Do Workplace Wellness Programs Work? What the Randomized Trials Found

At the University of Illinois, one wellness program earned two opposite verdicts from the same employees. Compared the usual way, participants against non-participants, it appeared to save about US$100 a month per participant in health spending (2016-2017). Compared by lottery, people offered it against people who were not, it showed no significant saving, and that figure was ruled out.¹

So do workplace wellness programs work? Mostly not in the ways they are sold. Two large randomized trials at US employers, published in 2019, found that comprehensive wellness programs did not measurably lower medical spending or absence, or improve job performance, over 18 to 30 months. Each moved an easier target: screening in one, habits people reported in the other.¹²

Three questions separate a real result from a flattering one, whoever presents it: compared with whom, which outcome moved, and for how long. The checklist further down turns them into a routine for any program you are offered.

Screenings, step challenges and classes are a small slice of what employee wellbeing covers beyond perks and programs. This page stays narrower: what the best trials found, why earlier results looked better, and what to check before paying.

Prefer to watch? The video covers the same ground as this article.

VideoDo Workplace Wellness Programs Work? What the Trials FoundWiserHours

Where the famous $3-for-$1 return came from

The case that wellness programs pay for themselves rests mostly on comparing employees who joined with employees who did not. A widely cited 2010 meta-analysis by Katherine Baicker, David Cutler and Zirui Song estimated that medical costs fell by about $3.27, and absence costs by about $2.73, for every dollar spent.³ Most studies behind such estimates were observational and open to selection bias, the authors of a later trial note.²

The problem has a name: self-selectionself-selection: When the people in a study chose to take part, or chose the thing being studied, rather than being picked or assigned. Volunteers and choosers can differ from everyone else, so the result may not apply more widely or may reflect who chose rather than what they chose.Full entry in the glossary. People decide whether to join, and that decision is not random. A colleague who already cycles to work signs up for the cycling challenge; the one working double shifts while caring for a parent does not.

A lunchtime running club’s members take fewer sick days the next year than colleagues who stayed at their desks. Did the club do that, or did fit people join a running club? A spreadsheet of members and non-members cannot tell you.

Myth

Wellness programs pay for themselves about three times over.

Fact

Most studies behind such estimates compared joiners with non-joiners. Randomized trials at US employers found no significant drop in medical spending within three years.

So when a savings figure lands on your desk, ask first: compared with whom? If the answer is “employees who did not take part”, the figure describes the joiners at least as much as the program.

Illinois: a lottery decided who got the offer

The Illinois Workplace Wellness Study let chance decide which employees of the University of Illinois at Urbana-Champaign were offered a two-year wellness program. Economists Damon Jones, David Molitor and Julian Reif reported in 2019 that screening rose, while medical spending, other health habits, productivity and self-rated health showed no significant change.¹

The study

Moderate evidence

A lottery decided: 4,834 University of Illinois employees and a two-year program

A lottery gave 3,300 of the 4,834 employees who joined the study the offer: weekly wellness activities, a yearly on-site health screening and online health risk assessment, paid time off to take part, and rewards of $50 to $650 over two years. The other 1,534 were not allowed to take part. Up to 30 months later, the offered and non-offered groups showed no significant difference in medical spending, campus gym visits, sick leave, promotion or job satisfaction. The results rule out 84 percent of the estimates on medical spending and absence reported in 112 earlier studies. US federal research bodies and non-profit groups funded the study.¹

The more useful finding concerns who took part. Employees who completed the screening had spent less on health care in the year or so before the study, and were more likely to have run in local races and used the campus gyms; those who stayed away were more likely to be in the lowest pay quartile. Analyzed the old way, participants against non-participants, the data produced the apparent saving described at the top, in line with earlier observational studies. The lottery comparison ruled it out.¹

1. Joiners compared with non-joiners: the people who sign up differ before the program starts; in the Illinois trial they had already spent less on health care in the year or so before, so this comparison showed a saving like those in earlier studies
2. Offered compared with not offered, by lottery: chance puts a similar mix of people in both groups, so any later difference belongs to the program; in the Illinois trial this comparison found no significant effect on medical spending

Bar heights are schematic, not data.

