Geographic Variation in Medicare Spending

Medicare data was the original source for research on geographic variation in health care utilization, cost and quality.  It continues to be a fertile source for such research and at the recent meeting of the Medicare Payment Advisory Commission, staff presented the latest analysis.   (MedPAC Presentation)   As the analysis initially points out, it is important to separate the effects of service use from price in looking at spending variation; they do not co-vary universally or consistently.  Service use variation results from volume and service complexity, as well as health status.  Even in Medicare, with its fiat prices, there are price differences depending on regional cost factors and add-on payments.  The latest analysis used data from 2013 and 2014.  They adjusted for add-on payments, demographics and health status.  484 geographic units were used.  Service use is strongly clustered around the national average; spending is more diffuse, especially on the right, or high, side of the distribution.  Per capita service use is close to equal in urban and rural areas.  A very large proportion of the variation is accounted for by post-acute care–skilled nursing care, home care, etc.  The ratio of 90th percentile spending to 10th percentile is 1.88 for post-acute care, 1.16 for inpatient care and 1.2 for ambulatory care.  Hence the drive to get bundled payments that include post-acute care.  Looking at those fee-for-service beneficiaries with drug coverage, they are more likely to be older, female, high higher spending and more medical conditions.  Geographic variation in drug use is less than the variation in drug spending.  Drug use is more concentrated among a subset of beneficiaries than is medical spending.  There was no systematic relationship between amount of drug use and total use of medical services or large categories of medical care in a particular geographic area.  Since the last analysis in 2011, the staff found that variation in medical service use had declined slightly and even post-acute care use variation was lower.  Some notable areas with high spending–Miami, McAllen, Texas–had average spending that declined, but was still much higher than the national average.