Objectives and hypotheses
The purpose of this study was to design and evaluate a multi-component weight gain prevention program in a workplace setting. Offering individual tools, providing information and persuasive messaging, and changing the social and physical environments over 12 months were hypothesized to result in healthier eating and increased physical activity. Consequently, it was expected that employees in the intervention condition would not gain weight or would gain less weight than the comparison group.
Study design
Using a quasi-experimental nonequivalent control group design [14], employees from a mid-sized healthcare system in the upper Midwest region of the United States were recruited. The intervention group consisted of employees from the hospital campus (including the main hospital, administrative offices, and several specialty outpatient clinics), whereas the comparison group consisted of employees from six primary care clinics located within 15 miles of the main hospital campus. This hospital system was chosen due to an existing working relationship with key hospital personnel. A non-randomized design was utilized for reasons of access, available resources, and the exploratory nature of some intervention components. Employees could not be randomized to conditions because of the nature of the intervention involving social networks and environmental changes made throughout the hospital. The primary care clinic employees were the best available comparison group as they worked autonomously from the hospital environment, yet still had the same employer (i.e., same benefits, wellness programming, health communications).
The study was approved by the University of Minnesota Institutional Review Board (reference #: 1002S78225) and the St. Luke’s Institutional Review Board (reference #: HSRC010-003). Informed consent was gathered from every participant in the study. Formative stages of the study took place in 2010, which involved identifying influentials as well as key leaders in the hospital to serve on an advisory board created to provide input on program development, implementation, and measurement. It was noted that previous wellness programs that involved sign-up and participation, such as Weight-Watchers groups, walking groups, and walking contests had poor response rates (less than 3 % of employee involvement). The intervention and associated data collection took place November 2010 to December 2011, and the data were analyzed from 2012 to 2014. This study was registered with clinicaltrials.gov (NCT01585480) on April 24, 2012.
Participant recruitment
All hospital and clinic employees were invited to participate, and were recruited through multiple methods, including email, direct mail, newsletter, and hospital intranet link (all sent a few days before data collection began). In exchange for participating in the assessments, hospital employees received a pedometer and up to $50 cash ($10 at baseline and $20 at the 6- and 12-month assessments). Unlike studies that enroll participants into “a program,” this study primarily recruited hospital employees to “complete assessments” to evaluate the effectiveness of an obesity prevention program being implemented throughout the hospital. That is, there was no expectation that employees in the study engage in any additional activities or attempt to lose weight as part of the study, even though they were informed that the Go! program was being implemented to help employees avoid weight gain. The rationale for this was that the purpose of the study was to determine the effects of hospital-wide changes on employee eating and physical activity behavior, rather than targeting only individuals ready and wanting to lose weight. That said, at the baseline assessment, enrollees received information and a pedometer that they were encouraged, but not expected, to use. After this point, no additional intervention components were given to the enrolled participants that were not also provided to the rest of the hospital employees, other than the data collection itself. Clinic employees were similarly asked to complete assessments, though informed that it was in order to examine changes in eating, physical activity, and weight of healthcare employees over time, without mention that they were a comparison group for an intervention occurring at the main hospital. This was done in order to maintain the status quo (i.e., no new wellness opportunities) as well as to minimize perceptions of unfairness, (i.e., that clinic employees were being deprived of a special program offered only to hospital employees). They received the same financial incentive, but not the pedometer.
All employees over the age of 18 were allowed to participate, regardless of employment status or intentions to remain employed, pregnancy status/intentions, health status, or desire to lose or maintain weight. Although such an inclusive approach resulted in greater attrition, we believed that involving the largest number of employees possible would have greater impact on social norms and subsequent intervention effects. Only employees who completed the first assessment were recruited for future assessments. Participants received up to three reminders to participate in follow-up measures.
