Do corporate wellness programs actually work?

What the randomised trials found

The BJ's Wholesale Club trial, published in JAMA in 2019, randomised 160 worksites covering 32,974 employees. Twenty sites got a program of health screenings and 8 modules on nutrition, physical activity and stress, delivered by registered dietitians; the rest got nothing. Employees completed an average of 1.3 modules. After 18 months there were no significant differences in 10 clinical markers, including cholesterol, blood pressure and body mass index, in 38 measures of medical spending and use, or in absenteeism, job tenure and job performance.

The Illinois Workplace Wellness Study randomised 4,834 University of Illinois employees. The treatment group was offered rewards and paid time off for on-site biometric screenings, health risk assessments and activities such as physical activity, smoking cessation and disease management. Over 56% of those offered the program took part. In the first year there were no significant effects on total medical spending, health behaviours, productivity or self-reported health. At 24 months there were still no significant effects on 16 clinician-collected biometric measures, on diagnoses of diabetes, hypertension or high cholesterol, or on use of medical care.

Both programs changed something. BJ's employees in the program reported more regular exercise (69.8% vs 61.9%) and more active weight management (69.2% vs 54.7%). Illinois employees in the program were more likely to report having a primary care physician (92.2% vs 86.1%). These gains are real but modest, and they are not the health and cost improvements most programs are sold on.

Why older studies looked better

Many earlier studies compared employees who joined a program with those who did not. The Illinois study shows why that misleads. Before the program began, participants already had lower medical spending and healthier habits, and were more likely to have visited the campus gym and entered local running events. When the researchers ran that kind of non-randomised comparison on their own data, it produced spending estimates in line with earlier observational studies, but their randomised estimate ruled them out. Its 95% confidence intervals excluded 83% of previous estimates for medical spending and absenteeism.

RAND reached a similar view in 2014. Its Wellness Programs Study, covering almost 600,000 employees at seven employers, found little if any immediate effect on what employers spend on health care. In 10 years of data from one Fortune 100 employer, the overall return was $1.50 per dollar invested, but most of the savings came from one part of the program.

What works better: intensity and personal follow-up

In RAND's Fortune 100 analysis, the part that paid off was disease management for employees who already had a chronic condition. It produced 87% of the savings: participants saved $136 per member per month, driven in large measure by a nearly 30% fall in hospital admissions, a return of $3.80 per dollar. Lifestyle management for people with risks such as smoking and obesity returned $0.50, although RAND notes it did reduce absenteeism and that lifestyle programs can reduce health risks.

Intensity matters for prevention too. In the Diabetes Prevention Program, published in NEJM in 2002, 3,234 adults with raised blood glucose were randomised to placebo, metformin or a lifestyle program aiming for at least 7% weight loss and 150 minutes of activity a week. Over an average of 2.8 years, the lifestyle group developed diabetes at 4.8 cases per 100 person-years, against 11.0 on placebo, a 58% reduction. About 7 people had to take part to prevent one case over three years.

The US Preventive Services Task Force recommends (grade B) offering or referring adults with cardiovascular risk factors, such as high blood pressure or abnormal blood lipids, to behavioural counselling on diet and physical activity. The programs it reviewed usually involved a median of 12 contacts and about 6 hours of contact time over 6 to 18 months, mostly delivered by nonclinicians such as nurses, registered dietitians, nutritionists, exercise specialists and lifestyle coaches. People who received them had fewer cardiovascular events (pooled relative risk 0.81), with small average falls in blood pressure, LDL cholesterol and weight. Compare that with the 1.3 modules the average BJ's employee completed.

What works better: incentives and commitment

Money can change specific behaviours when the design is right. A trial published in NEJM in 2015 randomised 2,538 CVS Caremark employees and their friends and family to usual care or one of four programs to stop smoking. Two paid rewards of about $800 for quitting. Two asked people to put down a refundable $150 deposit, plus $650 in rewards if they succeeded.

All four programs beat usual care on sustained abstinence at 6 months (9.4% to 16.0%, against 6.0%). Reward programs did better overall than deposit programs (15.7% vs 10.2%), mainly because 90.0% of people accepted a reward program while only 13.7% accepted a deposit. Among people who would accept either, the deposit programs worked better, by an estimated 13.2 percentage points. Commitment works for the people willing to make it; rewards reach far more people.

What works better: social competition

The STEP UP trial, published in JAMA Internal Medicine in 2019, enrolled 602 adults with overweight or obesity from 40 states, recruited from Deloitte employees. Everyone tracked steps with a wearable. Three groups entered a 24-week game with points and levels for hitting step goals, built around support, collaboration or competition; the control group got only feedback from the device.

Compared with controls, daily steps rose by 920 in the competition arm, 689 in the support arm and 637 in the collaboration arm. In the 12 weeks after the game ended, only the competition arm stayed significantly ahead, by 569 steps a day. Activity was lower in every arm once the game stopped, which suggests challenges need to keep running.

How to judge a program before you buy

The trials point to a short list of questions that separate programs likely to change health from programs likely to change only engagement figures.

• Does it measure clinical markers at the start and the end, not just sign-ups and app logins?
• How many personal contacts does each participant get, with whom, and over what period?
• Does it do more for people with existing conditions or clear risk factors, rather than offer the same content to everyone?
• Are incentives tied to specific actions or outcomes, and how many people are expected to accept them?
• Are challenges built around team competition, and do they keep running after launch?
• Is any claimed return on investment based on a randomised comparison, or on participants against non-participants?