In the past three decades, US employers sought ways to stem the

Transcription

1 CME AVAILABLE FOR THIS ARTICLE AT ACOEM.ORG Seven-Year Trends in Employee Health Habits From a Comprehensive Workplace Health Promotion Program at Vanderbilt University Daniel W. Byrne, MS, Ron Z. Goetzel, PhD, Paula W. McGown, MSN, MAcc, RN, FNP-BC, CPA, Marilyn C. Holmes, MS, RD, LDN, Meghan Short Beckowski, MPH, Maryam J. Tabrizi, MS, Niranjana Kowlessar, PhD, and Mary I. Yarbrough, MD, MPH, FACOEM, FACPM Objective: To assess long-term changes in health risks for employees participating in Vanderbilt University s incentive-based worksite wellness program. Methods: Descriptive longitudinal trends were examined for employees health risk profiles for the period of 2003 to Results: The majority of risk factors improved over time with the most consistent change occurring in physical activity. The proportion of employees exercising one or more days per week increased from 72.7% in 2003 to 83.4% in Positive annual, monotonic changes were also observed in percentage for nonsmokers and seat belt usage. Although the largest improvements occurred between the first two years, improvements continued without significant regression toward baseline. Conclusions: This 7-year evaluation, with high participation and large sample size, provides robust estimates of health improvements that can be achieved through a voluntary incentive-based wellness program. In the past three decades, US employers sought ways to stem the growth in health care costs, which were often at double-digit annual inflationary rates and incompatible with viable long-term business models. 1 This steep rise in health care spending changed the role of the employer from passively purchasing services for the treatment of disease to actively engaging in programs and benefit plan designs in attempts to control unnecessary utilization. 1 More specifically, employers have redirected their efforts at population health management and incorporating health promotion and disease prevention programs alongside more traditional case and disease management. Tools used by employers for population health management typically include extensive awareness building through health education, health risk assessments (HRAs), in-person risk reduction interventions, telephonic health coaching, Web-enabled communications, social networking, and establishment of data warehouses. 2 4 From Health & Wellness, Division of Administration (Mr Byrne, Ms McGown and Holmes, and Dr Yarbrough), the Division of General Internal Medicine and Public Health, Department of Medicine (Mr Byrne and Dr Yarbrough), the Department of Biostatistics (Mr Byrne), Vanderbilt University, Nashville, Tenn; the Institute for Health and Productivity Studies (Dr Goetzel), Emory University, Atlanta, Ga; and Thomson Reuters (Drs Goetzel and Kowlessar and Ms Beckowski and Tabrizi), Washington, DC. Authors Byrne, Yarbrough, McGown, Holmes, Short Beckowski, Tabrizi, Kowlessar, and Goetzel have no financial interest related to this study. Funding for this study was provided in part by Vanderbilt University Clinical and Translational Science Award grant UL1 RR from NCRR/NIH. The contents are the sole responsibility of the authors and do not necessarily represent the official views of Vanderbilt University, Emory University, or Thomson Reuters. The JOEM Editorial Board and planners have no financial interest related to this research. Supplemental digital content is available for this article. Direct URL citations appear in the printed text and is provided in the HTML and PDF versions of this article on the journal s Web site ( Address correspondence to: Mary I. Yarbrough, MD, MPH, FACOEM, FACPM, Health & Wellness, Department of Administration, Vanderbilt University, st Avenue South, Suite 640, Medical Arts Building, Nashville, TN (mary.yarbrough@vanderbilt.edu). Copyright C 2011 by American College of Occupational and Environmental Medicine DOI: /JOM.0b013e318237a19c Learning Objectives Discuss the characteristics of Vanderbilt University s workplace health promotion program, Go For The Gold (GFTG). Identify long-term effects on health risk factors such as physical activity, smoking, and seat belt use. Discuss factors leading to the overall health improvement and risk reduction among GFTG participants. When implemented appropriately, these tools are effective in three areas: risk identification; behavioral change facilitation; and creation of knowledgeable health care consumers. While some employers have adopted these programs with enthusiasm, others are looking critically at the evidence that expanded primary prevention can improve outcomes. For individual employers, assessing this evidence can be challenging because of the difficulty in establishing a stable, unbiased cohort within the workforce that has experienced consistent programming from which outcomes can be determined. Although the general consensus is that workplace wellness programs can help individuals improve modifiable risk factors, 5 the current literature is often limited to data collected at two time periods (before and after a health promotion program intervention), with the second period suffering from high attrition and low participation levels, many of which are less than 50%, 6 making it difficult to assess the true long-term trends in risk factors. The analysis described here attempts to overcome some of these limitations by tracking the experience of Vanderbilt University employees participating in an incentive-based wellness program called Go for the Gold (GFTG) over a 7-year period. The GFTG Program, initiated in 2003, aimed to engage faculty and staff at the university in identifying their lifestyle risks and maintaining or improving those risks through dynamic programming. The program, now in its eighth year of existence, has continued with stable management and consistent data collection. The longitudinal trends in lifestyle risks of faculty and staff who participated in this incentive-based program during the 7 years since its inception have been tracked and benchmarked against national and state results. The high annual participation rate in GFTG (averaging 75.5%) provides a valuable database to investigate changes in risk factors and healthy lifestyle characteristics over time and to compare the results to benchmark norms. METHODS Setting and Population This is a longitudinal descriptive analysis of Vanderbilt University employees health risk data for an aggregate and cohort population during the period of 2003 to Located on a 330-acre 1372 JOEM Volume 53, Number 12, December 2011

