Two Quick Comments on an Instant Classic: the Oregon Health Insurance Experiment
by Harold Pollack
I’ll have more to say about this study in another venue. (Until then, Naomi Freundlich, Jonathan Cohn, Ezra Klein, Gina Kolata, Austin Frakt, and Aaron Carroll have the study well covered. I envy the unique and timely access, opportunity, and the simple craftsmanship of this experiment. Hats off to the entire OHIE study team.)
For the moment, I want to address one bad argument and one good argument made by conservatives in response to OHIE's strong findings which document the value of Medicaid coverage for so many people....
OHIE demonstrates that Medicaid enrollment greatly improves the well-being of uninsured adults. Individuals have significantly improved self-assessed health status. They enjoy greater financial security. They are less likely to report depressive symptoms. They are hassled less often by collection agencies. Given the low incremental average costs of Medicaid enrollment for this group, this study provides very strong backing for the simple proposition that Medicaid improves life for millions of our fellow citizens. These findings are especially timely, given a Washington budget debate which seems focused in imposing punishing cuts on the single arguably underfunded component of our entire health care finance system.
Some commentators focus on the fact that there are no significant mortality differences between the treatment and the contrl group. Of course, there is no way that a mortality effect could be observed in these data. The combination of short duration and small sample size of relatively young individuals actually induced into health coverage provide no statistical power to observe or to reject a mortality effect. Overall mortality in the control group was 0.8 percent. Moreover, health insurance could only be expected to prevent or delay some of these deaths. If coverage had a powerful immediate impact on the causes of death it most plausibly effects, Back-of-the-envelope calculation suggests that the expected difference would only be a few deaths in this sample over this timeframe.
Twenty years ago, the investigative team within the RAND Health Insurance Experiment faced the same problem. To address it, they computed a predicted mortality measure. They indeed found that health insurance mattered. The lion’s share of the effect reflected improved hypertension control among low-income people with good insurance. An analogous study in 2011 would presumably add improved cholesterol and diabetes management given increased obesity incidence and the advent of statin medications. (Reduced stress and depression may reduce mortality, as well. It’s not unknown for people to commit suicide when they are hounded by bill collectors, either.)
If one examines Table 6 of Finkelstein et al's paper, they find that Medicaid enrollment increased hypertension screening by about 11 percentage points among those induced to take up coverage. The control mean is only 62.5%. This is a big effect. They find similar effects for blood glucose and larger effects for mammography and pap smears. One would have to run these preventive care gains through a credible mortality model, but it seems clear to me that one would find a noteworthy predicted decline.
I wouldn't expect the predicted mortality effects to be huge, but I would expect them to be valuable. We have even greater reason to believe that improved preventive care will reduce the incidence of avoidable disability and illness--that's important too. Improved health insurance coverage improves health. The ultimate mortality effects of insurance are fundamentally uncertain and are indeed easily oversold. I would not hang my hat on them. Nor does the policy question hinge on these. OHIE results are quite sufficient in my book to justify expanded coverage.
Conservatives such as Michael Cannon are correct about one matter: the non-impact of insurance on emergency department use. I continue to believe that it's a huge mistake to make reduced ED visits a liberal talking point or a metric of successful policy. There is a price elasticity of demand for these services. One should expect coverage to increase ED use, all else equal. Reducing both appropriate and inappropriate ED use is a complicated challenge of both social service delivery, medical management, and the culture of medical care use. For many reasons, I believe that the right goal should be to provide a solid financial and organizational foundation for ED services, not to reduce use.