Under the subject line, “Misleading Graphs of the Week,” Bill Jefferys sends along this:
I agreed with Bill’s colleague Helen Read who wondered why should the 90th percentile be some magic number? Just change it to 85% or 95% or whatever and all the graphs will look different. Also kinda horrible that they’re presenting percentages to 2 decimal places, but that’s not a graphics issue, it’s just plain old innumeracy.
“why should the 90th percentile be some magic number”
Some people just seem to have a psychological need for a “bright line” cutoff, I guess.
Maybe the have set for themselves the target of being in the top 10%. A completely arbitrary threshold,, but I don’t think plotting the full empirical distribution of scores (even if they knew it) would be better.
Typically, Blue Cross/Blue Shield of Vermont wouldn’t have access to the entire data set comparing all of the plans. They typically have their own data and a few percentile scores such as 90th, 75, 50, 25.
That said, there is a lot that HEDIS and similar quality organizations could do to improve the metrics collected and shared for health plans.
I really don’t understand what they are trying to show here: That they are slightly worse than the 90th percentile? I would think they would have wanted to pick some percentile over 50 that they could beat.
Also if the point is that they are close to the 90th percentile, why not expand the vertical axis to run from 0-100 so the bars look almost the same?