2 JOEM Volume 53, Number 12, December 2011 Vanderbilt University Workplace Health Promotion Trends campus in Nashville, Tennessee, Vanderbilt is a private research university and medical center containing 129 undergraduate, graduate, and professional degree programs in ten different schools. A total of 22,505 individuals (18,772 staff and 3733 faculty) were employed at the end of the fiscal year Of these, 95% were eligible for the GFTG Program (active, full-time, regular faculty and staff enrolled in the Vanderbilt Health Plan). Between 2003 and 2009, the number of eligible participants increased from 15,070 to 21,701, and the percent participating increased from 68% to 80%. The overall university population demographics, including the male to female ratio (34%:66%), faculty to staff ratio (15%:85%), and medical center to university employees ratio (76%:24%), remained relatively stable over the 7 years, as did the average age of employees (41 to 42 years). Intervention and Structure of the Wellness Program Health Plus Program Health Plus is one of the four Health and Wellness programs offered by Vanderbilt to support the health and productivity of faculty and staff; the other three programs are Occupational Health, Work/Life Connections-Employee Assistance Program, and the Child and Family Center. Health Plus provides awareness, health promotion, primary prevention, and environmental support for employee health, wellness, and productivity. Health Plus has a 17,000-square foot fitness facility; provides behavior change counseling, including health coaching; offers biometric testing; and makes available a variety of educational programs, including newsletters, Web tools, video and podcasting, lectures, workshops, and individual consultations. The cornerstone of Health Plus is the GFTG Program, a multiplatform, Web-based incentive program developed in 2003 by the Vanderbilt Health & Wellness team with input from a design team and outside consultants. The program is voluntary and multitiered. Tiers consist of (1) completing an HRA with the goal of identifying health risks, (2) completing a self-directed lifestyle management tool for setting goals to maintain health or improve health risks, and (3) viewing an annual educational video featuring local experts who discuss both the importance of reducing a particular health risk and ways to take charge of improving that risk. The Personal Wellness Profile Concise Plus Questionnaire version of the Wellsource HRA 7 was used consistently throughout the program. As an incentive for participation, up to $20 per month was added to an employee s paycheck during the following calendar year for completing all three tiers of the program. Other components, such as coaching and targeted risk-reduction programs, provide more personalized support. Employees with specific risks, based on their responses to the HRA, are notified of additional risk-reduction programming opportunities. Those with low scores on the HRA are also contacted and offered health coaching. The GFTG had three primary goals high participation ( 80%), increasing the percentage of participants who were low risk ( 80%), and increasing the overall HRA wellness score. During the course of the program, interventions and goals were refined based on the evaluation of participation data and health improvements observed. Primary Intervention Focus In 2003, GFTG staff identified specific modifiable lifestyle behaviors to target based on the evidence from medical literature, 8 10 the Healthy People 2010 Report, 11 university population demographics, and a health plan analysis. Lifestyle characteristics of primary, but not exclusive, focus were inadequate physical activity, overweight and obesity, smoking, too few fruits and vegetables consumed, and poor coping with stress. Go for the Gold Program Components The HRA used in GFTG has 39 questions focused on behavioral health risks, plus items asking for biometric and demographic information. The Wellsource wellness scores range from 0 to 100, with higher scores indicating better health behaviors. Scores are based on an algorithm that subtracts points for unhealthy behaviors, with adjustments for age. One-on-one feedback and coaching, as well as group sessions, are available to individuals at greatest risk. Employees, who only complete the HRA, without engaging in the other two program components, achieve bronze participation level. The Wellness Actions Log (WAL) is an on-line form that participants use to document healthy actions taken to maintain or improve health. Participants are awarded credit for completing a designated number of actions. The number of required actions increased from 5 of the 9 in 2003 to 7 of the 10 in 2009 to make the WAL broader in scope and more challenging. Employees who both complete the HRA and participate in the WAL achieve silver participation level. The Game Plan for Your Health videos feature interviews with Vanderbilt faculty and staff that provide practical advice on ways to take charge of one s health. Participants are awarded credits for completing a pretest, watching the video, and then completing a posttest. Employees who complete the HRA, participate in the WAL, and complete Game Plan requirements achieve gold level of participation. Cost of Program During the first year, an incentive was awarded to GFTG Program participants; the average annual cost to deliver the program, including incentives and operational costs for the program, was $157 per participant. The average annual cost per participant increased to $234 in Over the course of 7 years, the average cost per participant was $212. This cost excludes costs attributable to the fitness facility. Most costs were recovered in the design of the health plan. Definitions Participation in the program was defined as the proportion of benefits-eligible employees on the last day of the GFTG Program year who completed an HRA 7 during the previous 12 months. Biometric screening was not a mandatory element of the wellness program but offered throughout the year at various health promotion events. Participants could voluntarily take part in the biometric screenings and self-report their results into the HRA. Stratification of health risks were based on the work of Edington. 