Robin and others: this is the response from an Aussie medico. I suspect he
underestimates the risks to the poor in the USA, who I think might well die
or suffer drastically for lack of the treatment he assumes they'd get if
they were at grave enough hazard.
I don't disagree with Hanson's basic argument. Everyone knows that the
greatest effects on health have been sanitation/hygiene/nutrition and
>From: Robin Hanson <hanson@econ.Berkeley.EDU>
>
>A cryonet reader privately questions my claim that medicine has
>a low marginal value.
>
>The clearest evidence we have on this is the RAND Health Insurance
>Experiment [Newhouse et al, "Free for All?" '93]. [etc]
>======
However, there are a couple of things worth pointing out. Unfortunately I can't do this as well as I'd like because I couldn't pull up an extract for the RAND experiment. But judging from what's been said, I can say that there are still a couple of important provisos to the finding. The first is that the comparisons are between groups who ahve different levels of insurance. However, they still live in the same society and would (I hope!) have been randomised or case-matched so the the treatment groups were very similar. In this context, you wouldn't expect much difference in health outcomes because they are, for all intents and purposes, indistinguishable. No-one is going to miss out on life-saving surgery. No-one is going to have their heart medicine withheld. On the *key* issues, the groups would be much the same. The paper doesn't test the "effectiveness of medicine", because it's not comparing a group who receive medical care with a group who don't receive medical care (an experiment that I doubt will ever be performed). The paper doesn't actually test the "marginal effectiveness of medicine" so much as it tests the cost-benefits of different insurance options. And, not surprisingly, it found that most health indicators are the comparable, but people who don't pay for their health care use more services for very little benefit. <rant on> There is an obvious lesson here for the architects of Australia's Medicare, but since they are mostly brain-dead power-hungry public servants, there's very little in the way of argument that could ever persuade them to think about the effects of what they're doing beyond their personal feelings of cosmic significance. <rant off>
The second issue is that other studies often look only at mortality, but there's still an important role for treatment of morbidity in medicine. While the RAND paper checked a number of physiological and self-assessed wellbeing measures, it's not true that there were "no significant" differences between the groups. At least one paper I found on Medline tried to explain why eyesight was reportedly significantly improved in the high-insurance group. In fact, I dug up this quote: "In a randomized trial [the RAND trial] of the effects of medical insurance on spending and the health status of the nonaged, we previously reported that patients with limited cost sharing had approximately one-third less use of medical services, similar general self-assessed health, and worse blood pressure, functional far vision, and dental health than those with free care." (Effects of cost sharing on physiological health, health practices, and worry. Keeler EB; Sloss EM; Brook RH; et al. Health Serv Res, 22(3):279-306 1987 Aug.) The authors did conclude that "Overall, except for patients with hypertension or vision problems, the effects of cost sharing on health were minor." But this is a long way from Hanson's description. While I'd agree with Hanson's general argument, I think he has overstated the case and (on my limited information) tried to draw more succour from the literature than is actually there to be had.
The other paper mentioned must be a dud reference. The Lantz paper I picked up in that issue of JAMA doesn't actually deal with the topic Hanson uses it for. The paper tried to answer whether altering the health-risk behaviour of lower socioeconomic groups would lead to a significant improvement in health outcomes. What the analysis showed was that lowest SE group was at higher risk of mortality (hazard rate ratio of 3.2 compared to 2.34 for the highest SE group). But when the risk of modifiable behaviours, such as smoking, was factored out, there was still a large difference between SE groups (2.77 cf 2.14). This is indeed an interesting paper which shows that even if we could stop risk behaviour, there would *still* be a significant health risk in being low SE. Not really a surprise, but's it's good to see it experimentally verified. But this doesn't seem to have much to do with Hanson's argument. Perhaps he's thinking of another paper?
Damien Broderick