A Warning that Decreasing Variation Might Not Save As Many Dollars as We Hope

There is a food fight between researchers at Dartmouth and an economist at Wharton in the early December Health Affairs web edition.   The Dartmouth researchers have reviewed small area variation in Medicare claims data, and concluded that the states and regions with the highest cost also have lower quality.  Extrapolations show that if the lower efficiency areas could match the efficiency of the higher efficiency areas, we would need fewer physicians (by far) and we could save 30% (or more) of our health care bill.  This allows health care policy experts to conclude that we can save enough in health care to provide coverage for all Americans with no incremental cost.  

Katherine Baicker (now of Harvard School of Public Health) and Amitabh Chandra wrote a classic article in 2004 showing an inverse relationship between Medicare cost and quality. The more Medicare spends (think Louisiana and Florida), the worse the statewide quality scores are.   I’ve used their impressive graphics in my lectures, and I’ve seen them in many other talks.   I'm reproducing this correlation below , as pictured in the Cooper paper with explicit notice of different geographies which shows clustering of states by region.  If this is hard to read, double click on it to enlarge the graphic. 


Richard Cooper, a professor at Wharton, has made two pretty effective counter-arguments.  The first is that Medicare spending is often substantially different than non-Medicare spending state-by-state.  For instance, in Mississippi (a high Medicare cost state), 25% of Medicare beneficiaries are disabled, compared to only 10% in North Dakota (a low Medicare cost state).  Turns out that total cost of care in Mississippi is actually relatively LOW – because the non-Medicare payers are paying poorly.  Frankly, high Medicare costs might be a marker in the South for low non-Medicare spending, which is itself associated with the socioeconomic profile that is associated with lower apparent health care quality.


Below is Cooper's graphic showing a proportional relationship between non-Medicare spending and quality - with a slope inverse to that for Medicare spending. 

Cooper’s second point is that by his estimation low quality states consistently have low numbers of physicians – both primary care physicians and specialists.  They might have a worse primary care to specialist ratio – but they still have fewer specialists per capita.  He therefore argues against the assertion that substituting primary care physicians for specialists will necessarily lower costs or raise quality. 


All the analyses of primary care vs. specialists are confounded by rural vs. urban (few specialists in rural states), and the fact that primary care often considers only those trained in family medicine.  Family medicine represents a substantial portion of primary care in the Pacific Northwest and the Midwest – but a very small portion of overall primary care in most of New England.


There are a total of six articles in this series – and a wonkish argument about the validity of multiple regression analysis vs.  isolated correlations.  I went into this set of articles convinced of the Dartmouth Atlas argument that “more is less.”   I finished reading the articles feeling dubious about Cooper’s argument that “more is more,” but feeling far less confident that we can save as many dollars through decrease in variation than we all would hope.