Social Science & Medicine 71 (2010) 93e101

Contents lists available at ScienceDirect

Social Science & Medicine

journal homepage: www.elsevier.com/locate/socscimed

Short report

Individual and environmental correlates of dietary fat intake in rural

communities: A structural equation model analysis
April K. Hermstad a, *, Deanne W. Swan a, Michelle C. Kegler a, J.K. Barnette b, Karen Glanz a, c
a
Rollins School of Public Health, Emory University, 1518 Clifton Road, Atlanta, GA 30322, United States
b
Emory Prevention Research Center at the Southwest Georgia Cancer Coalition, United States
c
University of Pennsylvania, Schools of Medicine and Nursing, United States

a r t i c l e i n f o a b s t r a c t

Article history: Total dietary fat and saturated fat intake are associated with obesity, elevated cholesterol, and heart
Available online 13 April 2010 disease. This study tested a multi-group structural equation model to explore differences in the relative
influence of individual, social, and physical environment factors on dietary fat intake amongst adults
Keywords: aged 40e70 years. Participants from four rural Georgia, U.S., counties (n ¼ 527) completed a cross-
Dietary fat sectional survey that included questions about eating patterns and individual and social influences on
Eating patterns
healthy eating. Observational measures of nutrition environments in stores and restaurants in these
Psychosocial
counties also were completed. Models for both women and men found significant positive relationships
Food environment
Rural
between self-efficacy for healthy eating and perceived nutrition environments and family support for
USA healthy eating. The association between self-efficacy for eating a low-fat diet and frequency of eating out
Gender and grocery shopping was negative for both genders. The home nutrition environment was associated
with dietary fat intake for women but not men. The results indicate that the influence of individual and
environmental factors on dietary fat intake differs for men and women, with the home environment
playing a larger role for women in rural communities.
Ó 2010 Elsevier Ltd. All rights reserved.

More than two-thirds of Americans are overweight or obese thought to interact to influence health behavior. Several aspects of
(Ogden et al., 2004), and rates are higher in rural areas (Patterson, the physical environment, for example, may influence eating
Moore, Probst, & Shinogle, 2004). Total dietary fat intake and satu- behaviors (Glanz, Sallis, Saelens, & Frank, 2005), such as access to
rated fat intake are associated with obesity as well as unfavorable supermarkets as opposed to smaller grocery stores (Boehmer et al.,
cholesterol profiles and coronary heart disease (Boehmer, Lovegreen, 2006; Chang, Baumann, Nitzke, & Brown, 2005), availability of
Haire-Joshu, & Brownson, 2006; USDHHS, 2000; Williams, Wold, healthy foods in nearby stores (Liese, Weis, Pluto, Smith, & Lawson,
Dunkin, Idleman, & Jackson, 2004; WCRF/AICR, 2007). Research on 2007), and home nutrition environments (Patterson, Kristal,
the determinants of dietary behavior among rural Americans is Shannon, Hunt, & White, 1997; Satia & Galanko, 2007). Access to
limited, despite the fact that more than 20% of the U. S. population supermarkets has been shown to be associated with BMI and
lives in a rural area (US Census Bureau, 2000). Environmental factors several health conditions (Inagami, Cohen, Finch, & Asch, 2006;
that may influence dietary behavior in rural settings include social Morland, Diez Roux, & Wing, 2006), as well as lower fat intake.
and geographic isolation, limited access to transportation, limited The home nutrition environment also is associated with dietary fat
nutrition-related services (Ledikwe et al., 2003), and distance to the intake (Patterson et al., 1997; Satia & Galanko, 2007).
nearest supermarket (Boehmer et al., 2006). Consumer-related nutrition behaviors are also associated with
Ecologic models of health behavior posit a reciprocal relation- diet quality. Eating at fast-food restaurants has been positively
ship between behavior and multiple levels of influence including correlated with fat intake (Satia, Galanko, & Siega-Riz, 2004),
multiple types of environments (Sallis, Owen, & Fisher, 2008). whereas shopping at a supermarket rather than an independent
Ecological models include factors nested within increasingly broad grocer was associated with improved diet quality among women
levels of influence including intrapersonal factors, interpersonal or shoppers (Zenk et al., 2005). Echeverria, Diez Roux, and Link (2004)
social factors, physical environments, and policies, all of which are found that perceived availability of some healthful foods in stores
was significantly associated with purchase of those foods.
* Corresponding author. Tel.: þ1 404 712 9537; fax: þ1 404 727 1369. Social support for healthy eating has been associated with
E-mail address: [email protected] (A.K. Hermstad). reduced fat intake (Chang et al., 2005; Hagler et al., 2007;

0277-9536/$ e see front matter Ó 2010 Elsevier Ltd. All rights reserved.
doi:10.1016/j.socscimed.2010.03.028
94 A.K. Hermstad et al. / Social Science & Medicine 71 (2010) 93e101

