- On your point #1 on sample sizes, note that the study is effectively even smaller than you’re saying. Thinking of the high A1C sub-sample as containing 624 people is correct for the intent-to-treat analysis (i.e. for estimating the effect of offering insurance). But (net) take-up was only about 25 percent. Thus, for the IV analysis (which provide the estimates of the effect of insurance per se), the point estimates and variance are scaled up by a factor of 4=1/0.25. This means that the “RCT-equivalent” sample size is only 156=624/4. That’s really, really small, even before you get to the fact that insuring people does not automatically get them the appropriate care (your point about direct versus indirect interventions).
- In thinking about costs and benefits here, it’s really important to be clear about the fact that the (social) costs of the expansion is the incremental care consumed by these individuals. So the key question is whether that additional care consumed (additional prescriptions, office visits, preventative screenings, and maybe a tiny bit of additional inpatient care) is “worth it” in terms of the health benefits.
Trying to decide the “is it worth it” question on the basis of the Oregon study is crazy when we have a clinical literature actually powered to quantify the benefits of particular medical interventions. For the increases in utilization observed, I think the general message of the clinical literature would be that “yes” that care is worth it.
The only thing that would make appealing to the clinical literature suspect is if Oregon was giving clear evidence that the health benefits are not what we would have expected given the change in utilization. But for all the cases I’ve back-of-the-enveloped, the health responses are right in the range you’d expect based on the utilization changes and our prior estimates of treatment efficacy. They’re just insignificant because of the power issues.
Bottom line: Oregon tells us that Medicaid gets people more recommended care. It tells us (virtually) nothing about the efficacy of that care, so if we thought that recommended care was a good thing yesterday, we still should today. Properly interpreted, therefore, it’s a big win for Medicaid.