September 23, 2007

posted by Frank Pasquale

Frank: I'm curious why you didn't spend any time discussing last week's "fascinating NYT study" on the extreme weakness of epidemiological studies and the miserable state of knowledge about the real effect of various environmental inputs on human health. One would have thought this would have been right up your alley, except for the fact that the contents of that report were extremely disfavorable to many of the sorts of conclusions you offer in support of your prefered policies. A very telling ommission, given that the NYT is otherwise your number one go-to source.

I thought that was a fascinating article, and I agree that "the way epidemiologists frame the questions they ask can bias a measurement and produce an association that may be particularly misleading."
from
http://www.nytimes.com/2007/09/16/magazine/16epidemiology-t.html?_r=1&oref=slogin&pagewanted=print

I have critiqued health care "rankings" because I believe the "objective data" they are based on can result from a gaming of the system. For example, a heart hospital may have a low mortality rate because it only takes healthy patients.

If the private-investor-backed nursing homes discussed in the NYT story take a disproportionate share of the sickest and most vulnerable patients, then the study is flawed. But I did not see evidence of that in the Duhigg article. Do you have evidence that the private investor-backed homes take a disproportionate share of the sickest and most vulnerable patients?

The point was not to criticize this partticular article, but just to point out a glaring oversight in your NYT coverage, perhaps applicable here, but in any case relevant because you so conveniently picked up your NYT coverage after an "unexplainable" lapse last week. And your quote above from the article misses the much larger point about confounding variables that render many or all epidemiological studies suspect. It is not about gaming the system. It is about counseling some caution before leaping to conclusions based on these studies.

I am not sure I understand anon's larger point. Perhaps I am somewhat unfazed by the contention that a large number of scientific studies are poorly done insofar as I am relentless in perceiving science as a (human) practice, subject to the same prevalence of errors and fallacies as any other human practice. Given a Humean critique of induction (as well as contemporary philosophy of science critiques, pace Feyerabend, Rorty, etc.), it is equally unsurprising.

The idea that many epidemiological studies are suspect, of course, hardly implicates Frank's views alone. If correct, it means that the evidence-base for a number of public health policies based on epidemiological findings is shaky at best. This is also hardly a new development, as public health policies in general in the U.S. and beyond have typically rested on questionable evidentiary foundations, which is partly why many scholars maintain that cultural and social attitudes animate many of our public health policies in general.

Indeed, some of my own work argues that most scientific literature and public health policy is based on a patently dubious model of disease causality, on a fundamental misunderstanding of the largest causal factors of illness and health (and I am not alone in arguing this, either).

My point is that the potentially weak evidence base for public health policies hardly affects Frank's views alone. In fact, being somewhat familiar with his views in general, I'd go so far as to suggest such a premise is far more troubling for those who support the kinds of policies he typically opposes.

Having provided a link to the article in question at Daniel's Medical Humanities Blog I thought I might chime in as well. I gleaned several lessons from the article, for example: maintain a generally skeptical attitude toward epidemiological studies especially insofar as they are reported in an incrementalist and myopic fashion that tends to lose sight of the proverbial big picture, including the fact that ceteris paribus clauses loom large in such studies, as well as the (related) inevitable methodological simplification of possible causal variables and pathways intrinsic to most if not all epidemiological research. The author rightly notes that many of these studies provide hypothesis-generating evidence, not definitive or conclusive causal models of a disease or illness.

And the article notes some of the strengths of epidemiological research: "No one questions the value of these epidemiologic studies when they’re used to identify the unexpected side effects of prescription drugs or to study the progression of diseases or their distribution between and within populations." And thus it is no mean achievement that researchers have discovered that in the case of breast cancer, for instance, "we know...a carcinogenic factor of diet or lifestyle exists, waiting to be identified." Why? Well, "Breast cancer is not the scourge among Japanese women that it is among American women, but it takes only two generations in the United States before Japanese-Americans have the same breast cancer rates as any other ethnic group. This tells us that something about the American lifestyle or diet is a cause of breast cancer."

The media reporting of epidemiological research, perhaps skewed in part owing to researchers seeking the requisite publicity necessary to generate more research funding, often fails to appreciate the modest scope of specific epidemiological studies, not unlike research in genetics, in which case only now are we beginning to appreciate the manner in which the claims made on its behalf were chock full of so many false or misleading promises (of course some, like Richard Lewontin, long ago warned us of the oversell and hyperbole on behalf of genetic research) .

As Daniel makes clear, there is nothing in the article that in any way undermines the approach or conclusions of Frank's recent posts on health care, as his arguments were not relying in the main on this or that epidemiological study for their plausibility, soundness, or persuasiveness.

Ouch. The timing of this story couldn’t be worse for the Carlyle Group, which is busy finalizing the largest-ever buyout of a nursing home chain, its $6.3 billion buyout of HCR Manor Care. It would be nice if Carlyle used the buyout to prove everyone wrong and improve care, given that more than 85% (!) of Manor Care facilities already report CNA staffing levels below 2.8 hours per resident day—a standard identified in a Centers for Medicare and Medicaid Services study as necessary to properly care for residents. Unfortunately, so far it looks like more of the same – “more profits, less nursing” – the takeover will result in a windfall of as much as $254 million for top Manor Care execs, including as much as $186 million for CEO Paul Ormond, plus Carlyle will get millions in fees. Read more here.