Sunday, January 23, 2011
Could Medicine Learn Something From Sports Statistics?
In my teen years I was a baseball statistic junkie. I stayed up late playing Earl Weaver Baseball on my Amiga 500, played Pursue The Pennant and had a wicked collection of baseball cards, that now are all gone (*sniff*). I fell out of that phase but in a recent ESPN Magazine article I was inspired to think again about how medicine often struggles to be creative at times, especially when it comes to our data. We have tons of data in health care, but do we really analyze it in the best way possible?
So what is the stat which got me thinking? Not PAP, BABIP, VORP, WARP, wOBA or xFIP. And yes if you think medicine has cornered the market on acronyms you would be mistaken. The stat is Win Probability. From ESPN Magazine, Win Probability is the expected chance that a team will win a game at a particular moment in time given the situation it faces. And if you know anything about me, you know I am fascinated by prognosis and the lack of research into this very vital question. And Win Probability is probably the most innovative way to think about prognostication I have heard in a while.
Here is an example graph of the Sunday night's NFL playoff game between the New York Jets and the Pittsburgh Steelers:
If you go to the site Advanced NFL Stats you can actually move the cursor on the graph and see what event was happening and what the situation was that changed the Win Probability. What is nice about Win Probability is that it can help identify what the crucial plays (treatments/complications) are in a game (hospital course) which may influence the outcome. I think the ability for a family or patient to understand this may be much better than broad statements of prognostication or even the ranges so familiar to palliative practitioners.
The magazine article discusses that how we experience sports is really related emotionally and viscerally to the Win Probability. A big socre changes the odds in our teams favor. A successful surgery for a patient greatly improves her chances of survival. How you would actually implement Win Probability in health care deserves a day long conference and think tank session, but a football game is an unbelievably complex situation with chaos and rules and parts that are manageable and parts that are unexpected, just like the rest of life, or weather which contrary to common wisdom has greatly increased predictive ability.
Are there any other statistics from sports we could help us in medical decision making?
Other Win Probability sites: Baseball FanGraphs
Are there any other statistics from sports we could help us in medical decision making?
Other Win Probability sites: Baseball FanGraphs
Subscribe to:
Post Comments (Atom)
10 Responses to “Could Medicine Learn Something From Sports Statistics?”
January 24, 2011
Similarly, it hit me the other day that the modeling done on 538.com with its averaging of half-lifed, source-error-weighted data, and use of nearest-neighbor, historical-data analysis to fill in holes, etc. could be useful in prognostication.
It seems to me that these approaches are humblingly accurate in predicting outcomes, in the midst of our own real-time, experience-immersion, uncertainty and bias.
January 24, 2011
Fantastic post Christian, and comment by Brian.
I am admittedly a sports junkie - lover of fantasy sports & all things related to sports statistics. You know you have a problem when you've read a book called HockeyNomics.
You may have creasted a monster as I now will have to ponder any connections I can come up with between sports statistics that relate to medicine, pharmacy, pain & palliative care.
Well done!
January 24, 2011
And, to come full circle, let the record reflect that the brains behind 538.com, Nate Silver, got his start as a statistician and wrier for the blog Baseball Prospectus. http://www.baseballprospectus.com/
January 27, 2011
Wired magazine has recently been running a series on infographics (how to present information more effectively). In November they published an article on how to improve lab reports. You might also want to see their staff MD, Thomas Goetz, in a TEDMed talk on a similar subject. Fun!
January 27, 2011
Wired magazine has recently been running a series on infographics (how to present information more effectively). In November they published an article on how to improve lab reports. You might also want to see their staff MD, Thomas Goetz, in a TEDMed talk on a similar subject.
January 27, 2011
Tasha,
That is pretty funny that you mentioned that. I just watched Goetz's presentation last night. Will have to read the Wired magazine report. I thought the TED talk was fair, not the best I have seen. I wish he spent more time on the results and not so much on the lead up.
Thanks for posting!
January 27, 2011
Brian and Chris,
Thanks for posting you guys. I am dying to go see Edward Tufte at one of his conferences. (Tufte was mentioned by Goetz in the TED talk FWIW)
538.com is a great blog. Maybe with you giving them a shout out, we could get him to cover medical statistics.
The other way to look at data is crowd-sourcing, like the Hollywood Stock Exchange. ALthough it may be admittedly more complex and fraught with ethical challenges, the 'crowd' as defined by care providers, family and patient may best be able to figure out what the prognosis is, if everyone is willing to be honest and minimally biased.
January 28, 2011
Hi Christian,
Thanks for the nice post. It reminded me of two books: Moneyball and Super Crunchers. I'm sure there is a metric out there that can help with prognostication but as mentioned requires data via engagement of physicians, families, and patients etc...(but wait, we are in the age of engagement, right?)
I've been thinking Jamie Heywood's ALS talk and wondering how it would scale to EOL http://bit.ly/fE1eJZ
March 27, 2011
Wired magazine has recently been running a series on infographics (how to present information more effectively). In November they published an article on how to improve lab reports. You might also want to see their staff MD, Thomas Goetz, in a TEDMed talk on a similar subject. Fun!
March 27, 2011
Fantastic post Christian, and comment by Brian.
I am admittedly a sports junkie - lover of fantasy sports & all things related to sports statistics. You know you have a problem when you've read a book called HockeyNomics.
You may have creasted a monster as I now will have to ponder any connections I can come up with between sports statistics that relate to medicine, pharmacy, pain & palliative care.
Well done!
Post a Comment