12 Those with zero to two risks were defined as low risk, three to four as medium risk, and five or more as high risk. The definition of high risk for each factor is listed in Appendix A (see Supplemental Digital Content 1, To compare our results with national and state figures and goals, we contrasted our data with those of the Centers for Disease Control and Prevention s Behavioral Risk Factor Surveillance System, 13 the National Highway Traffic Safety Administration s National Occupant Protection Use Survey, 14 and Healthy People 2010 goals 11 (Figs.1to4). Statistical Analysis Approval for this project was obtained from Vanderbilt University s Institutional Review Board and university leadership. This was a single-center, observational, prospective cohort study. No sample size or power calculations were performed at inception since the goal was to include as many employees as possible. Descriptive statistics were used to present the results of the trends for both aggregate and paired analysis, based on a cohort of employees who completed an HRA in any given calendar year, and a cohort of employees who completed the HRA all 7 years. The aggregate group C 2011 American College of Occupational and Environmental Medicine 1373

3 Byrne et al JOEM Volume 53, Number 12, December 2011 FIGURE 1. Physical activity trends for the aggregate and cohort groups of Vanderbilt s GFTG Program and comparison to national and Tennessee norms from Behavioral Risk Factor Surveillance System. For the Vanderbilt data, the HRA question was How many days per week do you engage in aerobic exercise of at least 20 to 30 minutes duration (fitness walking, cycling, jogging, swimming, aerobic dance, or active sports)? FIGURE 3. Seat belt use trends for the aggregate and cohort groups of Vanderbilt s GFTG Program and comparison to national and Tennessee norms from National Occupant Protection use Survey. The HRA question was When driving or riding in a vehicle, how often do you wear a seat belt? FIGURE 2. Tobacco use trends for the aggregate and cohort groups of Vanderbilt s GFTG Program and comparison to national and Tennessee norms from Behavioral Risk Factor Surveillance System. The HRA question identified those who currently smoke cigarettes daily. Former smokers, pipe, cigar, and chewing tobacco were not included. of employees ranged from 10,248 individuals in 2003 to 17,335 in 2009; the cohort was composed of 3745 employees participating every year from 2003 to For comparison of participants versus nonparticipants, a chisquare test for categorical variables and the Mann-Whitney U test for continuous and ordinal variables were used on demographic data from the employee record, such as age, gender, race, job category, and length of employment. Attrition was assessed by using Cox proportional-hazard analysis. Health risks were compared descriptively, including annual differences and the average annual difference across all years. Several of the ordinal and continuous variables were dichotomized to allow for comparisons with state and national figures and previously published reports. McNemar s test was used for comparing the changes in the risk factors in the cohort between year 1 and year 7. A paired t-test was used for comparing the wellness score and body mass index (BMI) between these time points. FIGURE 4. Obesity trends for the aggregate and cohort groups of Vanderbilt s GFTG Program and comparison to national and Tennessee norms from Behavioral Risk Factor Surveillance System. Obesity was defined as a BMI 30. RESULTS Participation in the Go for the Gold Program A central goal of GFTG was to achieve high annual participation rates. Before GFTG was introduced, the annual HRA completion rate was less than 24%. This increased to 68% in the first year after introduction of the incentive program and continued to increase to 80% and more in years 4 to 7 (Table 1). Also during this time, a greater proportion of participants enrolled in the highest level, the gold level, compared with the silver and bronze levels. Of the 17,335 GFTG Program participants in 2009, 71% achieved gold (HRA, WAL, and Game Plan), 6% silver (HRA and WAL), and 23% bronze levels (HRA only) (Appendix B [see Supplemental Digital Content 2, Characteristics of Participants Participation in the GFTG Program was broad and fairly representative of the underlying employee population; however, the nonparticipants differed in several ways (Table 2). Nonparticipants were 1374 C 2011 American College of Occupational and Environmental Medicine

4 JOEM Volume 53, Number 12, December 2011 Vanderbilt University Workplace Health Promotion Trends TABLE 1. Characteristics of the Participants Aggregate and Cohort Data for 7 Years Year 1 (2003) 2 (2004) 3 (2005) 4 (2006) 5 (2007) 6 (2008) 7 (2009) Benefits-eligible 15,070 16,097 17,247 18,701 19,810 20,494 21,701 employees* Aggregate participants (n = 10,248) (n = 10,463) (n = 12,444) (n = 14,698) (n = 15,811) (n = 16,764) (n = 17,335) Participation rate 68% 65% 72% 79% 80% 82% 80% Age (yrs) 40.4 ± ± ± ± ± ± ± 11.7 (18 83) (18 79) (18 80) (18 81) (18 82) (18 82) (18 83) Gender Male 3,275 (32.0%) 3,260 (31.2%) 3,899 (31.3%) 4,611 (31.4%) 4,880 (30.9%) 5,153 (30.7%) 5,327 (30.7%) Female 6,973 (68.0%) 7,203 (68.8%) 8,545 (68.7%) 10,087 (68.6%) 10,931 (69.1%) 11,611 (69.3%) 12,008 (69.3%) Cohort participants (N = 3745), participation rate 48% (7,802 benefits eligible employees all 7 yrs) Age (yrs) 43 ± 9.4 (19 77) 44 ± 9.4 (20 78) 45 ± 9.4 (21 79) 46 ± 9.4 (22 80) 47 ± 9.4 (23 81) 48 ± 9.4 (24 82) 49 ± 9.4 (25 83) Gender Male 1,098 (29.3%) 1,098 (29.3%) 1,098 (29.3%) 1,098 (29.3%) 1,098 (29.3%) 1,098 (29.3%) 1,098 (29.3%) Female 2,647 (70.7%) 2,647 (70.7%) 2,647 (70.7%) 2,647 (70.7%) 2,647 (70.7%) 2,647 (70.7%) 2,647 (70.7%) *Total number of benefits-eligible employees defined as those eligible for health care coverage, as determined by Human Resources Benefits on the last day of the GFTG Program year (November 30, all active, full-time, regular faculty and staff). Participation in GFTG Program was defined as completing the HRA in that calendar year. Age is mean ± SD (range). TABLE 2. Comparison of Go for the Gold Participants and Nonparticipants, Year 1 (2003) and Year 7 (2009) Year 1 (2003) Year 7 (2009) Nonparticipants Participants Participation Rate a (%) P Nonparticipants Participants Participation Rate a (%) P Age (yrs) b 44.0 ± ± 10.8 <0.001* 45.1 ± ± 12.3 <0.001* Gender <0.001** <0.001** Female 3,153 6, ,653 11, Male 2,350 3, ,732 4, Job classification <0.001** <0.001** House staff Research Associate Staff 4,055 8, ,492 13, Faculty , Medical group Location 0.023** 0.001** Medical center 4,017 7, ,366 12, University 1,486 2, ,019 3, Race <0.001** <0.001** Asian , White 3,856 8, ,106 12, Hispanic African American 1,290 1, , American Indian Other *P values are based on the Mann-Whitney U test. **P values are based on the Pearson s chi-square test. a Participation rate is calculated by dividing the number of participants in a specific category (eg, female participants) by the sum of nonparticipants and participants in that category. b Age is mean ± SD (range). C 2011 American College of Occupational and Environmental Medicine 1375

