A Bradford Hill's 1965 paper about determining causation by association in epidemiological studies
Good reading on Hill's criteria of causation.http://www.epi-perspectives.com/content/1/1/3/abstract
-------------------------------------------------------------------------------------
I read this editorial on A Bradford Hill's 1965 paper about determining causation by association in epidemiological studies.
I can only approach the article's points of advice based on my own background knowledge and experiences, which is not epidemiology. Thus, I excuse my replies if I turn out to be "on a different hill" from the authors. It would have been nice had the authors provided specific epidemiologic or health policy examples to help drive the points; but then, people outside epidemiology are not their target audience.
For example, I would not encourage researchers to incorporate a great deal of health policy recommendations along with an outcome analysis of their own study. The way I see it, giving clinical researchers elevated political clout to present health policy advice will allow a strong influence for research bias. People should turn to the research more for facts rather than direction. Ideally, health policy is based on a degree of scientific consensus (peer review,) not on the conclusions of one group of researchers observations and outcomes.
Another point of this article is that statistical precision should not be mistaken for outcome validity. In my experience in the periphery of health policy development, this seldom occurs anyway. Peer review is often quick to slam the numbers based on other study design factors. Once in awhile it seems that a researcher has been self-fooled by attributing too much rigidity in the numbers. The study numbers often but serve as fodder for marketers to sell products to investors and consumers. Good numbers are an important key in determining a valid study, but not the only one. I believe most understand this. I must admit that based on experience, I am a little more skeptical of studies that appear to overly emphasize p values and statistical significance numbers.
Lastly, the article advocates that neither action nor inaction should occur despite lack of supporting or detracting scientific evidence. Again, I don't see this as a problem. There is much more than scientific understanding that dictate good and bad health policy. Health policy is often probably shamefully utilitarianistic (if that's a word.) There is only so much resources for paying for all of health care. Health policy cannot afford to support an intervention just because it appears to be benign, and can possibly benefit.
-------------------------------------------------------------------------------------
I read this editorial on A Bradford Hill's 1965 paper about determining causation by association in epidemiological studies.
I can only approach the article's points of advice based on my own background knowledge and experiences, which is not epidemiology. Thus, I excuse my replies if I turn out to be "on a different hill" from the authors. It would have been nice had the authors provided specific epidemiologic or health policy examples to help drive the points; but then, people outside epidemiology are not their target audience.
For example, I would not encourage researchers to incorporate a great deal of health policy recommendations along with an outcome analysis of their own study. The way I see it, giving clinical researchers elevated political clout to present health policy advice will allow a strong influence for research bias. People should turn to the research more for facts rather than direction. Ideally, health policy is based on a degree of scientific consensus (peer review,) not on the conclusions of one group of researchers observations and outcomes.
Another point of this article is that statistical precision should not be mistaken for outcome validity. In my experience in the periphery of health policy development, this seldom occurs anyway. Peer review is often quick to slam the numbers based on other study design factors. Once in awhile it seems that a researcher has been self-fooled by attributing too much rigidity in the numbers. The study numbers often but serve as fodder for marketers to sell products to investors and consumers. Good numbers are an important key in determining a valid study, but not the only one. I believe most understand this. I must admit that based on experience, I am a little more skeptical of studies that appear to overly emphasize p values and statistical significance numbers.
Lastly, the article advocates that neither action nor inaction should occur despite lack of supporting or detracting scientific evidence. Again, I don't see this as a problem. There is much more than scientific understanding that dictate good and bad health policy. Health policy is often probably shamefully utilitarianistic (if that's a word.) There is only so much resources for paying for all of health care. Health policy cannot afford to support an intervention just because it appears to be benign, and can possibly benefit.
posted by nawledge at
7:48 PM
0 Comments:
Post a Comment
<< Home