Intervention
All intervention components were delivered over 12 months and designed to increase physical activity, reduce calorie consumption, and manage weight using principles of energy balance. Specific messages and approaches were selected following a social ecological framework (i.e., targeting individual, interpersonal, social, environmental, and/or systems levels) and drew from three well-established multidisciplinary theories in health-related behavior change: extended parallel process model [15, 16], theory of planned behavior [17, 18], and social cognitive theory [19]. These theories suggest that knowledge, self-efficacy/perceived behavioral control, perceived risk, attitudes, and social norms all affect health behavior change. Each of these constructs was targeted through methods employed at the individual, interpersonal, social, and/or environmental level (i.e., use of the social ecological framework). See Fig. 1.
Nutrition labeling
Point-of-purchase calorie labeling on menus has been recommended to increase consumer knowledge and awareness of intake [20], yet few studies have fully incorporated such interventions into cafeteria or worksite settings [21]. A number of other approaches to labeling foods have been recommended, and to some degree, tested. In Europe, the Food Standards Agency has recommended that packaged foods include color-coded “traffic lights” on the front of packages, giving color ratings (green = go, eat frequently, yellow = caution, eat in moderation, red = stop, eat only occasionally) to foods [22]. Data suggest this particular labeling scheme results in the highest rate of consumer satisfaction and, more importantly, correct identification of healthful foods [22, 23]. However, it is not yet known whether such an approach influences actual purchases, and it has yet to be empirically tested in restaurant or cafeteria settings. In the present intervention, every food item in the hospital cafeteria and vending machines was labeled with calories, number of steps required to burn those calories (e.g., a slice of pizza = 690 calories and 13,800 steps), and a “traffic light” color rating (green = go, eat in large portions, yellow = caution, eat in moderate portions, red = stop, eat in small portions; see Fig. 2). Color codes were based on energy density, which is the number of kcalories per gram of food (green ≤ 1.0 kcal/g; yellow = 1.0 to 2.25 kcal/g; red > 2.25 kcal/g. This scheme was based on Volumetrics principles [24], but adapted to fit a 3-color traffic light model, rather than the four categories initially proposed by Rolls and Barnett. Most fruits and vegetables are “green foods,” whereas high-fat foods are “red foods” because fat has more than twice the number of calories per gram than carbohydrates or proteins. Step equivalents were based on the general principle of 100 kcalories expended per mile of walking (an individual weighing 175lbs, walking 4 mph, would burn 99 calories [25]). It was anticipate that such a labeling scheme at the point-of-selection would provide a highly salient and meaningful visual cue as to the availability of healthful foods, as well as affect employees’ eating behavior by changing their knowledge of and self-efficacy toward eating lower calorie food while increasing the perceived risk of eating high calorie foods. To our knowledge, no study has ever examined the effects of this particular labeling scheme.
Environmental changes in cafeteria
Over the course of the study, a number of changes were made to the hospital cafeteria. Due to the difficulty of making these types of environmental changes at an institution, most of these modifications did not occur until the latter half of the study. Changes included reducing the size of serving spoons (start of study), offering half portions at half price (at month 5), increasing the number of “green light” foods (month 7), moving the dessert case to a less visible area (month 8), establishing a highly visible “grab-n-go” cooler offering healthy foods (month 8), and decreasing portion sizes of some foods (month 8). These changes were designed to increase perceptions of access to and convenience of healthy foods, and to make the default the healthier option.
Pedometers
It has been increasingly recognized that to promote and maintain physical activity habits, individuals need to identify and make use of activities that can be readily adapted to their environments and schedules [26]. Because walking is a convenient and inexpensive form of physical activity, it offers a promising strategy for altering sedentary behaviors and managing weight. Worksite walking programs utilizing pedometers have produced increases in moderate intensity physical activity levels as well as improvements in BMI, systolic blood pressure, blood glucose, and cholesterol [27–29].
