Friday, November 20, 2009

Making Mammography Screening Guidelines: Two approaches

So how did the "experts" at the US Preventive Services Task Force (USPSTF) come up with the "new" mammography guidelines, anyway?

As you read in yesterday's post, the decision was based primarily (or fully) on the findings from epidemiological data (i.e. population-level health research, rather than medical case studies, case histories).

While an epidemiologist (...a "social epidemiologist," really) at heart, I am taking a decision sciences class that is completely fascinating and speaks to the pitfall of relying on critical research reviews and other ways to "understand the evidence" in a traditional way. For example, rather having the USPSTF make recommendations based on their understanding of epidemiological literature alone, it could use that literature to inform a decision sciences or decision analysis approach, instead.

Here's what I'm thinking...

So without getting to far into it, here's the basic decision science methodology, as I (a newbie) understand it. You create a lot of "trees" that reflect the probability of a particular outcome for a decision that you may make. For example, you can model the survival (the outcome) of a woman with breast cancer that receives a mammogram every year (decision #1) versus a woman that receives a mammogram every two years (decision #2). You can determine how survival is difference given certain predisposing characteristics (or risk factors -- family history of breast cancer, for example) and also consider what her survival would be if she tests positive and truly has breast cancer, as well as if she tests positive and does not have breast cancer.

There are problems with this approach, of course -- for example, if you are basing your decisions off of faulty probabilities or inaccurate scientific data. Nonetheless, decision analysis is well poised to deal with issues related to testing (or screening, such as mammography) where receiving a false positive result (you are told you have the disease when you really don't) is common or when a false negative result (you are told you don't have the disease when you really have it) is common.

I doubt USPSTF used a decision analysis approach given what I've read on their website. What do you think about it? Should we move in a decision analysis direction or should the traditional epidemiological approach be sufficient to make these public health decisions?

Now I get to complete my decision sciences mid-term. Yippee!

No comments: