruits and vegetables?

5
 Significance of the Study

Despite the fact that there has been a large-scale nutrition campaign designed to increase
fruit and vegetable consumption in this country for the past 10 years, it is clear that we are not
meeting the minimum goals of the nationally mandated nutrition objectives. This is true for all
Americans, but it is particularly problematic in the low-income population. Prevention of
chronic disease has never been more important because Americans are living longer. Thus,
delaying the onset of debilitating diseases not only make good health sense, it makes good
economic sense. This is especially poignant for those who live in poverty. The impact of
economic conditions on the health of the poor is especially dramatic. Health policies that
mitigate the negative influences of socio-cultural and economic environments on health are
important aspects of any society’s ability to help its members achieve better health.

The science base is strong for a decreased risk of chronic diseases when individuals
consume at least 5 daily servings of fruits and vegetables. A recent report estimated that “diets
containing substantial and varied amounts of vegetables and fruits will prevent 20% or more of
all cases of cancer” (WCRF/AICR, 1997). The authors estimated that "recommended diets,
together with maintenance of physical activity and appropriate body mass, can in time reduce
cancer incidence by 30-40%" (WCRF/AICR, 1997). This translates globally, to approximately
3-4 million cases of cancer per year that could be prevented by healthy eating and associated
lifestyle changes.

Current epidemiological evidence also suggests a strong protective effect of fruits and
vegetables on coronary heart disease and stroke. Two recent published reviews have come to
that conclusion for heart disease. A report by Klerk, et al. (1998) reviewed findings from 4
ecological, two case-control, and six cohort studies conducted after 1994. From this review, risk
reduction for coronary heart disease was estimated to be 20-40%. In the same review by Klerk
and colleagues (1998), the risk reduction for high fruit and vegetable intake on stroke may be up
to 25%.

These review studies reflect the preventive possibilities of diets high in fruits and
vegetables. In the United States, cancer is the second leading cause of death after cardiovascular
disease and accounts for one out of four deaths (McGinnis & Foege, 1993). In light of the wide
variety of protective substances in fruits and vegetables, the National Cancer Institute's 5 A Day
for Better Health Program and the American Cancer Society's 1996 nutrition guidelines advocate
eating at least 5 servings of fruits and vegetables daily (Heimendinger, et al., 1996).

The information derived from this survey analysis will be an important addition to the
scientific literature base on nutrition education behavior change in adults. It will also be highly
informative to the 5 A Day program in the context of overall national nutrition policy. The
policy relevant findings from this analysis will inform current and future policy regarding fruit
and vegetable nutrition programming at the state and national level. In addition, the findings
may inform the NCI and others as to what are the outstanding survey research needs in nutrition
surveillance in target populations.

6
 CHAPTER TWO
A REVIEW OF THE LITERATURE

This chapter will review the literature on nutrition education, particularly with respect to
increasing fruit and vegetable consumption in the low-income population. The material will be
organized around the following categories: 5 A Day Program overview, nutrition education
practices and approaches, psychosocial factors related to fruit and vegetable consumption, and
nutrition/health education in low-income populations.

5 A Day Program Overview

The National 5 A Day for Better Health Program, initiated in 1991, is a large-scale
public-private partnership between the fruit and vegetable industry and the United States’ (U.S.)
federal government, with the goal of increasing the average consumption of fruits and vegetables
in the United States to 5 or more servings every day. The long-range purpose is to reduce the
incidence of cancer and other chronic diseases through dietary improvements. The private side
of the partnership is coordinated by the Produce for Better Health Foundation, which is a non-
profit foundation composed of approximately 1000 members of the fruit and vegetable industry.
The National Cancer Institute of the National Institutes of Health in the U.S. Department of
Health and Human Services coordinates the public side of the partnership. As the federal
government specifically is concerned with populations with health disparities, the low-income
population is of special interest to the Program. The goal of the program is one of the national
health objectives for the nation and is also consistent with all other national dietary guidance
provided by the U.S. Government (USDA & DHHS, 2000; DHHS, 2000; USDA, 1992). The
specific program objectives are 1) to increase public awareness of the importance of eating 5 or
more servings of fruits and vegetables every day and 2) to provide consumers specific
information about how to incorporate more servings of fruits and vegetables into daily eating
patterns.

The 5 A Day national program structure is based on the concept of public and private
entities working together to positively impact the health of a population. The innovative and
highly successful 5 A Day Program structure has been used as an exemplary model of
government agencies working effectively with other government agencies and with private
industry to promote public health in the United States and abroad. The 5 A Day partnership has
a vision for modifying national dietary behaviors by capitalizing on the scientific credibility of
the NCI and the ability of the industry to reach the entire U.S. population at the point of sale.
Much of the nutrition and behavioral change research targeting increased fruit and vegetable
consumption has occurred since the initiation of the 5 A Day research initiative at the NCI.

Nutrition Education for Behavior Change: Philosophies and Approaches

Education Focus of the Dietetic Profession and in Practice

The American Dietetic Association (ADA) was founded in 1917, and established
professional, educational, and practice standards. A position paper on nutrition education for the
public (ADA, 1996) centered on the importance of behaviorally focused, evidence based,
consumer driven, nutrition education using research to guide interventions to improve food

7
choices, and to prevent disease and promote health. In order for “the public to achieve and
maintain optimal nutritional health, nutrition education should be an integral component of all
health promotion, disease prevention, and health maintenance programs, through incorporation
into all appropriate nutrition communications, promotion, and education systems” (p. 1183).

Evolution of nutrition education philosophy

Nutrition education has become more evidence-based and theory driven since the 1980’s.
Nutrition education has moved from the provision of knowledge and skills building activities to
combining nutrition education, research, and practice, in addition to guiding public nutrition
policy and fostering collaborative nutrition education efforts (ADA, 1996).

Early comprehensive reviews of nutrition education research by Whitehead in 1973 and

by Johnson and Johnson in 1985 found that nutrition education studies did not usually document
theoretical frameworks used in designing interventions. They did, however, report some
intermediary or outcome findings in terms of knowledge, attitudes, and dietary intakes. The
1973 review by Whitehead encompassing the previous seventy years found that “nutrition
education has been directed more toward the purpose of disseminating nutrition information than
toward the purpose of improving dietary habits (p. 99).” In fact, Whitehead notes that only
studies that set behavioral change as the goal of nutrition education and used educational
strategies to change behavior, actually achieved behavior change. Other nutrition education
efforts, which did not use strategies of behavior change, were typically successful only in
increasing knowledge.

The review by Johnson and Johnson (1985) used a meta-analysis of 303 studies. The
meta-analysis of the nutrition education interventions showed overall improvements of 33% in
knowledge, 19% in dietary practices, and 14% in attitudes. Johnson and Johnson concluded that
a shortcoming of most of the nutrition education research was that it was not based upon
theoretical models, and therefore was hard to determine the theoretical variables to be measured
as possible mediating factors to explain how the intervention impacted knowledge, attitudes, and
behaviors. It was difficult to determine what makes nutrition education effective, and to guide
development of future education efforts.

Achterberg and Clark (1992) reviewed 346 papers and abstracts for the purpose of
identifying the frequency of theory-driven nutrition education research and the types of theories
used from 1980 to 1990. Only 23.5% cited the use of a theory or model in the research. Of
those that utilized a theory driven approach, 57% focused on theories for the learner, such as
social learning theory, health belief model, and adult learning theory; and 22% utilized
curriculum development or testing theories.

In 1995, Contento, et al reviewed 217 nutrition education intervention research studies

for effectiveness and implications for policy, programs, and research. The general conclusion
was “the more effective programs are those that are behaviorally focused and based on
appropriate theory and prior research. Behaviorally focused nutrition education uses a set of
learning experiences to facilitate the voluntary adoption of food and nutrition-related behaviors
that are conducive to health and well being (p. 280).” Other elements that contribute to
effectiveness of nutrition education interventions were communication and educational strategies

8
for enhancing awareness and motivation, environmental interventions (changing the eating
environment to make healthy choices more available, accessible, and affordable), and
community activation and organization. The behavioral change strategies that add to
effectiveness included goal setting, tailoring intervention strategies for targeted groups, social
support, and enhancing self-efficacy.

This 1995 review found that the most effective and widely used behavioral change theory
utilized in nutrition education interventions was the social learning theory. Social learning
theory constructs or personal-level variables include self-efficacy (confidence in performing a
particular behavior), behavioral expectancies (perceived benefits), behavioral capabilities
(knowledge and skills to perform the behavior), and perception of the environment (Contento, et
al., 1995).

The public health approach to nutrition education encompasses educational and

behavioral change activities designed to move entire population groups to more healthful eating
patterns. A new conceptual framework for planning public health-focused nutrition education
interventions must address improving access to healthy foods, make it easier to eat healthy foods,
and assure greater awareness of nutrition messages (Johnson, et al., 2001). Interventions must be
developed that focus on the true determinants of nutrition behaviors and include effective social
marketing campaigns and communications (Johnson, et al., 2001).

A review of goal setting as a strategy for dietary behavior change in adults suggests that
goal setting is likely to lead to behavior change (Cullen, et al., 2001). Other reviews in nutrition
behavior change have revealed that successful programs target one or more personal, behavioral,
or environmental factors influencing the behavior of interest (Baranowski, et al., 1997,
Baranowski, et al., 1999).

Andragogical Approaches to Nutrition Education

The 1996 ADA position paper on nutrition education centers on adult education
(andragogical) approaches to nutrition intervention. People are more likely to make eating
behavior changes if they are involved in goal setting, and if their perspectives are incorporated
into the educational process.

80 of the 217 studies reviewed by Contento, et al. (1995) were conducted in the general,
healthy adult population. The more effective programs were those that were theory-based, in
particular, theory of reasoned action using persuasive communication strategies for increasing
awareness and motivation, and social cognitive theory. Behavioral change strategies in adults
that were likely to be effective were individualized interpersonal education, strategies
emphasizing the personal consequences of behaviors, and strategies based on social learning
theory and behavioral self-management (Contento, et al., 1995).

A study by Auld (1990) underscores the fact that people process information differently.
Even though both males and females seem to possess similar amounts of nutrition information
about fat and cholesterol; women tend to have a more complex base of nutrition knowledge, both
subjective and epistemological, that takes into account their own experiences and knowledge for

9
making food decisions. Based upon this research, Auld suggests that nutrition interventions be
tailored with gender and other socio-demographic characteristics in mind.

Kent (1988) describes the concept of empowerment through nutrition education. He

postulates that malnourished individuals suffer the consequences of those who are more powerful
and who may have other agendas than those of nutritional assistance. Kent supports the idea that
people should be supported to make their own analysis of their need for change and decide how
to make change happen in their own lives. The emphasis should be on creating a dialogue
among group members, and helping people uncover the barriers to behavior change, from the
sharing of experiences and interpretations.

Abusabha, et al. (1999) also approaches the use of empowerment in nutrition education
through facilitated group discussions with low-income adults. “By helping clients make
informed decisions about their diet, nutrition educators are empowering them to take control of
their eating habits (p. 73).” Rainey and Cason (2001) found these same andragogical theory
concepts in their qualitative research in planning nutrition education programs for low-income
elderly women. With input from the women regarding barriers (cost, availability, accessibility)
and limited understanding the health benefits of eating more fruits and vegetables, the
researchers were able to plan interventions that the women considered reasonable, realistic, and
achievable.

Other adult education theories in nutrition education

Nutrition education for social change, based on Paulo Freire's empowerment education
and principles of participatory research, was examined theoretically by Travers (1997) within the
context of the need to address inequities in health. Travers assumed that people make health
choices within the realm of available resources, and she proposed that traditional health
education strategies may have little impact on those who have limited access to health-
supporting resources such as adequate income for nutritious food, and culturally relevant health
education. Common strategies, such as dissemination of printed health information and targeting
culturally biased education programs to the socially disadvantaged groups, may actually increase
the differential between the advantaged and the disadvantaged by increasing the availability of
resources to those with literacy skills and income sufficient to enable them to act on
recommendations. In this way, the inequities are perpetuated. Simply transferring programs
that are inappropriate for low-income populations is what Travers equated to Freire's term
cultural invasion. In this way, the educator "seeks to penetrate another cultural-historical
situation and impose his system of values on its members" (Freire, 1973, p.113). Travers also
made the case for environmental changes, including legislation, as important for health education
to be effective for individuals and for social change. She stated that evidence suggests, "that
most major improvements in the public's health and their health behaviors have come through
movements based on public issues" (p. 345). She cited the antismoking campaigns as examples,
whereby education is combined with legislation to create environments more conducive to
healthy choices.

