STATS ARTICLES 2011
Literal heartbreak from losing the Super Bowl or silly statistics?
Rebecca Goldin Ph.D, February 3, 2011
What do you get when you combine an alarming study with the signature U.S. sporting event?
The media loves a good game. And with the Super Bowl coming up, they just couldn’t resist reporting on a the results of a new study published in Clinical Cardiology which appears to claim that losing the Super Bowl is bad for our hearts. But the media was gamed: it lost its critical eye to simple questions of study design, and even causation versus correlation.
The study compared death rates due to circulatory diseases in Los Angeles during the two weeks after the Super Bowl in 1980 with other days in late January and February in 1980, 1981, 1982, and 1983. In 1980, Los Angeles played against none other than the Pittsburgh Steelers, and L.A. lost.
The study made the same comparison with Super Bowl 1984, when Los Angeles played against the Washington Redskins, and won. The authors compared death rates on Super Bowl Sunday and the following two weeks in 1984 to other days in January and February in 1984, 195, 1986 and 1987. In 1984, the Los Angeles Rams won.
The findings were perhaps impressive at first glance. Death went up for people in Los Angeles in 1980, and went down for people in Los Angeles in 1984. But on closer inspection the study is riddled with problems:
- The authors compared 1980 with subsequent years (and similarly 1984 with subsequent years). Perhaps there are trends in the demographics or the health care that are driving changes over a period of years.
- There were many February days compared to January days. In 1980, the Super Bowl occurred on January 20. The comparison days (January 15-19, and February 4-29, 1980 and comparable days in 1981, 1982, 1983) included many more February days than January days. If there are more deaths in January than in February, then the study would be biased to see a result of increasing deaths. January is commonly thought to be riskier than other months, in part due to the holidays immediately before. It was for this reason that the authors excluded January 1-14.
- The authors tested many possible ways of slicing the data, such as male/female, older/younger. They did not adjust their statistics for multiple testing.
- They only looked at Los Angeles. Did they see a corresponding benefit in Pittsburgh when the Steelers won in 1980? Was there an increase in circulatory deaths in Washington DC when the Redskins lost in 1984?
- The authors insisted on causality, repeatedly referring to the Super Bowl as having “triggered” the increase in deaths.
- No one observed whether fans or even partners of fans were more affected than those who don’t give a hoot about the Super Bowl.
Nonetheless, the news media reacted like parakeets repeating what they hear: “The Super Bowl can cause a heart attack,” said the Toronto Star; led with the headline, “Super Bowl may increase rate of mortality from heart attacks,” headlined CNN, while The Atlantic Wire insisted, “Super Bowl madness could kill you.”
Should we really just skip the game and relax with a big steak? As far as this study is concerned, no doctor is going to tell you to sit it out.