Sunday, February 06, 2011

Publishing for News Events

It's always interesting the see the incredible media buzz a so-called "scientific study" can receive if its release is timed to some upcoming predictable news event. This weekend's story is, of course, Superbowl XLV.

Just in the nick of time came this study entitled "Role of Age, Sex, and Race on Cardiac and Total Mortality Associated With Super Bowl Wins and Losses" published online in an online publication called "Clinical Cardiology" on 31 January 2011.

Coincidence?

Unlikely.

The study reportedly went back and looked at LA county coroner death certificates:
from 1980 to 1988, covering only the period of January 15 through the end of February for each year. Data included total number of deaths as well as number of deaths due to diseases of the circulatory system, ischemic heart disease, acute myocardial infarction, heart failure, and congestive heart failure. Data on the entire population have been previously published.9 In the present study, we separately analyzed data on men, women, those aged < 65 years, those aged ≥ 65 years, those of white/Hispanic ethnicity, and those of nonwhite/non-Hispanic ethnicity, for each of the 2 Super Bowls. We compared death rates for days related to the 1980 Super Bowl (January 20, 1980 plus the following 14 days), which Los Angeles lost, with control days (all other days in January and February from January 15, 1980 to February 1983). To remove the impact of the known peak in total and cardiac death rates around the winter holidays, all analyses excluded data from January 1 to January 14. Death rates were reported as deaths per day per 100 000 population (using total deaths and average population). Parallel analyses were performed for Los Angeles's winning Super Bowl on January 22, 1984. During these time periods, Los Angeles County death certificates classified Hispanic patients as white; hence, white and Hispanic death-rate data are combined.
The authors ran some statistics on this narrow 14-day window from the various years and concluded:
The 1980 Super Bowl loss triggered more deaths in older patients compared with younger patients. It did trigger deaths in both men and women. In addition, there was a nonsignificant trend for a Super Bowl win to reduce death slightly better in older than in younger patients, and in women.
Needless to say, the media have had a field day with this article published just before the Superbowl. A quick search of Google at the time of this posting has identified 12,641 articles that have been written regarding this publication (excluding my current post). Without exception (at least as far as I could discern), all articles have assumed the article's conclusion is scientifically valid.

For medical students and residents, it might be helpful to dissect this study and its methodology so that they can enjoy the festivities without fear of dropping dead if their team loses.

First of all, this study is limited by a number of significant research bias's:

Selection bias (short window of dates, limited years studied)

Contamination bias (How do we explain the deaths of those who never saw the outcome of the game?)

Co-intervention bias (Recall that air polution standards in 1980 in Los Angeles were much worse in 1980 than later years and may have served as a co-intervention bias not explained by the researchers). The lack of a limitations section in this report to offer alternate explanations to the author's findings suggest an expectation bias and is a helpful sign of a poorly analyzed data set.

Finally, the use of diagnosis codes culled from death certificates hardly explains the mechanism of someone's death, but rather classifies (poorly) their most likely major underlying disease state. The retrospective nature of this trial can only point the way for future research, rather than imply causation.

So there you have it. Even if your team loses, you're not likely to have a heart attack.

Unless, of course, you have a hefty bet on the game.

-Wes

No comments: