The Continuing Saga of Geographic Spending Variation

A new study of sources of geographic spending variation is carried in the Quarterly Journal of Economics.  (QJE Study)   The researchers used the migration of Medicare beneficiaries from one location to another to attempt to disentangle patient and local health supply factors.  There has been ongoing disputation about the meaning of different utilization and spending patterns in various parts of the U.S., but understanding the factors behind utilization and spending decisions can help ensure that we reach the goal of ensuring appropriate but prudent delivery of medical care to optimize the health of patients.  Using data from 1998 to 2008 for a 20% sample of Medicare beneficiaries the authors tracked those who moved primary residence and analyzed changes in utilization and spending.  Removing geographic cost and price differences indicates the underlying utilization trends and any changes.  The primary research finding, both for patients moving from a high spending region to a low one and those moving from a low spending area to a high one, is that underlying patient characteristics appear to account for about half of the variation and local practice preferences about half.

Looking at different categories of utilization, the amount of regional variation that appears attributable to patients is higher in categories where patients have more ability to affect utilization, like preventive care and will be lower, and provider effects therefore higher, for ones which patients have less control over, like an inpatient hospital stay.  Particular provider characteristics, such as more for-profit hospitals and a higher percent of physicians who prefer aggressive care, tend to be associated with regions where location accounts for more variation than do patient characteristics.  There is also, however, some interaction between place and patient characteristics.  Regions with sicker patients and more of those from higher socioeconomic strata tend to have and use more health care resources.  And this study confirms what others have suggested; that unmeasured aspects of health status and health need account for a significant fraction, around 25%, of spending variation.

It should be noted that prior research finds different geographical spending patterns for Medicare versus commercially covered patients and that spending variation exists even within very small geographic areas, like a city.  This suggests that much variation is localized to patient and provider preferences as opposed to larger system-wide factors.  The results of this study indicate that in any given locale, we should expect a fair amount of utilization variation due to patient characteristics and preferences and that this variation will be most intense in areas like ER use and preventive care and less intense for inpatient care and diagnostic testing.  But we also should expect that without respect to those patient differences, we will see significant variation in care across regions due to provider characteristics and preferences.  So if our goal is all appropriate care all the time, we need to work on both the patient and provider sides of the equation.  And a lot of work yet needs to be done to connect outcomes to different utilization patterns, particularly if outcomes are to include patient satisfaction and preferences.