Development of an Instrument Measuring Perceived Environmental Healthfulness: Behavior Environment Perception Survey (BEPS)

ARTICLE IN PRESS Research Article Development of an Instrument Measuring Perceived Environmental Healthfulness: Behavior Environment Perception Survey (BEPS) Jade McNamara, PhD1,y; Melissa D. Olfert, DrPH, RDN2; Morgan Sowers, MBA, RDN3; Sarah Colby, PhD, RDN3; Adrienne White, PhD, RDN, FAND1; Carol Byrd-Bredbenner, PhD, RDN, FAND4; Kendra Kattelmann, PhD, RDN5; Lisa D. Franzen-Castle, PhD, RDN6; Onikia Brown, PhD, RDN7; Tandalayo Kidd, PhD, RDN8; Karla P. Shelnutt, PhD, RDN9; Tanya Horacek, PhD, RDN10; Geoffrey W. Greene, PhD, RDN11 ABSTRACT Objective: To create a tool to measure college students’ perception of the healthfulness of their environment. Design: (1) Item generation, (2) cognitive interview testing and exploratory factor analysis, (3) item refinement/modification, (4) factor structure validation, and (5) criterion validation. Setting: Ten college campuses. Participants: Time point 1 (n = 120 cognitive interviews; n = 922 factor analysis); time point 2 (n = 2,676), convenience sample of undergraduate students. Main Outcome Measures: Cognitive interviews and survey data were used to assess perceptions about the environment. Analysis: Exploratory factor analysis, structural equation confirmatory factor analysis, correlations, and regressions. Results: Item generation resulted in 93 items. Items were eliminated based on cognitive interviews, exploratory factor analysis of pilot data, and elimination of cross-loading or weak loading items. In confirmatory analyses, a 21-item, 5-factor structure was the best fit for the data (x2 = 3,286.77, degrees of freedom = 189; comparative fit index = 0.840; root-mean-square error of approximation = 0.078). Environmental factors include physical activity (a = 0.68, 4 items), healthful eating (a = 0.86, 5 items), mental health (a = 0.85, 5 items), barriers to healthful eating (a = 0.70, 4 items), and peer influences (a = 0.73, 3 items). There were significant associations between scales and validation criteria (P < .05). Conclusions and Implications: The Behavior Environment Perception Survey is a novel instrument measuring perceptions of the healthfulness of the campus environment. Strengths include a development process involving 10 different universities, strong psychometric properties, and breadth of constructs. Key Words: environment, health behavior, perception, survey methodology, young adult (J Nutr Educ Behav. 2019;000:1−10.) Accepted September 4, 2019.

1

School of Food and Agriculture, University of Maine, Orono, ME Department of Human Nutrition and Foods, West Virginia University, Morgantown, WV 3 Department of Nutrition, University of Tennessee, Knoxville, TN 4 Department of Nutritional Sciences, Rutgers, The State University of New Jersey, New Brunswick, NJ 5 Department of Health and Nutritional Sciences, South Dakota State University, Brookings, SD 6 Nutrition and Health Sciences Department, University of Nebraska-Lincoln, Lincoln, NE 7 Department of Nutrition, Dietetics, and Hospitality Management, Auburn University, Auburn, AL 8 Department of Food, Nutrition, Dietetics, and Health, Kansas State University, Manhattan, KS 9 Department of Family, Youth, and Community Sciences, University of Florida, Gainesville, FL 10 Department of Public Health, Food Studies, and Nutrition, Syracuse University, Syracuse, NY 11 Department of Nutrition and Food Sciences, University of Rhode Island, Kingston, RI yDr McNamara was affiliated with University of Rhode Island at the time this study was completed. Conflict of Interest Disclosure: The authors have not stated any conflicts of interest. Address for correspondence: Geoffrey W. Greene, PhD, RDN, Department of Nutrition and Food Sciences, University of Rhode Island, 125 Ranger Hall, 41 Lower College Rd, Kingston, RI 02881; E-mail: [email protected] Ó 2019 Society for Nutrition Education and Behavior. Published by Elsevier Inc. All rights reserved. https://doi.org/10.1016/j.jneb.2019.09.003 2

Journal of Nutrition Education and Behavior  Volume 000, Number 000, 2019

1

ARTICLE IN PRESS 2

McNamara et al

INTRODUCTION There is increasing awareness of the impact of environmental influences on health outcomes.1,2 College students are a group with historically unhealthful dietary behaviors and decreasing physical activity (PA), as well as a population that reports high levels of perceived stress associated with weight gain and increased risk for chronic disease.3 College is a vulnerable period for weight gain; students gain 3−4.3 kg in their first year,4−6 and it is estimated that about one third of college students are obese.7 Researchers showed that this rapid weight gain during college is associated with increased sedentary behavior and poor diet quality,8 both of which are influenced by the campus environment.9,10 College is a unique time point in life because it is often a place where students are both working and living, which heightens the need to create environments that facilitate healthy behaviors. Environmental influences that contribute to overall health include the built environment, peer support, access to programs, and policy initiatives.11 These different environmental influences are associated with PA (eg, neighborhood safety, opportunities for activities, peer influences [PIs]) and healthy eating (eg, availability of healthy food, perceived high cost of healthy food, time available to prepare food, and PIs).12 College students living on campus may be particularly susceptible to the characteristics of their environment because of limited transportation, financial constraints, PIs, the stresses of college courses, and living in a new environment.13 Previously, the healthfulness of college campus environments was examined with a specific focus on the domains of access and availability as well as resources and policy. Horacek et al14 measured the built environment of 13 college recreation facilities, which included both access and availability to the facilities and an audit of PA resources and policies on the campuses. They concluded that whereas campus recreation centers were available and accessible to students, none of the campuses that were audited had a PA policy.