10
Social Cognitive Theory
Social Cognitive Theory (SCT) is based in the field of psychology and is intended as a
framework for identifying factors that influence behavior (Glanz, et al., 1997). Bandura (1986)
conceptualized this theory as a reciprocal interaction of environmental factors, behavior, and
cognitive and other personal factors. This theory proposes that behavior, person factors (such as
self-efficacy and perceived benefits), and environmental factors (such as social support), interact
to explain and predict changes in behavior, and a change in one component will produce changes
in the others. It combines an emphasis on behavioral factors, with the recognition of the
importance of cognition. This kind of model is an appropriate framework for examining dietary
and health behaviors because it allows for the structuring and ordering of variables known to
influence dietary and health behaviors.

Self-efficacy is a major construct in Social Cognitive Theory. Self-efficacy is the

confidence a person feels about performing a particular activity, including confidence in
overcoming the barriers to performing that behavior (Glanz, et al., 1997). Bandura (1986)
proposed that self-efficacy is the most important prerequisite for behavioral change, because it
affects how much effort is invested in a given task and what level of performance is gained.
Efficacy expectations are influenced or learned from four major sources (Strecher, et al., 1986).
The first influence is performance accomplishments, which refers to learning through personal
experience and is reinforced by success of accomplishment. The second is learning that occurs
through observation of events and/or other people. The third is verbal persuasion, which pertains
to asking people to persevere in their behavior change efforts. The fourth influence is one’s
physiological state, whereby being nervous, for example, can hamper one’s success in
performing a behavior. Beliefs in personal efficacy can be changed by these different modes of
influence. Enhancement of efficacy beliefs leads to motivation for behavioral efforts (Maibach
& Murphy, 1995).

Social support, another SCT construct, is defined as aid and assistance exchanged
through social relationships and interpersonal transactions. Social support is always intended to
be helpful and not an intentional negative interaction. Cassel (1976) suggests that social support
serves as a key psychosocial “protective” factor that reduces individuals’ vulnerability to the
potentially harmful effects of stress on health.

Psychosocial factors and fruit and vegetable consumption

Dittus (1995) conducted a mail survey to examine attitudes toward nutrition and reported
fruit and vegetable intake among randomly sampled Washington state residents (n=1069).
Results of the regression analysis indicated that 16% of the variance in fruit and vegetable intake
was accounted for by attitudinal barrier variables (cost/time to prepare fruits and vegetables,
difficulty in changing eating habits, self-efficacy issues, and taste of fruits and vegetables).
Results suggested a relationship between attitudes about barriers to fruit and vegetable intake and
nutrition behaviors. A limitation of the study was that the sample was heavily skewed toward
upper income and education levels. The author concluded that since barriers are related to
nutrition behavior more strongly than benefits, nutrition education efforts should also concentrate
on identifying the barriers to behavior change for target audiences. Despite a very small sample

11
size of low-income respondents, Dittus notes that barriers are particularly important components
to address in nutrition education interventions for men and low-income/education audiences.

The baseline survey for the 5 A Day for Better Health Program included an analysis of
psychosocial factors associated with fruit and vegetable consumption (Krebs-Smith, et al. 1995).
Multiple regression analyses were performed to measure the net effect of the various
psychosocial factors on the intake of fruits and vegetables. Descriptive analyses were done to
determine the percentage of adults whose responses fell in various categories. Results included
that only 8% of adults thought five or more servings per day were needed for good health
(message awareness/knowledge). Most adults agreed strongly that they liked the taste of fruits
(82%) and vegetables (71%). About a third of adults felt strongly that family friends encouraged
them to eat fruits and vegetables. Of all the factors studied, the most important in determining
fruit and vegetable intake were the number of servings they thought they should have in a day,
whether they liked the taste, and whether they had been in the habit of eating many fruits and
vegetables since childhood. These few factors accounted for 15% more of the variation in fruit
and vegetable consumption than did demographic variables alone. Persons with less than a high
school education and those with incomes below 130% of the poverty level were less likely than
other groups to think that five or more servings per day of fruits and vegetables were needed.
The authors concluded that men, young adults, and persons with low levels of income and
education should be targeted in promoting fruits and vegetables.

Laforge and colleagues (1994) looked at the psychosocial factors influencing

consumption in a group of adults who are low consumers of fruit and vegetables (<2 servings
daily). In this study, males were twice as likely to be aware of the need to consume more fruits
and vegetables, as were women. Years of education was a strong predictor of those unaware of
the need consume more fruits and vegetables, with those with a high school education or less
being three times more likely to be unaware (Laforge, et al., 1994). Those unaware of the health
message reported lower fruit and vegetable consumption than those aware. Those unaware need
to be targeted with cognitive and experiential strategies when tailoring educational messages to
this group.

Two worksite research interventions assessed psychosocial correlates of fruit and

vegetable consumption. At the baseline of the Working Well Trial, beliefs/knowledge, perceived
benefits, and motivation were strong predictors of current diet (fat, fiber, fruits, and vegetables)
and intention to change (Kristal, et al., 1995). Also, at the baseline of the Next Step Trial,
beliefs/knowledge, perceived benefits, and motivation showed a stronger association with current
diet than did perceived barriers, norms or social support (Glantz, et al.,1998). This suggests that
attitudes, knowledge, and motivations may be more important than social factors, in the worksite
setting. Self-efficacy was not assessed in these trials. The intervention for the Next Step Trial
significantly increased attitude, knowledge, and motivation scales, and changes in these
correlates, along with changes in social factors were associated with significant dietary change
(Kristal, et al., 2000). In a side-by-side analyses of psychosocial factors from 7 large scale
community-based interventions (3 of which were worksite) to increase fruit and vegetable
consumption in adults, self-efficacy scores, fruit and vegetable consumption, and knowledge of
the 5 A Day message were positively associated with higher levels of stage of change, i.e. being

12
in the act of trying eat more fruits and vegetables or maintaining a high level of intake
(Campbell, et al., 1999).

A random sample telephone survey was completed in the Netherlands to analyze the
importance of psychosocial determinants of fruit and vegetable consumption. Self-efficacy and
attitudes (perceived benefits of performing a health behavior) were significantly associated with
consumption of fruit and vegetables (Brug, et al., 1995). It was concluded from this research that
nutrition interventions should focus on enhancing perceived benefits of performing a behavior
and self-efficacy expectations.

The relationship of cancer prevention-related nutrition knowledge, beliefs, and attitudes

to dietary behavior were assessed in a nationwide survey of over 12,000 adults (Harnack, et al.,
1997). After adjusting for covariates in this study, knowledge and awareness of key health
messages were predictive of dietary behavior (fat, fiber, fruit, and vegetables). In addition, it
was found that knowledge and awareness have less influence on dietary behavior for those with
lower educational levels, than for those with higher levels of education.

Qualitative research has been conducted with several groups of low-income women
receiving food assistance benefits. Reicks, et al. (1994) conducted focus group discussions with
low-income mothers participating in the USDA Expanded Food and Nutrition Education
Program (EFNEP), and found that women perceived fresh produce as expensive, and they
limited their purchases to sale items, and limited purchases due to lack of storage space.
Treiman, et al. (1996) conducted focus group discussions with low-income, primarily African
American mothers enrolled in the Supplemental Food Program for Women, Infants, and Children
(WIC). These mothers reported that a key motivator of adopting healthy eating patterns (more
fruits and vegetables) was being a good role model their children. Barriers included availability
of fruits and vegetables, time and effort to prepare, and preference for other foods. Some of
these same findings occurred in focus group discussions with Food-Stamp-Eligible Hispanic
mothers (Hampl & Sass, 2001). A recurrent theme with this group was that children greatly
affected what the women bought at the grocery store. Barriers included cost, and unfamiliarity
with the wide variety of fruits and vegetables in the U.S.

Another group described the multiple perspectives on nutrition education needs of low-
income Hispanics. (Palmeri, et al., 1998). Nine focus groups were conducted with low-income
Hispanics (n=65). Low-income Hispanics' primary concerns were their children's nutritional
habits and ways to prepare quick, nutritious meals and snacks. Relative to nutrition education
efforts, major barriers included lack of resources (time and money), family customs and
preferences, and confusion over conflicting nutrition messages. Consideration was given to the
idea of the use of abuelas (Hispanic grandmothers) as lay health advisors or paraprofessionals,
due to the great respect in the Hispanic community for elders. While this was ethnically a real
plus, the ongoing need for support and education of the abuelas may be prohibitive. The focus
group participants emphasized a desire for their families to adopt healthy eating habits, which
were described as consuming more fruits, vegetables and milk, and less red meat.

13
Nutrition Education interventions in low-income populations
Most of the published literature in health promotion strategies for the low-income
population describes actual programs and evaluation of such programs to reach the intended
audience. Few discuss the formative research, or the determination of barriers and underlying
constructs to behavior change in underserved populations. Several articles allude to the proposed
barriers to behavior change in low-income participants, but, other than focus group research,
little has been published on attempts to characterize salient mediating issues of dietary behavior
change from research in low-income individuals.

Several papers describe dietary and other health promotion interventions in low-income
populations. Nutrition education for cardiovascular disease prevention among low-income
individuals was addressed through the use of a physician-based model called the Food for Heart
Program (Ammerman, et al., 1992). Strategies used included low literacy education materials
emphasizing regional eating patterns, linking the dietary assessment with educational materials,
and organizational elements to facilitate physician counseling. Barriers that were addressed in
this low-income population included culturally determined eating practices, competing demands
and pressures, low literacy issues, and perceived or real costs of the recommendations. It was
noted that access to medical care is limited in low-income populations, physicians have little
time and are often lack training in nutrition; therefore physicians may not be the best purveyors
of nutrition education. Lay health providers with similar socio-cultural backgrounds may be
more appropriate to be able to reach, educate, and motivate the poor to make health promoting
behavior changes.

Calderon & Goerence (1998) conducted a short survey to determine Food Center
recipient's (n=207) perception of the quality of their own and their family's diet, and to assess
nutrition knowledge, in order to plan an appropriate nutrition education program.
Approximately 1/3 of the respondents rated their diet as good or excellent while the remaining
2/3 rated the quality of their diet as fair, poor or very poor. Their nutrition knowledge was
deemed very poor, with 70% of the respondents answering three or less of the eight questions
correctly. While this short survey questionnaire is quite rudimentary, it does point to the fact that
lack of nutrition knowledge is a definite problem in those seeking emergency food assistance.
Calderon notes in her discussion "those on a limited income need to discern what foods are most
cost effective in providing the necessary nutrients for health and well-being" (pg. 463).

Anderson, and colleagues (2001) evaluated the Michigan Farmer’s Market Nutrition
Program to determine the effect of vouchers and an educational program on fruit and vegetable
consumption behavior. The findings of this study showed that vouchers (price supports) directly
affected consumption behavior in a positive way, but did not affect attitudes directly or
indirectly. In contrast, education indirectly affected consumption behavior by positively
affecting attitudinal variables. The authors suggest that education, coupled with vouchers, can
achieve the maximum effects on consumption.

A 5 A Day educational demonstration project was done with low-income families

(Weaver, 1999). The results showed that cooking events were significantly more effective in
increasing awareness and knowledge, than was an advertising campaign. The authors felt that
the cooking events called attention to the message through colorful displays, the smell and taste

14
of food samples, and the opportunity to interact informally with outreach workers. These
participatory strategies utilize an andragogical approach and tie in with adult education theory.

Both the Black Churches United for Better Health Project and the Maryland WIC 5 A
Day research projects used a lay health advisor model to deliver the educational intervention.
The researchers in these projects point out that since the peer educators are selected from the
participant pool and share common language, norms, and cultural values, they provide social
support to the participants. In addition, they can make valuable contributions to program design,
and deliver the interventions effectively (Anliker, et al., 1999; Campbell, et al., 1999).

Summary of Literature Review

Fruit and vegetable consumption data in this country is consistently below

recommendations. Fruit and vegetable intake is particularly low in the poor population. Despite
a nationwide 5 A Day program to increase fruit and vegetable consumption, the progress is slow
with modest positive increases, and the low income population continues to lag behind the rest of
the country in consumption.

The scientific evidence that strongly links increased fruit and vegetable consumption with
decreases in the risk of the major chronic diseases in the U.S. is staggering. As the science base
increases and we begin to understand the biological mechanisms of fruits and vegetables that
fight disease, it is imperative that we speed up the rate at which healthy dietary behavior change
occurs.

Nutrition education philosophy in adults not only includes andragogical approaches to

education, but is becoming more theory-driven. The social cognitive theory seems to be the
most commonly used educational and behavioral change theory in the nutrition profession. The
public health approach of encouraging entire populations to adopt healthy eating behaviors,
encompasses the inclusion of a variety of educational, behavioral, and policy interventions
targeted to specific groups.

