A number of studies have pointed to the potential benefits of workplace wellness programmes — from increased gym use to decreased health-care spending. But what wasn’t clear was whether wellness programs are really making people healthier, or whether healthier people are more likely to enroll in them. A new study suggests it’s the latter.
Long before this insight was revealed by science, it was reflected in an episode of The Office, of all places. And the lesson learned might be applied to all sorts of other preventive health measures, from vitamin supplements to diet and exercise trends. That TV series wasn’t about health or science, but it did deal in the idiosyncratic nature of human behavior, and this matters a lot in medical studies — especially ones involving preventive health.
In the episode in question, the company offered a prize for the branch office that lost the most collective weight. Real workplace wellness presumably wouldn’t be this ill-conceived, but the fictional result was insightful. In the Scranton office, there were several characters who looked like they might have benefited from losing weight. They wanted nothing to do with the contest, and had no interest in going on a diet.
Needless to say, nobody’s state of wellness improved. But the story did illustrate something medical researchers say is important: There’s a difference between the people who want to participate and those who don’t. Healthier people are more likely to opt into wellness programs, exercise trends, vitamins and other preventive health measures. That effect can sometimes create the illusion of benefits that don’t really exist, especially in so-called observational studies, where subjects are divided depending on their own choices.
The new study, which involved close to 5,000 volunteers working for the University of Illinois Urbana-Champaign, was not an observational study, but what’s known as a randomised controlled trial. That means people were randomly divided in two groups, and one group was offered a workplace wellness program and one was not. When the researchers compared the two randomly divided groups they found no difference in gym attendance, health-care spending or other benefits that had been attributed to these programs.
Physician-writer Aaron Carroll explained in his New York Times column that because previous studies were observational studies, they can introduce selection bias. When the authors of the new study examined their data as if it had been an observational study, they saw that only about half the people in the intervention group opted in, and when those were compared to the ones who had the opportunity but didn’t take it, the data showed clear differences in gym visits, health-care and hospital spending, and participation in running races. But it’s unlikely the wellness program was the cause.
This doesn’t mean people shouldn’t do observational studies, or that the public should ignore them, said Alan Brookhart, a professor of epidemiology at the University of North Carolina, Chapel Hill. A few years ago, he and a pair of colleagues wrote a primer for physicians on interpreting observational studies.
They identified a number of patterns that might prove misleading. They called one the “healthy user effect” — the tendency for healthier people to use more preventive measures across the board — because they are more health conscious, in many cases, or more conscientious, or wealthier.
To illustrate the extent to which these effects can bias research, he pointed to a study that showed that people who adhered to a regime of heart-attack preventing statin drugs were much less likely to die in car accidents. Here it’s obvious that the statins aren’t protecting people on the road, but that the same kinds of people who would be responsible about taking a preventive drug would also be conscientious drivers and wear their seat belts.
In the primer, he and the other researchers argued that this same kind of bias led the medical community to wrongly assume that hormone replacement therapy prevented heart disease in post-menopausal women. Doctors began prescribing hormones to women as a routine preventive measure. Randomised controlled trials later showed no cardiovascular benefit and a range of previously unidentified harms.
On the other end of the spectrum, people who are very sick are less likely to partake of preventive health measures, and it can then look like their opting out caused them to die, when it was the fact that they were dying that caused them to opt out.
Brookhart said this explains some otherwise unlikely data showing that in an elderly population, people who got flu shots were 50% less likely to die than those who didn’t. This included deaths from all causes — not just flu. How could a flu vaccine have such an effect on non-flu deaths? To investigate, he said, he and his colleagues looked at the same correlation before flu season started. It was still 50%, but at that point there would have been very few cases of flu. The correlation was more likely caused by the fact that people who were near death, for a host of reasons, were much less likely to get a flu vaccine.
William Shrank, lead author on the observational studies primer and now chief medical officer for the University of Pittsburgh Medical Center insurance service, agreed that observational studies can be a great tool for medical research. They can allow researchers to study larger populations and longer-term effects, in some cases. But people using the studies, whether doctors or laypeople our journalists, have to be mindful of cause and effect. The correlation between statins and automobile fatalities was real, after all, but the interpretation needed some thought.
And despite the negative implications of this newest study, workplace wellness programs may still benefit employees. It may be that the health-conscious would still go to the gym, or yoga class, and would still get their flu shots and have their blood pressure checked, but they may still appreciate employers making it easier for them. Office-wide weight-loss contests, on the other hand, are probably not advisable.