Nothwehr, Snetselaar, & Wu, 2006; Walker, Pullen, Hertzog, Population and procedures
Boeckner, & Hageman, 2006; Watters & Satia, 2009), and lack of
social support for healthy eating from family members has been The study sample included AfricaneAmerican and Caucasian
identified as a barrier to healthy eating (Chang, Nitzke, Guilford, residents (aged 40e70) of four rural counties in Southwest Georgia,
Adair, & Hazard, 2008). Self-efficacy appears to play an important the United States, who lived in their county for at least five years
role in eating habits too, such that those with high self-efficacy for and lived with at least one other person. Purposive sampling
healthy eating, especially women, are more likely to eat healthfully methods were used to recruit approximately equal proportions of
(Kuppens, Eriksen, Adriaanse, Nijhuis, & Aaron, 1996; Nothwehr AfricaneAmerican and Caucasian men and women in each county.
et al., 2006; Walker et al., 2006). Watters and Satia (2009) found Because HRC2 is a study of cancer-preventive behaviors (diet,
in an African American population that greater self-efficacy for smoking, and physical activity), exclusion criteria included a cancer
eating less fat was associated with lower fat intake among women diagnosis or recurrence within two years prior to data collection.
but not men. Other studies that have looked at gender differences Participants were recruited in-person at local community sites in
in self-efficacy for healthy eating with respect to dietary fat found 2006e2007. Each participant completed a survey regarding health
no difference (Mosher et al., 2008; O’Hea, Wood, & Brantley, 2003). behaviors including diet, physical activity, and smoking; social and
In the research reported here, we sought to examine the relative physical dimensions of the home, church, and work environments;
influence of physical and socioenvironmental, as well as individual- and basic demographic and health information. Participants also
level factors on differences in dietary fat intake among residents of measured their waist circumferences using an unmarked tape.
rural communities in the Southeastern United States. We are Consumer nutrition environment data collection took place in the
specifically interested in determining whether relationships spring of 2007 and the winter of 2008, by five experienced trained
observed among urban men and women are corroborated in this raters using a reliability- and validity-tested instrument (Glanz,
rural population. This study is unique in that it explores simulta- Sallis, Saelens, & Frank, 2007). This research was approved by the
neously the nature of these relationships in a rural population, Emory University Institutional Review Board.
using a structural equation modeling technique. To improve the
eating habits and health status of rural populations, it is critical to Conceptual model
understand the determinants of behavior that may be unique or
universal to both geography and gender. We developed a conceptual model relating individual, socio-
environmental, and physical environment influences to the
Method dependent variable, dietary fat intake (Fig. 1). In the model, features
of the physical environment are posited to have direct and indirect
Design effects on socio-environmental and individual characteristics, as
well as individual behaviors, which in turn influence dietary fat
This analysis used baseline survey data from the Healthy Rural intake. Constructs were included in the model based on ecological
Communities 2 (HRC2) study, supplemented by observational models of health behavior (Sallis et al., 2008) and empirical
measures of neighborhood food environments. HRC2 is a longitu- evidence.
dinal descriptive study of the environmental and individual
determinants of these behaviors. For the present analysis, HRC2 Measures
survey data were linked by participant addresses to restaurant and
food store addresses that were geocoded using ArcMap 9.2 to Objective nutrition environment measures
assess distance between points (“as the crow flies”) (ESRI, 2006, Nutrition environment data were collected at two levels:
ArcGIS Version 9.2). We performed a multigroup structural equa- community and consumer (Glanz et al., 2005). The community
tion analysis to assess the relationship between individual- and nutrition environment data assessed the number, type, location,
environmental-level influences on dietary fat intake for men and and accessibility of food stores and restaurants in the study
for women. counties. Lists of all outlets possessing food and beverage licenses

Fig. 1. Conceptual structural model of environmental and behavioral influences on dietary behavior.
 A.K. Hermstad et al. / Social Science & Medicine 71 (2010) 93e101 95