Although there have been attempts to intervene in low-income groups, we need to know
more about the barriers and promoters of healthy eating habits in this population. Health
educators continue to grapple with planning effective, culturally appropriate educational
interventions in poor populations. Some policy mandated nutrition programs are in place for
selected populations, but many low-income persons are without safety nets when it comes to
encouraging healthy food consumption.

Survey and intervention research has been conducted in the effort to understand the
barriers and promoters of fruit and vegetable consumption. There is a dearth of information
examining psychosocial determinants of fruit and vegetable consumption by income group.
From the research studies looking at determinants or mediators of fruit and vegetables dietary
change, a few common themes emerge. Self-efficacy, awareness/knowledge of key health
messages, social factors, taste of fruits and vegetables, and barriers (costs, availability, and
accessibility) are the most often measured determinants of dietary behavior. Low income and
low educational level is associated with lower fruit and vegetable consumption. Likewise, there

15
is an inadequate body of knowledge from research efforts to help guide health education in the
low-income population.

For these reasons, it is important to better understand the real and perceived barriers, and
the factors that promote fruit and vegetable consumption.

16
 CHAPTER THREE
METHOD

The material in this chapter is arranged in the following way: research goals; survey
overview; sampling plan, and analysis procedures.

Research Goals

To effectively intervene in populations, it is important to understand the barriers and

promoters to eating healthfully. Effective nutrition education programs can help individuals to
improve their health and prevent diet-related chronic diseases. This research proposed to learn
more about the mediating variables to healthy eating. The information gleaned from this survey
may help practitioners to plan effective interventions and to understand the reasons why the low-
income population lags behind in fruit and vegetable consumption.

The specific research objectives are:

1. Via a secondary analysis of the 5 A Day for Better Health Program Follow-up survey,
characterize the barriers of cost, availability, lack of knowledge, and psychosocial
issues to consuming fruits and vegetables.
2. Identify the underlying latent structures of consumption, barriers to consumption, and
other constructs present in the survey data.
3. Characterize differences in consumption and related factors that are explained by
income level.
4. Based upon the underlying latent structures, determine a conceptual model for
influencing fruit and vegetable consumption in Americans.

Survey Overview

The National Cancer Institute (NCI) collected data from a representative sample of
American adults from July 9, 1997 through September 21, 1997 for the National 5 A Day for
Better Health follow-up Survey. The study was designed to measure six-year trends in fruit and
vegetable intakes, as well as in knowledge, attitudes and beliefs about diet and nutrition.
Approximately 2,600 interviews were completed using Computer Assisted Telephone
Interviewing (CATI) technology. 348 interviews were completed in the low-income category
(<130% of Poverty Index), 796 interviews in the 130-<300% Poverty Index Ratio (PIR) range,
and 1101 interviews were completed in the >300% PIR range. Over 300 respondents refused to
give their income information.

The study consisted of a Screener Survey to determine household eligibility and to

sample a respondent for the Extended Survey; and the Extended Survey, which gathered detailed
information on health status, fruit and vegetable consumption, and habits and attitudes towards
nutrition and health. Both components were conducted in English.

The Screener Survey was a random-digit-dial (RDD) survey of households in the United
States, conducted using CATI technology. A contractor called a nationwide sample of 13,521
telephone numbers during the survey. All residential units occupied by related individuals, or up

17
to six unrelated roommates, were considered eligible for participation in this study. When an
eligible household was contacted, the Screener questionnaire was administered to an adult (18
years of older) member of that household. The questionnaire asked the adult household member
to identify the household member with the most recent birthday and to answer questions about
that person's gender, age, and ethnicity.

The sample was composed of low and high minority strata. The CATI system selected
all respondents identified in the low minority stratum without regard to race. An additional
question was asked of each household in the high minority stratum to determine if it was a
minority household. If so, the respondent was selected with certainty. If not, a subsampling
algorithm determined whether to continue with the interview or resolve the household as
ineligible.

All interviewers assigned to the survey participated in training sessions and completed a
minimum of 12 hours of formal project-specific training. Quality assurance measures included
interviewer monitoring, callback procedures, and ongoing examination of the data during the
data collection phase. 3, 528 screener questionnaires were completed. Of these, 2,605 extended
interviews were completed. The overall response rate for screener and extended interview was
44.5%. Of the 2,605 completed interviews, 65.5% were completed with White respondents,
19.7% with Black respondents, and 10.3% were completed with Hispanic respondents.

Questionnaire Development

Since this was a follow-up survey to the 1991 baseline survey, the questionnaire was very
similar to the baseline questionnaire. Many of the questions were the same, with a few new
questions added. In order to comply with governmental clearance, the Food Frequency portion
of the questionnaire was abbreviated from 33 items to a 7-item frequency questionnaire. In an
open-ended response format, individuals reported the number of times per day, week, month, or
year they consumed each fruit and vegetable item. No questions were asked about portion sizes.
The seven-item fruit and vegetable frequency questionnaire and a similar instrument have been
validated in a number of U.S. populations, yielding correlations in the range of 0.47 to 0.56 with
longer food frequency questionnaires, multiple 24-hour recalls, and 3-day food records
(Thompson, et al., 2000; Campbell, et al., 1999; Hunt, et al., 1998; Serdula, et al., 1993).

Most of the psychosocial questions included a scale of 0-10, with a response of 0 =

“don’t agree at all” to 10 = “ strongly agree”. These questions were not assessed for reliability or
validity prior to administration of the survey.

Sampling Design

The sample design was a random digit dialing (RDD) telephone sample. A random
sample of telephone numbers was drawn from all working banks with all exchanges in all area
codes covering the United States. Banks are sets of one hundred telephone numbers with the
same area code and five-digit prefix. Exchanges were assigned to one of two strata: a high
minority stratum and a low minority stratum. An exchange was assigned to the high minority
stratum if 15 percent or more of the telephone households associated with the exchange were

18
black or Hispanic, according to the Genesys estimates. Otherwise the exchange was assigned to
the low minority stratum. A random sample of 12,000 telephone numbers was drawn from the
high minority stratum, and a random sample of 4,150 telephone was drawn from the low
minority stratum. The original sample included 8,000 telephone numbers respectively in the
high minority and low minority strata, but a shortfall in the number of completed interviews by
minority persons from expected yields required an increase in the relative sample size of high
minority stratum telephone numbers. The telephone numbers were selected using a stratified
sampling design: the working banks in the High and Low Minority Stratum were ordered
according to geography, metropolitan or non-metropolitan status, and by area code and five digit
prefix within the geographic regions. This ordered set of telephone numbers was the frame for
each of the two strata. The frame was partitioned into equally sized substrata using this ordering
of working banks. An independent equal probability sample of one telephone number was then
drawn from each of these substrata.

Analysis Procedure

With the Statistical Package for the Social Sciences (SPSS for Windows, version 9.0;
SPSS, Chicago, Illinois), the following analyses were performed: Exploratory Factor Analysis
with varimax rotation, one-way analysis of variance (ANOVA) with LSD Post Hoc Test, Chi
Square test, and multiple regression analysis. Prior to conducting the analysis, variables were
examined for normality. Although several variables were slightly skewed, servings of fruits and
vegetables and poverty ratio variables were severely skewed. These variables were transformed
using the Log 10 transformation operation in SPSS, version 9.0, resulting in a normally
distributed variable for fruit and vegetable consumption. The income variable resulted in a
bimodal distribution suggesting there are at least two distinct groups differing by income.

Exploratory factor analysis was used to initially analyze the data. Each measure in the
data set is considered to an observed indicator of one or more underlying factors. A single factor
is hypothesized to be expressed in a range of measures, and at times an individual item may load
on more than one factor. The factor analytical procedure (including varimax rotation and factor
scoring) used in SPSS was one of the analytic tools for research question 1, “What are the
relevant constructs in adult’s fruit and vegetable consumption that are being measured in the
1997 nationally representative survey for the National 5 A Day for Better Health Program?”

Research question 2, “What differences in consumption and related factors are explained
by income level?” was partially answered by one way analysis of variance (ANOVA) with an
LSD post hoc test which showed differences in awareness, total fruit and vegetable consumption,
mean total composite values of self efficacy, weak intent, and cost/quality scales between low
income and high income respondents. The question was further answered using multiple
regression modeling by income categories showing the percent of variance in total fruit and
vegetable consumption explained by demographic covariates, social cognitive factors, and
perceived barrier factors.

Research question 3, “What profile of demographic characteristics differentiates

individuals who a) are aware of the 5 A Day program, b) exhibit high level of self-efficacy, c)
have high level of perceived personal benefits of good health-related behaviors, and d) eat high

19
levels of fruits and vegetables?” was answered by one-way analysis of variance and chi square
analyses. This analysis simply provides a demographic profile of those with specific measures
of the behaviors listed. The profile is helpful in determining the variables for the hierarchical
regression analysis.

Multiple linear regression was done to determine the variance in fruit and vegetable
consumption as explained by psychosocial and demographic variables, by income group, eg.
Low-income, high income, and all income groups. This analysis contributed to research
questions 1 and 3.

Hierarchical linear regression was the final analysis to be completed and adds more information
to answer research question 1 on relevant constructs in adult’s fruit and vegetable consumption.
Psychosocial variables that were found to be significant in the ANOVA, Chi Square, and
multiple regression analysis and all the demographic variables were stepped into the hierarchical
regression equation, resulting in nine regression models. These promising variables were
included in several iterations of the analysis. If they were consistently not significant in the
regression, they were removed. For example, marital status, race/ethnicity, and cost/availability
variables were removed. In addition, for research question 3, the proportion of variance
contributed by the several latent factors was used to determine where further developmental
work for future surveys should be done. The factors discovered in this research will need to be
strengthened and new issues identified, both of which will lead to specific suggestions for future
research. Figure 1 shows the overall analysis scheme.

20
Figure 1. Analysis Scheme

Create Variables Variable Development & Selection Analytic Procedure

ANOVA
psychosocial
factors,
Low vs hi income

Multiple
Regression Hierarchical
Variance explained
Regression
Exploratory
Factor Most Relevant
Analysis predictors
ANOVA,
Chi Sq
Demographic
profile

Select Factor
Variables

21
 CHAPTER FOUR
RESULTS

The survey instrument has several methodological shortcomings and therefore the
findings need to be interpreted with caution. First, the fruit and vegetable intake findings are
based upon self-reported information of a short fruit and vegetable screener. Second, although
the survey instrument would have benefited from updating and from cognitive testing,
modification of individual questions was minimized in the Follow-up Survey to maintain
comparability to the Baseline Survey. Third, the psychosocial questions were not assessed for
reliability and validity prior to implementing the survey. Fourth, despite major efforts to
maximize the response rate, there was a less than optimal response rate of 44.5%. Even though
the Subar, et al. (1995) of the 1991 5 A Day Baseline Survey completed a rigorous nonresponse
survey and found that those nonresponsive to the survey did not differ significantly in fruit and
vegetable consumption or in responses to the psychosocial questions from those who completed
the survey, there remains a question of how representative of the U.S. population are the 1997
results. Nonresponse problems are common in national RDD surveys. The increased use of
answering machines and the increased magnitude of telephone solicitation have all increased the
survey burden in RDD surveys. In 1996, Massey, et al conducted a study of response rates in
large national RDD surveys, and found that 1/3 of the studies had response rates below 60%,
even though almost one-half of the surveys used methods that overestimated the response rate.
Massey, et al. (1997) concluded that considerable effort is required to obtain a response rate over
60%, that response rates have declined slightly over the past 10 years, calculation of response
rates is variable across surveys, and nonresponse in RDD surveys is an ongoing problem.

Analysis of Demographic Data

Table 1 shows the distribution of the sample (N and raw %) according to various
demographic characteristics by income group. Of the 2,605 adult participants, 14% are in the
low-income category, 31% are in the middle-income category, and 43% are in the higher income
category, with 12% not reporting income levels. These proportions by income category are
similar to the 1997 U.S. census.

The proportion of Whites increases in the higher income categories, while the proportion
of African Americans and Latinos decrease as income categories increase. A large percentage
(76%) of those with greater than a high school education are in the higher income category. The
proportion of smokers is highest (27%) in the low-income category. The greatest percentage of
those in the youngest (18-34 year old) age group and in the oldest (>65 year old) age group is in
the low-income category, while those aged 35-65 have the highest percentage in the higher
income group.