Journal of Nutrition Education and Behavior  Volume 000, Number 000, 2019 Similarly, Horacek et al15 evaluated dining facilities on 15 college campuses and found that healthy options were available to students, but there were also barriers to making healthy choices (eg, large portion sizes and all-you-can-eat dining halls) and few policies to secure the establishment as a healthful eating (HE) environment. Students often report that their built environment influences their behaviors. Greaney et al16 interviewed 115 college students about barriers to and enablers of healthful weight management and found that themes emerged related to perceived environmental circumstances, such as access to food served in the cafeteria, high availability of unhealthful food, and the high cost of healthful food. In that study, similar themes emerged regarding PA, in which students reported that PIs made PA more enjoyable, but barriers such as perceived time constraints prevented students from participating in PA. Nelson et al17 found similar results when asking students what they perceived to be determinants of weight gain in college. Students identified the availability of unhealthful food on campus, increased stress, and negative experiences with campus recreation facilities as determinants of weight gain. The studies underscored the need to capture students’ perceptions about access and availability, PIs, and resources and policies for HE, PA, and mental health (MH) on college campuses. Capturing perceived environmental factors will allow researchers to tailor health interventions better to address factors that students perceive enhance or impair their likelihood of engaging in healthful practices. Previous tools developed to measure perceptions of the college health environment include the Behavior, Environment, Changeability Survey (BECS)18 and the College Environment Perception Survey (CEPS).19 The BECS was designed to assess the perceived importance of environmental changeability (eg, access to healthier food environments and safe places to be physically active); CEPS was developed to mirror factors related to the built environment (eg, access to water, vending machine

options, availability of healthy foods, and opportunities to be physically active). The BECS and CEPS are both reliable and valid tools that capture different components of students’ perceptions accurately by evaluating access and availability. However, they do not capture PIs and resource or policy perception on college campuses that affect behavioral choices. Therefore, it is necessary to expand on these instruments and develop a more comprehensive instrument to assess additional domains of the campus environment. Items from the BECS and CEPS were used to develop the current survey, the Behavior Environment Perception Survey (BEPS), with additional items that measure constructs of MH, PIs, and resources and policy domains. The purpose of this instrument development study was to expand on previous tools to create a broader measure of college students’ perception of environmental constructs and domains that influence health.

METHODS Overview of BEPS Instrument Development A 5-step process of item development was implemented: (1) initial item generation, (2) cognitive interview testing and exploratory factor structure analysis, (3) item refinement and modification, (4) factor structure validation, and (5) criterion validation comparing factors with health behaviors in a separate sample.20 The researchers hypothesized that BEPS would (1) include scales assessing the constructs of HE, PA, and MH as well as the domains of access and availability, PIs, and resources and policy; and (2) have a valid psychometric structure demonstrated through internal reliability (coefficient a > 0.7), validate the latent constructs through strong factor loadings, and demonstrate criterion validity compared with measures of HE, PA, and MH.21

Sample Three independent samples at 2 stages were used in this study: 2 samples for stage 1 activities (ie, cognitive

ARTICLE IN PRESS Journal of Nutrition Education and Behavior  Volume 000, Number 000, 2019 interviews and exploratory factor analysis [EFA] of survey data) and 1 sample for stage 2 activities (ie, confirmatory factor analysis [CFA] of survey data and validation analyses) (Figure 1). All were convenience samples of college students aged 18−24 years and recruited from a group of 6 or 7 land grant universities across the US (University of Rhode Island, West Virginia University, University of Maine, Rutgers University, South Dakota State University, University of Tennessee, and University of Florida). Students were recruited through classes, campus announcements, and e-mails. Each university used different incentives to recruit students, including extra credit, raffles, and gift cards. Informed consent and data collection were provided verbally for the cognitive interviews and online for the surveys. The Institutional Review Board for the protection of human subjects at each university (University of Rhode Island, West Virginia University, University of Maine, Rutgers University, South Dakota State University, University of Tennessee, and University of Florida) approved the study. Students were excluded if they were aged >24 years or <18 years or were pregnant. For testing and administering the surveys, students were recruited at stage 1 (sample 1; n = 120)

Item generation fall 2016

from 7 universities to participate in cognitive interviews, along with a separate sample of students (sample 2; n = 922) from 6 universities to complete the initial BEPS for EFA. Students were recruited for stage 2 (sample 3; n = 2,676) from 7 universities, collected in fall, 2017 and spring, 2018, for CFA and criterion validation analyses. A subsample of participants from stage 2 was used for criterion validation that had complete BEPS, behavior, and anthropometric data (n = 2,209).

