3 ms·

This is what happens when too few MDs understand Bayesian inference

3 points by leot 15y ago

leot 15y ago
Extra true information can only ever be a good thing if handled properly.
- carbocation 15y ago
  Only if you're operating in a gamelike environment where logic, not the law, rules the day.
- drcode 15y ago
  Does anyone know if this observation has a name? There must be an official name for this logical fact. I ask because I'm an MD and made this point during a conversation with another MD a while ago and the other guy looked at me like I was a fool to believe this. He responded along the lines of "But a study has shown that people given test X for disease Y do worse than people not given the test." To him, this was a proof by negation, which of course it is not.
  - leot 15y ago
    I can't think of one. Hopefully there is -- if not, it needs coining. Your friend is wrong, though, if this extra information was "cheap" to obtain (caused no harm to the patient and was low cost). The argument used is that extra testing causes needless treatment, in the end causing more harm than good (we'll call a treatment "costly" if it has a large opportunity cost or causes harm). For the MD you mentioned, the test could cause people to "do worse" because either the test is harmful or the MD's use of its results is harmful. If the test is harmful, then it probably shouldn't be administered unless other evidence suggests it's necessary. If it's the MD's use fo the results that causes the harm, then it just means the MD should use the results differently. If the problem is the MD's reasoning, then one possible source of error is in giving a much higher prior probability to the patient's having the disease, or an ignorance of what, e.g., a "bad" result actually implies. In the case of the PSA test, it looks like the fee-for-service model is leading many doctors to ignore the relationship between a high PSA result and the probability of the patient's dying from prostate cancer, in addition to ignoring the harm that will be caused by performing a biopsy or by any further treatment. Perhaps more malpractice lawsuits that focus on needless treatment would put a dent in this problem ...
carbocation 15y ago
Would you mind editorializing in a comment rather than in the title?
drcode 15y ago
In my opinion, 100 years from now, our ancestors will view the lack of basic statistics literacy as one of the primary failings of our present society.

3 points by leot 15y ago

leot 15y ago

Extra true information can only ever be a good thing if handled properly.

carbocation 15y ago
Only if you're operating in a gamelike environment where logic, not the law, rules the day.
drcode 15y ago
Does anyone know if this observation has a name? There must be an official name for this logical fact. I ask because I'm an MD and made this point during a conversation with another MD a while ago and the other guy looked at me like I was a fool to believe this. He responded along the lines of "But a study has shown that people given test X for disease Y do worse than people not given the test." To him, this was a proof by negation, which of course it is not.
- leot 15y ago
  I can't think of one. Hopefully there is -- if not, it needs coining. Your friend is wrong, though, if this extra information was "cheap" to obtain (caused no harm to the patient and was low cost). The argument used is that extra testing causes needless treatment, in the end causing more harm than good (we'll call a treatment "costly" if it has a large opportunity cost or causes harm). For the MD you mentioned, the test could cause people to "do worse" because either the test is harmful or the MD's use of its results is harmful. If the test is harmful, then it probably shouldn't be administered unless other evidence suggests it's necessary. If it's the MD's use fo the results that causes the harm, then it just means the MD should use the results differently. If the problem is the MD's reasoning, then one possible source of error is in giving a much higher prior probability to the patient's having the disease, or an ignorance of what, e.g., a "bad" result actually implies. In the case of the PSA test, it looks like the fee-for-service model is leading many doctors to ignore the relationship between a high PSA result and the probability of the patient's dying from prostate cancer, in addition to ignoring the harm that will be caused by performing a biopsy or by any further treatment. Perhaps more malpractice lawsuits that focus on needless treatment would put a dent in this problem ...

carbocation 15y ago

Would you mind editorializing in a comment rather than in the title?

drcode 15y ago

In my opinion, 100 years from now, our ancestors will view the lack of basic statistics literacy as one of the primary failings of our present society.