There is a debate going on in the blogosphere between Ezra Klein, Arnold Kling, Karl Smith, Tim Carney, and others about, to put it crudely, whether health care really affects health that much. This is, in part, a proxy debate for whether it is worth it for the U.S. government to provide generous universal health-care financing for all of its citizens (or, I suppose, residents).
Either position can be caricatured. On one hand, no sane person would want to be without the advances of modern medicine. Recently, a little girl I know had scarlet fever. A century ago, this would very possibly have meant burying a small corpse; today, it implies a ten-day cycle of swallowing medicine at breakfast and dinner. There are few people on earth who have as much reason to be proud of how they spend their work week as pharmaceutical researchers.
On the other hand, the link from alternative methods of health-care finance, through the actual differences in provision of medical care these imply in the contemporary U.S., to the actual differences in health outcomes these treatment differences would cause, isn’t nearly so obvious. The net health effect of providing universal health-care coverage versus some alternative financing system is an empirical question, not a philosophy debate.
I’ve written a lot about why randomized experiments are so critical to understanding cause-and-effect relationships in social policy. In the case of health-care financing, the reason is that what system of health-care financing you have (high-quality “go to any doctor” plan; good HMO; catastrophic-only plan; VA; go to an emergency room because you are uninsured, etc.) is bound up with a myriad of other factors that influence health. A randomized experiment allows us to isolate the impact of the system of health-care financing.
To my knowledge, the only large-scale randomized experiment in the U.S. that has tested the actual effects on health of providing various kinds of health-care financing was the RAND Health Insurance Experiment (HIE). In this experiment, thousands of families were randomly assigned to one of five different health-insurance plans that ranged from something like a plan that provides free health care, to something like a pure catastrophic-only plan in which consumers pay out-of-pocket for day-to-day healthcare. The study tracked what exact health-care services each group used, and how their health varied over a period of 3–5 years.
Ezra Klein describes this experiment as “the best evidence we have,” and writes that it “suggests that health-care coverage does much more for the health of poorer people than it does for the health of well-compensated, highly educated people.” His statement is correct, but as a summary of the results of this experiment, seems to me to be radically incomplete. In fact, the experimenters wrote of the findings that “cost sharing reduced the use of nearly all health services,” but “the reduction in services induced by cost sharing had no adverse effect on participants’ health.” Think about that. Providing people coverage of their medical costs caused no average improvement in health.
Klein is correct that there appeared to be a net health benefit for the poorest participants, but this was for a tiny proportion of the population, and for a small subset of medical conditions. According to the study, “The poorest and sickest 6 percent of the sample at the start of the experiment had better outcomes under the free plan for 4 of the 30 conditions measured.” There are technical reasons why conclusions from such a experiment are not reliable for post hoc subgroups in the way that they are for average comparison of a test group versus a control group; but even if we were to accept this finding as valid, it’s not obvious to me that we would want to devise a health-care financing system for the United States around helping 6 percent of the population partially ameliorate about 10 percent of their potential health problems, as opposed to developing some specific supplementary programs for these issues, if they could be addressed feasibly.
Klein clearly has a very sophisticated take on the issue, and wrote in 2009 that health-care reform is not primarily about improving health, but in reducing how much we spend on it. As he put it, “The purpose of health reform, in other words, is to pay for health care — not to improve the health of the population.” Fair enough. But the real debate, then, would be about whether market forces or bureaucratic control would be better at reducing costs, not about which would be better at promoting health for the “poorest and sickest” or anybody else. It wouldn’t be about getting better health outcomes.
A single experiment like the RAND HIE is not definitive. Among other things: it finished in 1982, and we live in a different world; any such experiment requires replication; it might be that the important health effects take much longer than five years to materialize, and so on. But as an observer of the health-care debates, it always struck me as fascinating that the fact that the “best evidence we have” showed that providing health care coverage doesn’t actually improve average health wasn’t treated as more central.
Fascinating, but not surprising. In one social-policy topic after another, experts argue that some program will transform some area of public life, and solve persistent problems. They often have impressive theoretical arguments supported by complex empirical evidence. But what I believe randomized experiments have shown in many such areas — ranging from welfare to criminology to education — is that proposed policy interventions rarely work, and when they do, they tend to produce improvements that are very small as compared either to size of the problem or to the dreams of the advocates. This evidence is often ignored by those who have dedicated their lives to solving these problems, likely because it is so frustrating to almost everyone involved.