The original EnhanceWellness Program used a nurse—social worker team to review risk assessments with participants, develop an action plan, refer participants to appropriate programs, and revise the action plan as needed until maintenance was achieved. We thought a team approach might not be generalizable in worksites, so instead we used a single MPH-educated coach trained in principles of behavior change and motivational interviewing. We named this adapted version of EnhanceWellness the COACH program because we thought that name would have greater appeal to a working-age population.

The coach contacted participants to schedule and conduct an initial in-person Enhance-Wellness health-risk assessment, review the participant’s health-risk profile, discuss behaviors the participant was interested in adopting, and discuss the availability of health behavior change programs on or off campus. The coach reviewed the participant’s health-related goals and negotiated an action plan to meet those goals. The plan could be revised and expanded over time. During the first week after the assessment, the coach called participants by phone to ask them about their success in accessing resources needed to implement the plan. Participants who reported difficulty returned for a second meeting to revise the plan to reflect attainable goals. Thereafter, the coach contacted participants via e-mail or telephone biweekly during months 1 through 6 and monthly during months 7 through 12. During these contacts, the coach and participant reevaluated the plan, including the negotiation of other goals and related actions. The coach repeated the in-person assessment with participants at 6 and 12 months (and more frequently if needed).

Participants assigned to the Real Age group were sent an e-mail message to remind them to access a Web page set up by RealAge staff for use in this trial. The Web site contained the RealAge test, which participants were instructed to complete. The e-mail gave participants the login information needed to access the Web site and complete the test, which reviewed family history, diet and nutrition, PA, tobacco and alcohol use, satisfaction with family relationships, and use of preventive health care. After participants completed the test, the Web site generated individual risk profiles and indicated areas that could be worked on to improve health. Participants used the Web site to select behaviors to work on and create plans to meet behavioral goals. The RealAge site tracked each time a participant used it, and the site forwarded this information to the study team at regular intervals, enabling us to document use.

After the baseline interview and random study-arm assignment, participants in the light health education control group were personally handed printed health-promotion materials. These materials included a listing of health-promotion programs and services offered by the university and other community-based organizations.

Individuals younger than 40 years and faculty were excluded from the research. We originally set the minimum age at 50 years, but we lowered it to 40 years to boost recruitment. Participants were also required to have reliable access to the Internet at home or work.

We used methods outlined by Rochon¹⁵ to estimate ideal sample size. Under a wide variety of assumptions regarding attrition and covariance structures (e.g., AR-1 correlations varying from 0.4 to 0.7 and attrition varying from 10% to 30%), we found that a sample size of 150 in each of the 3 conditions (450 total) would enable us to detect between-group end-point differences of one quarter of a standard deviation, assuming 1-tailed tests and power of 0.8.

The primary independent variable was intervention assignment. Demographic variables were age, race/ethnicity, gender, income, type of health insurance coverage, educational attainment, job description, job title and department, and length of employment at the university. We used a self-report measure of comorbid conditions.¹⁶

We used SOC for specific health behaviors as a moderator in the outcome analyses. The transtheoretical model of behavior change states that persons who are in contemplation of or preparation for the adoption of a health behavior are more likely to actively engage in the behavior, at least in the short term. We used algorithms to classify individuals into stages of change at baseline on the basis of this model.^1,17-19 Participants were labeled as being in precontemplation, contemplation, preparation, action, or maintenance SOC for healthy eating,²⁰ dietary fat reduction,²¹ fruit and vegetable intake,²² PA, vigorous exercise,²³ moderate exercise,²³ and smoking cessation,²⁴ according to customary SOC criteria that were based on duration of participation in each specified behavior. For example, persons classified as maintainers had to engage in the behavior most days of the week and had to have been engaged at that level for longer than 6 months. The dietary SOC measures were significantly intercorrelated. Therefore, we selected a single indicator—SOC for healthy eating—as the moderator because the items in this scale appeared to be more global in nature. To assess SOC for stress management we used a question developed by the authors.

We measured changes in percent energy from fat and in fruit and vegetable intake from baseline to 6 months and 12 months. We used the National Cancer Institute’s Percentage Energy from Fat Screener questionnaire to assess participants’ usual daily intake of fat.²⁵ We used the National Cancer Institute All-Day Fruit and Vegetable Screener questionnaire to obtain self-reported data on fruit and vegetable intake during a typical day.²⁶

We assessed changes in vigorous activity, moderate activity, and level of exercise participation from baseline to 6 months and 12 months. We used the 7-item Behavioral Risk Factor Surveillance System scale to assess vigorous and moderate PA over a typical week,²⁷ and we used the Rapid Assessment of Physical Activity to assign participants to 1 of 5 categories of exercise participation.²⁸

We investigated change from baseline to 6 months and 12 months in 4 measures of stress. We used the Perceived Stress Scale²⁹ to measure overall stress during the preceding month. We used a 4-item scale developed by Lorig et al.³⁰ to assess health-related stress, and we used the Brief COPE³¹ to assess use of positive and negative coping behaviors.

Smoking cessation was defined as a minimum of 6 months of total abstinence from tobacco use at 6 months and 12 months after baseline, among participants who were current smokers at baseline or 6 months. We used a 3-item measure of past and current smoking status from the Behavioral Risk Factor Surveillance System to assess tobacco use.²⁵

Staff measured participants’ height without shoes (rounding to the nearest eighth of an inch) and weight without coat or shoes (rounding to the nearest pound), and we used this information to calculate BMI.³² Research staff used a tape measure to measure waist circumference. For all 3 measures we assessed change from baseline at 6 months and 12 months.

Of 150 participants assigned to the COACH arm, 146 (97%) completed the initial assessment. Of assessment completers, 68% chose to work on diet, 76% chose to work on PA, 14% chose stress, and 3% chose smoking cessation. Of the 146 COACH participants who developed a plan with the coach, 42% (n = 62) chose to work on 1 behavior, 51% (n = 75) chose to work on 2 behaviors, and 6% (n = 9) chose to work on 3 behaviors.

Follow-up contact with participants in COACH included telephone calls, e-mails, and in-person visits. Of 688 total communications, 634 took place via e-mail, of which 491 (77%) were initiated by the coach. Contacts also included 48 follow-up telephone calls and 6 in-person follow-up visits. On average, the coach spent 9.25 hours per participant over the 12-month period.

Participants randomized to RealAge were told at the baseline interview how to access the RealAge Web site and complete the health-risk profile. Research staff sent 1 e-mail after the baseline interview to remind participants to access the Web site and complete all 7 sections of the RealAge healthrisk assessment. Of 135 participants assigned to RealAge, 79 (59%) completed all 7 sections of the RealAge test, which participants had to complete to gain unlimited access to the Real Age Web site. The 79 participants who completed the assessment received generic e-mail health tips from RealAge, which provided information on various health topics as well as links to additional information. Of the 79 engaged RealAge participants, 72 (94% of the engaged RealAge participants; 53% of the total RealAge group) clicked at least 1 link to obtain additional information. The remaining 60 participants who did not complete all 7 sections of the test did not receive e-mail health tips.

We found a significant 3-way interaction for both percentage of energy from fat and fruit and vegetable consumption, such that those in the COACH group who were in the precontemplation, contemplation, or preparation SOC had greater changes over time. COACH participants reported a borderline-significant reduction in percentage energy from fat at 6 months (P = .063) and a significant reduction at 12 months (P = .027 Table 1). COACH participants also reported eating significantly more fruits and vegetables than did control group participants at 6 months (P = .026) and 12 months (P < .001). No significant differences were seen on either variable for RealAge participants at either time point. Figure A (available as a supplement to the online version of this article at http://www.ajph.org) presents descriptive data for participants in precontemplation, contemplation, or preparation at baseline in each of the 3 groups at all 3 time points for percentage energy from fat. Consistent with the findings from the random-effects analyses, COACH participants experienced the largest decrease in percentage energy from fat, RealAge participants experienced more modest (nonsignificant) declines, and control participants stayed the same over the 12-month period.

COACH participants reported significantly more minutes of moderate PA than did controls at 6 months (P = .05) and 12 months (P = .013; Table 2). No significant differences were seen for COACH participants on RAPA scores or on minutes of vigorous PA at 6 or 12 months, and no significant differences were seen on any of the PA variables at either time point for RealAge participants.

No significant differences were seen for COACH or RealAge participants on any of the 4 stress outcomes at either time point (Table B, available as a supplement to the online version of this article at http://www.ajph.org).

At 12 months, 2 of 16 COACH smokers, 4 of 16 RealAge smokers, and 3 of 21 control group smokers achieved maintenance of smoking cessation. We found no significant differences in rates of smoking cessation for smokers in COACH or RealAge relative to those in the control group with or without SOC for smoking at baseline as a covariate (data not shown). This finding probably reflects the small number of current smokers per group at baseline.

When we used baseline SOC for diet behaviors as a covariate, RealAge participants experienced a significant decline in waist circumference at 6 months (P = .05) that was maintained at 12 months (P = .018). Neither COACH nor Real Age participants experienced significant decreases in BMI or weight at 6 or 12 months compared with control participants (Table 3).