22
Table 1. Demographic Characteristics by Income Category

Low Incomeb Middle Higher All Income

Incomec Incomed
Demographics N % N % N % N %

Gender Males 115 32% 332 41% 521 47% 1068 41%
Females 249 68% 477 59% 596 53% 1536 59%

Race/Ethnicity Whites 170 47% 506 63% 800 72% 1700 65%
African American 114 31% 173 21% 173 16% 510 20%
Latinos 66 18% 86 11% 92 8% 267 10%
Other 34 4% 44 4% 105 4%

Smoking Status Non-Smoker 260 71% 607 75% 923 83% 2043 78%
Smoker 104 29% 203 25% 194 17% 562 22%

Age Category 18-34 148 41% 280 35% 363 33% 857 34%
35-49 95 26% 243 30% 451 40% 838 33%
50-64 40 11% 129 16% 215 19% 428 17%
>=65 79 22% 152 19% 79 7% 383 15%

Education <12 104 29% 92 11% 37 3% 285 11%

12 159 44% 297 37% 221 20% 778 30%
>12 97 27% 401 50% 843 76% 1497 57%

Martial Status Married 117 32% 395 49% 654 59% 1326 51%
Div/Sep/Widowed 151 42% 242 30% 218 20% 697 27%
Never Married 96 26% 172 21% 244 22% 574 22%
Total Population 364 810 1117 2605
a
314 did not report income
b
Low income defined as <130% Poverty Index Ratio (food stamp eligibility criteria); takes into account
income/size of household
c
Medium income defined as 130-300% Poverty Index Ratio
d
Higher income defined as >300% Poverty Index Ratio

23
 Factor Analysis

Exploratory factor analysis was done to address the first research question: What are the
relevant constructs in adult’s fruit and vegetable consumption, that are being measured by the
most recent survey for the National 5 A Day for Better Health program? Exploratory factor
analysis was completed with a grouping of variables addressing barriers to consuming fruits and
vegetables, and with a grouping of psychosocial variables addressing social/cognitive factors
involved in fruit and vegetable consumption.

Seventeen items of a social/cognitive nature were included in a factor analysis completed

on the entire sample population (Table 2), on the low-income group (Table 3), middle-income
group (Table 4), and higher income group (Table 5). A factor structure was identified that was
stable across income groups. Of these seventeen items, five loaded on factor one, and are
labeled as a self-efficacy factor since all five variables address the respondent’s confidence in
being able to do a particular behavior. Four items loaded on factor two and address friends and
family support issues, and therefore labeled “social support”. Three items addressing health
benefits of consuming vegetables and fruit loaded on factor three and were labeled as the
“perceived benefit” factor. The factor loading scores and the variance explained were very
similar across income levels. A “self efficacy” factor, a “perceived benefit” factor, and a “social
support” factor were saved for each of the three income groups and the entire sample.

Five additional variable items were eliminated from the factor analysis due to cross loading or
not loading on any particular factor. Items eliminated included questions regarding eating fruits
and vegetables by habit, fruits and vegetables make you feel better, how often stores display
nutrition information, and two questions about knowledge/awareness of the 5 A Day message
and program.

24
Table 2. Social/Cognitive Factors Related to Fruit and Vegetable Consumption: All Income Levels
(n = 1851)a

Factor loadings
Questionnaire item Self-efficacy Social support Perceived benefits
Can respondent eat 3 .789 .105 .147
servings of fruits and
vegetables daily
Can respondent eat 5 .780 .184 .119
servings of fruits and
vegetables daily
Can respondent eat .773 .089 .109
fruits and vegetables for
snacks
Can respondent eat .723 .144 .060
fruits and vegetables
outside the home
Can respondent drink .636 .033 .082
fruit juice daily
Friends urge respondent .035 .778 .095
to eat fruits and
vegetables
Friends eat lots of fruits .112 .714 .078
and vegetables
Family urges respondent .053 .660 .047
to eat fruits and
vegetables
Family eats lots of fruits .278 .644 .016
and vegetables
Fruits and vegetables .110 .092 .871
prevent heart disease
Fruits and vegetables .119 .129 .761
prevent cancer
Fruits and vegetables .141 .039 .748
help lose weight

Variance explainedb 24.1% 17.2% 16.4%

Chronbach Alpha 0.79 0.62 0.73

a
Extraction method: principal component analysis (forced into three factors). Rotation method: Varimax
with Kaiser normalization. In the rotated component matrix, the rotation converged in five iterations.
b
Total variance Explained = 57.7%.

25
Table 3. Social/Cognitive Factors Related to Fruit and Vegetable Consumption: Low Income (n =
231)a

Factor loadingsb
Questionnaire item Self-efficacy Social support Perceived benefits
Can respondent eat 5 .792 .169 .085
servings of fruits and
vegetables daily
Can respondent eat .782 .132 .128
fruits and vegetables for
snacks
Can respondent eat 3 .731 .070 .339
servings of fruits and
vegetables daily
Can respondent drink .721 .093 .053
fruit juice daily
Can respondent eat .650 .280 .095
fruits and vegetables
outside the home
Friends urge respondent .065 .768 .139
to eat fruits and
vegetables
Friends eat lots of fruits .120 .717 .187
and vegetables
Family urges respondent .054 .695 .057
to eat fruits and
vegetables
Family eats lots of fruits .353 .659 .102
and vegetables
Fruits and vegetables .052 .096 .853
prevent heart disease
Fruits and vegetables .236 .024 .681
prevent cancer
Fruits and vegetables .102 .022 .651
help lose weight

Variance explainedc 24.4% 18.0% 15.3%

a
Only low-income cases (<130% of the Poverty Index Ratio; takes into account the income and size of
household; <130% poverty is food stamp eligibility criteria) were used in the analysis.
b
Extraction method: principal component analysis (forced into three factors). Rotation method: Varimax
with Kaiser normalization. In the rotated component matrix, the rotation converged in six iterations.
c
Total variance = 57.7%.

26
Table 4. Social/Cognitive Factors Related to Fruit and Vegetable Consumption: Medium Income
(n = 578)a

Factor loadingsb
Questionnaire item Self-efficacy Social support Perceived benefits
Can respondent eat 3 .801 .160 .138
servings of fruits and
vegetables daily
Can respondent eat 5 .793 .197 .096
servings of fruits and
vegetables daily
Can respondent eat .782 .078 .117
fruits and vegetables for
snacks
Can respondent eat .721 .114 .069
fruits and vegetables
outside the home
Can respondent drink .678 .089 .104
fruit juice daily
Friends urge respondent .050 .752 .047
to eat fruits and
vegetables
Friends eat lots of fruits .073 .714 .048
and vegetables
Family urges respondent .050 .675 .086
to eat fruits and
vegetables
Family eats lots of fruits .260 .631 .025
and vegetables
Fruits and vegetables .158 .052 .861
prevent heart disease
Fruits and vegetables .142 .014 .777
help lose weight
Fruits and vegetables .081 .117 .755
prevent cancer

Total variance 24.9% 16.9% 16.6%

explainedc

a
Only medium-income cases (130–300% of the Poverty Index Ratio; takes into account the income and
size of household) were used in the analysis.
b
Extraction method: principal component analysis (forced into three factors). Rotation method: Varimax
with Kaiser normalization. In the rotated component matrix, the rotation converged in six iterations.
c
Total variance = 58.4%.

27
Table 5. Social/Cognitive Factors Related to Fruit and Vegetable Consumption: Higher Income (n
= 869)a

Factor loadingsb
Questionnaire item Self-efficacy Social support Perceived benefits
Can respondent eat 3 .784 .096 .073
servings of fruits and
vegetables daily
Can respondent eat .776 .071 .100
fruits and vegetables for
snacks
Can respondent eat 5 .772 .210 .136
servings of fruits and
vegetables daily
Can respondent eat .721 .128 .061
fruits and vegetables
outside the home
Can respondent drink .581 .055 .061
fruit juice daily
Friends urge respondent .024 .796 .108
to eat fruits and
vegetables
Friends eat lots of fruits .110 .730 .099
and vegetables
Family urges respondent .081 .656 .037
to eat fruits and
vegetables
Family eats lots of fruits .317 .621 .086
and vegetables
Fruits and vegetables .077 .129 .866
prevent heart disease
Fruits and vegetables .102 .164 .775
prevent cancer
Fruits and vegetables .133 .077 .762
help lose weight

Variance explainedc 23.5% 17.5% 16.7%

a
Only higher income cases (>300% of the Poverty Index Ratio; takes into account the income and size of
household) were used in the analysis.
b
Extraction method: principal component analysis (forced into three factors). Rotation method: Varimax
with Kaiser normalization. In the rotated component matrix, the rotation converged in five iterations.
c
Total variance = 57.8%.

28
 Ten barrier items were also included in a factor analysis completed on the entire sample
(Table 6), low-income sample (Table 7), middle-income sample (Table 8), and higher income
sample (Table 9). A factor structure was identified that was stable across income groups. Of
the ten items, four items loaded on factor 1, which is titled a “weak intent” factor because these
variables address issues of weak will power and effort. Three items loaded on construct 2, and
address expense and quality of fruits and vegetables. Three additional items were removed due
to not loading on any factor or cross loading on more than one factor. The items not included
were questions regarding confusing advice about healthy eating, needing preparation
information, and how often a person eats away from home.

A “weak intent” factor and “expense/cost” factor variable was saved by income groups.
Over all the income levels, the “fruits and vegetables not readily available” items did vacillate a
bit in the loading pattern. In the low-income group, this item firmly loaded with the “weak
intent” factor, and in the high-income group, this item loaded more strongly with “weak intent”.
Therefore, “fruits and vegetables not readily available” was important to include within the
“weak intent” factor. The factors were saved by income group to create factor variables specific
to those income levels. The income specific factor variables were used in the regression
equations.

Cronbach alpha (range of 0.62-0.79) was computed to evaluate the internal consistency of
the self-efficacy, weak intent, perceived benefits, cost/quality, and social support scales, and are
included in Table 2 and 6. The five factors (self-efficacy, social support, perceived benefits,
weak intent, and expense/cost) derived from the factor analysis represent underlying
psychosocial constructs related to fruit and vegetable consumption behavior and will be used in
further analysis to determine how they may predict these behaviors.

29
Table 6. Perceived Barriers to Eating Fruits and Vegetables: All Income Levels (n = 2389)

Factor loadingsa
Questionnaire item Weak intent Cost/quality
Preparing fruits and .886 .143
vegetables takes too much
planning
Preparing fruits and .873 .133
vegetables takes too much
time
Daily consumption takes .659 .210
willpower
Fresh fruits and vegetables are .125 .820
too expensive
Frozen and canned fruits and .078 .818
vegetables are too expensive
Fruits and vegetables are poor .289 .604
quality
Fruits and vegetables are not .443 .461
readily available

Variance explainedb 32.6% 28.6%

Chronbach Alpha 0.73 0.63
a
Extraction method: principal component analysis (forced into two factors). Rotation method:
Varimax with Kaiser normalization. In the rotated component matrix, the rotation converged in
three iterations.
b
Total variance = 61.2%.

30
Table 7. Perceived Barriers to Eating Fruits and Vegetables: Low Income (n = 338)a

Factor loadingsb
Questionnaire item Weak intent Cost/quality
Preparing fruits and .860 .137
vegetables takes too much
planning
Preparing fruits and .831 .102
vegetables takes too much
time
Daily consumption takes .674 .247
willpower
Fruits and vegetables are not .565 .300
readily available
Fresh fruits and vegetables are .157 .834
too expensive
Frozen and canned fruits and .139 .826
vegetables are too expensive
Fruits and vegetables are poor .347 .462
quality

Variance explainedc 33.8% 25.3%

a
Only low-income cases (<130% of the Poverty Index Ratio; takes into account the income and
size of household;<130% poverty is food stamp eligibility criteria) were used in the analysis.
b
Extraction method: principal component analysis (forced into two factors). Rotation method:
Varimax with Kaiser normalization. In the rotated component matrix, the rotation converged in
three iterations.
c
Total variance = 59.1%.

31
Table 8. Perceived Barriers to Eating Fruits and Vegetables: Medium Income (n = 748)a

Factor loadingsb
Questionnaire item Weak intent Cost/quality
Preparing fruits and .888 .196
vegetables takes too much
planning
Preparing fruits and .873 .181
vegetables takes too much
time
Daily consumption takes .705 .233
willpower
Fresh fruits and vegetables are .115 .822
too expensive
Frozen and canned fruits and .117 .788
vegetables are too expensive
Fruits and vegetables are poor .369 .571
quality
Fruits and vegetables are not .343 .520
readily available

Variance explainedc 33.3% 28.8%

a
Only medium-income cases (130–300% of the Poverty Index Ratio; takes into account the
income and size of household) were used in the analysis.
b
Extraction method: principal component analysis (forced into two factors). Rotation method:
Varimax with Kaiser normalization. In the rotated component matrix, the rotation converged in
three iterations.
c
Total variance = 62.1%.

