Visual Bias at Work

By Henry Woodbury

Last week I blogged about a Harvard Business Review article on the inherent biases in visualization. Visual information makes people overconfident of outcomes.

Today the New York Times offers a perfect example. In the debate around U.S. health care overhaul, the president’s budget director Peter Orszag argued that savings could be found by reforming the current system:

Mr Orszag displayed maps produced by Dartmouth researchers that appeared to show where the waste in the system could be found. Beige meant hospitals and regions that offered good, efficient care; chocolate meant bad and inefficient.

The maps made reform seem relatively easy to many in Congress, some of whom demanded the administration simply trim the money Medicare pays to hospitals and doctors in the brown zones. The administration promised to seriously consider doing just that. [my emphasis]

Unfortunately, the maps don’t show what they seem to show. While they show cost of care (a very specific kind of care it should be noted), they don’t show quality of care. Nor do the maps show anything about the demographics of the patients being cared for.

The Times compares the Dartmouth map (on the left) to Medicare’s own analysis of hospital quality (on the right) to show the disconnect. However, the Medicare map raises questions of its own. To start with, it shows a suspicious correspondence to U.S. population density.

Perhaps quality of care relates to the proposition that higher population density creates demand for more specialists which leads to better diagnoses. I’m sure I’m not the first person to think of this. Before anyone draws another map, let’s work on better analysis.