Disagreement Case Study: Hanson and Cutler

Michael Cannon, director of health policy studies at CATO, took my health economics course in spring 2006, wherein I elaborated my story that we see little correlation between variations in medicine and health.   Like most students he didn’t really believe me, and so last November Michael arranged to check on me by setting up a half hour group discussion between myself, star Harvard health economist David Cutler, Michael, and a few other CATO health folks. 

David Cutler is very pro-medicine.  For example, last August his New England Journal of Medicine article assumed half of lifespan gains have been due to medicine, and then concluded:

From 1960 through 2000, the life expectancy for newborns increased by 6.97 years, lifetime medical spending adjusted for inflation increased by approximately $69,000, and the cost per year of life gained was $19,900. … [This cost per year gained] was approximately $31,600 at 15 years of age, $53,700 at 45 years of age, and $84,700 at 65 years of age.

In our discussion, I set aside disputes over medicine’s average value of medicine to focus on medicine’s marginal value – how much more health we get from a bit more medicine.   Here David largely agreed.  Yes, "most ordinary people would be shocked" to hear, the thirty-year-old RAND experiment is our best data, which we should repeat today, and its results are confirmed by typical "cross section" results – regions that have more doctors or medical or cultural habits of using more medicine show little or no resulting difference in aggregate health. 

But David also hedged.  Had the RAND experiment followed subjects for ten years, he thinks it would have shown medical benefits.  Because some recent studies suggest patients are price sensitive about some cheap effective drugs, of which we have more today than before, David thinks we would see a different price effect today in a RAND Two.  And he says he sometimes tries to argue, though he admits he can’t quite convince himself, that this low marginal value is good because it pays for medical innovation. 

I’m happy with the main outcome here, that our "disagreement" seems minor compared to the agreement I would like to make clear to a wider audience.  On the point of disagreement we explored, whether this marginal value is different today, David pointed to none of the old or recent aggregate correlation studies that suggest positive benefits, nor to any evidence that the relative mix of help/harmful/useless medicine has changed recently.  This seems to instead be a matter of judgment for him.

My counter argument is that price sensitivity to cheap effective drugs was part of the effect seen RAND One. So unless the percentage mix of help/harmful/useless medicine has changed, we should expect similar results today.  Since total medical spending is three times what it was then, even if the mix hasn’t changed we should expect two thirds of cheap effective drugs to have been new since RAND One. 

I don’t know what David thinks of me, but I accept that he is clearly objectively more expert than I on this topic, given his prestigious position and many more years of focus on the topic.  But given the strong usual tendency to give medicine the benefit of the doubt, my impression that David gives medicine this benefit of doubt on other topics, and his inability to point to any concrete supporting evidence, I’m willing to attribute David’s more positive assessment here to such wishful thinking, rather than to his superior intuition on this matter.  How rational am I?

GD Star Rating
Tagged as: ,
Trackback URL:
  • Jor

    …relative mix of help/harmful/useless medicine has changed recently… — I’m not sure if this is in context of the mix on the margin or in general. If it’s in general, than it has _definitely_changed_. There are many, cheap, generic medications now that weren’t available 30 years ago, that each individually are proven to prevent cardiovascular disease.

    There is no clinical value of looking at the changes in the mix of medicines on aggregate. You could look at the # of prescriptions varying over time, take effect sizes from randomized trials, and model changes in mortality

  • Jor, yes there may be more good medicine, but there is also more medicine overall. The question is the relative fraction of the different types.

  • Buzzcut

    For example, last August his New England Journal of Medicine article assumed half of lifespan gains have been due to medicine,

    Why would you assume that?

    I would assume that it is because smoking has declined, the air and water have been cleaned up to an almost unprecedented extent (could LA have less smog now than in pre-industrial times?!?), people are heavier than in the past (overweight BMIs have the lowest death rates), the service economy is vastly safer and healthier than the industrial economy, etc.

  • Joseph Delaney

    I find this all very interesting (as an epidemiologist who studies drugs). What I wonder is whether medicine might be a classic case of diminsihing returns. The introduction of antibiotics, for example clearly had major benefits on life expectancy turning many common and lethal diseases into annoyances (sphyllis is a great example of this).