32
Table 9. Perceived Barriers to Eating Fruits and Vegetables: Higher Income (n = 1028)a

Factor loadingsb
Questionnaire item Weak intent Cost/quality
Preparing fruits and .901 .105
vegetables takes too much
planning
Preparing fruits and .891 .085
vegetables takes too much
time
Daily consumption takes .600 .149
willpower
Fruits and vegetables are not .447 .415
readily available
Fresh fruits and vegetables are .010 .839
too expensive
Frozen and canned fruits and .127 .807
vegetables are too expensive
Fruits and vegetables are poor .247 .657
quality

Variance explainedc 32.0% 28.6%

a
Only higher income cases (>300% of Poverty Index Ratio; takes into account the income and
size of household) were used in the analysis.
b
Extraction method: principal component analysis (forced into two factors). Rotation method:
Varimax with Kaiser normalization. In the rotated component matrix, the rotation converged in
three iterations.
c
Total variance = 60.6%.

33
 Psychosocial, Educational, and Behavioral Response Differences by Income

The five factors derived from factor analysis, plus the two knowledge/awareness
variables are examined by income group in Table 10 and 11. This analysis will partially answer
the second research question: What differences in consumption and related factors are explained
by income level? In general, care was taken in the interpretation of the significance (p value) of
differences found in the analysis to address how meaningful the differences are from a public
health perspective. Given the large sample size, even very small differences between income
levels were found to be significant or highly significant, but may not be meaningful.

One way analysis of variance (ANOVA) with an LSD post hoc test showed there are
significant differences in mean total composite values of self efficacy (p<.05), weak intent
(p<.001), and cost/quality (p<.001) scales between low income and high income respondents.
Higher income respondents have significantly higher scores in self-efficacy and lower scores
regarding perceptions of cost/quality barriers. Using a likert scale (1-10), the efficacy scores
were relatively low for all incomes, with no differences between income levels, for the very
important question, “How sure are you that you can eat at least 5 servings of fruits and
vegetables each day?” No significant differences were found between income groups in
composite social support factors or composite perceived benefits scales.

LSD post hoc tests revealed there are significant differences between low and higher
income respondents in two of the individual self-efficacy items and all of the cost/quality items.
Low-income respondents were significantly more likely to agree that vegetables and fruit
(canned, frozen, or fresh) were too expensive and that vegetables and fruit are of poor quality
where they shop. There was also a significant difference between income categories in the mean
weak intent composite score, with low-income respondents being significantly more likely to
report that vegetables and fruit are not readily available and that daily consumption takes will
power. There are significant, but small differences between low and higher income respondents
regarding knowledge of the 5 a day message and awareness of the 5 a Day Program, with high-
income respondents have higher knowledge/awareness than low income respondents. Even
though there are differences in awareness by income level, it is important to note that both low
and higher income respondents have overall low awareness of the 5 A Day message.

34
Table 10. Psychosocial Response Differences between Low Incomea Respondents and Higher
Incomeb Respondents: Social/Cognitive Factors

FACTORS Mean Score (SD)

higher
Scales F Low incomea
incomeb P
Value n=310-364
n=993-1116 valuec
Items
SOCIAL/COGNITIVE FACTORS

Self Efficacyd,e 5.90 32.63 (12.5) 34.06 (10.7) 0.041

Can eat 5 servings Fruit/Veg daily 1.60 4.88 (3.4) 5.18 (3.1) 0.119
Can eat 3 servings Fruit/Veg daily 5.39 6.99 (3.3) 7.36 (2.9) 0.047
Can drink juice daily 4.85 7.23 (3.4) 7.46 (3.1) 0.240
Can eat Fruit/Veg for snacks 2.62 7.38 (3.2) 7.40 (2.7) 0.909
Can eat Fruit/Veg outside home 49.08 6.10 (3.5) 6.61 (2.8) 0.005
Social Supporcd,f 1.54 22.06 (10.4) 21.03 (8.6) 0.086

Family urges to eat Fruit/Veg. 0.59 5.79 (3.9) 5.63 (3.6) 0.497
Friends urge to eat Fruit/Veg. 0.82 3.59 (3.8) 3.42 (3.2) 0.401
Family eats lots of Fruit/Veg. 2.88 6.96 (3.3) 6.55 (2.7) 0.021
Friends eat lots of Fruit/Veg. 0.21 5.27 (3.3) 5.37 (2.4) 0.553
Perceived Benefitsd,g 0.94 23.60 (6.0) 23.73 (5.2) 0.744
Fruit/Veg. Prevent Heart Disease 0.76 7.69 (2.6) 7.86 (2.1) 0.258
Fruit/Veg. Prevent Cancer 0.97 7.47 (2.9) 7.49 (2.3) 0.946
Fruit/Veg. Help Lose Weight 1.65 8.32 (2.4) 8.24 (2.2) 0.558
Awareness/Knowledge
Servings Fruit/Veg. a person should eath 14.97 2.83 (1.5) 3.29 (1.8) 0.001
Aware of 5 A Day Program 8.12 1.19 (0.4) 1.28 (0.4) 0.001

Items are paraphrased

a
Low Income defined as <130% Poverty Index Ratio; takes into account the income and size of
household; <130% poverty is food stamp eligibility criteria; medium income analyzed, but not reported
b
Higher Income defined as >300% Poverty Index Ratio
c
LSD Post Hoc Test
d
Response 0 = don’t agree at all to 10=strongly agree
e
Composite score of 5 self efficacy items
f
Composite score of 4 social support items
g
Composite score of 3 perceived benefit items
h
Knowledge: number of servings fruits and vegetables respondent thinks are needed for good health

35
Table 11. Psychosocial Response Differences between Low Incomea Respondents and Higher
Incomeb Respondents: Perceived Barrier Factors

FACTORS Mean Score (SD)

higher
Scales F Low incomea
incomeb P
Value n=310-364
n=993-1116 valuec
Items

PERCEIVED BARRIER FACTORS

Weak Intentd,e 7.63 10.35 (10.6) 8.25 (8.1) 0.001

Preparation takes too much planning 0.55 2.34 (3.3) 2.17 (2.6) 0.322
Preparation takes too much time 1.74 2.78 (3.4) 2.44 (2.9) 0.065
Daily Consumption takes will power 2.40 2.69 (3.6) 2.29 (2.9) 0.033
Fruits vegetables not readily available 22.55 2.55 (3.7) 1.39 (2.5) 0.001
Cost/Qualityd,f 35.15 11.91 (8.6) 8.18 (7.0) 0.001
Fresh Fruits/Veg. too expensive 30.49 4.52 (3.8) 3.09 (3.0) 0.001
Frozen/Canned Fruits/Veg. too expensive 29.23 4.20 (3.9) 2.81 (3.0) 0.001
Fruits/Veg. poor quality 16.48 3.23 (3.7) 2.17 (2.8) 0.001

Items are paraphrased

a
Low Income defined as <130% Poverty Index Ratio; takes into account the income and size of
household; <130% poverty is food stamp eligibility criteria; medium income analyzed, but not reported
b
Higher Income defined as >300% Poverty Index Ratio
c
LSD Post Hoc Test
d
Response 0 = don’t agree at all to 10=strongly agree
e
Composite score of 4 weak intent items
f
Composite score of 3 cost/quality items

36
 Examination of Relationships between Total Fruit and Vegetable Consumption,
Psychosocial Factors, and Demographic Characteristics

The goal of these analyses was to address research question #3, which was to determine
the profile of demographic characteristics that identifies those who a) have a high fruit and
vegetable consumption pattern, b) exhibit high level of self-efficacy, c) have high level of
perceived personal benefits from consuming more fruits and vegetables, and d) are aware of the
5 A Day Program and message. These analyses do not adjust for multicollinearity, therefore
their significance is limited to beginning to determine a profile of demographic characteristics
with respect to fruit and vegetable consumption, self-efficacy, perceived benefits, and
awareness/knowledge.

Total Fruit and Vegetable Consumption by Demographic Characteristics (Table 12)

There are significant differences (p<.001) in total fruit and vegetable consumption by
gender, smoking status, and education level, and significant differences by income category
(p<.01) and age group (p<.05). Non-smokers consume almost a full serving of fruits and
vegetables more than smokers (p<.001), and females consume almost 0.8 servings more than
males (p<.001). Those with greater than a High School education consume 0.5 servings more
fruits and vegetables than those with a High School education or less (p<.001). The respondents
>= 65 years of age consumed slightly more fruits and vegetables than those in the lower age
categories <50 years of age (p<.05). Increasing income level is also correlated with higher fruit
and vegetable consumption. Generally as income increases in the total sample, mean total fruit
and vegetable intake increases, from 3.83 total daily servings in the low income group to 3.90
total daily servings in the middle income group, to 4.20 total daily servings in the higher income
group. This pattern of increasing intake among increasing income categories is generally true
even when broken down by gender and race/ethnicity groups (data not shown).

Self-Efficacy: Differences by Demographic Characteristics (Table 13)

Significant differences (p<. 001) in mean self-efficacy scores were found by gender,
smoking status, and educational level; with females, nonsmokers, and more highly educated
respondents with higher self-efficacy scores. The F scores were particularly high by gender and
smoking status, which signifies an important significant difference between males and females in
self-efficacy scores, and between non-smokers and smokers, with females and non-smokers
having a higher self-efficacy score. Significant differences (p<. 01) were also seen by income
level and race/ethnicity. Those respondents in the higher income level had a significantly higher
self-efficacy score than those in the low and middle incomes. African Americans have a self-
efficacy score significantly higher than all other race/ethnicity groups.

37
Table 12. Total Fruit and Vegetable Consumptiona: Differences by Demographic Characteristics

Characteristic Mean (SE) servings F Value P value

Gender 60.35 .000
Male 3.63 (.08)
Female 4.36 (.06)
Race/ethnicity 5.62 .001b
White 4.10 (.06)
African-American 3.89 (.12)
Latino 4.04 (.18)
Other 4.33 (.29)
Smoking status 102.18 .000
Nonsmoker 4.27 (.06)
Smoker 3.29 (.10)
Age group (yr) 7.08 .000c
18–34 3.91 (.09)
35–49 3.99 (.08)
50–64 4.21 (.12)
≥65 4.34 (.11)
Education (yr) 23.131 .000d
<12 3.67 (.14)
12 3.78 (.09)
>12 4.25 (.06)
Marital status 7.331 .001e
Married 4.12 (.07)
Divorced/widowed/separated 4.07 (.01)
Never married 3.91 (.11)
Poverty Income Ratio 13.00 .000f
<130% 3.83 (.14)
130%–300% 3.90 (.09)
>300% 4.20 (.08)
a
Servings per day estimated from fruit and vegetable frequency questions, including fruit, vegetables (no
fried potatoes), and 100% juice. Total fruit and vegetable consumption variable was log transformed for
analysis.
b
LSD post hoc test reveals significant differences between whites and blacks (p=.000), and whites and
Latinos (p=.017).
c
LSD post hoc test reveals significant differences between the 18–34 and 50–64 age groups (p = .014),
between the 18–34 and ≥65 age groups (p = .000), and between the 35–49 and ≥65 age groups (p = 003).
d
LSD post hoc test reveals significant differences between <12 and >12 years of education (p = .000) and
between 12 and >12 years of education (p = .000).
e
LSD post hoc test reveals significant differences between married and never married (p=.000).
f
LSD post hoc test reveals significant differences between <130% and 130%-300% Poverty Income Ratio
(p = .022) and <130% and >300% (p=.000), between 130%–300% and >300% Poverty Income Ratio (p =
.001).

38
Table 13. Self-Efficacya: Differences by Demographic Characteristics

Self-efficacy F Unadjusted
Characteristic mean (SE) value P value
Gender 58.83 .001
Male 31.13 (.36)
Female 34.69 (.30)
Race/ethnicity 3.99 .008b
White 32.90 (.28)
African-American 34.81 (.55)
Latino 32.74 (.70)
Other 32.15 (1.18)
Smoking status 56.66 .001
Nonsmoker 34.13 (.25)
Smoker 29.96 (.54)
Age group (yr) 1.78 .149
18–34 32.63 (.39)
35–49 33.20 (.40)
50–64 34.21 (.57)
≥65 33.31 (.65)
Education (yr) 9.45 .001c
<12 32.30 (.77)
12 31.92 (.45)
>12 34.06 (.29)
Marital status 3.24 .039d
Married 33.52 (.31)
Divorced/widowed/separated 33.53 (.47)
Never married 32.11 (.50)
Poverty Income Ratio 5.90 .003e
<130% 32.62 (.67)
130%–300% 32.31 (.43)
>300% 34.06 (.32)
a
The self-efficacy scores were derived by summing the scores of five self-efficacy questions.
b
LSD post hoc test: significant difference between Whites and African-Americans (p < .001) and
between African-Americans and Latinos/Others (p < .05).
c
LSD post hoc test: significant differences between <12 and >12 years of
education (p = .023) and between 12 and >12 years of education (p < .001).
d
LSD post hoc test: significant differences between Married and Never married (p < .05) and between
Married and Divorced/widowed/separated (p < .05).
e
LSD post hoc test: significant differences between <130% and >300% Poverty Income Ratio (p < .05)
and between 130%–300% and >300% Poverty Income Ratio (p < .05).

