• What is Evidence?

    Continuing on what is likely to be a controversial topic, at least in the rarefied strata of blogosphere wonkery if not beyond, let me add a few more less literature-based thoughts on the connection between health insurance and mortality.

    • The bulk of the uninsured are relatively young. Since mortality rates are very low for the non-elderly death is a particularly difficult outcome to work with when studying the effects of uninsurance.
    • If lack of insurance does kill you, it does so slowly. It isn’t as if you are dramatically more likely to die if you go without insurance for a month. Chronic lack of insurance or inconsistent coverage or poor coverage is what is likely to do damage to longevity. This also exacerbates the problems with death as an outcome measure.

    So, to the extent to which death has unfortunate properties for use as an outcome measure, one might expect conflicting and weak evidence in the literature. (I’m not convinced that is in fact what one should conclude about the literature. I haven’t had time to review it yet. I may be able to say more later.)

    Anyway, what to do? Well, statistical models of the causal effect of insurance on mortality are not the only form of evidence. Insisting that there ought to be a strong, measurable effect before one will believe it exists is silly. We’re smarter than that.

    We do know that death is related to poor health, right? Well then, all we have to do is relate insurance to health and we have an indication of its relationship to death. That’s what I reviewed  earlier. And there is sufficient evidence to be confident that lack of insurance leads to bad health outcomes, and particularly so for individuals with chronic illness (the very ones who struggle to obtain insurance in the marketplace due to pre-existing condition exclusions and medical underwriting).

    OK, suppose we’re convinced that lack of insurance has a role in increased mortality. Now ask, how big an effect? That’s harder to say with precision for the very same reasons given above. But we know the effect is not zero. It can’t be (read the previous paragraph). So, lack of insurance causes some number of deaths, but it is hard to say exactly how many. Some have tried. Others object.

    Recognize that this is a debate over the size of an effect that is exceedingly hard to measure, but that we know exists. Since it is clear there is a problem, I don’t agree with those that suggest that we need a “better handle on the case” before we address it.

    Share
    Comments closed
     
    • It isn’t a matter of insisting that there ought to be a strong, measurable effect before one believes in the causal effect of insurance. It’s that the search for that strong, measurable effect has shown that there isn’t one. We’re not dealing with an absence of evidence here, we’re dealing with evidence that contradicts what you know to be true. And you’re turning away from that evidence.

      • @Thomas – No. I’m saying one should not expect strong evidence on mortality. But that is not a good reason to believe there isn’t an important relationship. Anyway, hold your fire until you see the more thorough literature review in a forthcoming post.

    • Will do. I hope you address the Kronick study (which Ezra never mentions).