Abstract
Care Continuum Alliance (CCA) spearheaded the creation of a database on wellness programs and incentives with data from CCA member companies and their clients. The CCA database was used by RAND Corporation in the study.
The study analyzed data spanning 5 years from 7 employers with over 1.7 million eligible employees between the ages of 18 and 64. Most of the employers included were large scale; one had 15,000 employees. The database includes eligibility and medical claims (including pharmacy), health risk assessment data for over 200,000 unique individuals, and wellness/disease management program participation data.
The study's methodology and findings aroused some important questions among industry stakeholders, which CCA received and studied in collaboration with a group of researchers and experts, and released the RAND Report on Workplace Wellness: What Employers Must Know. This discussion is important enough to bring some of the most salient issues to this column.
The main takeaway? This study shows statistically significant and clinically meaningful effects on health risks and approximate cost neutrality for wellness programs. The industry continues to evolve. Since the data used in this study were compiled, there is greater use of technology, new approaches to improve engagement, and more robust measurements. Most importantly, wellness programs are most successful when offered as part of an organization's overall culture of health. Workplace wellness involves complex interventions, and success always will depend on the particular intervention within the particular context and the particular measures of outcomes used.
Earlier meta-analyses may have exacerbated rosy expectations in terms of cost savings and the overall immediate impact of these programs. But we must look at what the wellness programs are actually doing. It is not unrealistic that cost neutrality is a positive result because these programs are intervening in a workforce population that is not sick. These programs are a preventive effort to avoid future health care costs; therefore, achieving cost neutrality and better health is a positive result for both employers and employees.
To determine participation, the study used the definition from each respective data contributor, including their definition of minimal threshold of participation, for the comparison group. The effect that the definition might have had on the analysis was considered and participant engagement was assumed to be the program's responsibility. With that technique, researchers arrived at a real-world estimate of the program's impact. Further research is needed to include a dose-response curve and to examine the efficacy of continuous participation.
Wellness programs that include outcomes-based interventions remain rare, so there were not enough good data for meaningful statistical assessment. The researchers did not have enough data in the sample, and none of the employers' programs were tied to outcomes in order to assess specifically the impact on health care costs. A fine line exists between shifting risk and cost to more vulnerable employees and dependents and making employees feel compelled to take advantage of the programs offered. There is a very clear need for more research and data to find the right balance between appropriate risk sharing and inappropriate cost shifting.
Although incentives are commonly used for wellness program participation, the RAND Corporation researchers believe that the public debate is significantly ahead of the actual state of the field when it comes to incentives tied to health outcomes. It was surprising to find that high-powered incentives tied to health outcomes are much less common than the literature would have us think. Furthermore, those employers that use incentives stay well within the parameters defined in the Affordable Care Act.
Incentives were associated with significant improvements in smoking, body mass index (BMI), and exercise, yet the effect size was small. The challenge with this estimation was that the employers included in the analysis had little variation with incentives offered, both across employers and within each employer over time. Therefore, the study had a large sample size with little variation to run the regression analysis. The data indicate that higher incentives impact reduction in BMI with significant effect. When translated to pounds, the effect was still significant. However
The study's reach and level of granularity were limited because of time and funding; therefore, not all the components – online, in-person counseling, content, classes, support, communications – of wellness programs were examined. The researchers' wish list for follow-up study includes amount of exposure, level of interventions, and level of exposure.
The study did include medical and prescription drug costs, but not work loss, workers' compensation, or disability. Rather, this was a high-level look at overall costs without attempting to attribute them at a more granular level. However, the study did deduct years in which an employee was pregnant or participating in a case management program for high-cost, high-risk conditions.
Although other risk factors such as glucose, blood pressure, alcohol consumption, overall health status, and fruit and vegetable consumption were provided in the data, they were not included in the report. The required length of the report made it impossible to include all risk factors, so the list was prioritized to include the most common. Also, because of the variation in programs between employers, some of the data were not uniform and could not be included in the analysis.
Many have suggested the possibility of motivation selection bias because the study was not a randomized controlled trial. Although the difference-in-difference method that was used is powerful in observing such bias, we cannot rule out unobservable differences between participants and nonparticipants. It is possible that nonparticipants were exposed to and derived benefit from health promotion activities (eg, better food offerings, exercise messaging) at the worksite but didn't participate in a personalized counseling program and would not be identified by the analysis. Although the benefits gained by the nonparticipants could lead to other savings, the estimates presented in the report reflect the marginal impact of a lifestyle management component. Changes in environment would impact both participants and nonparticipants, but the study did not have means of comparing to other employers to determine that effect.
Some suggested that an established list of wellness-sensitive conditions would allow outcomes to be studied more specifically. In essence, we know many conditions are sensitive to health behaviors. The former Surgeon General attributes 75% of health care costs to behavior. Many cancers, and potentially asthma and chronic obstructive pulmonary disease, can be tied to obesity. But we must find a way to quantify the strength of that relationship in order to call a condition “wellness-sensitive.” To begin to unpack the “black box,” we should define wellness-sensitive conditions and look more closely at what can be done for a quicker response versus what can be expected to happen later in the process.
Press coverage made hay of the apparent insignificance of a 1 lb/year weight loss in a simulated hypothetical cohort of participants. However, this finding of a 1 lb/year reduction is significant because it represents the average weight loss of participants compared to nonparticipants on a population level (ie, 1 lb times the number of participants). Furthermore, participants continued to lose an average of 1 lb/year in the first and second years after the year of participation.
Future research should seek more granularity with related versus unrelated costs. Another area to consider is different types of outcomes, such as productivity and other work-related impacts. A larger sample size may show a significant effect on cost, so more employers in the database and a longer time series may indicate at what point the curves converge and reach statistical significance. Also, researchers in the population health management industry need to begin to unpack the black box. We cannot assume that all programs are effective or all programs are ineffective, and we need to understand the distinctions; specifically, how do employer (eg, culture, support) and employee (eg, health literacy, age, sex, ethnicity) characteristics drive changes so that targeting interventions become more effective?