39
 Perceived Benefits by Demographic Characteristics (Table 14)
Significant differences in perceived benefits of eating 5 servings of fruits and vegetables
were found by gender (p<. 001), with females perceiving a higher level of health benefits from
eating vegetables and fruit, and by smoking status (p<.01), with non-smokers having a
significantly higher perceived benefit score. No significant differences were noted by
race/ethnicity, income category, or other demographic characteristics.

Table 14. Perceived Benefits of Eating 5 A Daya: Differences by Demographic Characteristics

Perceived benefitsa
Category mean (SE) F value P value
Gender 56.85 .001
Male 22.45 (.20)
Female 24.34 (.15)
Race/ethnicity 2.36 .070b
White 23.40 (.15)
African-American 23.62 (.31)
Latino 24.48 (.40)
Other 23.40 (.55)
Smoking status 9.98 .002
Nonsmoker 23.78 (.14)
Smoker 22.83 (.29)
Age group (yr) 1.09 .350
18–34 23.36 (.20)
35–49 23.68 (.20)
50–64 23.93 (.34)
≥65 23.22 (.43)
Education (yr) .419 .657
<12 23.47 (.51)
12 23.39 (.24)
>12 23.64 (.15)
Marital status 2.32 .099
Married 23.84 (.17)
Divorced/widowed/separated 23.28 (.28)
Never married 23.32 (.24)
Poverty Income Ratio .943 .390
<130% 23.60 (.36)
130%–300% 23.34 (.23)
>300% 23.73 (.17)
a
The perceived-benefits scores were derived by summing the scores of three perceived-benefits questions.
b
LSD post hoc test: significant difference between Whites and Latinos (p = .008).

40
Knowledge/Awareness of the 5 A Day Message by Demographic Characteristics
Females, whites, non-smokers, those with greater than a high school education and in the
higher income categories were significantly more aware of the 5 A Day message than males,
non-whites, smokers, high school education or less, and lower incomes (Table 15). Those
respondents with more than a high school education were twice as likely to have knowledge of
the need to consume at least 5 servings of vegetables and fruit for good health, than those with
less than a high school education. Income level is correlated with knowledge of the 5 A Day
message. Those in the high-income group were more than twice as likely to know the 5 A Day
message as those in the low-income group.

Table 15. Knowledge/Awareness of the 5 A Day Messagea by Demographic Characteristics

%
Chi-square
Characteristic Aware Unaware DF statistic P value
Gender 1 83.995 .001
Male 11.4 88.6
Female 26.0 74.0
Race/ethnicity 3 55.241 .001
White 24.1 75.9
African-American 12.6 87.4
Latino 9.4 90.6
Other 16.2 83.8
Smoking status 1 39.200 .001
Nonsmoker 22.6 77.4
Smoker 10.7 89.3
Age group (yr) 3 8.948 .030
18–34 18.3 81.7
35–49 23.1 76.9
50–64 20.1 79.9
≥65 16.9 83.1
Education (yr) 2 52.582 .001
<12 10.5 89.5
12 14.4 85.6
>12 24.8 75.2
Marital status 2 2.270 .321
Married 21.2 78.8
Div/wid/sepb 19.2 80.8
Never married 18.5 81.5
Poverty Income Ratio 2 35.440 .001
<130% 11.3 88.7
130%–300% 17.2 82.8
>300% 24.4 75.6
a
Awareness/Knowledge that 5 or more servings of fruits and vegetables are needed for good health.
b
divorced/widowed/separated.

41
Awareness of the 5 A Day Program by Demographic Characteristics

Non-smokers, whites, more educated, higher income, and those in the younger age
categories were more likely to be aware of the 5 A Day program (Table 16). Again, those
respondents with more than a high school education were more than twice a likely to have heard
of the 5 A Day for Better Health Program than those with less than a high school education. In
addition, the high-income respondents were much more likely (27.8% vs. 19.2%) to have heard
of the 5 A Day Program than those in the low-income group.

Table 16. Awareness of the 5 A Day Program by Demographic Characteristics

%
Chi-square
Characteristic Aware Unaware DF statistic P value

Gender 1 5.887 .015

Male 21.1 78.9
Female 25.2 74.8
Race/ethnicity 3 11.017 .040
White 25.3 74.7
African-American 21.8 78.2
Latino 18.4 81.6
Other 15.3 83.8
Smoking status 1 14.740 .001
Nonsmoker 25.2 74.8
Smoker 17.4 82.6
Age group (yr) 3 45.442 .001
18–34 28.8 71.2
35–49 25.1 74.9
50–64 20.9 79.1
≥65 12.0 88.0
Education (yr) 2 31.912 .001
<12 12.7 87.3
12 21.0 79.0
>12 27.1 72.9
Marital status 2 26.619 .001
Married 24.7 75.3
Div/wid/sepa 17.0 83.0
Never married 28.9 71.1
Poverty Income Ratio 2 16.151 .001
<130% 19.2 80.8
130%–300% 21.4 78.6
>300% 27.8 72.2
a
divorced/widowed/separated

42
Variance in Fruit and Vegetable Consumption Explained by Various Factors

A multiple regression model was used to examine associations between demographic

characteristics/psychosocial constructs and fruit and vegetable consumption. Tables 17 and 18
give results of multiple regression analysis by income categories showing the percent of variance
in total fruit and vegetable consumption explained by demographic covariates, social cognitive
factors, and perceived barrier factors, and further addresses the second research question, “What
differences in consumption and related factors are explained by income level?”

Multiple regression analysis was used to calculate the unique variance explained by
psychosocial factors and demographics. Table 17 shows the variance in total fruit and vegetable
consumption explained by these factors in the total sample population was 35%. Of the 35%,
8% was explained by demographic characteristics, 25% was explained by social cognitive
factors, and 2% was explained by perceived barrier factors. Self-efficacy was by far the
strongest single variable contributor to the total variance, explaining 17% of the variance. Other
single variables which contribute substantial variance include knowledge/awareness of the 5 A
Day message (5%) and smoking status (4%).

Table 18 shows some interesting differences by income category in variance explained by

independent variables. Combined, the demographics, social cognitive factors, and perceived
barrier factors explained 37% of the variance in total fruit and vegetable consumption in the low-
income group, and 34% in the higher income group. Self-efficacy factors explained 11% of the
variance in the low-income group and 14% of the variance in the higher income group. This
factor measure of self-efficacy is very strongly associated with fruit and vegetable intake.
Combined, the demographic characteristics explained 13% of the variance in the low-income
group and 9% of the variance in the higher income group. Knowledge of the 5 A Day message
alone accounted for 7% of the variance in the low-income group, and 5% of the variance in fruit
and vegetable consumption in the higher income group, indicating that knowledge/awareness of
the need to consume at least 5 servings of fruits and vegetables is strongly predictive of fruit and
vegetable consumption.

Individually, perceived barrier factors were significant predictors in both the low and
higher income groups, but explained only 2% of the variance. Cost/quality was not a significant
predictor in the low income, but was in the higher income group. Program awareness was not a
significant predictor in the higher income group, but was significant in the low-income group.

43
Table 17. Variance in Total Fruit and Vegetable Consumptiona (R2) Explained by Psychosocial
Factors and Demographic Characteristics: All Income levels

Fruit and vegetable intakesa (servings/day)

All income (n=1,463)

Independent
variables R2∆ Cumulative P value
R2

Demographic characteristicsb .081 .081 .000

Income Levelc .014 .000
Smoking Status .042 .000
Educational Level .011 .000
Gender .020 .000
Age .003 .006
Race/Ethnicity .001 .175
Marital Status .003 .012

Social cognitive factorsd .249 .330 .000

Self-efficacy .166 .000
Social support .024 .000
Perceived benefits .005 .001
Knowledge/awareness of message .051 .000
Program awareness .003 .020

Perceived barrier factorse .015 .345 .001

Weak intent .015 .000
Cost/quality of fruits and vegetables .000 .464
____
Full model .345
a
Servings per day estimated from fruit and vegetable frequency questions, including fruit, vegetables (no
fried potatoes), and 100% juice. Total fruit and vegetable variable was log transformed.
b
Demographics listed do not add up to total variance of .081, due to missing values when total model is
analyzed.
c
Income variable was log transformed.
d
Factor analysis scores from tables 3 and 5 for self-efficacy, social support, and perceived benefits all
income categories used in regression models.
e
Factor analysis scores from tables 7 and 9 for weak intent and cost/quality of fruits and vegetables all
income category used in regression models.

44
Table 18. Variance in Total Fruit and Vegetable Consumptiona (R2) Explained by Psychosocial Factors and Demographic
Characteristics: Low vs. High Income Levels

Fruit and vegetable intakesa (servings/day)

Low incomeb (n=203) Higher incomec (n=761)

Independent
variables R2∆ Cumulative P value R2∆ Cumulative P value
R2 R2

Demographic characteristicsd .126 .126 .000 .087 .087 .000

Income Levelc .003 .318 .010 .001
Smoking Status .071 .000 .051 .000
Educational Level .002 .396 .009 .002
Gender .008 .094 .024 .000
Age .000 .706 .003 .051
Race/Ethnicity .001 .562 .001 .346
Marital Status .015 .023 .002 .180

Social cognitive factorse .250 .346 .000 .227 .314 .000

Self-efficacy .110 .000 .144 .000
Social support .028 .007 .015 .000
Perceived benefits .001 .709 .014 .000
Knowledge/awareness of message .068 .000 .054 .000
Program awareness .012 .064 .001 .395

Perceived barrier factorsf .023 .369 .036 .024 .338 .000

Weak intent .023 .010 .021 .000
Cost/quality of fruits and vegetables .000 .811 .003 .062
____ ____
Full model .369 .338
a
Servings per day estimated from fruit and vegetable frequency questions, including fruit, vegetables (no fried potatoes), and 100% juice. Total
fruit and vegetable consumption variable was log transformed.
b
Low-income population defined as <130% Poverty Income Ratio. cHigher income population defined as >300% Poverty Income Ratio.

45
d
Demographics listed do not add up to total variance of .126 and .087, due to missing values when full model is analyzed.
e
Factor analysis scores from tables 3 and 5 for self-efficacy, social support, and perceived benefits by low and high income category used in
regression models.
f
Factor analysis scores from tables 7 and 9 for weak intent and cost/quality of fruits and vegetables by low and high-income category used in
regression models.

46
 Most Relevant Predictors of Fruit and Vegetable Consumption

To further examine the set of demographic and psychological variable main effects as
predictors of fruit and vegetable consumption, a hierarchical regression analysis was done to
determine the most relevant predictors of adult’s fruit and vegetable intake (Table 19).
Beginning with those psychological variables that were shown to be highly significant predictors
in the regression analysis of the full model (Table 17); awareness, self-efficacy, social support,
then income along with the rest of the demographic variables were stepped into the model, one at
a time until nine models were analyzed. Race/Ethnicity was not included because it does not
contribute to the variance as shown in Table 17. After rigorously adjusting for shared variance,
the variables that remained highly significant in the ninth model included awareness, self-
efficacy, perceived benefit, educational level, smoking status, age, and weak intent. These data
show that the standardized beta coefficients for awareness/knowledge, self-efficacy, social
support, smoking status, and weak intent remain robust even after adjusting for the other
demographic variables.

Awareness/Knowledge of the 5 A Day message accounted for 14.4% of the variance,

self-efficacy accounted for 13.2%, social support accounted for 3.1%, smoking status accounted
for 0.7%, and weak intent accounted for 1.4% of the variance in fruit and vegetable
consumption. Although not shown, several iterations with variables stepped in a rotated manner
were completed to approximate interaction terms. These iterations revealed that
awareness/knowledge, self-efficacy, social support, weak intent and smoking status maintained
their significance and the change in the beta coefficient remained roughly similar regardless of
order entered. This suggests that these variables are not related in any meaningful way to the
other variables. These psychosocial variables are an interesting set of main effects, that may
have policy-related interpretations.

Income status was a borderline significant predictor after adjusting for all variables,
suggesting that income is important. The beta coefficient dropped from model #5 (when income
level was stepped in) to model #9 in Table 19, suggesting that income shares variance with other
variables. Future research should include examining these interactions.

