Do Workplace Wellness Programs Work? What the Evidence Says

Evidence grade: Strong. Several independent studies with credible causal identification point the same way. This grade describes the published evidence. It is not a prediction about your workforce and not a result RetainScore has produced.

Strong evidence, and it says no

Wellness programs are unusual among retention interventions: they have been tested properly, twice, at scale, with random assignment and follow-up out to three years. The grade is strong because the trials are good and agree with each other. What they agree on is that a wellness program does not change whether people stay, how often they are absent, what they cost to insure, or their measured health, within the horizon anyone has looked at. It changes some of what they say they do.

If work means improve a few self-reported health behaviours, the answer is yes, modestly. If it means any of the outcomes a wellness program is bought for, the answer, so far, is no.

The two trials

A retailer, 160 worksites, three years

Song and Baicker ran the largest test on record. "This clustered randomized trial was implemented at 160 worksites from January 2015 through June 2016": twenty worksites of a large US warehouse retailer were assigned a dietitian-led program of nutrition, activity and stress modules, 140 were not, and 32,974 employees were followed through claims, employment records, surveys and biometric screenings.

At 18 months two things had moved. Employees at program sites were more likely to report exercising regularly, "for engaging in regular exercise (69.8% vs 61.9%; adjusted difference, 8.3 percentage points", and more likely to report managing their weight. Nothing else did: not the other 27 self-reported outcomes, not the ten clinical markers, not the 38 spending and utilization measures, and not the "3 employment outcomes (absenteeism, job tenure, and job performance)".

The three-year follow-up, published in Health Affairs in 2021, is the important one for a retention question, because three years is long enough for a program to have kept anyone. It did not. "No significant differences were found in self-reported health; clinical markers of health; health care spending or use; or absenteeism, tenure, or job performance". Tenure here is not a survey answer; it is "Tenure (days employed during the treatment period)" from the employer's records for every worker ever employed at the sites. The three-year difference was 5.43 days, inside a confidence interval running from minus 6.44 to plus 17.29. The authors' own summary is that the results "temper expectations of substantial improvements in health outcomes or financial returns on investment from wellness programs up to a three-year horizon".

A university, 4,834 employees, randomized one by one

Jones, Molitor and Reif built a comprehensive program at the University of Illinois, screening plus incentivised wellness activities, and randomized eligibility at the level of the individual employee, which removes the objection that whole worksites differ. More than half of eligible employees took part.

On the employer's own records, "we do not detect statistically significant effects on any of the three outcomes that are administratively measured: annual salary, the probability of job termination" after twelve months, or sick leave. In a footnote they are precise about retention: "Our causal effects analysis finds positive, albeit small and insignificant, effects of the intervention on retention after one year". At 24 months, the medical results published in JAMA Internal Medicine were the same shape: the program "had no significant effects on measured physical health outcomes, rates of medical diagnoses, or the use of health care services after 24 months", though more employees had a primary care physician and thought better of their own health.

Two further findings from Illinois matter more than the nulls. First, the program selected. Before it began, "program participants had lower medical expenditures and healthier behaviors than non-participants". People who join wellness programs were already the healthier, cheaper, likelier-to-stay employees, which is exactly why every comparison of participants with non-participants ever published looks good. Second, the one thing the program clearly changed on the employment side was a belief: treated employees were 5.7 percentage points more likely "to believe that management places a priority on health and safety". That is a real effect and a cheap one, and it is not retention.

Where the return-on-investment figure came from

The number that sells these programs is from a 2010 meta-analysis by Baicker, Cutler and Song: "medical costs fall by about $3.27 for every dollar spent on wellness programs and that absenteeism costs fall by about $2.73 for every dollar spent". It is "Authors' calculations based on 36 studies", most of which compared participants with matched or unmatched groups rather than assigning anyone at random, which is the design the selection finding above breaks.

The Illinois trial tested those numbers directly. "Our 95% confidence intervals rule out 83 percent of previous estimates on medical spending and absenteeism". It is worth noticing who ran the retailer trial that confirmed this: the same Song and Baicker. The authors of the return-on-investment figure went and did the experiment that undercut it, and published the result. That is how the literature is supposed to work, and it is why the circulating figure is now a decade out of date.

Why the grade is strong

Strong is the grade for several independent studies with credible causal identification pointing the same way. Two trials, one randomizing worksites and one randomizing individuals, with follow-ups at 12, 18, 24 and 36 months, agree on retention, absence, spending and clinical health. Strong evidence of no effect is still strong evidence, and it is rarer in this field than strong evidence of an effect.

The honest boundaries

• Two programs, two US employers. A dietitian-led module program at a retailer and a screening-plus-incentives program at a university. Other designs are not tested to this standard, and RAND's 2013 review for the US government noted that "Work-related outcomes, such as absenteeism, productivity, and retention, are of critical importance to employers" and largely unmeasured in the evaluation literature it reviewed.
• Small effects are not excluded. The tenure interval in the retailer trial runs to about 17 days over three years. A program could keep people a fortnight longer and the trial would not see it. It could not keep them a year.
• Participation was partial. Roughly a third to a half of employees took part, which is normal for these programs and is part of what is being tested. The trials estimate what offering a program does to a workforce, which is the decision an employer actually makes.
• Three years is the horizon. Nobody has followed a randomized wellness program longer. Claims about decade-long returns are unmeasured, not disproven.

If you run one anyway

• Run it for the reasons the trials support. More employees exercising, managing weight, holding a primary care physician, and believing that management cares about their health. Those are real. They are not a retention line.
• Do not book a return. The $3.27 figure is from studies the trials have since ruled out. A business case built on it will not survive a finance director who reads the 2019 paper.
• Never compare participants with non-participants. Participants were healthier and likelier to stay before the program existed. Compare sites, or cohorts before and after eligibility, or accept that you cannot tell.
• If burnout is the problem, name it. A wellness program is frequently bought as a response to exhaustion, which is one of the stronger predictors of leaving in the meta-analytic literature. The retention diagnostic scores that directly, and the burnout page covers what the research finds drives it, which is workload and management rather than the absence of a step-count challenge.

Read the companion grade on whether employee engagement surveys work, another intervention whose measured effects sit some distance from the outcome it is bought for.