As part of the Go! study, participants who completed the baseline assessment received a pedometer (Digi-Walker SW-200) that tracked number of steps per day, and were given a brochure containing information about its use, including its relation to the Go! program. Participants were informed about the labeling scheme in the cafeteria that was about to be initiated, and told how the pedometer could help them monitor activity in relationship to food intake (i.e., energy balance). In addition to providing an increase in awareness/knowledge of one’s physical activity level, access to pedometers were expected to (a) create more positive attitudes towards walking, such that employees would view walking as a convenient and inexpensive form of physical activity; and (b) increase self-efficacy toward maintaining a walking program by giving employees a tool to easily self-monitor and adjust their amount of walking, especially in response to calories consumed (as marked in the cafeteria). These changes in attitude and self-efficacy were, in turn, expected to motivate employees to increase their actual (walking) behavior. For instance, employees with a pedometer may be more motivated to take the stairs instead of the elevator at the worksite. Complementary informational and motivational messages were offered to help employees reduce barriers to walking at work (i.e., signage pointing out the stairs, mapping walking routes).
Participants were provided with a personalized magnet providing energy balance facts and the number of calories each individual needed per day to maintain their current weight. It was hoped that the magnet would be placed in a frequently viewed location (refrigerator door, work filing cabinet), thereby increasing familiarity with energy balance principles, serving as a prompt to consider healthy eating and activity behaviors, and facilitate branding of the Go! program.
Environmental strategies to increase physical activity
A stairwell campaign incorporated stair use signs adapted from the CDC’s StairWELL campaign [30] as well as signs created specifically for Go! that were placed in and around stairwells and elevators. Two indoor walking routes were created, along with posters and flyers of maps, distances, and step information. These materials developed for this intervention component were designed to target the key constructs of knowledge, self-efficacy, social norms, access, and convenience.
Persuasive messaging
Strategic, theoretically derived, health communication messages aimed at changing employee attitudes, risk perceptions, and efficacy were offered in a variety of mediums including posters, table toppers, and a website. Messages were placed in and around stairwells, the cafeteria, break rooms, and hallways. There were three phases in message design and dissemination. In the initial phase, messages focused on educating employees about the meaning of the "traffic light" labels affixed to foods in the cafeteria as well as the calorie content of popular foods. Phase 1 messages were designed to motivate employees to eat more green light foods and cut back on red light foods. The goal of the second phase was to educate employees about the meaning of energy balance and the importance of portion sizes. In this phase, messages were also designed to motivate employees to use the stairs, pedometer, gym, and walking routes and to encourage self-efficacy with tag lines such as “even small changes, make big differences.” The final phase focused on underscoring the role of social support in losing and maintaining weight-loss both as a provider and recipient of social support. This phase was designed to coincide with the influential component of the intervention (described below). An additional file depicts examples of messages used in the Go! campaign (see Additional file 1).
Influentials
Another novel component of the Go! study was the use of “influentials” or natural helpers to influence social norms and enhance social support within the worksite. Influentials are well-respected, knowledgeable, socially well-connected and persuasive employees [31, 32] who are able to affect others’ attitudes and behaviors [33], thus providing a sustainable change to social norms. Studies point to the importance of strong and supportive social ties in influencing health-related behaviors, especially in relation to weight control [34] and dietary change [35]. A number of worksite health promotion programs also note the importance of social support and the social culture in which eating takes place in enabling health behavior change [3, 36].
In this study, influentials were identified and trained to provide support, reinforce healthful messages, change perceptions of risk, influence positive attitudes, and increase self-efficacy of employees, thereby complementing the messages on receipts and other materials. Notably, influentials provide an advantage over mediated messages in that influentials are able to deliver more personalized and tailored messages to each person in their worksite social network. Through informal channels of communication, the influential can address and clarify misconceptions or gaps in information that particular people may have and encourage others to change behaviors.
Influentials for the Go! study were identified in two primary ways. The first involved an employee survey completed by 530 out of 1565 hospital employees (34 % response rate). The survey included a validated measure that included the Maven scale (measuring health knowledge), the Persuader scale (measuring persuasiveness), and the Connector scale (measuring connectedness to others in the hospital) [37]. Participants were also asked to list fellow employees who met the description of a connector, persuader, and/or maven. Secondly, supervisors and directors (n = 61) were asked to identify employees who exemplified these three key characteristics.
