Monday, June 7, 2010

Efficiency Graphs (Well, Not Really)

The New York Times today printed a few graphs comparing health care spending (in percentage of the growth domestic product) versus "result metric". They show the numbers of various countries for CT scanners, angioplasties (both per million people), life expectancy at birth, deaths from cancer (per 100,000 people), annual consultations per capita, and hospital beds (per 1,000 people).

"Oh, great - efficiency graphs!" I thought.

The cost graph is very interesting. It supposedly shows that the U.S. is a real outlier in terms of the country's health care spending. However, the U.S. also has a higher GDP per capita compared to other large developed countries (such as France, Britain, and Turkey). Then, the total health care spending is even more of an outlier (see this graph).

However, I have two "real" takes on this: First, life expectancy at birth is the only outcomes measure. All the other measures are are not easy to interpret with additional information.

For example, the number of cancer deaths: The Times mentions that economists think this is an indication that the spending is out of line with what quality we get for that money. In other words: we get a low value. However, without the denominator (how many people have cancer?), this information is a mix of an epidemiological and an outcomes measure.
Or the number of angioplasties (Germany certainly has a lot of them) - are they all warranted? Are German and American cardiologists simply more aggressive to do this procedure? Does it lead to better or to worse outcomes? The Times does mention that the number of deaths from heart attacks is not much lower. However, there might be a higher number of heart attacks in the U.S. compared to other countries.

And then second, it would make a lot of sense to put these graphs together in an X-Y scatter plot to compare how cost and "results" relate to each other. For example, the real outcomes measure life expectancy at birth over GDP per capita. I looked for OECD data and came up with this:
Now, as for the value (cost over effectiveness or vice versa) and if the spending is "out of line": you could do a linear regression analysis. Here is what that would look like graphically:
As you can see, the U.S. has one of the largest residuals. However, maybe another function would fit the data better, what do you think?

By the way, if you are interested in health care cost then you probably know already the blog "Managing Health Care Costs". If not, I would like to highly recommend that you follow it. The blog is written by my former Professor Dr. Levin-Scherz who teaches a course on this topic. 


Post a Comment