I hope this thread receives some further discussion as currently (in US at least) there is tremendous enthusiasm (and expected bonanza in funding) related to the projects based on data mining of electronic medical records (EMR) (mostly within new Comparative Effectiveness Research program).
Steve wrote: "Remember that medical records constitute observational data, and suffer all the limitations of
observational data." However, this argument has not diminished the enthusiasm (and marketing) for ever increasing number of data-mining projects (as, for example, witnessed by increasing number of the quality measurements/performance projects, demands for public reporting of health outcomes, studying all kind of associations such as the link between statins and dementia, etc, etc). So, the question is: can we define circumstances/types of the projects (design, "rules of evidence" etc) when data-mining is more or less acceptable? (Is there "user guide" how to design and critically appraise data-mining reports?)
My second comments relates to the 6S model (pyramid of evidence) (Systems -- the computerized decision support that occurs automatically within the system being used and not requiring external searching is the
most efficient): this is an aspirational goal. I am not aware that such systems have been developed (outside some specific, decision-support systems) and very much doubt that they can be developed to tackle the entire field of medicine (despite tremendous progress in artificial intelligence, computing etc)
ben djulbegovic
________________________________________
From: Evidence based health (EBH) [[log in to unmask]] On Behalf Of Steve Simon, P.Mean Consulting [[log in to unmask]]
Sent: Tuesday, September 21, 2010 11:48 PM
To: [log in to unmask]
Subject: Re: Data mining in Medical Records?
On 9/21/2010 3:53 PM, Ernesto Barrera wrote:
> I have known a data mining software that calculates predictive models
> without much statistical knowledge. I wonder if the automatic
> incorporation of predictive models to computerized medical records,
> patient risk modeling based on regional or local data could be a future.
> In that case, could we say that we move towards a model of 7S
> (predictive Statistics) in the pyramid of evidence?
>
> I appreciate your views on this subject.
I know a videotape that you can watch that will allow you to perform
surgery without much knowledge of anatomy. How well do you think that
would work?
Predictive models based on computerized medical records are a laudable
goal, but only if they incorporate the appropriate statistical
expertise. Maybe it will change in a decade or two, but today's
computers are not smart enough yet to dispense with expert knowledge.
Even an intelligently built predictive model based on medical records,
though, will have serious limitations. Remember that medical records
constitute observational data, and suffer all the limitations of
observational data.
Full disclosure: I am a professional statistician who will be put out of
work as soon as the computers get smart enough that they can build good
predictive models without much statistical knowledge.
--
Steve Simon, Standard Disclaimer
Sign up for The Monthly Mean, the newsletter that
dares to call itself "average" at www.pmean.com/news
|