Can’t Get Enough of Those Correlations
We see a pattern. Take a look at the list of the top attention-getting stories on the JAMA network for 2019. Because if you do, you will see that most of them are about correlations. Or associations. Or links. In other words, they’re not about experimental evidence of causality or effects. It seems that we can’t get enough of those correlations.
Are some of them spurious? We suspect that the risk is greater than zero.
Push-ups for Heart Health
Far and away, the number one paper of 2019 in the JAMA network was a study of push-ups. More than half a million people read this paper. The authors found that firefighters who could do at least 11 push-ups had fewer heart attacks, less heart failure, and fewer sudden cardiac deaths. To their credit, they didn’t claim that push-ups guarantee a healthy heart and a long life. They merely suggest that push-up capacity might be a cheap measure of cardiovascular disease risk.
A useful correlation
Correlations Get the Clicks
It used to be that money talks. But now, clicks do the talking for ad sales. And it’s the correlation studies that get all the clicks. Thus, those stories are the source of viral health news buzz. Our good friend Gary Foster pointed out to us that nine of the top ten most viral stories on JAMA Network Open were about an association. These results are based on Altimetric scores. They’re a measure of how much attention research gets in both traditional and social media.
The buzziest stories land on top.
Contempt for Familiarity
Familiarity breeds contempt and we must admit it. Even though some of these correlations can provide important insights, we are weary with too many of them. That’s because repeated reporting on a correlation – for example between diet soda and poor health – creates an illusion of causality.
This is called a familiarity bias. And yes, we know. You’re immune to such bias. Research tells us that everyone thinks they can rise above their biases. But guess what. None of us have that superpower. We’re all human.
So we suggest that everyone think twice about how much we venerate and depend upon correlations.
Click here for the review of the top-ten most-read stories on the JAMA network in 2019.
Dots, photograph © Barbara W / flickr
Subscribe by email to follow the accumulating evidence and observations that shape our view of health, obesity, and policy.
December 24, 2019
December 24, 2019 at 10:16 am, David Brown said:
Good post. Correlation science is mainstream. Experimental evidence is backwater. Why? It’s complicated and decidedly political. Here is what experimental science is up against.
“BIOLOGIST SANJOY GHOSH is a rebel with a cause. Sure, his rebellion is of the archetypal University Professor kind—moral, accountable, collegial. But as a nutritional scientist, he’s decidedly antiestablishment.”
“The global food economy—a formidable regime, what with its cereal mountains, factory farms, and deep-fried GMOs—looms large over Dr. Ghosh and his international research team at UBC’s Okanagan campus.”
“Ghosh leads the Dietary Interventions and Better Exercise through Experimental Science Centre (D.I.A.B.E.T.E.S. lab). His research team develops new studies that investigate the detriments of polyunsaturated fats (PUFA)—a cheap, go-to fat in mass-produced consumer products.”
https://ourstories.ok.ubc.ca/stories/sanjoy-ghosh/
December 29, 2019 at 6:24 am, Giovanni Tarantino said:
Correlation does not mean causation, nevertheless is one of the most powerful technique to study any phenomenon…. if third factors influencing it are excluded!