in each county were obtained from the Georgia Department of whether a large selection of these foods was available in their
Agriculture. The lists were then updated using online business neighborhood.
directories.
The consumer nutrition environment data described the avail- Home nutrition environment. The Home Nutrition Environment
ability, cost, and quality of healthy food items in grocery and was measured using items from Patterson’s inventory measure
convenience stores in the study area (Glanz et al., 2005). The (Patterson et al., 1997). Participants were asked whether three key
Nutrition Environment Measures Survey for Stores (NEMS-S) high-fat items were available in their home in the last week:
(Glanz et al., 2007), an established observational tool with high regular bacon or sausage, regular whole milk, regular hot dogs.
reliability, was used to assess the availability of key categories of
foods in grocery and convenience stores. For the present analysis Consumer nutrition behavior. Items to measure this construct were
focusing on dietary fat intake, NEMS-S scores were computed based adapted from the Continuing Survey of Food Intake by Individuals
on availability of low- or reduced-fat milk, lean ground beef, and (USDA, 1996). These questions assessed how often participants
light or fat-free hot dogs, to correspond with HRC2 survey variables. shopped at large supermarkets or convenience stores, or ate out at
NEMS-S scores were summed for all stores (convenience and sit-down or fast-food restaurants when shopping for food or dining
grocery stores) less than 1 and within 1e5 miles of participants’ out. A 4-point scale was used, ranging from 1 (Never or rarely) to 4
homes. Higher NEMS-S scores on a scale of 1e9 reflected greater (Almost always).
availability of lower-fat foods in grocery and convenience stores in
the participant’s local food environment. For example, if a partici- Dietary fat intake behavior. Dietary fat intake, the dependent vari-
pant has one grocery store and two convenience stores within one able, was measured using six items from the Percent Energy from
mile of their home, and the grocery store received a score of 9 and Fat Screener from the Johnson, 2002; National Cancer Institute
both convenience stores a score of 7, then the nutrition environ- (NCI) (1996). The measure of dietary fat intake is a composite
ment score for 1-mile is 23. For restaurants, information was calculated based on the frequency of self-reported consumption of
collected on the number of fast-food or sit-down restaurants and six food items: sausage or bacon, cheese or cheese spread, French
their locations in each of the four counties but not on menu fries or hash browns, mayonnaise, salad dressings, and margarine,
selections at those restaurants. Distances from participants’ homes butter, or oil. Responses were converted to daily servings for
to fast-food and sit-down restaurants were mapped and partici- analysis and interpretation. Internal consistency of the dietary fat
pants were classified as living less than 1 mile, 1e5 miles, or more intake score was high (a ¼ 0.71).
than 5 miles from any fast-food restaurant. The same categories
were used for sit-down restaurants. Higher Community and Statistical analysis
Consumer Nutrition Environment scores indicated greater access to
stores and restaurants. The goal of the present analysis was to examine direct and
indirect effects of the objectively measured neighborhood nutrition
Self-reported measures environment, the home nutrition environment, the social envi-
Sociodemographic characteristics. Items included gender, age, race, ronment, self-efficacy and behavior. In addition, we examined
marital status, number in household, educational attainment, and whether these relationships were different across genders. To
annual household income. investigate the relationships between these constructs, we fit
a multi-group structural equation model (SEM) to the data (Bollen,
Risk factors. Body Mass Index (BMI) was computed based on self- 1989). Because it was a multi-group analysis, we first tested the
reported weight and height. BMI was corroborated with waist model for all participants and then fit the model separately for
circumference, which was highly correlated (r ¼ 0.79). A second women and men.
risk factor indicator was the number of chronic disease diagnoses
(CDC, 2005). Structural equation modeling
SEM, an extension of the general linear model, is a regression-
Self-efficacy for healthy eating. Self-efficacy for healthy eating was based technique that incorporates factor analysis for the creation of
measured by a six-item scale (a ¼ 0.78) using a subset of items from latent constructs. Data were examined prior to modeling to ensure
a scale developed by Sallis (Sallis, Pinski, Grossman, Patterson, & that they met assumptions for the technique. A full set of descrip-
Nader, 1988). Participants were asked to use a 5-point Likert scale tive statistics is available from the authors upon request.
from 1 (I definitely cannot) to 5 (I definitely can) to rate their Models were estimated using LISREL, version 8.71 (Jöreskog &
confidence to consistently maintain low-fat eating patterns in Sörbom, 2004). First, a measurement model using confirmatory
difficult situations. factor analysis (CFA) was used to confirm the relationship between
the latent variables (i.e., theoretical constructs) and their indicator
Family support for healthy eating. Family Support was measured by (observed) variables. In addition to assessing measurement error,
three items from a 6-item scale adapted from Sallis (Sallis, CFAs in multi-group models also determine group effects on the
Grossman, Pinski, Patterson, & Nader, 1987) (a ¼ 0.78). Partici- latent variables, indicating where loadings may differ across
pants indicated how often their family discussed eating habits groups. The measurement model was tested first to the full sample
changes, reminded them not to eat high-fat/salt foods, or com- and then to each group (women and men) separately.
mented if they went back to their old eating habits. A 4-point scale Second, the structural model was tested to estimate the strength
was used, ranging from 1 (Never or rarely) to 4 (Almost always). of the relationships between latent variables. It also allows for the
examination of direct and indirect effects among the constructs (i.
Perceived nutrition environment. The Perceived Nutrition Environ- e., latent variables) in the model. Because this was a multi-group
ment scale was based on four items from Echeverria et al. (2004). analysis, the structural model was fit first to the full sample and
Using a 5-point Likert scale from 1 (strongly disagree) to 5 (strongly then to each group separately.
agree), participants were asked to rate their level of agreement In the process of fitting a multi-group structural model, different
with statements regarding how easy it was to purchase low-fat paths are either constrained to be equal across both groups or
products and fruits and vegetables in their neighborhood, and allowed to vary. When a constraint is released, a different
96 A.K. Hermstad et al. / Social Science & Medicine 71 (2010) 93e101

coefficient is estimated for each of the groups, with the inference Table 1
that the relationship between the constructs is different for the Demographic characteristics by gender.

groups. Constraints across groups were determined using both Men Women Overall
empirical justification (i.e., modification indices) and judgment of N % N % N %
whether it was consistent with the conceptual model and health
243 284 527
behavior theory. Constraints were released sequentially, assessing Age (in years)
the statistical significance of the largest change in fit each time and 39e49 years 113 46.5% 136 47.9% 249 47.2%
stopping when releasing additional paths no longer improved fit. 50e59 years 83 34.2% 106 37.3% 189 35.9%
60e70 years 47 19.3% 42 14.8% 89 16.9%
Parameters for paths where constraints were released are inferred
to differ across the groups; parameters with constraints are inter- Race/ethnicity
preted as being equal for both groups. White/Caucasian 130 53.5% 136 47.9% 266 50.5%
Black/AfricaneAmerican 113 46.5% 148 52.1% 261 49.5%
Although data collection could be considered a cluster sample
from counties, multi-level techniques were not used in data anal- BMI/weight status*
ysis. Using such techniques with a small number of clusters (in this Normal or underweight 60 24.7% 75 26.4% 135 25.6%
(BMI < 25)
case, n ¼ 4) and little between-cluster variability (all ICCs < 0.01) Overweight (BMI 25e29.9) 97 39.9% 77 27.1% 174 33.0%
tends to result in inadmissible solutions (Hox & Maas, 2001). The Obese (BMI  30) 86 35.4% 132 46.5% 218 41.4%
counties used in this study were selected because they were similar
Marital status*
in key demographic variables according to 2000 U.S. Census data. Married/with partner 173 71.2% 155 54.6% 328 63.0%
Not married 68 28.0% 125 44.0% 193 37.0%
Model fitting (widowed, divorced,.)
Multiple indices of goodness of fit were used to assess model fit Number in household
(Bollen, 1989; McDonald & Ho, 2002): the chi-squared fit statistic Two (self þ 1) 79 32.5% 89 31.3% 168 31.9%
(c2:df ratio), the Bentler-Bonnett non-normed fit index (NNFI), the Three (self þ 2) 60 24.7% 92 32.4% 152 28.8%
comparative fit index (CFI) and the root mean square error of Four (self þ 3) 48 19.8% 59 20.8% 107 20.3%
Five or more 56 23.0% 44 15.5% 100 19.0%
approximation (RMSEA).
SEM requires a complete dataset with no missing values for Education
Less than 12th grade 33 13.6% 31 10.9% 64 12.2%
model testing. Overall, missing data were not a problem in the
High school/GED 74 30.5% 83 29.2% 157 29.8%
dataset for the variables used in the present analysis. Of the 527 Some college/tech 67 27.6% 97 34.2% 164 31.2%
respondents to the baseline questionnaire, 87.3% had complete College graduate 68 28.0% 73 25.7% 141 26.8%
survey data for the observations in the present analysis. To retain or post-graduate
the maximum number of cases for the analysis, a hotdeck impu- Income*
tation algorithm was used to impute missing values using PROC IML Less than $10,000 18 7.4% 48 16.9% 66 15.0%
in SAS version 9.1. $10,001 to $25,000 38 15.6% 57 20.1% 95 21.5%
$25,001 to $50,000 59 24.3% 64 22.5% 123 27.9%
$50,001 or more 86 35.4% 71 25.0% 157 35.6%
Results
Note: * Group difference for chi-square, p < 0.05.

