Overcoming Bias

Anxiety about what is true

Robin links to an article about apparent manipulation of the medical process by drug manufacturers, dealing particularly with the drug Fosamax which is supposed to improve bone strength. The article raises the possibility that many modern diseases don't exist at all but are the creation of the pharmaceutical industry to give them tools to sell more drugs.

My wife used to take Fosamax, so this was a topic of interest to me. To learn more, I did a search on scholar.google.com for "fosamax hip-fracture". It looks to me like most of the hits are pretty favorable to the drug, but how much does that really prove?

Brownlee on Selling Anxiety

Shannon Brownlee in the Post on drug companies inventing new diseases:

It turns out that much of what we -- and our doctors -- think we know about many health problems has been shaped by drugmakers and their marketers. ... Osteoporosis and osteopenia aren't really diseases. Before the 1990s, doctors decided that you had osteoporosis if you were elderly and you broke a bone. When the pharmaceutical company Merck came up with its anti-bone-loss durg Fosamax, it wanted a broader market than just elderly fracture patients. The solution? The company helped fund a panel of medical experts to create diagnostic criteria for osteoporosis so that a diagnosis could be made before the patient actually broke a bone.

The panel's first step was to define "normal" bone density as that of the average 30-year-old woman. Next, the experts chose as their cutoff for osteoporosis a statistical point that was slightly below the bone density of their normal 30-year-old -- a definition they admitted was "somewhat arbitrary." Finally, they came up with a completely new disease -- osteopenia -- for bone density that fell somewhere between that normal 30-year-old and their arbitrary definition of osteoporosis.

Voila -- 30 percent of post-menopausal women suddenly had a disease that needed to be treated early in order to prevent a problem -- hip fracture -- that wouldn't occur for many years, if ever. According to the new guidelines, millions more women now had osteopenia, which their doctors needed to watch like hawks so that their patients could be treated once they progressed to osteoporosis. Merck then took the added step of helping doctors buy DEXA scanners, X-ray machines needed to scan your bones to get that all-important diagnosis. ...

Fosamax ... can cause necrosis (death) of the jawbone. What's more, there aren't any valid scientific studies to show that treating osteoporosis early will prevent fractures down the road. The drug can also trigger serious heartburn.

Added 6Apr: More from Brownlee:

Showing That You Care

My ambitious theory paper, which attempts to explain diverse health behavior puzzles with just a few assumptions, has finally been published in Medical Hypotheses.  (Print copies were mailed today.)  The abstract: 

Human behavior regarding medicine seems strange; assumptions and models that seem workable in other areas seem less so in medicine. Perhaps, we need to rethink the basics. Toward this end, I have collected many puzzling stylized facts about behavior regarding medicine, and have sought a small number of simple assumptions which might together account for as many puzzles as possible.

The puzzles I consider include a willingness to provide more medical than other assistance to associates, a desire to be seen as so providing, support for nation, firm, or family provided medical care, placebo benefits of medicine, a small average health value of additional medical spending relative to other health influences, more interest in public than private signals of medical quality, medical spending as an individual necessity but national luxury, a strong stress-mediated health status correlation, and support for regulating health behaviors of the low status. These phenomena seem widespread across time and cultures.

Eternal Medicine

Most fans of modern medicine do not realize how similar in appearance was ancient medicine:

While the Greeks left a vast legacy of medical texts in a familiar language, we know of only 12 from the time of the pharaohs - written on papyrus in a vanished language that scholars are still grappling with. From their descriptions of diseases and treatments, the texts have left little doubt that the ancient Egyptians had considerable medical skills, but weighing up their pharmaceutical knowledge has proved trickier: although the papyri include some 2000 prescriptions, doubts surround the identity of many of the ingredients listed. ....