    But there has not been anything else that I can think of that is as potent at reducing all-cause mortality.

    In the same sense, the first drugs to treat a “type of condition” tend to be really important and later drugs add little extra benefit. Once you have one effective anti-inflammatory, the marginal benefit of the next one is much less (but you still pay the full development costs).

    All that being said, even if I agree that the marginal benefit is small (and I need to think a lot more about this question), I think it is important to realize that we might get to a point on the cost-benefit curve where reduction of medical access could have rather surprisingly large adverse effects.

  • Stuart Armstrong

    How rational am I?

    You seem rational, parsing and analysing David’s arguments, and pointing out where they come up short and where they didn’t. But there’s a problem with that:

    These disagrement analysis are premised on the fact you can’t “agree to disagree“, so any disagrement must be irrational or dishonest. But that cuts both ways; if two honest and rational people always did agree simply by comparing their priors, then they would end up embracing a position without evidence to back it up. All they would be able to say is “someone I know, who is rational and honest, has priors in this direction”.

    However, your comments “matter of judgment for him” and “his inability to point to any concrete supporting evidence” imply that you put great store by supporting evidence. Would you have thought his arguments stronger if he had simply said “I asked the opinion of someone knowledgeable and rational, and we agreed to agree, hence my position”?
    David’s “judgement” may have been formed in just such conversations.

  • Stuart, an interesting question is whether the fact that someone cannot point to any particular evidence supporting his view should be considered a bad signal about his view.

  • joe

    I think there is a huge difference between “cannot” and “did not”. I would guess that he is probably not used to having to cite sources supporting his view, though this in and of itself is probably an artifact of the “usual tendency to give medicine the benefit of the doubt.”

  • Joe, Cutler did in fact cite sources for several other claims during our conversation – he just not cite a source for that particular claim.

  • Stuart Armstrong

    an interesting question is whether the fact that someone cannot point to any particular evidence supporting his view should be considered a bad signal about his view.

    If he’s meeting you in a debate, and is an expert on the subject, it’s definitely a bad signal – if there was evidence, he should know it (though having evidence isn’t automatically a good signal, because of selection bias).

    But in general, there are many reasons for people to adopt positions with indirect evidence – such as believing that the market price is the right one for a stock. Prediction markets are other indirect evidence, as are some “wisdom of crowds” situations and respecting certain experts.

    Indirect evidence is sometimes in error, but it’s very hard to know when. It’s also not cumulative in the same way that direct evidence is. Two thousand successful experiments tell you more than one thousand experiments; finding a thousand more biologists who endorse evolution doesn’t tell you much more (which why indirect evidence isn’t generally allowed in those cases where a quasi-certainty is sought: science and courtrooms).

    So if the person you’re talking to is rational, articulate and generally knowledgeable, but not an expert in the subject, the fact he can’t point to any direct evidence is “noise”. It dilutes the value of his position but does not contradict it. So it’s only a bad signal if the width of the noise is an issue – i.e. if you already nearly agree with him.

  • Quick quote from an excellent overview article from The New York Times:

    “An obvious explanation is that wealth buys health. And it seems plausible. Poorer people, at least in the United States, are less likely to have health insurance or access to medications.

    But Dr. Fuchs says, then why don’t differences between rich and poor shrink in countries where everyone has health care?

    “All you have to do is look at the experience of countries like England that have had health insurance for more than 40 years,” he says. “There is no diminution in the class differentials. It’s been the same in Sweden. It’s true everywhere.”


    But more importantly, increased spending isn’t touching the astonishing rise in chronic illnesses from cancer through eczema (now more North Americans have this once rare autoimmune disorder than don’t. Of course, diabetes was once very rare, too, two hundred years ago.)

    Many of these illnesses are tied to metabolic disorder, but it’s increasingly looking as though obesity and changes in appetite are a a downstream event in this disorder, not the cause. Successes in treatment are masking an outrageous growth in (chronic) disease, so concentrating on longevity is highly misleading. Spending isn’t touching this epidemic; I suspect because excess light exposure and consequent mitochondrial dysfunction isn’t getting much attention yet. (See photoperiodeffect.com) Until you know where to strike, having a fine hammer is of little use, and two hammers are no more use than one.