A while back I blogged about some important work being done by Dr. Jesse Polansky and colleagues:
I'm pleased to report that their work has now been published in the Journal of General Internal Medicine--that's the good news. The bad news is that most people reading the JGIM paper would miss the important connections between their findings and drug industry marketing. (My guess is that the editorial review process might have toned down some of the more pointed comments in the original text.) You might be able to find the article at http://www.springerlink.com/content/x266735764406234/fulltext.html-- but I suspect it is available to journal subscribers only.
Anyhow, here's a brief reminder summary plus the missing piece. There are two versions of the Framingham risk calculator that is recommended by most guidelines for physicians to use to decide how high-risk a patient is for coronary disease, which then leads to the decision to prescribe a statin, and what target cholesterol level to aim for. One directly calculates risk and the other assigns points. So long as you are doing paper-pencil calculations, the point system is much easier to manage; but if you're computerized, it's just as easy to use the one as the other.
The point system turns out to be less accurate than the full calculator, and the errors are not random. About twice as many people will be classified by the point system erroneously as needing more statins, compared to those misidentified as needing less.
What got this whole thing going was the observation that when drug company money was somewhere in the neighborhood (for instance, Epocrates, an educational medical-reference website that is heavily supported by industry advertising), the on-line risk calculator was much more likely to be based on the inaccurate point system, rather than the more accurate "full" calculator. But there is no way for the average doc to be aware that this bias is present when she's rushing to calculate a patient's risk score so that she can get that patient out of the office and get on to the next patient.
The published article goes into all the details about why the two calculators are different and just how many patients in each general category are likely to be erroneously classified when using the point-based model. Sadly, it soft-pedals all the links to industry funding and marketing. So docs now know that there's two models of the Framingham calculator out there and that using one of them is more likely to lead to error. What they don't know from this article is that drug industry funding is a fairly good predictor of whether they'll run into the less accurate (but more profitable for statin sales) calculator.
Gordon WG, Polansky JM, Boscardin WJ, Fung KZ, Steinman MA. Coronary risk assessment by point-based vs. equation-based Framingham models: significant implications for clinical care. Journal of General Internal Medicine, DOI 10.1007/s11606-101-1454-2; published online 8 Sept. 2010