What do Workplace Wellness Programs do? Evidence from the Illinois Workplace Wellness Study

Abstract

Workplace wellness programs cover over 50 million U.S. workers and are intended to reduce medical spending, increase productivity, and improve well-being. Yet limited evidence exists to support these claims. We designed and implemented a comprehensive workplace wellness program for a large employer and randomly assigned program eligibility and financial incentives at the individual level for nearly 5,000 employees. We find strong patterns of selection: during the year prior to the intervention, program participants had lower medical expenditures and healthier behaviors than nonparticipants. The program persistently increased health screening rates, but we do not find significant causal effects of treatment on total medical expenditures, other health behaviors, employee productivity, or self-reported health status after more than two years. Our 95% confidence intervals rule out 84% of previous estimates on medical spending and absenteeism.

Figure I

Experimental Design of the Illinois Workplace Wellness Study

Figure II

Employee Participation Rates in the iThrive Workplace Wellness Program Participation rates are measured as a fraction of the treatment group (N = 3,300).

Figure III

Preintervention Medical Spending among Treatment Group, by Participation Status Data are from claims covering the period July 2015–July 2016 (N = 2,188). The first two bins ($0 and (0−25]) include 25% of those not screened. The remaining five bins were defined to include 25%, 25%, 15%, 5%, and 5% of those not screened, respectively. The null hypothesis of the Pearson’s chi-squared and the nonparametric Kolmogorov-Smirnov tests is that the two samples are drawn from the same distribution.

Figure IV

Preintervention Salary among Treatment Group, by Participation Status Salary was measured on June 1, 2016 (N = 3,257). The six bins were defined to include 25%, 25%, 25%, 15%, 5%, and 5% of employees not screened, respectively. The null hypothesis of the Pearson’s chi-squared and the nonparametric Kolmogorov-Smirnov tests is that the two samples are drawn from the same distribution.

Figure V

Postintervention Medical Spending by Treatment Status Results based on health care claims over the 12-month period August 2016–July 2017. The null hypothesis of the Pearson’s chi-squared and the nonparametric Kolmogorov-Smirnov tests is that the two samples are drawn from the same distribution.

Figure VI

Comparison of Experimental Estimates to Prior Studies Each panel shows the distribution of N point estimates from prior workplace wellness studies. Panel A plots intent-to-treat (ITT) and treatment-on-the-treated (TOT) estimates for medical spending. Panel B plots corresponding estimates for absenteeism. The point estimates from our own study (RCT Estimate), and their associated confidence intervals, are taken from Online Appendix Table A.11, column (3) for medical spending and Table III, column (3) and Table IV, column (2) for absenteeism. Our randomized controled trial (RCT) estimates and confidence intervals are plotted to demonstrate the share of prior study point estimates we can rule out. Online Appendix Table B.1 provides the full details of this meta-analysis.