I recently attended a presentation by Bernard Black about a study he is working on where he evaluates the effect of health insurance on overall mortality and health among the near-elderly through an observational study using the Health Retirement Survey. He found that health insurance, in general, was not associated with differences in health outcomes, except for public insurance, which he found to be associated with higher mortality.
Black concluded that we would be better off without health insurance because it has either no effect, or perhaps even a negative effect, on population health. I’m not sure the observational study design he relies on supports such a strong causal inference, despite the sophisticated econometrics he employs. But I’ll leave the discussion about the internal and external validity of Black’s findings to the statisticians. I want to focus on his research question itself.
I find the question “what is the effect of health insurance on the health status of the near elderly” somewhat puzzling. In my view, health insurance is not a public health intervention or a healthcare service that is primarily meant to improve the health of any recipient; it’s really just a financial tool. Unlike iodizing salt or health education campaigns, we have no reason to expect that health insurance itself will improve the health of the average person.
Most people (including the near elderly) simply don’t need medical care in any given year, whether or not they have health insurance. Some people (i.e. those who can afford it) who need to go to the doctor in any given year will do so, whether or not they have health insurance. Most of the time, going to the doctor does not have an impact on your health that would show up on in a survey study (ex. it’s hard to measure functional or quality-of-life improvements). All of this suggests that having health insurance will not make a measurable difference on the overall health status in a population.
I’ll be more explicit: The way I see it, the purpose of health insurance is (1) to protect poor, credit-constrained individuals who become seriously ill from having to endure preventable health outcomes for reasons of affordability of care, and (2) to smooth the overall long-run income in a population.
The intuition: (1) Insurance will not improve your health outcomes if you do not become seriously ill, or if you do but can afford effective healthcare; (2) effective healthcare for serious illness is so expensive that even wealthy (or non-credit-constrained) individuals will face undesirable shocks to their income if they become ill.
I therefore find it puzzling that researchers seek to evaluate whether we should promote health insurance by asking whether it has an impact on population health. In fact, health insurance would not work if everyone who got insured needed it and used it. Health insurance, like any other kind of insurance, only works if most of the people who are insured don’t actually end up needing/using it. People buy insurance because they don’t know if they will end up needing it or not, but insurance works only if a large portion of those who do purchase it end up not actually needing it. So in any insurance pool, you should expect most of the people in that pool are in fact healthy and will not benefit from having insurance.
Therefore, asking about the effect of health insurance on overall health outcomes in a population (and not asking about the effects of health insurance on their financial outcomes) stacks the deck against insurance–it’s asking only about something which health insurance does not purport to do. A more fair research question would ask about the effect of health insurance on the health outcomes of those who we anticipate would benefit from health insurance (i.e. credit-constrained individuals who cannot afford healthcare and are likely to become seriously ill), and/or about the effect of health insurance on overall financial outcomes.
Of course, it would certainly be very interesting if it turned out that health insurance, by itself, did have an impact on the health of the average person in any given population! But I believe it it is wrong to conclude that we would be better off without health insurance if it doesn’t.
It’s sort of like asking “are lawyers good at sophisticated econometric analyses?” It would certainly be most interesting if it turns out they are, as with Professor Black. But it would be unfair to conclude that we would be better off without lawyers if it turns out most of them aren’t.
Julian (if I may),
This is a great post. But note that one need not even accept your reasoning to agree with the conclusion, for which there is a veritable mountain of epidemiologic evidence: access to health care services is simply not a significant determinant of health and its distribution in human populations. We can quibble over effect size, and, depending on how we operationalize “health care services” we can get larger or smaller estimates therein, the point put well by Lantz et al. in a 2007 people remains: lack of access to health care is not the fundamental cause of health vulnerability.
It’s ex juvantibus reasoning as well, as Ichiro Kawachi points out in _Unnatural Causes_: just b/c aspirin treats fever does not imply that the cause of fever is lack of aspirin.
As I said, I don’t think these kinds of arguments are controversial given the weight of the evidence (at least not in some of the public health circles in which I move). But nevertheless, the conflation of health with health care continues apace among both lay and professional, which I think is a serious problem (especially among the latter). Whatever the moral case for expanding access to health care — which I support completely — we have little reason for thinking that doing so would substantially improve population health and compress health inequities. In contrast, we have lots of evidence regarding what kinds of interventions would be likely to satisfy those goals (unsurprisingly, interventions targeted high up the causal pathway, at the structural determinants of health).
But instead we spend all of our time and energy on a category of intervention that is only going to move the needle a bit.
Frustrating, IMO.
Julian, many people have claimed that health insurance has an impact on health. Summarizing this literature, the 2009 IOM report said the lack of health insurance “results in needless illness, suffering, and even death.” Perhaps your point is best understood as a more methodological one, saying that the health-effect of health insurance is only going to be seen in the small proportion of individuals who will have catastrophic healthcare in a given year. But since these are only say 5-10% of the population, the effect gets swamped by population-level studies like Black’s.
Christopher (if I may),
Indeed, there is little doubt that insurance has some impact on population health. When thinking about priorities and categories of interventions, the question is one of effect size. And IMO there is limited evidence that the variable for health insurance is larger than many other important variables (I know there are some studies that suggest big numbers, but those studies virtually all have significant methodological problems, IMO).
Julian,
I agree with you. If the insurance is to have a significant positive impact on health it seems that the insurance would have to be used by large numbers. If the insurance is use in these large numbers the insurance is sure to collapse. I think they should have asked a different question.
I also find this quite a strange conclusion from a strange study. As you mention, insurance wouldn’t work if everyone received the benefit of the said insurance. The only thing that insurance does is that it transfers risk to the insurance company for the premium you pay, and I find it absurd that someone could conclude that this type of insurance has a negative effect on health. I liken this to saying that house insurance cause more fire, or car insurance causes more accidents. Crazy.