Item Generation and Cognitive Interviews The 5-stage process of development began with item generation (fall, 2016). Initial item generation was based on existing instruments, BECS and CEPS. New items were generated by teams of undergraduate and graduate students as well as researchers from 10 universities using roundtable brainstorming discussions. A total of 100 items were selected (n = 15) or modified from existing instruments (n = 3) or were newly written (n = 82). Items were then sorted and reviewed by experts at the 10 universities by construct (HE, PA, and MH) and domains of those constructs (access and availability, peers, and resources

Round table discussions and Q sort

Sample 1: Cognitive interviews

Development process of BEPS

Stage 1 Data collection spring 2017

McNamara et al

Sample 2: Exploratory factor analysis

3

and policy) using a modified Q-sort procedure.22 After the Q-sort, 7 items were discarded because they fit on >1 construct. This left 93 items selected to represent 3 constructs of environmental healthfulness: HE, PA, and MH. All 3 constructs had items that represented the domains of access and availability, PI, and resources and policy. A 5-point Likert response scale was used, from 1 (strongly disagree) to 5 (strongly agree). The survey was administered using Qualtrics software (Provo, UT, 2018). During stage 1 of data collection, cognitive interviews (n = 120) were conducted at 7 universities. Students were excluded if they were health majors (eg, nutrition, kinesiology, premedical, nursing) to include only students with limited knowledge and potentially less exposure to health behavior terminology (eg, food preparation equipment).23 Interviews were stratified by gender to assess the appropriateness and clarity of items. All cognitive interviews followed a standardized data collection protocol that included a script with prompts to ensure uniformity.24 Cognitive interviews were recorded and transcribed for thematic analysis to assess the face validity of the items.25 Items that participants deemed unclear were eliminated or rewritten to

10 universities

7 universities (n = 120)

6 universities (n = 922)

URI, WVU, UME, RU, SDSU, UT, KSU, AU, UF, UNL

URI, WVU, UME, RU, SDSU, UT, UNL

URI, WVU, UME, RU, SDSU, UT

Sample 3: Confirmatory factor analysis Stage 2 Data collection fall 2017 and spring 2018

7 universities (n = 2 676)

URI, WVU, UME, RU, UT, UF, SU

Sample 3: Criterion validation analyses

Figure 1. Stepwise process of item generation and psychometric property testing of the Behavior Environment Perception Survey (BEPS) among 10 universities. AU indicates Auburn University; KSU, Kansas State University; RU, Rutgers University; SDSU, South Dakota State University; UF, University of Florida; UME, University of Maine; UNL, University of Nebraska-Lincoln; URI, University of Rhode Island; UT, University of Tennessee; WVU, West Virginia University.

ARTICLE IN PRESS 4

McNamara et al

improve clarity using explicit feedback (ie, change the word cafeteria to dining hall) before the stage 2 survey was administered.

Exploratory and Confirmatory Factor Analyses The researchers conducted EFA using maximum likelihood with promax and varimax rotations for more replicable results.26 The ideal number of factors was identified using minimum average partial test21; items were eliminated based on crossloadings (> [0.30]) or weak loadings (< [0.40]).27 Internal consistency was also addressed. Items were removed if there was the potential for redundancy or multicollinearity (r > 0.90)21 or if the subscale would be enhanced if the item were deleted (Cronbach a).27 All analyses were conducted using SPSS software (version 24, IBM Corp, Armonk, NY, 2016). Confirmatory factor analysis structural equation modeling was conducted with EQS for Windows (version 6.0, Multivariate Software, Inc., Encino, CA, 2006)28 using a separate sample. CFA model fit indices were compared with macrolevel fit indices that indicated an excellent model fit: the chi-square to degrees of freedom ratio (4:1), comparative fit index (CFI) > 0.90, and rootmean-square error of approximation (RMSEA) < 0.05.21 Model fit for the final BEPS instrument was assessed and compared with the model fit indices.

Demographic Variables Self-reported age, gender, and university were assessed in the exploratory and confirmatory surveys. Students also reported whether they lived on or off campus. For the demographic items, means were generated for continuous variables, and frequencies for categorical variables.

Criterion-Related Validation Measures Behavior scales to assess HE, PA, and MH were selected for criterion validation because the scales were hypothesized to be significantly correlated

Journal of Nutrition Education and Behavior  Volume 000, Number 000, 2019 with the BEPS factors. The Dietary Screener Questionnaire was administered to assess HE. The Dietary Screener Questionnaire is a 26-item self-report food frequency questionnaire that collects dietary intake data from the past 30 days.29 Cups of fruits and vegetables per day and teaspoons per day of sugar from sugarsweetened beverages were used to validate criteria for this study to compare behaviors with the perception of HE on campus, and with the other BEPS scales. The International Physical Activity Questionnaire, a validated instrument that captures estimates of selfreported PA over the past 7 days, was administered to assess PA.30 PA estimates for the current study focused on total minutes per week of activity to compare behavior with the perception of PA opportunities on the college campus, along with the other BEPS scales. The Centers for Disease Control and Prevention’s Healthy Days Symptom Module31 was used to collect overall quality-of-life data from the sample. This instrument includes items about mental and physical health over the past 30 days and how many days per month mental and physical health affected daily activities. One item from the instrument was used as a continuous variable to capture students’ MH status: During the past 30 days, for about how many days have you felt very healthy and full of energy? This item was compared with the MH perception scale to assess relations between MH and perceptions of MH access, availability, and resources on the college campus environment, as well as the other BEPS scales. The researchers calculated body mass index (BMI) using students’ selfreported height (in inches) and weight (in pounds) with the equation: 703.1 £ (weight [lb] / height [in]2). Body mass index was used as a continuous variable to compare self-reported anthropometric data with perceptions of health on the college campus. A BMI of 18.5−24.9 was considered normal weight, 25−29.9 was considered overweight, and ≥30 was considered obese.32 The researchers first tested BEPS items and self-reported behaviors for distribution and central

tendency. Pearson correlation analysis was conducted to assess the relation between the BEPS scales and selfreported behaviors. Hierarchical linear regressions were also conducted to determine whether the independent BEPS scales significantly predicted college students’ reported behaviors. At step 1 of the regression, demographic variables were entered (gender, age, and university); at step 2, 1 of the BEPS scales was entered as the independent variable to determine how well each scale predicted each behavior after controlling for demographic variables. All analyses were performed using the SPSS software.