Healthy rural communities 2 participant characteristics

regular fat foods in the home (M ¼ 1.9 vs. 2.1). These differences
Of all 750 individuals screened, 79.3% were eligible, and 88.6% of provide empirical support that there may be differences in the
eligible persons completed the survey. Table 1 shows baseline relationship between the environment and dietary fat intake
background characteristics of participants (n ¼ 527). Just over half of behavior across gender.
participants in the sample were White (50.5%) and female (53.9%). The overall fit of the measurement models for both women and
Nearly half of the participants were aged 39e49 years (47.2%), and men were adequate based on fit indices (Women’s measurement
the mean age was 51.1 years. Most participants resided in small model: RMSEA ¼ 0.048, CFI ¼ 0.93; Men’s measurement model:
towns (51.8%) and were married or living with a partner (63.0%), had RMSEA ¼ 0.047, CFI ¼ 0.93). Additionally, the full multi-group
at least some college education (58.0%), and more than one-third measurement model was adequate, RMSEA ¼ 0.064 (90% CI ¼ [0.059,
(36.5%) reported an annual household income of $25,000 or less. A 0.069]), NNFI ¼ 0.83, CFI ¼ 0.84. Final standardized and unstan-
majority of participants were overweight (BMI 25e29.9) or obese dardized factor loadings for the measurement models by gender are
(BMI 30 or greater): 33.0% and 41.4%, respectively. Most participants in Table 3. There were minor differences between genders on the
(63.8%) reported having none or only one of the eight common risk loadings of some of the indicators onto latent variables. This suggests
factors or chronic conditions queried in the survey. High blood that there may be subtle differences in how the latent variables, such
pressure was the most common chronic condition reported. as self-efficacy for healthy eating, manifest based on gender, but the
latent variables are conceptually similar across groups.
Measurement model Two latent variables, dietary fat intake and home nutrition
environment, were each represented with a single-item indicator
There were nine instances of multicollinearity between indi- variable. Because of this, the loading of the factor on the indicator
cator variables in the measurement model. Because all of these variable was set to one and the error variance was set to zero. Prior
inter-correlations involved items intended to measure aspects of to fitting the structural model, each endogenous latent variable
the same construct and there was no problem of collinearity with multiple indicators was scaled by setting the factor loading for
between factors in the structural model, all items were kept in the one of the indicator variables to one.
model. Table 2 presents descriptive statistics for all indicators,
overall and by gender; tables of correlations for all indicators are Structural model
available by request from the authors. When compared to men,
women had a significantly lower percentage of calories from fat Before the model was tested for individual groups, the overall
(women vs. men: M ¼ 33.4 vs. 33.9) and reported lower levels of model was tested for all participants. The structural model for all
 A.K. Hermstad et al. / Social Science & Medicine 71 (2010) 93e101 97

Table 2
Descriptive statistics for indicator and outcome variables by gender.

Men Women Overall p-value

M SD M SD M SD
n¼ 243 284 527

Community & consumer nutrition environment

NEMS-S score, fat, convenience store, <1 mile (range: 0e9) 1.6 2.07 1.4 1.69 1.5 1.87 0.246
NEMS-S score, fat, convenience store, 1e5 miles (range: 0e17) 6.3 4.72 5.8 5.01 6.0 4.89 0.242
NEMS-S score, fat, grocery store, <1 mile (range: 0e23) 3.0 5.00 2.4 4.50 2.7 4.73 0.216
NEMS-S score, fat, grocery store, 1e5 miles (range: 0e26) 8.4 8.63 8.0 8.71 8.2 8.67 0.588
NEMS-R distance to sit-down restaurant (miles) 2.3 1.42 2.3 1.53 2.3 1.48 0.913
NEMS-R distance to fast-food restaurant (miles) 2.7 1.53 2.7 1.64 2.7 1.59 0.748

Perceived nutrition environment

Scale (range: 1e5) 4.0 0.98 3.8 1.15 3.9 1.08 0.061

BMI & risk factors

BMI 29.1 5.26 30.8 7.15 30.0 6.41 * 0.002
Number of disease diagnoses 1.3 1.30 1.3 1.25 1.3 1.27 0.776

Shopping & eating out behaviors (range: 1e4)

Dine at a sit-down restaurant 2.3 0.80 2.3 0.85 2.3 0.83 0.809
Dine at a fast-food restaurant 2.2 0.73 2.1 0.70 2.1 0.71 0.179
Shop at a large grocery store 3.0 0.88 3.1 0.92 3.0 0.90 0.306
Shop at a convenience store 1.7 0.88 1.7 0.91 1.7 0.89 0.726