Antidepressant Publication Bias

Medical publication bias was real:

The makers of antidepressants like Prozac and Paxil never published the results of about a third of the drug trials that they conducted to win government approval, misleading doctors and consumers about the drugs' true effectiveness, a new analysis has found. ... The new analysis, reviewing data from 74 trials involving 12 drugs, is the most thorough to date. And it documents a large difference: while 94 percent of the positive studies found their way into print, just 14 percent of those with disappointing or uncertain results did. ... In the study, a team of researchers identified all antidepressant trials submitted to the Food and Drug Administration to win approval from 1987 to 2004.  The studies involved 12,564 adult patients testing drugs like Prozac from Eli Lilly, Zoloft from Pfizer and Effexor from Wyeth.

Fortunately, you needn't worry; the problem has been fixed:

Alan Goldhammer, deputy vice president for regulatory affairs at the Pharmaceutical Research and Manufacturers of America, said the new study neglected to mention that industry and government had already taken steps to make clinical trial information more transparent. "This is all based on data from before 2004, and since then we've put to rest the myth that companies have anything to hide," he said.

Can you guess what he will say if the same results are found in data through 2007, but not published until 2011?

Does Healthcare Do Any Good At All?

The RAND experiment showed that people with more generous health insurance consumed a lot more health care than those with less generous insurance, but didn't have much (or maybe anything) in the way of better health outcomes.  The natural interpretation of this is that everyone, including those with less generous insurance, chooses to get all of the high-value treatments, and that the extra treatments consumed only by those with more generous insurance aren't worth much.*  If this was true, then Robin's suggestion to radically cut health care would follow directly; it would be the low-value marginal treatments that would get cut while the high-value infra-marginal treatments would remain.  This would also be consistent with the evidence that it is damaging to one's health to have no insurance at all (everyone in the RAND experiment had insurance of some kind), as people with no insurance would be missing out on (at least some of) the high-value infra-marginal treatments along with the low-value marginal ones.

The problem is that some of the other evidence from the RAND study is not really consistent with this story.  It seems that the marginal care consumed only by people with more generous insurance is not just low-value stuff.  The marginal treatments consumed only by those with more generous insurance, in the opinion of expert doctors, looks a lot like the infra-marginal treatments consumed by everybody.  But if that's true, doesn't it have to mean that all health care is of little value?  If the marginal care looks just like the infra-marginal care, and the marginal care is of little value, then doesn't the infra-marginal care have to be of little value too?  I don't think anybody seriously believes that, which makes me think that there is something wrong with the studies that say that the marginal care is just like the infra-marginal care.  Does anyone have any other ideas?

Doctor Hypocrisy

The fact that your life would be easier if you could trust someone does not make that person trustworthy.  Doctors are a good example.  Wednesday's Post:

The first-of-its-kind survey of more than 1,600 physicians, published today in the Annals of Internal Medicine, found that 45 percent said they did not always report an incompetent or impaired colleague to the appropriate authorities -- even though 96 percent agreed that doctors should turn in such people.

Moreover, 46 percent said they had failed to report at least one serious medical error that they knew about, despite the fact that 93 percent of doctors said physicians should report all significant medical errors that they observe. ...

A majority said they would refer patients to an imaging facility in which they had a financial interest, but only 24 percent would inform patients of that financial tie.  Yet 96 percent told researchers that doctors should put their patients' welfare above their own financial interests.

Also, more than a third of physicians, 36 percent, said they would order an unneeded MRI (magnetic resonance imaging) test if it were requested by a patient with low back pain, though most doctors say they do not want to waste scarce resources.

And while 93 percent said doctors should provide necessary medical care regardless of a patient's ability to pay, only 69 percent currently accept uninsured patients who are unable to pay.

I doubt doctors are much different from other professionals in succumbing to such temptations.  The problem is that people want to believe that doctors are somehow different, and can be trusted just because they are doctors.  Which lets them get away with ...

Treatment Futures

A key problem in medicine is: what general process or institution can ordinary sick patients and concerned loved ones rely on to choose the best treatments (or none)?  They could rely on a doctor's advice, but then how do they pick him or her, or be assured he or she has sufficient incentives to find and choose the best? 