RESULTS Sample Characteristics Stage 1 cognitive interview participants (n = 120) were evenly split between women (50%) and men (50%) and aged 18−24 years. Stage 1 survey participants (n = 922) were mostly female (70.1%) and lived on campus (55.1%) and had a mean age of 19.8 (§ 1.4) years (Supplementary Data). Stage 2 participants (n = 2,676) were mostly female (71.1%) and lived off campus (64.1%) and had a mean age of 20 (§ 1.4) years. Using stage 2 data, a subsample (n = 2,209) was used to validate criteria. For this, participants were primarily female (72.1%), with an average age of 20 (§ 1.4) years; the majority reported living off campus (69.3%).

Stage 1 Results Of the 93 items generated for the BEPS, 17 were removed (1) because of cognitive interview feedback or (2) because the item had ≥30 missing responses on the survey and/or similarly worded items were significantly (P < .05) correlated. Based on the cognitive interview data, items were removed if students considered them to be multifaceted or if the items were difficult to rephrase in students’ own words. Cognitive interview data also were employed to improve the clarity of items and to use more common vernacular (eg, dining hall instead of cafeteria).

ARTICLE IN PRESS Journal of Nutrition Education and Behavior  Volume 000, Number 000, 2019

McNamara et al

5

Table 1. Descriptive Statistics and Internal Consistencies of BEPS by Sample Time Points Exploratory Sample (n = 922) BEPS Factors

Mean (SD)

Physical activity Healthful eating Mental health Peer influences Barriers to healthful eatingb Meal preparation BEPS total scorec

Cronbach aa

4.1 (0.54) 3.2 (0.63) 3.8 (0.60) 3.4 (0.76) 2.8 (0.79) 3.7 (1.23) 3.5 (0.44)

0.82 0.68 0.78 0.75 0.67 0.87 0.71

Confirmatory Sample (n = 2,676) Mean (SD)

Cronbach aa

4.1 (0.57) 2.9 (0.83) 3.8 (0.69) 3.3 (0.81) 2.9 (0.83)

0.68 0.86 0.85 0.73 0.70

3.4 (0.51)

0.87

BEPS indicates Behavior Environment Perception Survey. a Cronbach a: 0.60−0.70 = moderate; 0.71−0.80 = good21; and 0.81−0.9 = very good; bBarriers to healthful eating items 16−18 were reverse-scored (1 = strongly agree and 5 = strongly disagree). A lower score indicates more perceived barriers; cBEPS total score = (physical activity + healthful eating + mental health + barriers to healthful eating + peer influences) / 5. Notes: The exploratory sample was from stage 1 data collection, used for exploratory factor analysis. The confirmatory sample was from stage 2 data collection, used for confirmatory factor analysis. Responses were scored using a Likert scale in which 1 = strongly disagree and 5 = strongly agree. Empty cells for meal preparation under confirmatory sample indicate factor not included in final model. The researchers performed EFA using the 76 items remaining in the survey. Items were removed that did not load on 1 of the identified 6 factors (n = 22) or had a low loading (<0.40) or cross-loading (>0.30) (n = 23); this resulted in a 6-factor structure with 31 items. The 6 factors were PA (8 items), HE (6 items), MH (5 items), PIs (4 items), barriers to HE (BHE) (6 items), and meal preparation (MP) (2 items) (Table 1). The PA scale captured students’ perceptions about the opportunities and availability of being physically active on campus. The HE scale assessed perceptions about the availability and access to healthful foods on campus. The MH scale asked about the resources and access to supports around campus related to stress, abuse, and emotional or psychological problems. The BHE scale measured perceptions that made healthy eating more difficult, such as time management and stress. Three of the 4 items that comprised the BHE scale were reverse-scored, meaning that a lower score represented greater perceived BHE. The PI scale assessed the perceptions of the role that peers had in HE and PA. Finally, the MP scale asked about perceptions related to preparing meals where students lived. The overall BEPS mean was calculated by adding the PA, HE, MH, PI, BHE, and MP factors together and then dividing by 6 to reach a mean score.

Stage 2 Results The 31-item instrument underwent CFA using a separate sample. After the researchers reviewed the CFA results, 2 items were deleted based on further expert review, 4 because of low loadings (<0.40),21 and 2 items because of a similarity to other items; the MP factor was removed owing to the limited breadth the 2 items provided in fully identifying the factor. A reduction of items to a 21-item, 5-factor model was determined to be the best fit for the instrument based on macrolevel results of the CFA (chi-square ratio, CFI, and RMSEA), item loadings, Cronbach a for scale improvement, and further expert review of the items. The final 5-factor model included PA (4 items), HE (5 items), MH (5 items), BHE (4 items), and PIs (3 items) (Figure 2). The final model fit indices were as follows: x2 = 3,286.77, degrees of freedom = 189; CFI = 0.840; and RMSEA = 0.078. All factors showed adequate internal reliability with Cronbach a of 0.68−0.85.21 Mean BEPS score for the final instrument was 3.41 (§ 0.51) out of a possible score of 5, with an overall Cronbach a of 0.87. For the final list of items and total BEPS scoring formula, see Supplementary Data.