47
Table 19. Most Relevant Predictors in Adult’s Fruit and Vegetable Consumptiona:
Hierarchical Regression in All Income Levels

Standardized
Coefficients
Model R2 Change Beta T value P value
1 Awareness/knowledge .144 .380 15.71 .000
of 5 A Day message
2 Awareness/knowledge .283 12.29 .000
of 5 A Day message
Self Efficacy .132 .376 16.34 .000
3 Awareness/knowledge .280 12.46 .000
of 5 A Day message
Self Efficacy .374 16.63 .000
Social Support .031 .177 8.16 .000
4 Awareness/knowledge .274 12.10 .000
of 5 A Day message
Self Efficacy .377 16.75 .000
Social Support .178 8.20 .000
Perceived Benefit .003 .055 2.53 .012
5 Awareness/knowledge .265 11.66 .000
of 5 A Day message
Self Efficacy .376 16.79 .000
Social Support .180 8.32 .000
Perceived Benefit .056 2.58 .010
Income Levelb .005 .072 3.28 .001
6 Awareness/knowledge .257 11.34 .000
of 5 A Day message
Self Efficacy factor .370 16.57 .000
Social Support .167 7.63 .000
Perceived Benefit .053 2.45 .014
Income Levelb .062 2.84 .005
Smoking Status .007 -.085 -3.83 .000
7 Awareness/knowledge .253 11.04 .000
of 5 A Day message
Self Efficacy .370 16.53 .000
Social Support .166 7.62 .000
Perceived Benefit .054 2.51 .012
Income Levelb .048 2.01 .045
Smoking Status -.080 -3.58 .000
Educational Level .001 .034 1.38 .167
8 Awareness/knowledge .249 10.62 .000
of 5 A Day message
Self Efficacy .364 16.30 .000
Social Support .160 7.31 .000
Perceived Benefit .052 2.39 .017
Income Levelb .043 1.75 .080
Smoking Status -.074 -3.31 .001
Educational Level .044 1.77 .077
Age .064 2.91 .004
Gender .005 .024 1.06 .291

48
Table 19 con’t. Most Relevant Predictors in Adult’s Fruit and Vegetable Consumptiona:
Hierarchical Regression in All Income Levels

Standardized
Coefficients
Model R2 Change Beta T value P value
9 Awareness/knowledge .244 10.53 .000
of 5 A Day message
Self Efficacy .341 15.17 .000
Social Support .161 7.43 .000
Perceived Benefit .052 2.42 .016
Income Levelb .039 1.63 .103
Smoking Status -.075 -3.36 .001
Educational Level .045 1.83 .068
Age .073 3.32 .001
Gender .021 0.94 .345
Weak Intent .014 -.123 -5.65 000
a
Total fruit and vegetable consumption variable analyzed as a log transformed variable.
b
Income level analyzed as a log transformed variable.

49
 CHAPTER FIVE
SUMMARY, CONCLUSIONS, AND RECOMMENDATIONS

The evidence that vegetable and fruit consumption may contribute to the prevention of
cancer and heart disease is drawn from hundreds of epidemiological and biological studies
(Block, 1992, Steinmetz & Potter, 1991a, 1991b, 1996, van’t Veer, 2000). The overall weight
of the evidence has led to a major focus on vegetable and fruit consumption in U.S. federal
nutrition policy, with Dietary Guidelines for Americans 2000 and Healthy People 2010
identifying vegetables and fruit for increased attention and monitoring.

The 5 A Day for Better Health Program was established in 1991 as a nutrition education
campaign designed to increase awareness of the need to consume more vegetables and fruit, and
to increase vegetable and fruit consumption in the United States to an average of 5 or more daily
servings. In the first seven years of the program, fruit and vegetable intake increased on average
by about ¼ of a serving in the total adult population, and increased even less in the low-income
population (Stables, et al., in press). At this rate, it will take decades to move the U.S.
population to the minimum recommendations of 5 servings daily. Awareness of the 5 A Day
message increased significantly from 8% to 20% (1991 to 1997) in the adult population.

Given the lower fruit and vegetable consumption in the low-income population and the
health disparities among this population, it is important to investigate the barriers and promoters
to fruit and vegetable consumption among the low-income. Some research has been done to
investigate underlying psychosocial constructs to fruit and vegetable consumption in the total
population, however, less has been investigated by income status, and may help to inform
targeted nutrition education interventions in particular income groups and by race and ethnicity.
More data may also help to inform policy decisions in the low-income population.

The limitations of the survey instrument included a low response rate, and other issues.
The survey used is a follow-up survey to the 1991 5 A Day Baseline Survey, with a nationally
representative sample of U.S. adults’ (Subar, et al, 1995). Due to a lower than optimal response
rate (43.8%) in 1991, an additional Non-respondent Survey was conducted. This survey
provided data showing that non-respondents did not differ significantly from respondents in their
consumption of fruits and vegetables or in the number of servings they believed they should
consume. Due to the low overall response rate in the Baseline Survey, several strategies were
employed to improve response rates for the Follow-up Survey. These methods included a
minimum of 20 attempts (vs. four attempts in the 1991 survey) to screen a randomly selected
household and identify multiple telephone lines, at least 12 additional attempts to reach a
respondent at a working telephone number, and two attempts at refusal conversion. A computer
algorithm scheduled calls over different days and included day, evening, and weekend attempts
over a minimum of four weeks. In addition, the National Cancer Institute’s name (vs. the
research contractor’s name in the Baseline Survey) was used in the introduction of the Follow-up
Survey to evoke a credible health authority. The survey length was decreased from 30 minutes
to 15 minutes. Oversampling in high minority strata added to the nonresponse problem, as it was
noted that the screener response rate in the high minority strata was about 10% lower than in the
low minority strata.

50
 Another major weakness is the small amount of variance in fruit and vegetable
consumption (30-35%) accounted for in the survey instrument. The instrument is not capturing
the vast majority of factors that account for consumption. Even though the findings should be
interpreted with caution, the analysis is valuable in that it reveals provocative findings that point
towards further research needs.

Summary

The objective of this study was to investigate the underlying psychosocial determinants
of fruit and vegetable consumption by income via secondary analysis of a nationally
representative RDD survey of adults. First, a factor analysis was done to analyze the
interrelationships among a large number of items, and for data reduction purposes. Certain
variables correlated highly among themselves, therefore measuring the same general construct.
Select question items were reduced via exploratory factor analysis to 5 factors or underlying
latent constructs, which were then used in multiple regression models, along with
awareness/knowledge variables and demographic characteristics, to determine the variance in
fruit and vegetable consumption explained by these covariates. In addition, chi square analysis
and analysis of variance was done to determine a demographic profile of respondents by income
category in consumption, self-efficacy, perceived benefits, and awareness/knowledge.

The research questions that the study was designed to address were:
1. What are the relevant constructs in adult’s fruit and vegetable consumption that are
being measured by the 1997 nationally representative survey for the National 5 A
Day for Better Health program?
2. What differences in consumption and related factors are explained by income level?
3. What profile of characteristics differentiates individuals who a) are aware of the 5 A
Day program and message, b) exhibit high level of self-efficacy, c) have high level of
perceived personal benefits of good health-related behaviors, and d) eat high levels of
fruits and vegetables?

The relevant psychosocial constructs found via exploratory factor analysis that were
measured by the 5 A Day for Better Health Survey are self-efficacy, perceived benefit, social
support, weak intent, and cost/availability of fruits and vegetables. Other constructs are
awareness/knowledge of the 5 A Day health message. These underlying constructs are
consistent with constructs found in similar research. De Vries, et al found that attitude
(perceived benefits), social influences, and self-efficacy were predictors of behavioral intentions
(De Vries, et al., 1995). Attitudinal variables, including perceived benefits, barriers (cost,
availability, social support, etc) were assessed in a state-wide survey addressing fruit and
vegetable behaviors (Dittus, et al., 1995). A group in the Netherlands used focus groups to study
psychosocial factors in determining increased fruit and vegetable consumption and found these
same underlying constructs (Brug, et al., 1995). Two large worksite studies used predisposing
factors (perceived benefits, beliefs, and motivation) and enabling factors (barriers including
weak intent, norms including availability, and social support) as conceptual constructs of their
dietary worksite interventions (Kristal, et al., 1995; Glanz, et al., 1998; Kristal, et al., 2000).

51
 Awareness and knowledge of the health message to eat at least 5 servings of fruits and
vegetables for good health is associated with increased consumption of fruits and vegetables and
is considered an underlying construct in several studies ( Krebs-Smith, et al., 1995; Stables, et
al., in press, Harnack, et al., 1997; Patterson, et al., 1997; Neill, et al., 2000). Lack of knowledge
and awareness of key health messages has communication and educational implications.
Educational interventions, including media interventions, designed to increased knowledge and
awareness of the importance of consuming at least 5 servings of fruits and vegetables, may have
an impact on the process of dietary change.

Income level differences

What differences in consumption and related factors are explained by income level? This
research question was answered by using analysis of variance with LSD post hoc tests to
determine psychosocial, education, and behavioral response differences between low income and
high-income respondents. Very little research has been published on the psychosocial,
education, and behavioral factors regarding fruit and vegetable intake by income level.

In an unadjusted analysis, a significant but small difference was found between low and
higher income respondents in the composite self-efficacy scale and two of the individual items,
with the higher income respondents having greater self-efficacy scores. There was a significant
modest difference with respect to eating at least 3 servings daily, however no difference between
income status with respect to eating at least 5 servings daily. Even though eating at least 5
servings daily is a stated health objective, the self-efficacy scores for this behavior were lower
than all other social cognitive, social support, or perceived benefit items, suggesting that this a
behavior perceived to be difficult to master. The scores for confidence in being able to eat fruits
and vegetables outside the home were highly significantly different between low and high
income, with high-income respondents being more confident. It can be speculated that higher
income respondents may frequent family style or white table cloth restaurants where menus
include more variety, including more fruits and vegetables, versus lower cost fast food-type
restaurants where fruits and vegetables are scarce.

Scores for Social Support and Perceived Benefit Scales and items were not significantly
different between low and high income, with the exception that low income respondents were
more likely to agree that family members eat lots of fruits and vegetables. The findings between
income level on the perceived benefit scales is consistent with another study whereby there was
no difference by income level on an attitude scale regarding benefits of fruit and vegetables
intake (Dittus, et al., 1995).

Individual knowledge/awareness items were significantly different between low income

and higher income respondents, with low-income respondents being less aware of the need to
consume at least 5 servings of fruits and vegetables for good health. The difference in
knowledge/awareness between income groups is small, and awareness is below optimal for all
incomes. Knowledge that persons should eat at least 5 servings of fruits and vegetables for good
health is associated with increased fruit and vegetable consumption in populations. In this
research, the higher income population on average, responded that people should eat more fruits
and vegetables than did low income individuals. Indeed, higher income individuals actually
consumed more fruits and vegetables. In moving along the continuum of processes of change,

52
awareness of key health messages is an important step in behavior change (Prochaska, et, al.,
1992). Other research has shown that awareness and knowledge of dietary recommendations are
significant predictors of increased intake (Patterson, et al., 1996; Neill, et al., 2000) and that
parental knowledge and awareness of the need to consume more vegetables and fruit are
independent predictors of children’s fruit intake (Gibson, et al., 1998). In addition, a strong
relationship was found between 5 A Day message awareness and stage of change in 5 A Day
community-based research in adults (Campbell, et al., 1999).

Higher income respondents were slightly, but significantly more aware of the 5 A Day
Program, than were low-income respondents. These income differences in both message
awareness and program awareness, have several implications. It is possible that supermarkets in
low-income areas do not display 5 A Day nutrition education materials. Access to interventions
to increase fruits and vegetables may not be available to the low-income. The 5 A Day for Better
Health Program media outreach may not be targeted to low-income, or simply is not reaching the
low-income.

There were significant differences between income levels in weak intent and cost/quality
composite scores, and in most of the corresponding individual questionnaire items. The low-
income were more likely to agree that fruits and vegetables were too expensive, of poor quality
and not readily available, and that it takes too much will power to consume more fruits and
vegetables. These kinds of cost, quality, and availability barriers have major policy implications
for the low-income.

Variance in Fruit and Vegetable Consumption Explained by Psychosocial Factors

Using a multiple regression model, the factors that significantly contributed to the
variance in total fruit and vegetable consumption in the total sample population included
demographic characteristics (with the exception of race/ethnicity), all social cognitive factors,
and perceived barrier factors. Of the 34.5% of variance explained by these variables,
demographic characteristics, self-efficacy and knowledge/awareness of health message all
contributed the majority of variance. The self-efficacy variable was responsible for half of the
explained variance.