Home nutrition environment e fat

Number of regular-fat items (range: 0e3) 2.1 1.15 1.9 1.20 2.0 1.18 * 0.026

Self-efficacy for healthy eating

Scale (range: 1e5) 3.5 0.90 3.7 0.87 3.6 0.89 0.085

Family support for healthy eating

Scale (range: 1e4) 1.9 0.77 1.8 0.80 1.9 0.78 0.589

Dietary fat intake behavior

Percent of calories from fat 33.9 2.02 33.4 2.29 33.6 2.18 * 0.009

Note: * Group difference for one-way ANOVA, p < 0.05.

participants with standardized coefficients is in Fig. 2. The model fit The models for both genders had significant paths from self-
well with the data (NNFI ¼ 0.92, CFI ¼ 0.93, 90% CI for RMSEA efficacy for healthy eating to consumer nutrition behavior (women:
[0.051, 0.060]). There were significant relationships between self- t ¼ 2.52; men: t ¼ 2.92), perceived nutrition environment (women:
efficacy for healthy eating and perceived nutrition environment, t ¼ 3.01; men: t ¼ 3.91), and family support for healthy eating
family support for healthy eating, and shopping and eating out (women: t ¼ 2.42; men: t ¼ 3.22). Only one of the constrained paths
behavior. There were also significant relationships between shop- (equal across groups) was significant, from consumer nutrition
ping and eating out behaviors with the home nutrition environ- behavior to home nutrition environment (t ¼ 2.50).
ment for fat. Finally, there were significant direct effects on dietary Additionally, women exhibited relationships among the
fat intake from both shopping and eating out behavior and the constructs that men did not. The model for women had a significant
home nutrition environment for fat. path from neighborhood nutrition environment to risk factors
The final multi-group structural model with coefficients for each (t ¼ 2.02). This model also suggests that there were significant
group is in Fig. 3. inter-relationships between women’s consumer nutrition behavior,
Gender-specific analyses without constrained parameters for their immediate food environments, and their dietary behavior.
factor loadings and parameter estimates between the predictors (i. There were significant direct effects on dietary fat intake from
e., the fully unconstrained model) showed several differences consumer nutrition behavior (t ¼ 2.20) and from the home nutri-
between women and men in the theorized paths. These differences tion environment (t ¼ 4.62). In combination with the shared
were also evident in the correlations between the theoretical significant path, consumer nutrition behavior to home nutrition
constructs (i.e., latentvariables) for the different genders. Both environment, this pattern of relationships indicates that the effect
genders exhibited relationships between self-efficacy for healthy of women’s consumer nutrition behavior on dietary intake is
eating and several other factors. A relationship between the mediated by the home food environment.
neighborhood nutrition environment and BMI and risk factors was
only found for women (r ¼ 0.13), for whom there was also a rela-
tionship between the home nutrition environment and dietary fat Discussion
intake (r ¼ 0.28). These relationships were also born out in the
structural model. This study examined the potential of a social ecological model to
The structural models support the relationships among the explain dietary fat intake in a rural population. Results support that
physical and social environments and dietary behavior. In Fig. 3, multiple levels of influence are operating to explain dietary fat
coefficients are presented for the paths that varied across groups. intake for people living in a rural community. The overall model
Paths that were fixed (i.e., equal) across groups are represented by predicted dietary fat intake well, suggesting that dietary fat intake
thin lines. Coefficients for these paths are the same as for the overall is influenced by a complex interplay of individual, social and
model. The final multi-group structural model had an adequate fit environmental factors. In addition, these relationships were found
to the data (indices presented in the box, Fig. 3). to be stronger for women than for men living in a rural community.
98 A.K. Hermstad et al. / Social Science & Medicine 71 (2010) 93e101

Table 3
Unstandardized (b) and Standardized (b) coefficients for measurement model by gender.

Factor indicator Men Women

b b SE b b SE
Community & consumer nutrition environment
NEMS-S score, fat, convenience store, <1 mile 0.80 1.61 0.37 0.70 1.19 0.51
NEMS-S score, fat, convenience store, 1e5 miles 0.30 1.39 0.91 0.52 2.67 0.73
NEMS-S score, fat, grocery store, <1 mile 1.00 4.88 0.56 2.51 0.69
NEMS-S score, fat, grocery store, 1e5 miles 0.38 3.11 0.86 0.44 3.72 0.81
NEMS-R distance to sit-down restaurant (miles) 0.66 0.91 0.22 0.88 1.38 0.18
NEMS-R distance to fast-food restaurant (miles) 0.88 1.32 0.57 0.91 1.50 0.23

Perceived nutrition environment

Ability to buy low-fat foods 1.00 1.00 1.00 1.00
Variety of low-fat foods 0.48 0.53 0.77 0.55 0.64 0.70
Ability to buy fruits and vegetables 0.40 0.41 0.84 0.51 0.56 0.74
Variety of fruits and vegetables 0.35 0.35 0.88 0.47 0.52 0.78

BMI and risk factors

BMI 1.00 1.00 1.00 1.00
Number of disease diagnoses 0.25 0.94 0.18 0.04 0.97

Shopping & eating out behaviors

Dine at a sit-down restaurant 0.01 0.01 1.00 0.05 0.06 1.00
Dine at a fast-food restaurant 1.00 1.00 1.00 1.00
Shop at a large grocery store 0.02 0.02 1.00 0.11 0.15 0.99
Shop at a convenience store 0.16 0.21 0.97 0.15 0.20 0.98