Criterion Validation Scales to establish the validity of criteria were included during stage 2 using

a subsample for BEPS and criterion validation instruments (n = 2,209) (Table 2). All behavioral items and BEPS factors were normally distributed, as measured by skewness and kurtosis. Average fruit and vegetable intake was 2.3 (§ 0.9) cups/d, and an average of 2.2 (§ 0.9) tsp/d of sugar came from sugar-sweetened beverages. Participants reported an average of 132.6 (§ 85.3) min/wk of total exercise and an average of 8.9 (§ 9.0) d/mo of feeling happy and full of energy. Mean BMI was 24.2 (§ 4.3); 22.4% of participants were overweight and 11.6% were obese. After controlling for university, age, and gender, the Pearson correlations indicated that there were significant positive associations between the behavior items and the BEPS PA, HE, PI, and reverse-scored BHE factors (Table 3). The MH factor had a significant positive relation to the number of days feeling healthy and full of energy and a negative correlation with PA (metabolic equivalent [MET] minutes per week) (Table 3). In addition, BMI was significantly negatively correlated with the PA, reverse-scored BHE, and PI factors (Table 3). Before conducting the hierarchical regressions, the researchers calculated intraclass correlation coefficients for the dependent behavior items and university to assess whether university should be considered a nesting variable. All intraclass correlation coefficients were <0.05, indicating that

ARTICLE IN PRESS 6

McNamara et al

a nested model using university was unnecessary. 21 Results from hierarchical linear regression analysis indicated that after considering university, gender, and age, the individual BEPS factors of PA, HE, reversescored BHE, and PI were significant predictors of fruit and vegetable intake, PA, sugar from sugar-sweetened beverages, and healthy/full-ofenergy days (Table 4). The MH factor was a significant predictor of the number of healthy/full-of-energy

Journal of Nutrition Education and Behavior  Volume 000, Number 000, 2019 days and MET minutes of PA per week. In addition, BMI was significantly predicted by PA, BHE, and PI.

DISCUSSION The aim of this research was to develop and validate the BEPS instrument as a comprehensive measure of students’ perceptions about the healthfulness of their environment. The BEPS builds on previous instruments and provides a broader measure

There are sports (intramural or club) available to play on campus.

0.55

Physical Activity α = 0.68

that includes 5 environmental constructs that affect the overall health of college students: PA, HE, MH, BHE, and PIs. Items within the instrument represent the domains of access and availability as well as resources and policy, to capture an extensive overall picture of college students’ perceptions of the healthfulness of their environment. The rigorous instrument development process showed strong psychometric properties, and items were validated using a separate sample.

There are plenty of opportunities on campus to be moderately or vigorously active.

0.69 0.52

I feel welcome to use the recreation center on campus. 0.6

2

I see people being physically active on campus.

9

It is easy to find healthy foods on campus.

0.70

It is easy to find fruits and vegetables on campus.

0.7

Healthful Eating α = 0.86

0.82

There are a variety of healthy foods available on campus. 0.8

1

My campus makes it easy to eat healthy. 0.5

8

It is easy to live a healthy lifestyle while living on campus.

4

0.7

0.82

Mental Health α = 0.85

0.76

There are programs on campus that offer stress management.

My campus has a system of support for emotional or psychological problems.

There are resources on campus for a person who needs help managing stress.

0.6

4

0.

68

There are resources on campus for a person who is in an abusive relationship.

I can get an appointment with a mental health professional.

Figure 2. Factor structure by subscale for 21-item Behavior Environment Perception Survey.

ARTICLE IN PRESS Journal of Nutrition Education and Behavior  Volume 000, Number 000, 2019

0.57

Barriers to Healthful Eating α = 0.70

0.59

McNamara et al

7

My class schedule makes it easy to eat healthy meals.

It is hard to eat healthy because of all the stress at school.

0.69

I do not have enough time to pack healthy snacks for myself. 0.5 9

I cannot afford to eat healthy.

0 0.5

Peer Influence α = 0.73

0.75

The people I eat with make it easy to choose healthy foods.

Friends motivate me to workout.

0.8 4

Friends have a positive influence on my physical activity.

Figure 2. Continued.

Table 2. Self-Reported Dietary Behavior and BMI Using a Subsample of Stage 2 Participants With Complete Behavior Environment Perception Survey and Criterion Data (n = 2,209) Characteristic

Mean (SD)

Age, y BMIa,b Fruit and vegetable intake, cups/dc Sugar from sugar-sweetened beverages, tsp/dc Physical activity, total min/wkd Days of feeling happy/healthy, d/moe

20.3 (1.4) 24.2 (4.3) 2.3 (0.9) 2.2 (0.9) 132.6 (85.3) 8.9 (9.0)

BMI indicates body mass index. a Calculated using self-reported height (inches) and weight (pounds) by the equation: 703.1 £ weight (lb) / height (in)2; bMissing data for BMI (n = 2,186); cMeasured with the Dietary Screener Questionnaire28; dMeasured with the International Physical Activity Questionnaire29; eMeasured with the Centers for Disease Control and Prevention’s Healthy Days Symptom Module30.