A month ago I described one solution: health plans that "feel your pain" via payments designed to match your health value.  Today I'll outline another solution: "treatment futures," i.e., decision markets where speculators can bet on your health, conditional on treatment decisions.   

Imagine a surgeon had recommended heart surgery for you, but you had doubts.  You could post an anonymized health record to the web, and let people bet on how many more years you will live if you did the surgery as suggested, and how many years if you did nothing for now.  Market estimates of those year numbers would tell you which option speculators thought best. 

You could use quality-adjusted years to make sure speculators considered disability and pain, and you could compare many options, such as different kinds of surgery or drugs and switching to a different surgeon.  And you could let your doctors, their associates, and your friends bet on you, as long as you made sure they kept a positive interest in your doing well.  As with college choice futures, most bettors would probably bet on bundles of patients, such as all 40 year old men with certain symptoms.   

Now while a heart surgery might have a big effect on years to live, most treatments have too small an effect to see clearly over market noise.  But you could bundle lots of small decisions into a big decision with a larger effect.  For example, you could ask about the choice of a health plan or doctor for the next year(s).  Also, if we bundled up decisions about many different patients who were comparing two particular doctors, hospitals, or plans, that could give us a good evaluation of the relative quality of those doctors, hospitals, or plans.   

The main problem I see is people being unwilling to believe the likely market advice that they should get a lot less treatment than most people now do.  I expect decision market advice will have to prove itself well in other areas before people will consider its advice about medicine, and even then I'm not sure people will listen. 

This post in response to a question by Alan Garber.  I first presented this concept at the RWJF Health Policy Scholar annual conference in 1999. 

Hospice Beats Hospital

If a doctor declares that the normal progression of your condition gives you less than six months to live, you can opt for hospice, where they don't even try to cure your condition, but just try to make your remaining days as comfortable and meaningful as possible.   According to this paper (ungated here) from the March Journal of Pain and Symptom Management, choosing a hospice makes you live longer. 

We studied the difference of survival periods of terminally ill patients between those using hospices and not using hospices. ... We analyzed the survival of 4493 patients from a sample of 5% of the entire Medicare beneficiary population for 1998-2002 associated with six narrowly defined indicative markers. For the six patient populations combined, the mean survival was 29 days longer for hospice patients than for nonhospice patients. The mean survival period was also significantly longer for the hospice patients with CHF, lung cancer, pancreatic cancer, and marginally significant for colon cancer (P = 0.08).  Mean survival was not significantly different (statistically) for hospice vs. nonhospice patients with breast or prostate cancer.

I'm not entirely happy with they way they did their analysis, but it is certainly intruiging.  Hat tip to my wife, who works in a hospice. 

A Valid Concern

In the October Journal of Health Politics, Policy and Law, John Nyman questions the RAND experiment:

Of the 1,294 adult participants who were randomly assigned to the free plan, 5 participants (0.4 percent) left the experiment voluntarily during the observation period, while of the 2,664 who were assigned to any of the cost-sharing plans, 179 participants (6.7 percent) voluntarily left the experiment. ... The explanation that makes the most sense is that the dropouts were participants who had just been diagnosed with an illness that would require a costly hospital procedure. ... If they dropped out, their coverage would automatically revert to their original insurance policies, which were likely to cover major medical expenses (such as hospitalizations) with no copayments ... As a result of dropping out, these participants' inpatient stays (and associated health care spending) did not register in the experiment, and it appeared as if participants in the cost-sharing group had a lower rate of inpatient use.

This is a valid concern, which the original work tried to address, but perhaps not sufficiently.  While this is the best of the health-medicine studies, it was far from the only one.  But this seems all the more reason to redo the experiment today.  Hat tip to John McDonough, Alex Tabarrok, and Michael Cannon.

Added 10/24: The RAND Experiment folks have responded with vigor.  HT to Alex Tabarrok.