Home nutrition environment e fata

Number of regular-fat items 1.00 1.00 1.00 1.00

Self-efficacy for healthy eating

Stick to low fat, low salt foods when depressed. 1.00 1.00 1.00 1.00
Stick to low fat, low salt foods when dining with. 0.78 0.82 0.39 0.73 0.70 0.47
Eat smaller portions at dinner 0.45 0.44 0.80 0.42 0.40 0.83
Eat meatless entrees for dinner. 0.32 0.38 0.90 0.29 0.34 0.92
Substitute low or non-fat milk. 0.45 0.55 0.80 0.41 0.49 0.83
Eat poultry and fish instead of red meat. 0.57 0.54 0.68 0.63 0.63 0.60

Family support for healthy eating

Discussed eating habit changes. 1.00 1.00 1.00 1.00
Reminded you not to eat high-fat. 0.58 0.59 0.66 0.59 0.60 0.66
Commented if you went back to old. 0.45 0.43 0.80 0.49 0.48 0.76

Dietary fat intakea

Percent of calories from fat 1.00 1.00 1.00 1.00
a
Denotes a one-indicator factor. Loading is set to 1, error variance is set to zero.

Fig. 2. Overall structural model. Standardized path coefficients (b) are indicated (*p < 0.05).
 A.K. Hermstad et al. / Social Science & Medicine 71 (2010) 93e101 99

Fig. 3. Multi-group structural model. Bold paths represent paths allowed to vary across groups; thin paths represent equality constraints across groups. Standardized path coef-
ficients (b) are indicated for women (above or left of path) and men (below or right of path) for all paths that vary across groups (*p < 0.05).

We found that the home nutrition environment partially supermarkets with greater fruit and vegetable consumption
mediated the relationship between self-efficacy for healthy eating (Inagami et al., 2006; Zenk et al., 2005).
and dietary fat intake among women. Women who were confident We found no relationship between objectively measured
they could eat healthy foods reported less high-fat food in the home neighborhood nutrition environment and dietary fat intake. This is
and lower levels of dietary fat intake. Other studies have shown in contrast with Boehmer et al. (2006), who found that distance to
a direct relationship between self-efficacy with regard to fat the nearest supermarket does influence dietary habits in rural
restriction and reduced fat intake (Mosher et al., 2008) as well as areas. We did, however, observe a modest association between
the home nutrition environment and dietary intake (Patterson neighborhood nutrition environment and BMI/risk factors among
et al., 1997; Satia & Galanko, 2007). women. More research is needed to understand this relationship.
Although the relationship between self-efficacy for healthy The counties in this study have large numbers of convenience
eating and family support for healthy eating was significant, stores relative to grocery stores, and the overall choice in the food
particularly for men, family support for healthy eating was not environment is limited. This lack of variability is consistent with
associated with dietary fat intake for either gender. Others have prior research that has shown grocery stores are limited in rural
reported greater family support for healthy eating among men than areas (Liese et al., 2007; Sharkey & Horel, 2008).
women (Nothwehr et al., 2006) and a positive relationship between Compared to similar analyses that positioned dietary fat intake
social support for healthy eating and dietary behavior among men as the dependent variable, similar findings were noted. Chang et al.
(Hagler et al., 2007). Chang et al. (2008) found that women (2005) found positive relationships between individual-level
perceived their family’s lack of support as a barrier to healthy eating factors (e.g., intentions and mood) among all participants, and
and providing healthy foods for the family. It is possible that this environmental factors (e.g., food cost, accessibility to purchase
greater sense of social support for healthy eating among men is food) for the obese group only. Similar to our findings, Anderson,
because women family members often do the food shopping and Winett, and Wojcik (2000, 2007) observed a positive relationship
meal preparation for them. between self-efficacy for decreasing fat and healthy eating.
We examined the relationship of consumer nutrition behavior Anderson et al., (2007) also found that higher social support
to self-efficacy for healthy eating. Self-efficacy for healthy eating operated indirectly through self-efficacy to lower fat intake, as well
was negatively associated with consumer nutrition behavior for as gender differences similar to the present findings.
women and men. Among women, consumer nutrition behavior had
a direct effect on dietary fat intake, and operated indirectly through Limitations
the home nutrition environment. Compared to other women in the
study, women who grocery shopped more frequently and who This study has several limitations. First, the data are cross-
frequently ate at fast-food restaurants were more likely to report sectional thereby limiting our ability to make causal inferences.
high-fat foods in the home and higher levels of dietary fat intake. Second, this is one of numerous possible models that could explain
This relationship was not significant for men, perhaps because men dietary fat intake using these same variables. Third, our ability to
do less grocery shopping and food preparation and may be less operationalize variables in the model was limited to variables on
familiar with what foods are in the home (Watters & Satia, 2009). the Healthy Rural Communities survey, which was relatively short
The consumer nutrition behavior variable combined frequency to minimize respondent burden. Consequently, several key
of grocery shopping and eating out. Future studies should examine variables were assessed incompletely or not at all. For example,
the effects of frequenting different types of stores and restaurants. socio-economic status, was removed due to estimation issues
For example, eating at fast-food restaurants is associated with arising from missing data on those indicators; SES is an important
higher dietary fat intake (Satia et al., 2004) and shopping at determinant of dietary fat intake. Age and number of children in the
100 A.K. Hermstad et al. / Social Science & Medicine 71 (2010) 93e101