As 5 factors that captured different health domains, the BEPS showed strong criterion validity with measures of PA, fruit and vegetable intake, sugar from sugar-sweetened beverages, and the number of days students felt healthy and full of energy. This instrument has applicability for researchers who are interested in measuring students’ perceptions of the healthfulness of their college environment. Development of the BEPS was based on previous instruments

constructed and tested to measure college students’ perception of their environment.18,19 However, the BEPS expands on environmental perceptions that were not a part of the scope of those tools. Through the input of experts and graduate and undergraduate students from 10 different universities, items were generated and evaluated as important health perceptions to capture on college campuses and were then fit into specific domains and constructs. Using this approach to generate items resulted in

a diverse assortment for the 5 constructs and domains of the BEPS. After undergoing a rigorous 5-step process of instrument development, the BEPS instrument was found to be psychometrically sound, as evidenced by the macrolevel fit indices, good factor loadings, and reliability assessment. Instrument validity was also demonstrated through the significant correlations and linear relations between the scales and corresponding behavior scales, along with anthropometric data. This 21-item instrument provides a large breadth of items for evaluating college students’ perceptions of their environment. Few other researchers assessed college students’ perceptions of the healthfulness of their environment or investigated how those perceptions relate to self-reported behaviors and anthropometrics.18,19 Results from this research indicated that overall, students perceived their environment to be conducive to positive health behaviors; an average overall BEPS score was 3.41 (§ 0.51) out of a possible score of 5, indicating that students were neutral about whether their environment was healthy. When the researchers examined the means of the different constructs, the PA environment was perceived to be

Sugar From Sugar-Sweetened Beverages, tsp/da

Fruit and Vegetable Intake, cups/da

Factors Barriers to healthful eating Healthful eating Mental health Physical activityf Peer influences

0.06 0.08 0.03 0.11 0.09

Physical Activity, Metabolic Equivalent, min/wkb

0.06 0.09 0.03 0.13 0.11

Healthy and Full of Energy, d/moc

0.11 0.11 0.05 0.13 0.20

Anthropometric Measure BMId,e 0.08 0.04 0.00 0.12 0.13

0.17 0.14 0.12 0.30 0.17

P value <.001 .09 .91 <.001 <.001

Self-Reported Anthropometrics

Self-Reported Behavior Physical Activity, Metabolic Equivalent, min/wkb

Fruit and Vegetable Intake, cups/da BEPS Scales Physical activity Healthful eating Mental health Barriers to healthful eatingf Peer influences

Adjusted R2

b

0.06 2.70 .01 0.09 3.96 <.001

0.06 0.07

0.12 0.11

0.03 1.43 .15 0.11 5.13 <.001

0.06 0.07

0.09 4.14 <.001

0.07

b

T

P value

T

P value

Adjusted R2

Sugar From Sugar-Sweetened Beverages, tsp/da b

T

P value

Healthy and Full of Energy, d/moc

Adjusted R2

b

T

P value

BMId,e

Adjusted R2

b

T

P value

Adjusted R2

5.41 <.001 4.94 <.001

0.05 0.05

0.06 3.02 <.001 0.09 4.30 <.001

0.06 0.06

0.17 0.15

7.99 <.001 6.76 <.001

0.05 0.04

0.08 3.64 <.001 0.04 1.67 .10

0.02 0.02

0.05 2.26 .02 0.13 6.06 <.001

0.04 0.06

0.03 1.62 .11 0.13 6.41 <.001

0.05 0.07

0.12 5.80 <.001 0.29 14.37 <.001

0.04 0.11

0.01 .43 .67 0.12 5.76 <.001

0.00 0.03

9.40 <.001

0.08

0.10 4.95 <.001

0.06

0.17

8.17 <.001

0.05

0.13 5.98 <.001

0.03

0.19

BEPS indicates Behavior Environment Perception Survey; BMI, body mass index; R2, multivariate coefficient. a Measured with the Dietary Screener Questionnaire28; bMeasured with the International Physical Activity Questionnaire29; cMeasured with the Centers for Disease Control and Prevention’s Healthy Days Symptom Module30; dCalculated using self-reported height (inches) and weight (pounds) by the equation: 703.1 £ weight (lb) / height (in)2; e Missing data for BMI (n = 2,186); fBarriers to healthful eating items 16−18 were reverse-scored (1 = strongly agree and 5 = strongly disagree). A lower score indicates more perceived barriers. Notes: The 5 BEPS scales were used independently in the hierarchical liner regression to predict each behavior and BMI. Demographic variables (gender, age, and university) were entered in step 1; for step 2, 1 of the BEPS scales was entered. Self-reported behavior and BMI were used as the dependent variables in each of the regression analyses for each scale. P < .05 was considered statistically significant.

ARTICLE IN PRESS

Table 4. Hierarchical Regression Results for BEPS Scales Predicting Behavior or BMI Using Subsample With Complete BEPS and Criterion Data for Stage 2 Participants (n = 2,209)

Journal of Nutrition Education and Behavior  Volume 000, Number 000, 2019

BEPS indicates Behavior Environment Perception Survey; BMI, body mass index; tsp, teaspoon. a Measured with the Dietary Screener Questionnaire28; bMeasured with the International Physical Activity Questionnaire29; cMeasured with the Centers for Disease Control and Prevention’s Healthy Days Symptom Module30; dCalculated using self-reported height (inches) and weight (pounds) by the equation: 703.1 weight (lb) / height (in)2; eMissing data for BMI (n = 2,186); fBarriers to healthful eating items 16–18 were reverse-scored (1 = strongly agree and 5 = strongly disagree). A lower score indicates more perceived barriers. Note: P < .05 was considered statistically significant.

