• Optimal design of healthcare organization and finance

    An interesting paper from Ted Ruger’s seminar at Penn Law, by Jessie Smith Nibley, looks at health outcomes across OECD countries (ssrn, free registration).  Of course, you’ve seen that from Aaron before.  What makes this paper novel is her method of categorizing countries by salient health system structural features:


    The conclusion (emphasis added):

    First, there is a rough bell- or upside-down-bell-curve in most of the charts—countries in the outer-most groups generally did poorly, and countries in the center groups generally performed well. In terms of the country groups’ characteristics, extremes in reliance on public or private delivery and insurance appear to be bad for health outcomes, and moderation or a mix of public and private regulation appears to be good for health outcomes. Second, Groups 3 and 4, representing countries with no gate-keeping, performed consistently well across all categories, often the best of all the groups. This trend suggests that gate-keeping may lead to relatively poor health outcomes and may be a better explanation of the curve than the extreme/moderate attributes of the groups. Finally, the high expenditure averages in Groups 1 and 2 suggest that systems based mainly on market regulation with private insurance are more expensive than mainly public systems.

    Here’s the charts:

    • I wish we had access to all of Brian Greene’s parallel universes (The Hidden Reality: http://www.amazon.com/Hidden-Reality-Parallel-Universes-Cosmos/dp/0307265633/ref=sr_1_1?s=books&ie=UTF8&qid=1302689575&sr=1-1) so we could get an N > 2 in Group 4.

    • Agree that this is an interesting classification scheme. Always get a little nervous when outcomes data only report means without std dev, especially with tiny n’s.

    • I wonder how much of the performance in Group 1 is dominated by the US outcomes? I have direct experience with both the US and Swiss systems. They are very different, though full implementation of the ACA will bring them closer. I can’t speak to health outcome differences. But the patient experience, with physicians, hospitals, and insurers, is vastly superior in Switzerland than in the US. This includes things like wait times for appointments, time doctors spend with patients, amount of interaction with hospital staff, amount of haggling over insurance coverage decisions, etc.

    • @David – the paper also presents the charts with the US taken out, I just didn’t put them in the blog post

      @ Larry & Bill – as for the small n, I agree, but that’s inevitable with comparative international health policy in OECD countries

    • This doesn’t seem to be a very good way of trying to generalize about health outcomes. You really have a mixed bag of outcomes in each of the category and your “averages” are really a mixture of extremes.
      If you take your Category 1 (“worst health” – Germany, Netherlands, Switzerland, US) you have a wide variation in the indicators you cite. For instance, you cite a life expectancy of 80.5 years for the group but the US is actually 77.9 and Germany is 81.7 years in the same group. The Category 3 (“best health”) contains Japan (82.6) and South Korea (78.6). So the grouping is meaningless for life expectancy. (If you do the same analysis for the other indicators, you will find similar disparities in health indicators which make the “averages” meaningless).
      I just don’t see how the authors could come to any conclusions about financing mechanisms and health outcomes when they have grouped countries with such a wide variation in health outcome indicators together.

    • Not sure the U.S. system should be described as extreme in its reliance on private insurance. Taking Medicare and Medicaid into account, it is decidedly mixed.

    • Too small of a data set because there could be confounding variables.

    • As a computational biologist I find it strange that an a priori classification scheme was placed on the OECD countries and then summary stats for these categories were computed and used to make broad conclusions about the countries in the groups without much attention paid to within group variation. A more objective way to go would be to take all of the health indicators from all countries as use clustering methods to bin countries into similar groups (k-means clustering or decision trees or whatever). Then the groupings could be inspected to try to detect the main ‘shared’ properties of the medical systems within the countries that are within the group in contrast to those outside the group. Then the a priori hypotheses on the importance of gate-keeping versus no-gate-keeping and private versus public could be objectively assessed as to whether they correlated with the observed clusters of countries. Other features like income uniformity across population, average income, GDP etc could then also be examined to see if they are predictive of the clusters observed.