home were not but should be included (Anderson et al., 2000). Echeverria, S. E., Diez Roux, A. V., & Link, B. G. (2004). Reliability of self-reported
neighborhood characteristics. Journal of Urban Health, 81, 682e701.
Transportation barriers are a common issue in rural settings, so
ESRI. (2006). ArcGIS 9. Redmond, CA, USA: Environmental Systems Research
distance to resources and access to transportation should be Institute.
considered (Pucher & Renne, 2004). Fourth, the associations Glanz, K., Sallis, J. F., Saelens, B. E., & Frank, L. D. (2005). Healthy nutrition environments:
observed in this study were modest, which should be considered concepts and measures. American Journal of Health Promotion, 19, 330e333.
Glanz, K., Sallis, J. F., Saelens, B. E., & Frank, L. D. (2007). Nutrition environment
for interpretation. Last, the findings reported here may not gener- measures survey in stores (NEMS-S): development and evaluation. American
alize to rural communities with lower rates of obesity. Journal of Preventive Medicine, 32, 282e289.
Hagler, A. S., Norman, G. J., Zabinski, M. F., Sallis, J. F., Calfas, K. J., & Patrick, K. (2007).
Psychosocial correlates of dietary intake among overweight and obese men.
Implications for research American Journal of Health Behavior, 31, 3e12.
Hox, J. J., & Maas, C. J. M. (2001). The accuracy of multilevel structural equation
modeling with pseudobalanced groups and small samples. Structural Equation
This study also has implications for interventions to promote Modeling, 8, 157e174.
healthy eating. For women, food in the home directly influences Inagami, S., Cohen, D. A., Finch, B. K., & Asch, S. M. (2006). You are where you shop:
dietary fat intake. Future interventions should focus on improving grocery store locations, weight, and neighborhoods. American Journal of
Preventive Medicine, 31, 10e17.
the foods kept in the home and the subsequent impact on diet. The Johnson, R. K. (2002). Dietary intakedhow do we measure what people are really
current study affirms the need to understand the role of gender eating? Obesity Research, 10, 63Se68S.
differences in planning and implementing health promotion Jöreskog, K., & Sörbom, D. (2004). LISREL version 8.7. Lincolnwood, IL: Scientific
Software International.
programs. For example, self-efficacy operated indirectly on dietary
Kuppens, R., Eriksen, M. P., Adriaanse, H. P., Nijhuis, F. J. N., & Aaron, J. C. (1996).
fat intake for women in this sample. In a study of cancer survivors, Determinants of fat and fiber consumption in American rural energy workers.
Mosher et al. (2008) found a greater improvement in self-efficacy Preventive Medicine, 25, 212e217.
Ledikwe, J. H., Smiciklas-Wright, H., Mitchell, D. C., Jensen, G. L., Friedmann, J. M., &
for fat restriction and reduced fat intake as a percentage of calories
Still, C. D. (2003). Nutritional risk assessment and obesity in rural adults: a sex
post-intervention among women as compared to men. Interven- difference. American Journal of Clinical Nutrition, 77, 551e558.
tions should address both environmental and individual factors. Liese, A., Weis, K., Pluto, D., Smith, E., & Lawson, A. (2007). Food store types,
The results of this research suggest that focusing on modest availability, and cost of foods in a rural environment. Journal of the American
Dietetic Association, 107, 1916e1923.
changes in the environment alone as an intervention strategy may McDonald, R. P., & Ho, M. R. (2002). Principles and practice in reporting structural
not have the desired impact unless supplemented with social and equation analyses. Psychological Methods, 7, 64e82.
behavioral strategies to change consumer nutrition behavior. Thus, McLeroy, K. R., Bibeau, D., Steckler, A., & Glanz, K. (1988). An ecological perspective
on health promotion programs. Health Education Quarterly, 15, 351e377.
the design of effective interventions should include multi-level Morland, K., Diez Roux, A. V., & Wing, S. (2006). Supermarkets, other food stores,
strategies, e.g., self-efficacy for healthy eating and the home and obesity: the atherosclerosis risk in communities study. American Journal of
nutrition environment (McLeroy, Bibeau, Steckler, & Glanz, 1988). Preventive Medicine, 30(4), 333e339.
Mosher, C. E., Fuemmeler, B. F., Sloane, R., Kraus, W. E., Lobach, D. F., Snyder, D. C., et al.
By focusing on individual, family, and societal factors influencing (2008). Changes in self-efficacy partially mediates the effects of the FRESH START
dietary behavior, gender preferences for behavior change strategies intervention on cancer survivors’ dietary outcomes. Psycho-Oncology,17,1014e1023.
can be accommodated while also improving nutrition environ- National Cancer Institute (NCI). (1996). Quick food scan. http://riskfactor.cancer.
gov/diet/screeners/fat/ Retrieved August 22, 2008, from.
ments at the micro and macro levels.
Nothwehr, F., Snetselaar, L., & Wu, H. (2006). Weight management strategies
reported by rural men and women in Iowa. Journal of Nutrition Education and
Behavior, 38, 249e253.
Acknowledgements
Ogden, C. L., Carroll, M. D., Curtin, L. R., McDowell, M. A., Tabak, C. J., & Flegel, K. M.
(2004). Prevalence of overweight and obesity in the United States, 1999e2004.
Funding for this research was provided through cooperative Journal of the American Medical Association, 295, 1549e1555.
O’Hea, E. L., Wood, K. B., & Brantley, P. J. (2003). The transtheoretical model: gender
agreement #U48 DP 000043 from the Centers for Disease Control
differences across 3 health behaviors. American Journal of Health Behavior, 27(6),
and Prevention (CDC) for the EmoryPrevention Research Center 645e656.
(EPRC). Partial support for Karen Glanz’s effort came from a Georgia Patterson, P. D., Moore, C. G., Probst, J. C., & Shinogle, J. A. (2004). Obesity and
Cancer Coalition Distinguished Scholar Award. The authors physical inactivity in rural America. Journal of Rural Health, 20, 151e159.
Patterson, R. E., Kristal, A. R., Shannon, J., Hunt, J. R., & White, E. (1997). Using a brief
acknowledge the contributions of the EPRC’s Community Advisory household food inventory as an environmental indicator of individual dietary
Board to the design and conduct of this research and the Southwest practices. American Journal of Public Health, 87, 272e275.
Georgia Cancer Coalition for coordinating data collection. We also Pucher, J., & Renne, J. L. (2004). Urbanerural differences in mobility and mode choice:
Evidence from the 2001 NHTS. Bloustein School of Planning and Public Policy,
wish to thank our NEMS raters, baseline HRC2 interviewers, and RutgersUniversity.
study participants for their valuable contributions to this research. Sallis, J., Owen, N., & Fisher, E. B. (2008). Ecological models of health behavior. In K. Glanz,
B. K. Rimer, & K. Viswanath (Eds.), Health behavior and health education: Theory,
research, and practice (4th ed.). (pp. 465e486) San Francisco, CA: Jossey-Bass.
References Sallis, J. F., Grossman, R. M., Pinski, R. B., Patterson, T. L., & Nader, P. R. (1987). The
development of scales to measure social support for diet and exercise behav-
Anderson, E. S., Winett, R. A., & Wojcik, J. R. (2000). Social-cognitive determinants of iors. Preventive Medicine, 16, 825e836.
nutrition behavior among supermarket food shoppers: a structural equation Sallis, J. F., Pinski, R. B., Grossman, R. M., Patterson, T. L., & Nader, P. R. (1988). The
analysis. Health Psychology, 19(5), 479e486. development of self-efficacy scales for health-related diet and exercise behav-
Anderson, E. S., Winett, R. A., & Wojcik, J. R. (2007). Self-regulation, self-efficacy, iors. Health Education Research, 3, 283e292.
outcome expectations, and social support: social cognitive theory and nutrition Satia, J. A., & Galanko, J. A. (2007). Comparison of three methods of measuring
behavior. Annals of Behavioral Medicine, 34(3), 304e312. dietary fat consumption by AfricaneAmerican adults. Journal of the American
Boehmer, T. K., Lovegreen, S. L., Haire-Joshu, D., & Brownson, R. C. (2006). What Dietetic Association, 107, 782e791.
constitutes an obesogenic environment in rural communities? American Journal Satia, J. A., Galanko, J. A., & Siega-Riz, A. M. (2004). Eating at fast-food restaurants is
of Health Promotion, 20, 411e421. associated with dietary intake, demographic, psychosocial and behavioural
Bollen, K. A. (1989). Structural equations with latent variables. New York, NY: Wiley. factors among AfricaneAmericans in North Carolina. Public Health Nutrition, 7,
Centers for Disease Control and Prevention (CDC). (2005). Behavioral risk factor 1089e1096.
surveillance system survey data. Atlanta, Georgia: U.S. Department of Health and Sharkey, J., & Horel, S. (2008). Neighborhood socioeconomic deprivation and
Human Services, Centers for Disease Control and Prevention. minority composition are associated with better potential spatial access to the
Chang, M., Baumann, L. C., Nitzke, S., & Brown, R. L. (2005). Predictors of fat intake ground-truthed food environment in a large rural area. Journal of Nutrition, 138,
behaviors differ between normal-weight and obese WIC mothers. American 620e627.
Journal of Health Promotion, 19, 269e277. U.S. Census Bureau. (2000). Urban and rural: total population. Retrieved
Chang, M., Nitzke, S., Guilford, E., Adair, C. H., & Hazard, D. L. (2008). Motivators July 1, 2008, from. http://www.factfinder.census.gov/servlet/DTTable?_bm¼y&-
and barriers to healthful eating and physical activity among low-income context¼dt&-ds_name¼DEC_2000_SF1_U&-mt_name¼DEC_2000_SF1_U_P002&-
overweight and obese mothers. Journal of the American Dietetic Association, CONTEXT¼dt&-tree_id¼808&-redoLog¼true&-geo_id¼01000US&-search_results¼
108, 1023e1028. 01000US&-format¼&-_lang¼en.
 A.K. Hermstad et al. / Social Science & Medicine 71 (2010) 93e101 101