McNamara et al

Behavior Measures

8

Table 3. Correlations Among the 5 BEPS Factors and Behavior Measures Using a Subsample With Complete BEPS and Criterion Data From Stage 2 Participants (n = 2,209)

ARTICLE IN PRESS Journal of Nutrition Education and Behavior  Volume 000, Number 000, 2019 the highest regarding healthfulness, and students perceived their eating environment to be the least healthy. Barriers to HE contributed to the low overall mean score. Although negatively worded items were reversescored for the BHE subscale (strongly agree = 1 instead of 5), the low score on this scale suggests that students perceived significant BHE to exist in their environment. This finding is similar to the qualitative findings of Greaney et al16 and Nelson et al.17 The significant correlations between the BEPS and behavior scales were consistent with those found by Schembre et al,20 who developed and tested an instrument using similar methods. The criterion analysis also suggested significant relations between the BEPS factors and selfreported behavior and BMI. Based on the regression results, the individual factors of BEPS were significant predictors for different health behaviors. For example, the BHE factor significantly predicted MET minutes of PA, cups of fruit and vegetable intake, number of healthy days per month that were full of energy, and BMI. In contrast, the MH factor was significant in predicting the number of healthy days per month that were full of energy but was negatively correlated with PA (MET minutes per day). The MH factor is especially relevant to today’s college students. With issues related to MH growing significantly over the past 5 years33 and an estimated 12% to 18% of students in higher education diagnosed with an MH disorder,34 students’ perceptions of MH access and resources on college campuses are an important environmental factor to measure. In further instrument validation, researchers might explore the MH factor and how it relates to other self-reported behaviors such as stress management or self-efficacy in dealing with stress. Two factors emerged from the analysis that were not originally expected: BHE and PI. BHE emerged as an independent factor separate from the HE scale. Previously, researchers found that barriers such as time management and not having enough time in general affected eating behavior in college students,35 which may explain why this factor emerged as an

independent scale. Moreover, PI was intended to be a domain in each scale, but through EFA and CFA, it emerged as an independent scale. This scale has items that relate to both HE and PA. In previous research, friends were shown to have an important role in PA36 and food choices through PI and peer pressure.37 Assessing college students’ perceptions of PI in their environment may prove to be an important mediating factor for making and changing health behaviors. Further validation of the BHE and PIs could be investigated to better understand how they relate to health decisions. Study results were strengthened by the recruitment of a large sample from multiple universities across the US, which allowed for greater power in analyzing the psychometric properties. Furthermore, involving multiple perspectives in item generation led to the development of a diverse instrument to measure multiple factors related to the healthfulness of college campuses. The 5-step development process used, with separate sample collection for EFA and CFA, allowed for greater confidence in the robustness of the BEPS. Future validation studies are needed to assess the association between BEPS and other behavioral measures, as well as prospective studies to assess change in scores in response to changes in environmental factors.

McNamara et al

9

properties to assess students’ perceptions of their campus environment, demographic variables did not affect the reliability or validity of the factors. However, future research is needed to assess demographic variables to determine whether they demonstrate structural invariance or whether modifications are needed before global generalization can be established. In addition, as discussed earlier, data were cross-sectional, and prospective studies are needed to assess the sensitivity of this instrument to change and whether scores are predictive of change in behavior or BMI.

IMPLICATIONS FOR RESEARCH AND PRACTICE The BEPS is a brief instrument measuring college students’ perceptions of the healthfulness of their environment. Strengths of the instrument include a comprehensive instrument development process involving a large sample from multiple universities, breadth of constructs, and strong psychometric properties. Future steps include validation studies to assess how factors relate to other selfreported behaviors and demographic variables. Future studies are necessary to confirm the sensitivity, reliability, and validity of the BEPS in applied studies.

Limitations

SUPPLEMENTARY DATA

A large sample was recruited, but although the criterion analyses were significant, they did not demonstrate large effect sizes.21 However, the effect sizes were comparable to those in previous findings when similar methods were used for instrument development.20 Although the study assessed whether students were living on or off campus, their year in school was not considered, which may have influenced responses and students’ familiarity with their environment. In addition, race and ethnicity were not assessed, which limits the ability to generalize the final outcomes of perceptions and their relation to behaviors. Because the main study objective was to develop an instrument with strong psychometric

Supplementary data related to this article can be found at https://doi. org/10.1016/j.jneb.2019.09.003.

REFERENCES 1. Bandura A. The self system in reciprocal determinism. Ann Psychol. 1978;37:122– 147. 2. Gase LN, Glenn B, Kuo T. Self-efficacy as a mediator of the relationship between the perceived food environment and healthy eating in a low income population in Los Angeles county. J Immigr Minor Health. 2016;18:345–352. 3. Cluskey M, Grobe D. College weight gain and behavior transitions: male and female differences. J Am Diet Assoc. 2009;109:325–329.

