2009/12/15
John Allen Paulos (professor of mathematics at Temple University): “If screening catches the breast cancers of some asymptomatic women in their 40s, then it would also catch those of some asymptomatic women in their 30s. But why stop there? Why not monthly mammograms beginning at age 15?”
Jonathan Bush (CEO of Athenahealth) on health reform: “I still have to keep going to Washington and sucking up… Because the problem is when you have a baby with an Uzi, right, they might accidentally mow you down. But here’s the thing . . . they’re brilliant people. It’s just that the idea of a market in health care never occurred to them.”
Jonathan Bush (CEO of Athenahealth) on stimulus money for health I.T.: “It is kind of too bad that all these software companies that we’re really close to putting out of business, these terrible legacy companies, with code that was written in the ’70s, are going to get life support. …What it is, basically, is a federally sponsored sale on old-fashioned software.”
Professor Paulos is correct that some women might be helped by greater screenings. However, the small benefit to a few women would be offset by the immense cost and inconvenience of increased screenings. That why the task force makes recommendations.
This is an interesting perspective on breast cancer. I have never thought about it that way.
The comment on health IT is really interesting. This is the first time I have seen this point of view.
The quote on mammograms is excellent. It ;should be obvious that there is no right number. The choice should be made based on individual preference — including attitudes toward risk and the willingness to trade off money against risk reduction.
Prof. Paulos is giving an specific illustration of a general truth: The smaller the actual incidence of X in a population, the bigger the problem of false positives. Furthermore, we generally define statistical significance with respect to the Type II (beta) error, which is 5% in Prof. Paulos’ example. The false positive (Type I or alpha) error can only be reduced by increasing the specificity (alternatively, reducing the power or sensitivity) of the test, which increases the Type II error.
Indeed, we could go to the other extreme than than suggested by Prof. Paulos: Outlaw all mammograms. That would eliminate false positives! Of course, the result is that 100% of the population would live in ignorance, which is surely stressful.
We can go around all day like this. Far more important are the consequences of the errors. If the treatment has limited side effects, and failure to treat infected patients has catastrophic consequences, then why not accept a large number of false positives? It would seem to be a good policy to leave the individual to decide how to deal with the results of the tests.
I am bowled over by the Jonathan Bush quote on health IT. We are using stimulus money to prop up outdated technology that no one should be using in the first place? How crazy is that?