5 Byrne et al JOEM Volume 53, Number 12, December 2011 more likely to be older, male, and African American. For the aggregate population, the age of employees was bimodal and the average increased slightly from 40.4 in year 1 to 41.2 in year 7, with a range of 18 to 83 (Table 2). Participants averaged 4 years in the program. The cohort population over the 7-year period (3745 employees) was similar to the aggregate population in terms of age and gender (Table 1). Trends for Specific Risk Factors Aggregate Group Analysis Aggregate data collected between 2003 and 2009 were analyzed for trends in specific risk factors. Employees overall risk levels across the seven study years are displayed in Table 3. The largest decrease in the percent high risk occurred between years 1 and 2 ( 1.2%) and then remained relatively stable for the next 5 years. The mean overall wellness score increased from 52.5 to 57.9 from year 1 to year 7. From year 1 to year 2 (2003 to 2004), dramatic improvements were observed in the overall wellness score and percentage of low risk. A similar effect was seen between the first and second years for those who entered the program in later years (data not shown). Several of the biometric risk factors showed an increase in the percentage at high risk over time, reflecting the increased awareness of existing conditions. The percent with high cholesterol in year 1 was 9.7%, which increased annually (on average) by 0.1% to 10.4% in year 7. Self-reported hypertension also increased in the 7-year period, from 14.8% to 17.0% (an increase of 2.2%). The percentage of employees taking medication for their elevated cholesterol or blood pressure also increased during the 7-year period. Obesity rates showed a plateau effect; the percent of obese in 2003 was 37.6%, which increased only slightly to 38.4% in Eleven behavioral and psychosocial risk factors decreased across the 7 years: physical inactivity; poor nutrition; smoking; lack of seat belt use; excess alcohol use; high stress; poor perception of health; life dissatisfaction; job dissatisfaction; use of drugs for relaxation or sleep; and five or more illness days (Table 3). The percentage of sedentary showed the greatest improvement over the 7-year period, both in the percentage decrease and the average annual percentage difference. At baseline, the proportion of Vanderbilt University employees who reported not exercising one or more days per week was 27.3%. The greatest decrease occurred between years 1 and 2 when this rate dropped from 27.3% to 20.9%. The percent of lacking adequate physical activity decreased from year 1 to year 7 by 10.7% percentage points (a 1.8 percentage point annual average decrease). Poor nutrition, smoking, safety belt nonusage, and high stress followed a similar pattern to physical activity as risks decreased over the 7-year study period, with percentage point reductions of 4.4%, 3.3%, 7.9%, and 3.5%, respectively, across the entire time period. In a separate analysis, the percentage at high risk for smoking was stratified by work location, comparing smoking rates for medical center and university employees. The year 1 smoking rate of medical center employees was 12.1%, while the rate was 9.7% for university employees. The gap remained during the first 5 years of the study period, but in year 6, the year after a smoking ban was introduced at the medical center, the difference between smoking rates of the two groups narrowed, with 8.8% of medical center employees and 8.2% of university employees reporting that they smoked. The gap narrowed again in year 7 to 8.3% for medical center employees and 8.0% for university employees. Cohort Group Analysis There were 7802 employees eligible to participate in all 7 years of the program. A total of 3745 Vanderbilt University employees elected to complete an HRA each year of the GFTG Program between 2003 and 2009, thus achieving a participation rate of 48% (3745/7802). These 3745 comprised the cohort group. This group represented 36.5% of the employees who completed a year 1 HRA and 18.8% of employees who were eligible to participate on average in each of the 7 years. Table 4 displays the prevalence of risk factors for the cohort group across the seven study years. The percent of high risk (five or more risks factors) for the cohort was 7.0% in year 1. This figure dropped to 3.8% in year 2 but then gradually increased to 4.5% by year 7. Despite the gradual increase, the percentage at high risk from year 1 to year 7 showed a net decrease of 2.5 percentage points. Similar to the aggregate group, most biometric risk factors for the cohort group showed an increase over time. The percent of the members of the cohort group who were obese was 27.3% in year 1. The percentage of obese decreased by 2.1 percentage points from year 1 to year 2 but then increased each year thereafter to 28.0% in year 7. Over the 7-year period, the percent of obese employees increased by 0.7 percentage points. The behavioral and psychosocial risks at baseline with the highest percentage of employees at risk were poor nutrition (90.5%), physical inactivity (26.8%), frequent use of drugs for relaxation or sleep (10.6%), high blood pressure (15.5%), five or more illness days, (11.8%), and seat belt nonuse (11.3%). The percentage for poor nutrition, sedentary, smoking, and seat belt nonuse, all decreased from year 1 to year 7; percentage point decreases for these factors were 11.9%, 15.5%, 3.5%, and 7.7%, respectively. The sedentary risk factor showed the greatest improvement over the 7-year period, both in the percentage decrease and the average annual percentage difference. Many of the findings observed in the aggregate group were also observed in the cohort group over the 7-year study period. Both aggregate and cohort population demonstrated an overall decrease in the proportion of participants at overall high risk (five or more high risks) from 2003 to 2009, 1.8 percentage points and 2.5 percentage points, respectively. Physical inactivity, poor nutrition, smoking, high stress, and seat belt nonuse were the risk factors that demonstrated the greatest decrease in percentage at risk for both the aggregate and cohort populations. Findings of an increase in percentage at risk over 7 years for biometric factors in the aggregate group corresponded to similar findings in the cohort group. National and State Comparisons In 2003, year 1 of the GFTG Program, the percentage of adults who reported exercising at least one day a week was 77.3% in the United States and 70.2% in Tennessee. 13 The Vanderbilt University aggregate group and cohort group prevalence for physical activity were between these two benchmark values at 72.7% and 73.2%, respectively. Whereas the United States and Tennessee populations have made progress toward achieving the Healthy People 2010 goal of 80% or greater exercising at least one day a week, this goal was reached by the Vanderbilt aggregate group in year 3 (2005) and by the Vanderbilt cohort group in year 2 (2004). The percent of adults who reported smoking was 22.0% in the United States and 25.6% in Tennessee in 2003 (the baseline year for the study), while the Vanderbilt University aggregate group and cohort group percentages were considerably lower at 11.5% and 10.5%, respectively. While the United States and Tennessee populations reduced their smoking rates and made progress toward achieving the Healthy People 2010 goal of 12% or lower, the Vanderbilt community already had fewer than 12% of adults reporting smoking in year 1. The 2009, smoking rates in the United States (17.9%) and Tennessee (22.0%) remained considerably higher than those observed in either the Vanderbilt aggregate (8.2%) or cohort groups (5.8%). The percent of adults who reported always wearing a seat belt was 79.0% in the United States and 68.5% in Tennessee in 2003, C 2011 American College of Occupational and Environmental Medicine

6 JOEM Volume 53, Number 12, December 2011 Vanderbilt University Workplace Health Promotion Trends TABLE 3. Risk Factors Aggregate Data for 7 Years* Percentage Point Difference Years 7 1 Average Annual Percentage Difference Year 1 (2003) Year 2 (2004) Year 3 (2005) Year 4 (2006) Year 5 (2007) Year 6 (2008) Year 7 (2009) Participants n = 10,248 n = 10,463 n = 12,444 n = 14,698 n = 15,811 n = 16,764 n = 17,335 Overall wellness score 52.5 ± ± ± ± ± ± ± Median (range) 49.0 (10 96) 53.0 (10 99) 49.0 (10 98) 49.0 (10 98) 49.0 (10 97) 60.0 (10 99) 64.0 (10 98) Overall risk level Low risk (0 2 high risks) 7,587 (74.0%) 8,274 (79.1%) 9,926 (79.8%) 11,749 (79.9%) 12,655 (80.0%) 13,524 (80.7%) 13,989 (80.7%) Medium risk (3 4 high risks) 2,049 (20.0%) 1,685 (16.1%) 1,939 (15.6%) 2,285 (15.5%) 2,477 (15.7%) 2,526 (15.1%) 2,616 (15.1%) High risk (5 or more high risks) 612 (6.0%) 504 (4.8%) 579 (4.6%) 664 (4.5%) 679 (4.3%) 714 (4.3%) 730 (4.2%) Biometric risks High total cholesterol 998 (9.7%) 1081 (10.3%) 1,238 (9.9%) 1,516 (10.3%) 1,640 (10.4%) 1,765 (10.5%) 1,797 (10.4%) Not taking medication 493 (49.4%) 481 (44.5%) 519 (41.9%) 628 (41.4%) 651 (39.7%) 657 (37.2%) 647 (36.0%) Taking medication 505 (50.6%) 600 (55.5%) 719 (58.1%) 888 (58.6%) 989 (60.3%) 1,108 (62.8%) 1,150 (64.0%) High high-density lipoprotein 103 (5.6%) 47 (3.8%) 47 (3.7%) 52 (3.2%) 130 (7.3%) 96 (5.0%) 144 (7.0%) cholesterol (n = 103/1829) (n = 47/1238) (n = 47/1277) (n = 52/1606) (n = 130/1789) (n = 96/1919) (n = 144/2043) High blood pressure 1,520 (14.8%) 1,534 (14.7%) 1,882 (15.1%) 2,293 (15.6%) 2,573 (16.3%) 2,771 (16.5%) 2,950 (17.0%) Not taking medication 299 (19.7%) 246 (16.0%) 316 (16.8%) 413 (18.0%) 493 (19.2%) 506 (18.3%) 543 (18.4%) Taking medication 1,221 (80.3%) 1,288 (84.0%) 1,566 (83.2%) 1,880 (82.0%) 2,080 (80.8%) 2,265 (81.7%) 2,407 (81.6%) Body mass index, average 27.4 ± ± ± ± ± ± ± (range) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) Body mass index category <18 underweight 211 (2.1%) 199 (1.9%) 198 (1.6%) 217 (1.5%) 252 (1.6%) 277 (1.7%) 272 (1.6%) normal weight 4,383 (42.8%) 4,563 (43.6%) 5,443 (43.7%) 6,246 (42.5%) 6,626 (41.9%) 6,955 (41.5%) 7,251 (41.8%) overweight 2,919 (28.5%) 3,133 (29.9%) 3,680 (29.6%) 4,478 (30.5%) 4,809 (30.4%) 5,131 (30.6%) 5,239 (30.2%) obese 2,735 (26.7%) 2,568 (24.6%) 3,123 (25.1%) 3,757 (25.6%) 4,124 (26.1%) 4,401 (26.2%) 4,573 (26.4%) Behavioral risks Physical inactivity 2,794 (27.3%) 2,187 (20.9%) 2,429 (19.5%) 2,834 (19.3%) 2,864 (18.1%) 2,835 (16.9%) 2,877 (16.6%) <5 fruits and vegetables per day 9,387 (91.6%) 8,935 (85.4%) 10,851 (87.2%) 12,817 (87.2%) 13,534 (85.6%) 14,518 (86.6%) 15,116 (87.2%) Smoking behavior 1,179 (11.5%) 1146 (11.0%) 1,281 (10.3%) 1,491 (10.1%) 1,560 (9.9%) 1,452 (8.7%) 1,421 (8.2%) Excess alcohol use 101 (1.0%) 106 (1.0%) 99 (0.8%) 115 (0.8%) 136 (0.9%) 152 (0.9%) 149 (0.9%) Lack of safety belt usage 1,404 (13.7%) 1,125 (10.8%) 1,087 (8.7%) 1,243 (8.5%) 1,180 (7.5%) 1,146 (6.8%) 998 (5.8%) Psychosocial risks High stress 1,176 (11.5%) 940 (9.0%) 1,073 (8.6%) 1,222 (8.3%) 1,243 (7.9%) 1,298 (7.7%) 1,385 (8.0%) Poor/fair perception of health 500 (4.9%) 402 (3.8%) 467 (3.8%) 540 (3.7%) 596 (3.8%) 634 (3.8%) 685 (4.0%) Life dissatisfaction 400 (3.9%) 349 (3.3%) 434 (3.5%) 444 (3.0%) 481 (3.0%) 536 (3.2%) 515 (3.0%) Job dissatisfaction 1,172 (11.4%) 985 (9.4%) 1,095 (8.8%) 1,277 (8.7%) 1,420 (9.0%) 1,487 (8.9%) 1,616 (9.3%) Other Existing medical problem 707 (6.9%) 710 (6.8%) 867 (7.0%) 1,077 (7.3%) 1,191 (7.5%) 1,270 (7.6%) 1,371 (7.9%) Frequent use of drugs for 1,089 (10.6%) 1,036 (9.9%) 1,141 (9.2%) 1,390 (9.5%) 1,565 (9.9%) 1,559 (9.3%) 1,676 (9.7%) relaxation or sleep 5 illness days 1,290 (12.6%) 1,158 (11.1%) 1,326 (10.7%) 1,530 (10.4%) 1,616 (10.2%) 1,775 (10.6%) 1,813 (10.5%) C 2011 American College of Occupational and Environmental Medicine 1377 *Overall risk level is based on the number of risk factors in Appendix A. These risk factors are based on the research by Edington, et al. 12

7 Byrne et al JOEM Volume 53, Number 12, December 2011 TABLE 4. Risk Factors Cohort Data for 7 Years Participants (N = 3745) Year 1 (2003) Year 2 (2004) Year 3 (2005) Year 4 (2006) Year 5 (2007) Year 6 (2008) Year 7 (2009) Average Annual Percentage Difference Percentage Point Difference (Years 7 1) P Overall wellness score 53.5 ± ± ± ± ± ± ± <0.001* Median (range) 49.0 (10 95) 66.0 (10 97) 65.0 (10 96) 66.0 (10 96) 66.0 (10 96) 68.0 (10 97) 69.0 (10 98) Overall risk levels Low risk (0 2 high risks) 2,700 (72.1%) 3,025 (80.8%) 3,046 (81.3%) 3,038 (81.1%) 3,029 (80.9%) 3,009 (80.3%) 2,981 (79.6%) <0.001 Medium risk (3 4 high risks) 784 (20.9%) 576 (15.4%) 551 (14.7%) 561 (15.0%) 564 (15.1%) 573 (15.3%) 595 (15.9%) High risk (5 or more high risks) 261 (7.0%) 144 (3.8%) 148 (4.0%) 146 (3.9%) 152 (4.1%) 163 (4.4%) 169 (4.5%) Biometric risks High cholesterol 402 (10.7%) 453 (12.1%) 455 (12.1%) 482 (12.9%) 527 (14.1%) 548 (14.6%) 584 (15.6%) <0.001 Not taking medication 203 (50.5%) 208 (45.9%) 185 (40.7%) 181 (37.6%) 180 (34.2%) 169 (30.8%) 179 (30.7%) Taking medication 199 (49.5%) 245 (54.1%) 270 (59.3%) 301 (62.4%) 347 (65.8%) 379 (69.2%) 405 (69.3%) High high-density lipoprotein 44 (6.5%) 13 (2.4%) 15 (3.1%) 11 (1.9%) 33 (5.0%) 32 (4.5%) 55 (6.5%) cholesterol (n = 44/682) (n = 13/533) (n = 15/490) (n = 11/580) (n = 33/664) (n = 32/707) (n = 55/848) High blood pressure 580 (15.5%) 616 (16.4%) 657 (17.5%) 694 (18.5%) 763 (20.4%) 813 (21.7%) 878 (23.4%) <0.001 Not taking medication 109 (18.8%) 88 (14.3%) 109 (16.6%) 108 (15.6%) 129 (16.9%) 131 (16.1%) 126 (14.4%) Taking medication 471 (81.2%) 528 (85.7%) 548 (83.4%) 586 (84.4%) 634 (83.1%) 682 (83.9%) 752 (85.6%) Body mass index, average (range) 27.6 ± ± ± ± ± ± ± * ( ) ( ) ( ) ( ) ( ) ( ) ( ) Body mass index category <18 underweight 74 (2.0%) 62 (1.7%) 53 (1.4%) 40 (1.1%) 49 (1.3%) 45 (1.2%) 45 (1.2%) < normal weight 1,584 (42.3%) 1,627 (43.4%) 1,579 (42.2%) 1,559 (41.6%) 1,514 (40.4%) 1,488 (39.7%) 1,455 (38.9%) overweight 1,063 (28.4%) 1,112 (29.7%) 1,148 (30.7%) 1,155 (30.8%) 1,185 (31.6%) 1,174 (31.3%) 1,196 (31.9%) obese 1,024 (27.3%) 944 (25.2%) 965 (25.8%) 991 (26.5%) 997 (26.6%) 1,038 (27.7%) 1,049 (28.0%) Behavioral risks Physical inactivity 1004 (26.8%) 643 (17.2%) 563 (15.0%) 550 (14.7%) 517 (13.8%) 497 (13.3%) 424 (11.3%) <0.001 <5 fruits and vegetables per day 3,388 (90.5%) 3,026 (80.8%) 3,088 (82.5%) 3,013 (80.5%) 2,942 (78.6%) 2,926 (78.1%) 2,944 (78.6%) <0.001 Smoking 331 (8.8%) 313 (8.4%) 277 (7.4%) 255 (6.8%) 228 (6.1%) 204 (5.4%) 200 (5.3%) <0.001 Excess alcohol use 31 (0.8%) 27 (0.7%) 24 (0.6%) 22 (0.6%) 29 (0.8%) 25 (0.7%) 28 (0.7%) Lack of safety belt usage 424 (11.3%) 321 (8.6%) 245 (6.5%) 207 (5.5%) 176 (4.7%) 173 (4.6%) 134 (3.6%) <0.001 Psychosocial risks High stress 393 (10.5%) 306 (8.2%) 259 (6.9%) 261 (7.0%) 244 (6.5%) 225 (6.0%) 217 (5.8%) <0.001 Poor/fair perception of health 134 (3.6%) 114 (3.0%) 108 (2.9%) 110 (2.9%) 109 (2.9%) 112 (3.0%) 127 (3.4%) Life dissatisfaction 137 (3.7%) 113 (3.0%) 102 (2.7%) 101 (2.7%) 92 (2.5%) 102 (2.7%) 86 (2.3%) <0.001 Job dissatisfaction 379 (10.1%) 338 (9.0%) 303 (8.1%) 294 (7.9%) 319 (8.5%) 306 (8.2%) 318 (8.5%) Other Existing medical problem 246 (6.6%) 257 (6.9%) 287 (7.7%) 324 (8.7%) 360 (9.6%) 378 (10.1%) 406 (10.8%) <0.001 Frequent use of drugs for relaxation or sleep 355 (9.5%) 341 (9.1%) 326 (8.7%) 320 (8.5%) 355 (9.5%) 339 (9.1%) 385 (10.3%) illness days 443 (11.8%) 392 (10.5%) 415 (11.1%) 410 (10.9%) 392 (10.5%) 424 (11.3%) 428 (11.4%) *Paired t-test comparing year 1 with year 7. McNemar s Bowker test comparing year 1 with year 7. McNemar s test comparing year 1 with year C 2011 American College of Occupational and Environmental Medicine

8 JOEM Volume 53, Number 12, December 2011 Vanderbilt University Workplace Health Promotion Trends whereas the Vanderbilt University aggregate and cohort group percentages were considerably higher at 86.3% and 88.7%, respectively. Here again, while the United States and Tennessee populations seat belt utilization rates increased toward the Healthy People 2010 goal of 90% or greater, this goal was reached by the Vanderbilt aggregate group in year 3 (2005) and by the Vanderbilt cohort group in year 2 (2004). Finally, at baseline, the percent of adults who were obese was 22.9% in the United States and 25.0% in Tennessee. 13 Contrary to the previous comparisons, the Vanderbilt University aggregate group and cohort group percentages were higher than national and state norms at 26.7% and 27.3%, respectively. Over time, the United States and Tennessee obesity rates increased each year and reached 26.9% and 32.8%, respectively, in 2009, while the Vanderbilt obesity rate in the aggregate group was lower than these norms at 26.4% and the cohort group was in between these two prevalence rates at 28.0%. While none of the populations studied (national, state, or Vanderbilt employees) achieved the Healthy People 2010 goal of less than 15% obese, Vanderbilt did achieve the more realistic goal set for Healthy People 2020 set at less than 30% obese. DISCUSSION In this 7-year evaluation of the Vanderbilt GFTG Wellness Program, we observed a gradual increase in participation rates along with a decrease in the proportion of employees at high health risks. The pattern of risk reduction was found for the aggregate employee population by taking snapshots of risk prevalence for each of the study years, as well as for a cohort group that was followed in each of the 7 years observed. The most consistent improvements were found in increased physical activity, decreased poor nutrition, decreased smoking rates, and increased seat belt usage. For these four risk factors, the results suggest that health improvements are sustainable for at least 7 years without regression to the baseline. Specific findings, by risk category, are summarized later. The percentage exercising one or more days per week increased by 6.4 percentage points in the first year after the GFTG Program was introduced and then increased at a more gradual but consistent rate in subsequent years. During the 7 years of the GFTG Program, the net increase in the rate of physical activity was 10.7 percentage points. In contrast, during this same period, the rate of exercising at least one day a week for the United States and Tennessee populations decreased slightly. At Vanderbilt, several initiatives introduced during the study period may have prompted greater physical activity among employees. Physical activity was a question on the HRA, and on the WAL, employees were asked to report whether they had maintained recommended exercise levels, improved, or set a goal to exercise more for a defined period. Other programs supported these goals, including upgrading the fitness facility on campus, obtaining discounts to fitness facilities in the area, and designating Vanderbilt a walking campus with the creation of safe walking trails. Programs such as the American Heart Association s START! Physical Activity Program, a step tracker converting all exercise to steps, Walking Wednesdays, leadership walks, and the Lap It Up Swim Club were initiated to encourage various forms of exercise. The smoking rate at Vanderbilt decreased by 0.5 percentage points per year or 3.3 percentage points from 11.5% in 2003 to 8.2% in This rate of decrease was steeper than that of the United States (0.4 point annual decrease) and Tennessee (0.3 point annual decrease). In addition to the introduction of a medical center smoking ban in 2008 and the HRA and WAL, the university provided personalized smoking cessation support through the Quit Rx Smoking Cessation Program consisting of medical resources for nicotine replacement, coordination with primary care providers, one-on-one wellness coaching, a nicotine-anonymous group, and self-help tool kits. The percentage of Vanderbilt employees always wearing seat belts increased by 7.9 percentage points from 86.3% to 94.2% between 2003 and The rate of increase for the United States closely paralleled the increase at Vanderbilt, but the rate of increase in seat belt usage in Tennessee was steeper and resulted in a 13 percentage point increase from 68.5% to 80.6% during the same time period. In addition to being a focus of the HRA and the WAL, a Tennessee seat belt law, making it a primary offense for not wearing a seat belt, was enacted in 1986, with enforcement initiated in After an initial year 1 to year 2 decrease, obesity rates at Vanderbilt flattened during the 7 years of the GFTG Program, starting at 26.7% and 27.3% in year 1 for the aggregate and cohort groups, respectively, and remaining at a similar level (26.4% and 28.0%) in year 7. Changes in obesity rates at Vanderbilt (0.3 percentage point decrease in the aggregate group and 0.7 percentage point increase in the cohort group) were modest when compared to a 4.0 percentage point increase in the United States and a 7.8 percentage point increase in Tennessee during the same 7-year period. The HRA and WAL supported weight management as well as did several educational and environmental nutritional initiatives, Weight Watchers at Work, defined weight programs and a Game Plan for Your Health feature. The analysis here, and for normative populations, highlights the need for long-term monitoring of health risk factors to develop realistic expectations of a population s progress toward controlling weight. The Healthy People 2010 goal of an obesity rate less than 15.0% was not reached by the United States, Tennessee, or any other state. 13 The Healthy People 2020 goal of less than 30.3% of the population classified as obese may be more realistic but will require strong public health efforts at a state and national level. Employer efforts to manage overweight and obesity can support national and state initiatives addressing weight management, physical activity, and nutrition. At Vanderbilt, not all risk factors demonstrated a consistent change from year to year. As noted previously, the largest improvements occurred in the first 2 years of the program, particularly for the overall wellness score, percentage of low risk, poor nutrition, exercise, and seat belt usage. This effect was seen in both the aggregate and cohort groups. This is a key finding, which, to our knowledge, has not been previously reported since most published reports only describe two time points or have low participation in a follow-up result. These reports often make the assumption that each of the following years will result in similar dramatic improvements. We speculate that the year 1 to year 2 effects may be the result of a wake-up call when participants first learn about their heightened health risks and seek help to change a behavior. 15 The dramatic change from year 1 to year 2, which was seen each year in the new program cohorts, may also reflect an individuals readiness to change. For employees in contemplation or preparation stages, the program provided the extra push needed to move to action required for change. In subsequent years, increasing effort is necessary to take additional action, as well as maintain the changes made in year 1. Given competing professional and personal priorities, frequent programming changes were required to retain employee interest and continue to move them through the stages of change. Beyond the specific program elements, what might have accounted for overall health improvement and risk reduction among participants? Certainly, a powerful financial incentive was used to attract the attention of employees and enroll them in programs. The intent of the incentive was to draw their attention to workplace health promotion programs (ie, I can earn money by doing this ) and eventually transform the external incentive to one that is internalized by the individual when the new behaviors become intrinsically rewarding. To receive the incentive, employees were required to do more than just simply answer a 15-minute survey to earn their reward. They were also provided the opportunity to participate in a variety of follow-up health improvement programs, built on evidence-based behavior change theory, which would, in turn, reinforce long-term C 2011 American College of Occupational and Environmental Medicine 1379

9 Byrne et al JOEM Volume 53, Number 12, December 2011 healthy habit formation. Thus, the philosophy behind the program was to offer employees a compelling reason to first become involved and then ensure that the program was so engaging and motivational that participants naturally adopted a healthier lifestyle. The decision was made to develop the Vanderbilt wellness program internally because of the availability of expertise in preventive medicine, occupational health, and general preventive/public health. Health Plus and the GFTG programs were developed out of the occupational medicine program utilizing technology, communications, biostatistics, research (evaluation), and clinical expertise available at the university. By using internal resources, it was cost-effective and efficient to deliver in-house programming. In addition, having the program reside in house made it feasible to introduce frequent changes to the program and develop positive relationships and trust with employees. The university made the decision to address prevention in the employee population as part of its health care mission, as well as to address employee satisfaction, health care costs, and improved worker productivity. Future Research Since this intervention was not randomized and lacked a control group, future research should test whether similar changes occur using a prospective, randomized crossover study design with a true control group. There are several challenges in conducting such a study. First, the ethical issues in randomizing some employees to a control group must be addressed. Second, there are legal constraints related to offering wellness benefits to some employees and not others. Third, the control group would need to complete an HRA each year and receive no follow-up programming to have a comparison group. Fourth, the control group would need to be at a completely separate worksite to avoid contamination from the social networking effects of the intervention. The crossover design, in which half of the employees are assigned to the control group for the first few years and then switched over to the intervention arm, might make it feasible to conduct such a study, especially if an employer is unable to initially fund a full-scale wellness program. Exploring the feasibility of such a study design is warranted. Limitations This article examines the trends in self-reported risk factors, which are potentially biased. Such a bias, however, would likely be constant throughout the entire study period, and there is a substantial body of literature supporting the reliability of self-reported data for these purposes. 16 Furthermore, incentives were provided for participation not outcomes and therefore, there is little reason to believe that people intentionally reported healthier results. In 2003, we compared self-reported weight and height with that measured through biometric testing at the fitness facility and found good agreement (data not shown). Although the GFTG Program had high participation, another limitation is related to the potential for dropout bias. Of the 3293 participants in year 1 who did not participate in the following year, one third no longer worked at Vanderbilt (33.7%) and another 5.2% became part-time or temporary employees and were no longer eligible to participate in GFTG. Of those who participated in the wellness program in year 1, and were also eligible in year 7, survival analysis showed that the following factors were associated with dropout: younger age (those in their 20s at baseline were more likely to drop out); males; smokers; those not wearing a seat belt; and employees with hypertension. Thus, the bias associated with this attrition could be producing trends that are overly optimistic. The consistent results found in the aggregate and cohort analyses, however, suggest that the bias is minimal. Study findings might be the result of a Hawthorne effect in which participants who know that they are being studied change their behaviors or give an expected answer based on what they think the researcher wants to hear. This is an unlikely explanation for the results observed for several reasons. First, we would expect a regression toward baseline if there was a Hawthorne effect. Second, continually offering socially desirable responses would have produced large improvements in years 2 through 7 similar to those found when comparing years 1 and 2 results, which did not occur. Third, no significant improvements were observed in several other self-reported risk factors, such as hypertension, hypercholesterolemia, and BMI, which would have been impacted if there was a strong Hawthorne effect. The current study also lacks a true control group to provide information on trends in health risk factors for employees not participating in the GFTG Program. Trends observed in the aggregate and cohort groups may be due to employees participation in GFTG, but without a true nonparticipant control group, a direct cause effect relationship may not be inferred. In the absence of a control or comparison group, the Vanderbilt data were contrasted with national and state norms for specific risk factors. CONCLUSIONS At Vanderbilt University, results show that an incentive-driven voluntary wellness program can drive participation and health risk improvement in an employee population over a long time horizon of 7 years. Reductions were observed for many risk factors measured, while others such as obesity did not worsen over time. In the absence of a control group, Vanderbilt trends were compared to national and state estimates. Of particular note, risks associated with physical inactivity, poor nutrition, smoking, and seat belt usage were reduced demonstrably over the course of 7 years. While improvements were most dramatic in the first 2 years of the program, continuous improvements were observed throughout the course of the study. Further improvements in employees health and health risks will require creating more intensive individualized risk-reduction programs, an even healthier social and physical work environment, and aligning incentives in the health care system with health improvement outcomes. ACKNOWLEDGEMENTS The authors thank Larry Chapman, Brad Awalt, Lisa Connor, Stacey Kendrick, Lori Rolando, Kimberly Tromatore, DeMoyne Culpepper, Madeline Garr, Lori Cowan, and Keith Franklin for their contributions to the program. REFERENCES 1. Whitmer RW, Pelletier KR, Anderson DR, Baase CM, Frost GJ. A wake-up call for corporate America. J Occup Environ Med. 2003;45: Christakis NA, Fowler JH. The spread of obesity in a large social network over 32 years. N Engl J Med. 2007;357: Christakis NA, Fowler JH. The collective dynamics of smoking in a large social network. N Engl J Med. 2008;358: Christakis NA, Fowler JH. Connected: the Surprising Power of Our Social Networks and how they Shape our Lives How Your Friends Friends Friends Affect Everything You Feel, Think and Do. New York, NY: Little, Brown and Company; Chapman LS. Proof Positive: An Analysis of the Cost Effectiveness of Worksite Wellness. Seattle, WA: Chapman Institute; Robroek SJ, van Lenthe FJ, van Empelen P, Burdorf A. Determinants of participation in worksite health promotion programmes: a systematic review. Int J Behav Nutr Phys Act. 2009;6: Personal Wellness Profile. Clackamas, OR: Wellsource, Inc. http//:www. wellsource.com. Accessed June 10, Reeves MJ, Rafferty AP. Healthy lifestyle characteristics among adults in the United States, Arch Intern Med. 2005;165: McGinnis JM, Foege WH. Actual causes of death in the United States. JAMA. 1993;270: C 2011 American College of Occupational and Environmental Medicine