Two large randomized trials tested whether workplace wellness programmes cut medical spending. Neither found savings. What the evidence actually supports — and why insurers keep funding prevention anyway.
The pitch is intuitive: pay a little for prevention now, save a lot on claims later. Wellness programmes — screenings, coaching, gym benefits, step challenges — have been sold to employers and insurers on that arithmetic for decades. The strongest evidence we have says the arithmetic, at least in its simple form, does not hold. When wellness programmes were finally tested with randomized controlled trials rather than before-and-after comparisons, they changed some behaviours and changed essentially nothing about medical spending within the periods studied.
That does not make prevention worthless. It makes the honest question sharper: what exactly are insurers buying when they fund it?
What the randomized trials found
For most of the industry's history, wellness ROI claims rested on observational studies: compare employees who joined a programme with those who did not, observe that joiners cost less, credit the programme. Two large randomized trials in the United States dismantled that method by removing its central flaw.
The first, published in JAMA in 2019 by Zirui Song and Katherine Baicker, randomized worksites of the retailer BJ's Wholesale Club — 20 treatment sites with about 4,000 employees against 140 control sites with nearly 29,000 — over 18 months. Employees at programme sites reported meaningfully better health behaviours: regular exercise was 8.3 percentage points higher, and active weight management 13.6 points higher. But across ten clinical markers such as blood pressure and cholesterol, dozens of spending and utilisation measures, and employment outcomes including absenteeism and job performance, the trial found no significant differences.
The second, the Illinois Workplace Wellness Study, randomized roughly 5,000 university employees over two years and added the decisive finding: selection. Employees who chose to participate already had lower medical spending and healthier behaviours before the programme existed. The programme itself produced no significant effect on medical spending, diagnoses or health behaviours at 24 months — and the study's confidence intervals were tight enough to rule out the large majority of savings estimates published by the older observational literature. What the observational studies had been measuring was not wellness changing people; it was healthy people choosing wellness.
The selection finding is the important one. Wellness programmes do not primarily make members healthy — they primarily attract members who already are.
Why the savings fail to appear
Three mechanisms explain the gap between intuition and evidence. Timing: the conditions wellness aims to prevent — cardiovascular disease, diabetes complications — develop over decades, while trials, budgets and insurance contracts run in years; even a genuinely effective programme would show its claims impact long after the cohort has churned to another insurer or employer. Detection: screening, the most common wellness component, finds conditions, and found conditions generate claims before they save any — near-term spending can rise, not fall. And dilution: participation skews toward the already-healthy, so the members whose claims a programme might actually move are the least likely to show up.
So why do insurers keep funding prevention?
Because ROI-on-claims was never the only return, and the defensible returns are real even where the claimed one is not.
- Selection as strategy: if wellness engagement attracts healthier members, a wellness-branded product can build a healthier portfolio — the same selection effect that invalidated the old studies, working in the insurer's favour at acquisition.
- Engagement and retention: insurance is a low-contact product; a wellness layer creates touchpoints, and members who interact with a product are plausibly likelier to renew — a distribution benefit, not a claims one.
- Data: engagement generates consented signals that can inform group-level pricing and product design, within the bounds privacy rules allow.
- Targeted clinical value: the trials tested broad-population programmes. Narrow, clinically-anchored interventions — chronic disease management for diagnosed members, medication adherence, maternity pathways — target people whose claims are already in motion, and are a different economic proposition than step challenges for the fit.
The practical conclusion for anyone designing or buying these products: fund broad wellness as marketing, engagement and selection, and hold it to marketing metrics; fund targeted disease management as clinical intervention, and hold it to clinical metrics. The category error is funding the first while claiming the metrics of the second.
Limitations
The two trials cited are US studies of employer populations over 18 to 24 months; they say little about longer horizons, other health systems, or clinically-targeted programmes, and evidence on those narrower interventions is more mixed and harder to generalise. Behaviour change itself — which both trials partly found, and one measured clearly — has value this analysis does not price. The claim here is specific: the evidence does not support expecting broad workplace wellness programmes to reduce medical claims within a typical policy horizon. Anyone selling them on that promise is selling ahead of the data.