In the low-income population, the factors that significantly contributed to the variance,
included demographic characteristics (smoking status, gender, and marital status) most of the
social cognitive factors, and weak intent. Of the 36.4% of variance explained by these variables,
demographic characteristics, self-efficacy and knowledge/awareness of health message
contributed the majority of the variance. Demographic characteristics accounted for almost 13%
of the variance versus 9% in the high income, with smoking status showing the most difference
between low and high-income. Social support explained 3% of the variance in the low income,
but only 1.5% in the high income. Interestingly, perceived barrier factors explained only 2% of
the variance in fruit and vegetable consumption in both the low and the high-income
respondents.

In the high-income population, self-efficacy alone contributed almost 15% of the

variance, with knowledge/awareness of the health message and demographic characteristics
contributing 5% and 9%, respectively.

53
 This research suggests that, regardless of income category, social cognitive factors, in
particular self-efficacy and knowledge/awareness of the health message, are stronger predictors
of fruit and vegetable consumption than are perceived barrier factors. This is in contrast to the
findings by Dittus, et al., (1995), whereby barriers were the largest contributor to explaining
variability in fruit and vegetable consumption. However, in this study, perceived barrier factors,
do explain significant variance in fruit and vegetable consumption in both income groups.

Profile of characteristics which differentiates individuals who a) are aware of the 5 A Day
program and message, b) exhibit high level of self-efficacy, c) have high level of perceived
personal benefits of good health-related behaviors, and d) eat high levels of fruits and
vegetables
One-way analysis of variance with LSD post hoc tests were done to determine
differences by demographics in total fruit and vegetable consumption, self-efficacy, and
perceived personal benefits of good health-related behaviors. The demographic profile of those
who consume larger amounts of fruits and vegetables are female, non-smoker, >50 years of age,
more educated, and in the higher income category (Table 12). The demographic profile of those
with higher levels of self-efficacy are female, African American, non-smoker, >50 years of age,
more educated, and in the higher income category (Table 13). The demographic profile of those
with higher perceived benefits of eating more fruits and vegetables are female non-smokers
(Table 14).

In addition, chi square analysis was done to determine demographic differences in

knowledge/awareness of 5 A Day message and program. The demographic characteristics of
those who are aware of the need to consume 5 or more servings of fruits and vegetables daily are
female, white, nonsmoker, more educated, and higher income (Table 15). The demographic
profile of those who are aware of the 5 A Day program are female, white, nonsmoker, those 18-
34 years of age, more educated, and in the higher income category (Table 16).

The ANOVA and Chi square analyses establish a profile of demographics, but it would
be misleading to assume that this is everybody’s profile. The unadjusted, overall demographic
profile of those who consume a high level of fruits and vegetable and have high levels of the
mediating or predictive factors to consumption, are white, nonsmoking females who are usually
older, more educated, and in the highest income category.

Most Relevant Predictors of Fruit and Vegetable Consumption

A hierarchical regression analysis that adjusted for multicollinearity was done too further
analyze the data to determine the most relevant constructs in fruit and vegetable eating behavior
and to explore the possibility of shared variance among variables (Table 19). The overall
findings reveal that awareness/knowledge of the health message and self-efficacy are the best,
most stable predictors of fruit and vegetable consumption. This suggests that further research to
devise even better measures and scales of awareness and self-efficacy, and to determine the
contributors to awareness and self-efficacy would result in more meaningful predictors of
consumption. These factors, along with social support and weak intent, could potentially be
addressed by educational interventions designed and tailored to specific demographic groups.

54
 Policy relevant variables that are significant or borderline significant predictors of fruit
and vegetable consumption include knowledge/awareness, income level, and educational level.
Policies that affect these variables could include broad nutrition education efforts to increase
awareness and importance of the need to consume at least 5 servings of fruits and vegetables,
price supports for fruits and vegetables for all income levels, and potentially increasing minimum
wage to increase buying power for fruits and vegetables.

Conceptual Framework of Fruit and Vegetable Consumption Behavior

The conceptual model (figure 2) developed from these analyses, is a model incorporating
domains from social cognitive theory and based on the findings from the analysis of
psychosocial response differences between low income and high income respondents, and on the
multiple regression model examining associations between psychosocial constructs and fruit and
vegetable consumption. Due to the likely possibility of interactions between income status and
other demographic variables, income is portrayed as having an influence on other demographic
characteristics, and vice versa. In addition, there were many differences in psychosocial
responses between low-income respondents and high-income respondents via analysis of
variance

The main factors that emerged from the multiple regression analysis to predict fruit and
vegetable consumption were demographic characteristics, awareness/knowledge of the 5 A day
message and self-efficacy, with social support and barriers also responsible for substantial
variance in intake. The implications of the model is that a person’s health behavior (fruit and
vegetable consumption) can be changed by changing awareness/knowledge of key health
messages, a person’s perceptions of social support, self efficacy expectations, and barriers.
Although barriers in these analyses did not contribute substantially to variance in intake, perhaps
due to an inadequate survey instrument, conceptually, barriers play a role in behavior change.
Demographic characteristics are expected to influence behavior through the behavioral
determinants.

The design of the conceptual model show that some of the psychosocial constructs are
more significant contributors to the variance in fruit and vegetable consumption than others.
The model reflects those constructs that contribute significantly to the variance in total fruit and
vegetable consumption. The variance explained by both self-efficacy and awareness ranges
between 18% and 20%, low income vs. high income, respectively.

A likely next research step could include a path analysis (structural equation analysis) to
further investigate causal relationships between the observed variables. This next step would
serve to test the postulated relationships developed from the analyses completed thus far, and
included in the conceptual model. Awareness/Knowledge of the 5 A Day message and self-
efficacy accounted for the largest single variable variance in fruit and vegetable consumption in
the regression modeling. It would be especially important to determine the causal relationship
via a path analysis, between these two variables and among the other variables.

55
Figure 2. Theoretical model of fruit and vegetable consumption behavior
determinants

AVAILABILITY
COST/QUALITY
WILL POWER

INCOME
AWARENESS
SELF-EFFICACY FRUIT &
OTHER (both constructs account VEGETABLE
DEMOGRAPHIC for 18-22% of variance CONSUMPTION
in F & V consumption)
CHARACTERISTICS

SOCIAL SUPPORT

PERCEIVED
BENEFITS

56
 Implications for Practice

The findings from this study add support to research suggesting that specific psychosocial
constructs are positively associated with increased fruit and vegetable consumption. These
constructs and degree of association vary by income group. Self-efficacy and
awareness/knowledge of key health messages account for the greatest variance in consumption
for both low and higher income participants. Self-efficacy is a more important predictor for
high-income respondents, while awareness/knowledge was more important for low-income
versus higher income respondents. Social support was twice as important predictor of fruit and
vegetable consumption in the low-income group versus the high-income group.

The implications of these study results are that nutrition education interventions to
increase fruit and vegetable consumption in adults should include strategies aimed at increasing
awareness/knowledge of key health messages, and aimed at increasing self-efficacy toward
eating fruits and vegetables. Strategies to increase self-efficacy would include breaking the
desired skills down into achievable steps, provide education to achieve these results, and then
monitor and reward achievement of these skills. Strategies to increase social support for this
behavior, and to affect weak intent are also important. The many differences between
psychosocial, educational, and behavioral responses between income groups underscore the need
for tailoring educational interventions by income group for greater impact. The design of
nutrition education interventions by income category should include formative research to
increase our understanding of awareness/knowledge, self-efficacy, barriers, and social support
issues in this population.

The findings from the ANOVA analyses are suggestive that issues relating to
cost/quality/availability of fruits and vegetables may be more salient in the low-income
population. Nutrition education in the low income could include cost strategies (in-season, sale
items, farmers market fruits and vegetables), and accessibility/availability strategies (keep frozen
and canned fruits and vegetables on hand). Hierarchical regression suggests that income level
interacts with other variables and is maintained as a significant predictor of fruit and vegetable
consumption even after adjustment for shared variance.

Clearly, the policy issues of income, age, and educational level should be addressed.
Price supports for fruits and vegetables are possible policy interventions. Providing additional or
double coupons for fruits and vegetables for WIC clients has been tried on a small scale and is
successful in increasing consumption among that population. Another possible intervention is to
increase the redeemable value of food stamps to buy fruits and vegetables. Monetary incentives
for schools to incorporate fruit and vegetable bars into cafeterias is another possibility.
Including low-fat snacks at a reduced price in vending machines has also been shown to be
successful in increasing consumption of healthy foods in school settings (French, et al., 2001). It
may be possible to send a package of fruits and vegetables home with those who eat at senior
congregate meal centers. There are many creative ways to provide price supports for purchase of
fruits and vegetables, if there is political will to do so.

57
 Policies enacted to increase awareness/knowledge of the need to include 5 servings of
fruits and vegetables would mandate broad-scale nutrition education efforts through a variety of
intervention channels, including media, point of sale, worksites, faith institutions, and schools.
Certainly, the policy issues should be examined further for this population. However, given the
relatively low overall variance that is explained in this analysis, it is not possible to make policy
recommendations based upon this study.

Future Research Needs

Further research on psychosocial mediators of fruit and vegetable consumption is needed

to aid in designing interventions that focus on increasing awareness/knowledge and self-efficacy.
Constructing and validating scales for mediating variables of fruit and vegetables consumption is
needed to accurately measure change. These measures could be used in a well designed
longitudinal study to assess contributors to dietary behavior. By using the information found in
this research, better tailoring may be done in developing and delivering nutrition education
interventions by income level.

A research focus on determining common terminology for the psychosocial constructs for
nutrition education is needed. For example, attitudes were often interchangeable with barriers,
across research papers. In addition, a research focus on common measures would also improve
generalizability of findings across studies. In some studies self-efficacy was determined by a
single question, whereas in others, a composite score of many questions addressing self-efficacy
were used.

In keeping with the above research needs, a better understanding of the relationship of
mediating variables, such as the psychosocial constructs used in this study, to outcomes is
needed to determine the effectiveness of educational interventions. As was true in this study,
most models account for approximately 30- 35% of the variance in fruit and vegetable intake.
This leaves the majority of variance unaccounted for, and therefore, work needs to be done to
increase the projected predictability of behavioral and educational mediators.

Policy research to address the problems of availability, accessibility, and quality and to
determine the impact of price supports on fruits and vegetables needs to be done with regard to
the entire population. The change in fruit and vegetable consumption in this country has been
entirely too slow. Policy changes and changes in the food environment hold promise to speed up
the velocity of dietary change. Especially in the low-income population, systems level changes
designed to make fruits and vegetables more affordable, and more available, have potential in
narrowing the consumption gap, between the low-income population and the high-income
population, and to move the entire population to recommended levels of fruits and vegetables.
Alternative policy decisions, such as redeemable Farmer’s Market coupons for fruits and
vegetables for low-income seniors, or doubling of food stamps used to purchase fruits and
vegetables, could be examined in small settings for efficacy and effectiveness. The low-income
population certainly has many competing stressors in life that take priority over food
consumption decisions. However, policy changes that make it easier for the low-income to
purchase and consume fruits and vegetables will help to make a healthier diet more attainable.

58
 Finally, this investigation may bolster the efforts to build a set of theories specific to
nutrition education for behavior change in adults. The social cognitive theory tends to include
many of the constructs that are relevant to nutritional behavior change, but given the large
variance unaccounted for when using these constructs, this theory needs to be augmented by
new, creative constructs, yet to be discovered, or better ways to measure the existing constructs.

Conclusions

Even though the 1997 5 A Day for Better Health Program survey instrument had major
methodologic limitations, implications for practice and future research needs were identified.
The psychosocial factors of awareness/knowledge and self-efficacy were the most robust
predictors of variance in fruit and vegetable consumption. Further work in construction of
reliable and valid scales of these two factors and other promising mediators of nutrition behavior
change is warranted, to increase the amount of variance that can be explained. In terms of
practice, this research suggests the importance of developing interventions that promote
increased self-efficacy and awareness/knowledge of the 5 A Day message. Given the observed
differences between low and higher income groups in mediators of behavior change, efforts to
tailor interventions to demographic characteristics could be an important strategy in nutrition
education.

The observation of potential interactions between income level and other demographic
characteristics suggests policy implications that should be investigated further. Price supports
and enhanced provision of fruits and vegetables to all individuals, especially the low-income,
through existing nutrition entitlement programs and other novel programs are possibilities for
further policy interventions and research. Making systems level changes for a more conducive
environment to increase fruit and vegetable consumption in all Americans is a policy change that
would enhance disease prevention and health promotion efforts.

59