Of the 83 eligible employees identified as meeting the characteristic of an influential, 54 agreed to participate (65 % acceptance rate). An initial 4-h training emphasized listening and persuasion skill-building in addition to providing basic education of energy balance principles and team building. Three, 1-h follow-up sessions included training in motivational interviewing, identification of barriers to healthy eating and physical activity among employees, and development of strategies to reduce barriers. The influentials were encouraged to develop a plan to move forward, beyond the 12-month study timeframe. This resulted in the formation of action committees who chose to target environmental and policy changes (e.g., policies limiting free lunches from pharmaceutical representatives, initiating exercises classes on site). Such broader changes were seen after the conclusion of the study period. Further details of the selection procedures, trainings, and assessment related to this component were published elsewhere [38].
Fidelity checks
Intervention fidelity was assessed in a number of ways. All cafeteria food labels were checked by study staff for accuracy at the start of each weekday lunch period (the busiest period of use among employee patrons) and continued through the entire 12 months. Vending machine labels were verified weekly and any new items were added to the labels within one week. Weekly to biweekly fidelity checks were conducted to assure that mediated messages were posted (posters hung and visible in stairwells, elevators hallways, and break rooms; table tents were on each cafeteria table and in break rooms).
Measures
Employees completed assessments at baseline, 6-months, and 12-months. Questionnaires were sent out (hard copy and/or electronically with link to online version) in advance of data collection and took approximately 30 min to complete. Participants then underwent anthropometric measurements completed by trained research personnel in a private clinic room or an enclosed area within the cafeteria. Several data collection sessions were scheduled over a 3-week period, offering a variety of times to accommodate hospital work shifts. Many of the measures were selected to be consistent with those in the NHLBI-funded studies [6], so comparisons could be made.
Anthropometric measures
Weight was measured in light clothing and without shoes to the nearest 0.2 lbs using a standard scale equipped with a leveling bubble on the platform (Seca 869). Height was measured without shoes to the nearest 0.25 in. using a calibrated portable stadiometer (Charder HM200p Portstad). BMI was calculated as weight (kg) / height2 (m). Waist circumference was measured to the nearest 0.1 cm using a standard tape equipped with a tension calibration mechanism (Gulick II). Measurement procedures were adapted from the National Health and Nutrition Examination Survey III - Anthropometry Procedures Manual [39]. Two waist measurements were taken, and the average used. If the difference was greater than 0.5 cm, a third measurement was taken, and the average of the two closest scores was used.
Dietary intake
The National Cancer Institute Multifactor Screener [40] is a 17-item self-report food frequency questionnaire that estimates daily fruit and vegetable intake, grams of fiber, and percent of energy from fat. Out of 9 frequency options ("never" to "4 or more times per day"), employees selected the one that best estimated specific food intake over the past month. Scores on all three variables have been found to be highly correlated (r = 0.5–0.8) with estimates of actual intake [41]. This screening measure was selected because it was expected to be most sensitive to changes in eating behavior targeted by the intervention (i.e., fruits and vegetables are low-calorie, low energy-dense foods and thus encouraged, whereas high fats foods were discouraged due to their higher calories and energy density). Other foods specifically targeted in Go! (including donuts/muffins, cookies/bars, sugary beverages, and ice cream, all served in the cafeteria) were added to the survey. These items were analyzed separately and not included in the subscale calculations.
Physical activity
Two self-report measures were used to gather data about physical activity. Found to be both reliable and valid [42], the Godin Leisure Time Exercise Questionnaire (GLTEQ) [43] is a brief measure that asks an individual about weekly frequencies of strenuous, moderate, and light activities, and yields an overall intensity-weighted frequency score [metabolic equivalent per week (METs)]. The International Physical Activity Questionnaire (IPAQ) [44] walking subscale was also used. Based on the work of French and colleagues [45], modifications were made to shorten the instructions and simplify response formats. Given that increasing walking as part of one’s daily routine was a specific target in the Go! intervention, the measure was further shortened to four items that only assess minutes per day of walking across work, home, and leisure settings (with a total score calculated). Participants were asked to select one of eight frequencies (e.g., less than 10 min, 10–20 min…more than 2 h).
