Lying With Statistics
December 8th, 2008 by Candice Chen (email author)In 1954, Darrell Huff wrote a book entitled, “How to Lie with Statistics” which suggests the way statistical analyses are chosen and presented can bend the truth. This month’s Health Affairs web exclusive suggests we could all use a review of this lesson.
On Dec. 4, Health Affairs released a package of web exclusives - a sort of point-counterpoint examining the relationship between physician supply, workforce composition and health care quality. Buz Cooper’s article: States with More Physicians Have Better Quality of Health. Baiker and Chandra’s response: Cooper’s Analysis is Incorrect.
For the non-statistician, Baiker and Chandra’s arguments are at times confusing and Cooper’s graphics appear compelling. Cooper finds that states with more physicians per capita generally rank higher in state health care quality rankings. He also finds that states with more specialists are associated with higher state incomes per capita.
But, does this mean that more specialists result in better quality (Figure A) or does it mean that higher income results in better quality and incidentally more specialists (Figure B)? Where is the true causal relationship? Are more specialists only a representation of higher income and otherwise unrelated to quality?
In this case, income might be considered a “confounding factor” – in other words, income is related to both quality and the number of specialists, and the failure to control for income in the analysis may lead to an inaccurate conclusion that more specialists equals better quality. These are hypothetical arguments. I have not done the analysis either to support or negate Cooper’s arguments. However, Baiker and Chandra find that when they compare areas with the same number of total physicians, areas with higher generalist to specialist ratios have better quality and lower costs.