U.S. Department of Agriculture (USDA), Agricultural Research Service, Beltsville Human Watters, J. L., & Satia, J. A. (2009). Psychosocial correlates of dietary fat intake in
Nutrition Research Center, Food Surveys Research Group (Beltsville, MD). (1996). AfricaneAmerican adults: a cross-sectional study. Nutrition Journal, 8(15).
Continuing survey of food intakes by individuals 1994e1996. Retrieved May 5, Williams, A., Wold, J., Dunkin, J., Idleman, L., & Jackson, C. (2004). CVD prevention
2006, from USDA Agricultural Research Service, Food Surveys Research Group strategies with urban and rural AfricaneAmerican women. Applied Nursing
website. http://www.ars.usda.gov/SP2UserFiles/Place/12355000/pdf/Hhold.pdf. Research, 17, 187e194.
U.S. Department of Health and Human Services (USDHHS). (2000). Healthy people World Cancer Research Fund / American Institute for Cancer Research (WCR/AICR).
2010: Understanding and improving health (2nd ed.). Washington, DC: U.S. (2007). Food, nutrition, physical activity, and the prevention of cancer: A global
Government Printing Office. perspective. Washington, DC: AICR.
Walker, S. N., Pullen, C. H., Hertzog, M., Boeckner, L., & Hageman, P. A. (2006). Zenk, S. N., Schulz, A. J., Hollis-Neely, T., Campbell, R. T., Holmes, N., Watkins, G.,
Determinants of older rural women’s activity and eating. Western Journal of et al. (2005). Fruit and vegetable intake in AfricaneAmericans: income and
Nursing Research, 28, 449e468. store characteristics. American Journal of Preventive Medicine, 29, 1e9.