ARTICLE IN PRESS 10

McNamara et al

4. Nelson MC, Story M, Larson NI, Neumark-Sztainer D, Lytle LA. Emerging adulthood and college-aged youth: an overlooked age for weight-related behavior change. Obesity (Silver Spring). 2008;16:2205–2211. 5. Lloyd-Richardson EE, Bailey S, Fava JL, Wing R, Tobacco Etiology Research Network (TERN). A prospective study of weight gain during the college freshman and sophomore years. Prev Med. 2009;48:256–261. 6. Gropper SS, Simmons KP, Gaines A, et al. The freshman 15-a closer look. J Am Coll Health. 2009;58:223–231. 7. American College Health Association. American College Health Association National College Health Assessment Spring 2006 Reference Group data report (abridged). J Am Coll Health. 2007;55:195–206. 8. Hebden L, Cook A, van der Ploeg HP, Allman-Farinelli M. Development of smartphone applications for nutrition and physical activity behavior change. JMIR Res Protoc. 2012;1:e9. 9. Kapinos KA, Yakusheva O. Environmental influences on young adult weight gain: evidence from a natural experiment. J Adolesc Health. 2011;48:52–58. 10. Kapinos KA, Yakusheva O, Eisenberg D. Obesogenic environmental influences on young adults: evidence from college dormitory assignments. Econ Hum Biol. 2014;12:98–109. 11. Booth ML, Owen N, Bauman A, Clavisi O, Leslie E. Social-cognitive and perceived environment influences associated with physical activity in older Australians. Prev Med. 2000;31:15–22. 12. Booth KM, Pinkston MM, Poston WSC. Obesity and the built environment. J Am Diet Assoc. 2005;105:110– 117. 13. Strong KA, Parks SL, Anderson E, Winett R, Davy BM. Weight gain prevention: identifying theory-based targets for health behavior change in young adults. J Am Diet Assoc. 2008; 108:1708–1715. 14. Horacek TM, White AA, Byrd-Bredbenner C, et al. PACES: a physical activity campus environmental supports audit on university campuses. Am J Health Promot. 2014;28:e104–e117. 15. Horacek TM, Erdman MB, ByrdBredbenner C, et al. Assessment of the

Journal of Nutrition Education and Behavior  Volume 000, Number 000, 2019

16.

17.

18.

19.

20.

21.

22.

23.

24.

25.

26.

dining environment on and near the campuses of fifteen post-secondary institutions. Public Health Nutr. 2013;16: 1186–1196. Greaney ML, Less FD, White AA, et al. College students’ barriers and enablers for healthful weight management: a qualitative study. J Nutr Educ Behav. 2009;41:281–286. Nelson MC, Kocos R, Lytle LA, Perry CL. Understanding the perceived determinants of weight-related behaviors in late adolescence: a qualitative analysis among college youth. J Nutr Educ Behav. 2009;41:287–292. Walsh JR, Hebert A, Byrd-Bredbenner C, et al. The development and preliminary validation of the behavior, environment, and changeability survey (BECS). J Nutr Educ Behav. 2012;44: 490–499. Sowers MF, Colby S, Greene GW, et al. Survey development to assess college students’ perceptions of the campus environment. Am J Health Behav. 2017; 41:701–709. Schembre S, Greene G, Melanson K. Development and validation of a weight-related eating questionnaire. Eat Behav. 2009;10:119–124. Harlow LL. The Essence of Multivariate Thinking: Basic Themes and Methods. Mahwah, NJ: Lawrence Erlbaum Associates; 2005. Nahm AY, Rao SS, Solis-Galvan LE, Ragu-Nathan TS. The Q-sort method: assessing reliability and construct validity of questionnaire items at a pre-testing stage. J Mod Appl Stat Methods. 2002;1:15. Noar SM. The role of structural equation modeling in scale development. Struct Equ Model Multidiscip J. 2003;10: 622–647. Carbone ET, Campbell MK, HonessMorreale L. Use of cognitive interview techniques in the development of nutrition surveys and interactive nutrition messages for low-income populations. J Am Diet Assoc. 2002;102:690–696. Willis GB, Royston P, Bercini D. The use of verbal report methods in the development and testing of survey questionnaires. Appl Cogn Psychol. 1991; 5:251–267. Costello AB, Osborne J. Best practices in exploratory factor analysis: four

27.

28.

29.

30.

31.

32.

33.

34.

35.

36.

37.

recommendations for getting the most from your analysis. Pract Assess. 2005; 10:1–9. Velicer WF. Determining the number of components from the matrix of partial correlations. Psychometrika. 1976;41: 321–327. Bentler PM. EQS 6 Structural Equations Program Manual. Encino, CA: Multivariate Software Inc; 2006. Thompson FE, Midthune D, Kahle L, Dodd KW. Development and evaluation of the National Cancer Institute’s Dietary Screener Questionnaire scoring algorithms. J Nutr. 2017;147:1226– 1233. Craig CL, Marshall AL, Sj€ ostr€ o m M, et al. International physical activity questionnaire: 12-country reliability and validity. Med Sci Sports Exerc. 2003; 35:1381–1395. Centers for Disease Control and Prevention. Measuring Healthy Days: Population Assessment of Health-Related Quality of Life. Atlanta, GA: Centers for Disease Control and Prevention; 2000. Berrington de Gonzalez A, Hartge P, Cerhan JR, et al. Body-mass index and mortality among 1.46 million white adults. N Engl J Med. 2010;363:2211– 2219. Pedrelli P, Nyer M, Yeung A, Zulauf C, Wilens T. College students: mental health problems and treatment considerations. Acad Psychiatry. 2015;39:503– 511. Kruisselbrink Flatt A. A suffering generation: six factors contributing to the mental health crisis in North American higher education. Coll Q. 2013;16:n1. Pelletier JE, Laska MN. Balancing healthy meals and busy lives: associations between work, school, and family responsibilities and perceived time constraints among young adults. J Nutr Educ Behav. 2012;44:481–489. Deliens T, Deforche B, De Bourdeaudhuij I, Clarys P. Determinants of physical activity and sedentary behaviour in university students: a qualitative study using focus group discussions. BMC Public Health. 2015;15:201. Deliens T, Clarys P, De Bourdeaudhuij I, Deforche B. Determinants of eating behavior in university students: a qualitative study using focus group discussions. BMC Public Health. 2014;14:53.