Nothing about the program differed between the two analyses; only the comparison did. A result that follows participants alone mostly tells you who they were. A fair test compares people offered the program with people not offered it, with chance deciding which is which. Illinois was one employer and one design, which is why a second trial matters.

The BJ’s Wholesale trial: habits reported, health and costs unchanged

Zirui Song and Katherine Baicker ran the second trial at BJ’s Wholesale Club, a US warehouse retailer, and published it in JAMA in 2019. Eighteen months in, staff at the worksites given the program were more likely to say they exercised regularly and were actively managing their weight, but clinical measures, health spending, absence and job performance showed no significant difference.²

This trial assigned whole worksites rather than individuals, which the authors suggest may help change workplace culture and increase social support. The program, eight modules on nutrition, physical activity and stress reduction led by registered dietitians, resembled what many US employers buy. The US National Institute on Aging, the Robert Wood Johnson Foundation and J-PAL North America funded it, with logistical support from BJ’s. Why would reported habits move while clinical measures and costs stayed put? The authors offer two readings: behavior change may come before health gains that take longer to show, or it may simply be easier to achieve.²

Waiting longer did not change the picture. At three years, reported habits at program sites were still better, with no detectable improvement in clinical, economic or employment outcomes.⁴

1. 1What people say they doregular exercise, weight management: moved in the BJ's trial
2. 2What a clinical test measuresblood pressure, cholesterol, body mass index: no significant change
3. 3What the organization recordshealth spending, absence, job performance: no significant change

“More employees say they exercise” is a weaker result than “fewer have high blood pressure”, which is weaker again than “absence fell”. When someone reports a program’s success, ask which of the three it reached.

Mindfulness apps and resilience courses: Fleming’s UK analysis

Individual mental wellbeing offers fared no better in a large UK dataset. In a 2024 study of survey data from 46,336 workers in 233 organizations, William Fleming of the University of Oxford’s Wellbeing Research Centre found that workers who took part in resilience training, mindfulness, wellbeing apps and similar offers appeared no better off than those who did not, across several measures of how they rated their own wellbeing. This was a cross-sectionalcross-sectional: Describes a study that measures everyone in its sample at a single point in time. Because the possible cause and the outcome are recorded together, it can show that two things occur together but not which of them came first.Full entry in the glossary comparison, not a trial, and selection can cut both ways: people who are struggling may be the ones who sign up, which could hide a real benefit. The study examined that possible bias, and Fleming reads the results as these offers not supplying the resources people need to meet the demands of their jobs.⁵

The evidence is contested rather than settled: stress-management courses, such as ones based on mindfulness, receive a conditional recommendation on low-certainty evidence in WHO’s 2022 guidelines.⁶ A nurse offered a meditation app while working a rota with no break cover may find it helps her unwind, but it cannot give her the break. Treat such offers as a possible extra.

Do workplace wellness programs work? The evidence at a glance

The two randomized trials of full programs are the strongest tests of the question, and they found the least.

One or two null trials do not prove that every program fails. A 2021 meta-analysis in The Lancet Public Health, pooling 121 controlled studies of multicomponent programs from North America, Europe and elsewhere, linked them to small average gains in diet and body weight; improvements in blood pressure and cholesterol lost statistical significance once the authors corrected for likely publication biaspublication bias: The tendency for studies with striking or statistically significant results to be published, while quieter ones stay in the drawer. Since the missing studies are mostly the unimpressive ones, a pooled result tends to overstate the effect until it is corrected.Full entry in the glossary.⁸ The large trials do shift the burden of proof onto whoever claims a program works.

Testing a wellness program’s claims before you pay for it

The two trials hand you a ready test for any program’s evidence, whether it comes from a vendor, a consultant or your own HR team. The most useful question is whether the program was judged on everyone it was offered to, against a similar group that was not offered it.

Next, check which outcome moved and for how long: more sign-ups or better self-reported habits after a few months say little about health or absence two years later. Ask who paid for the evaluation, and whether the people you most want to reach, such as those with the heaviest workloads or the lowest pay, took part.

If you run a program anyway, build in a fair test. The BJ’s trial chose program worksites at random;² a larger employer can do something similar by starting at randomly chosen sites or teams and comparing them with the rest. If you change the work itself, as the next section suggests, WHO’s guidelines add that monitoring should track work and mental health outcomes without identifying any individual’s health.⁶

Questions to ask about any wellness program's results

• Was it compared with a similar group not offered it, chosen by chance?
• Does the result count everyone offered the program, not just the people who joined?
• Which outcome changed: self-reported habits, clinical measures, or spending, absence and performance?
• How long was the follow-up: months, or a year or more?
• Who paid for the evaluation, and who ran it?
• Did the people under most strain take part, or mainly the already healthy?
• Can we start at randomly chosen sites or teams first and compare?
• Will monitoring use grouped data only, so no individual can be identified?

Download PDF ↓

Get this checklist as a one-page PDF. We'll email you the download link.

Your emailAlso send me the weekly newsletter (launching soon)Email me the PDF

We use your address to send this link and, only if you tick the box, the newsletter. Privacy policy.

What the evidence points to instead: workload, control and management

WHO’s mental health at work guidelines (2022) list changes to the work itself first: organizational measures against psychosocial risks such as work overload and inflexible schedules, including ones designed with workers’ participation. WHO grades the evidence for these as very low certainty. It strongly recommends training managers in supporting their staff’s mental health, backed by moderate-certainty evidence.⁶

Other evidence is suggestive. A 2007 systematic review led by UK researchers found 18 studies of reorganizations that gave employees more participation and control, none of them randomized. Most controlled studies found some health benefits, especially for mental health, when control improved, but two participatory changes made alongside redundancies were followed by worse health.⁹ A say in the work does not offset generally poor conditions.

Why aim at the work when that evidence is also thin? Because an opt-in program only helps those who opt in. The Illinois authors suggest one explanation for their results: those who might benefit most, such as people who smoke or have high medical costs, may decline to join even when offered large rewards.¹ Fix the work, and the fix lands on everyone doing it, including people who would never sign up for anything. The hazards to start with, from overload to bullying, are set out in how to spot and manage psychosocial hazards.

A claims team’s targets rose while staffing did not. A wellness budget could buy a meditation app. The same attention could let the team choose the order of its queue, cover breaks at peak times and cut a weekly report to monthly. Neither option has strong trial evidence, but only the second changes the working conditions themselves.

Fix the work, and the fix lands on everyone doing it, including people who would never sign up for anything.

If work strain is affecting your health: where to get help

Neither a wellness program nor a better-run job replaces health care. The urgency tiers below are our own signposting; the routes come from NIMH, a US government research institute, and from the UK’s NHS, and serve as examples. Readers elsewhere should look up their own emergency and health services.

• Now: where a life may be at stake, contact your local emergency number first. As a US example, NIMH points to 911 or the closest emergency room when life is threatened, and to the 988 Suicide & Crisis Lifeline, by phone or text, for suicidal feelings or emotional distress.¹⁰ As a UK example, the NHS directs people to 999 or A&E straight away when a person’s life is in danger, and also if you feel you cannot keep yourself or someone else safe; its list of free listening services includes Samaritans, on 116 123.¹¹
• Soon: stress, low mood or anxiety that is not lifting deserves an appointment with a family doctor. In England, the NHS offers two routes for urgent mental health needs: 111, or an urgent appointment with a GP.¹¹
• Routine: strain that returns again and again is worth mentioning at a normal check-up. NIMH also suggests checking with HR for an employee assistance program, a service it describes as free and confidential, with the employer paying.¹⁰

The bottom line

Judged by lottery at two large US employers, comprehensive wellness programs lifted screening and reported habits but left medical spending, absence and job performance without significant change for up to three years. The larger savings reported earlier mostly reflected who signed up. Before paying for a program, ask for results on everyone offered it against a comparable group that was not, and give at least as much attention to workload, control and management, where WHO’s guidelines start.