MedPac Analysis Finds Less Geographic Spending Variation

A number of researchers, most prominently those associated with the Dartmouth Atlas project, have suggested that Medicare spending per enrollee has wide geographic variation which is not correlated with better health outcomes.  The conclusion these researchers draw is that a lot of inappropriate care is being delivered in the high spending areas and that reducing that inappropriate care could significantly cut Medicare outlays.  A new report by the Medicare Payment Advisory Commission supports the notion that there is significant variation, although not as great as some researchers have suggested.  (MedPac Report)

MedPac first noted the importance of distinguishing between Medicare spending and Medicare service use.  The former varies in part because Congress has intentionally set up formulas to pay different amounts for the same service in different parts of the country.  Raw spending is 55% higher per enrollee in the geographic areas at the 90% level of spending versus those areas at the 10% level of spending.  After making a series of adjustments to account for the geographic payment rate differences, health status and other factors, MedPac arrives at a service use variation metric, which it views as a more accurate reflection of actual variation in medical care delivery.  The service use variation is 30% higher per enrollee at the 90% level versus the 10% level.  One puzzling question is why MedPac doesn’t just look at pure utilization on a health status adjusted basis–how many units of various services per enrollee?

MedPac’s analysis shows that there is substantial variation in service use within states and across providers in the same geographic area.  MedPac also looked at rates of growth, as opposed to just the absolute spending or service use.  Interestingly, rates of growth are not strongly correlated with current level of service use.  Some high service use areas have low growth rates and some low service use areas have high growth rates.  Overall the correlation between service use and growth rate is slightly negative.  This might imply that over time either service use variation will lessen or that the pattern of geographic distribution of high and low service use areas shifts.  MedPac recites the usual possible explanations for variation–physician practice patterns, beneficiary preferences, resource availability, etc.–but adds that in some of the extreme high service use areas, fraud and abuse is a likely contributing factor.

The ongoing problem with all the Medicare spending variation research is a lack of detailed studies at the individual physician and hospital level, correlated with the individual patient level.  Only then could you make a firm judgment about the appropriateness of the care and care delivery patterns.  What is frustrating is that Medicare has this data in its possession now  and could have performed these analyses long ago.  That level of analysis would also help create a clearer set of options for how to address any inappropriate care.

Significant variation in care patterns also exists in the private insurance world.  This variation creates opportunities for software and service vendors to assist in identifying and eliminating inappropriate care.