Knowledge
A 6-item measure designed to assess knowledge about principles of energy balance was created for this study. Items included individualized information (e.g., “How many calories do you think YOU need each day to maintain your current weight?”), general information (e.g., number of calories needed to cut in order to lose one pound), and calorie content in foods commonly purchased in the hospital cafeteria (e.g., regular fat salad dressing).
Perceptions of support
From the Worksite Health Climate Scale [46], the 4-item Organizational Support Scales of Employer’s Health Orientation assessed how employees perceived employer concern about the health and well-being of employees (e.g., “[Hospital Name] values workers who have healthy lifestyles.”, and the 3-item Job Flexibility to Exercise subscale assessed perceptions that one’s job is flexible enough to allow exercise at work (e.g., “Í can make time to exercise at some point during normal work hours.”). The items were rated on a 5-point response format (1 = strongly disagree to 5 = strongly agree) with a mean item value calculated for each of the subscales. Internal consistency was adequate in the present study (Commitment, α = 0.87–0.89; Flexibility, α = 0.70–0.77).
Perceived coworker support
This 3-item measure, adapted from the Route H study [45], asked employees about support they received from coworkers with respect to eating a low-calorie/low-fat diet, exercise, and weight management (e.g., “How supportive are your coworkers of your efforts to manage your weight?”). Scores ranged on a 6-point scale from “not applicable,” “not at all supportive” to “very supportive.” The three items were added for a total score. The measure demonstrated good internal consistency in this study (α = 0.93).
Information
Three items assessed access to information about healthy eating, physical activity, and weight management on a 7-point scale (e.g., “There is a lot of information on healthy eating at [hospital name]”. The scale was adapted from the Route H study [45] and had high internal consistency (α = 0.92–0.96).
Reactions to Go!
On the final survey, participants were asked questions about their attitude toward, use of, and desire for certain components of the intervention to continue.
Weight and health history
Twenty-three questions regarding weight goals, weight loss history, and health and medication status were also included.
Statistical analysis
Data were analyzed using SAS software, version 9.4 and SPSS, version 21. Missing data for outcome variables and covariates were imputed using fully conditional (MCMC) multiple imputation. Imputations were run separately for sets of variables expected to strongly related to each other and separately for each campus. SAS procedure MIANALYZE was used for combining of analyses across multiply imputed data sets.
For the intervention group, the number of missing values that were multiply imputed increased across the three time points (1.7, 20.3 and 21.9 %, respectively); and were almost entirely from complete missing assessments at follow up time points. For the comparison group 1.2, 10.7 and 8.5 % of data values imputed at each of the time points, respectively. When residuals from outcome variables were highly skewed (i.e., weight, BMI, waist circumference, IPAQ, GLTEQ, dietary screen variables), the natural log was used for analyses. Results are reported in the original scale means or originals scale geometric means for ease of interpretation.
Comparisons of baseline differences were made between participants in the intervention versus comparison group using t-tests for continuous variables and either Chi Square or Fisher’s exact test for categorical variables. These same tests were used to compare participants who completed two or three anthropometric assessments (completers) versus those who completed only baseline assessments (noncompleters).
The impacts of the intervention on the outcome variables were assessed using repeated measures ANCOVAs (adjusted for known correlates of BMI and other reported measures, including age, sex, education, income, smoking status, job category, perceived job flexibility, weight loss goal, and being on medications for diabetes, depression, hyperlipidemia, or hypertension). For covariates that could change over time, the covariate value at each time point was used as the covariate for the dependent variable at that time point.
For all repeated measures analyses, data were also analyzed using nonimputed, reported data only, and by using the last observation carried forward (LOCF) method for handling missing data as an additional intent-to-treat approach. The results of all six analyses (nonimputed, multiply imputed, and LOCF; each with and without covariates) were similar with a few differences for which comparisons were slightly above or below p = 0.05. The analyses reported here are from the multiply imputed data with covariate adjustments. For all outcomes, changes in adjusted and unadjusted means were compared over time within groups, and changes from baseline were compared across groups with interaction type contrasts (i.e. Time 2 - baseline for intervention group versus Time 2 - baseline for comparison group, and similarly for Time 3 - baseline.