Geographic Variation in Medicare Spending and Service Use

Today, MedPAC released a report on regional variation in both spending and service use under Medicare, updating a previous analysis from 2011. Regional variation in unadjusted Medicare spending reflects many factors, such as differences in beneficiaries’ health status, Medicare payment rates, service volume, and service intensity. In contrast, regional variation in use of Medicare services reflects differences only in the volume and intensity of services that beneficiaries with comparable health status receive. There is little evidence that higher service use results in higher quality. Moreover, it is likely that health care service use and spending could be substantially reduced without harming quality if service-use patterns of high-use areas were brought into line with those of lower-use areas.

Some findings we present in this report were similar to those in our previous geographic variation report (2011). Areas with the highest (or lowest) service use are often different from those with the highest (or lowest) unadjusted spending; and post-acute care (PAC) — particularly home health care— is the primary driver of variation in service use.

But there are several important differences between our current findings and those from our previous work. Service variation is slightly lower now than we reported in 2011. Also, Miami, FL, and McAllen, TX—the highest use areas in our 2011 study—had substantial decreases in their use relative to the national average, but were still above average.

How did MedPAC measure regional variation in spending and service use under Medicare?

Our analysis looks at about 38 million beneficiaries who were enrolled in fee-for-service (FFS) Medicare in Part A, Part B, or both for at least one month in the year of analysis (2013 or 2014) and were never enrolled in a Medicare Advantage plan in that year. To examine the relationship between variation in drug use and variation in the use of medical services, we analyzed a subset of beneficiaries who were enrolled in Part D’s stand-alone prescription drug plans (PDPs) and received their medical services under Part A and Part B of Medicare. We aggregated the data into 484 geographic areas (which we call “MedPAC areas”) based on where the beneficiaries reside.

To determine medical service use, we adjusted beneficiary-level spending for regional differences in wages and special payments to certain hospitals and physicians (e.g., indirect medical education payments to teaching hospitals). We also adjusted for beneficiaries’ health status and demographic characteristics.

To estimate drug use we used gross drug payments to pharmacies for covered drugs excluding dispensing fees, sales tax, and any retrospective rebates and discounts from manufacturers and pharmacies. We again adjusted for beneficiaries’ health status and demographic characteristics.

What regional variation in medical spending and service use under Medicare Parts A & B did MedPAC find?

Overall, we found greater regional variation in Medicare spending than in the use of Medicare services. One way to examine this variation is by comparing spending and use by beneficiaries in high areas (which we defined at the 90th percentile of areas) with spending and use by beneficiaries in low areas (which we defined at the 10th percentile of areas). Spending in high areas was 47 percent higher than spending in low areas. By comparison, service use was 24 percent higher in high areas than in low areas.

Medicare Part A and Part B service use had less regional variation than Medicare Part A and Part B spending but large differences remained, 2013 and 2014

Note: “Spending” is per capita Medicare Part A and Part B spending among fee-for-service beneficiaries in each area. “Service use” is per capita Part A and Part B service use among fee-for-service beneficiaries in each area. We defined geographic areas as the metropolitan statistical areas (MSAs) of the core-based statistical areas for urban counties and rest-of-state non-MSAs for nonurban counties. If an MSA crosses state borders, we divided the MSA into multiple areas based on state borders. The units of analysis for determining the percentile distribution were the 484 MedPAC geographic units, weighted by beneficiary count.

Source:  MedPAC analysis of 2013 and 2014 beneficiary-level spending from the Medicare Beneficiary Summary Files and Medicare inpatient claims.

Our data also enabled us to look at regional variation for different categories of services. We created the following three groups:

• The acute hospital inpatient sector, which combines the short-term inpatient and inpatient psychiatric sectors;

• The ambulatory sector, which combines the services from claims processed by carriers (primarily for physician services) and the hospital outpatient (OPD) sector; and

• The post-acute care (PAC) sector, which combines the home health, skilled nursing facility, long-term care hospital, and inpatient rehabilitation facility sectors.

Overall, PAC had the greatest variation and also had a disproportionate effect on the overall variation in service use. The ratio of service use at the 90th percentile to service use at the 10th percentile was 1.88 for the PAC sector, 1.20 for the ambulatory sector, and 1.16 for the inpatient sector. Looking at the ratio of maximum and minimum values, the PAC sector had the largest ratio at approximately 5.7, while the ambulatory sector was at about 1.6, and inpatient was at about 1.5.

The post-acute sector had the greatest variation in service use, 2013 and 2014

                     
Note: “Service use” is per capita Medicare Part A and Part B service use among fee-for-service beneficiaries in each area. We defined areas as metropolitan statistical areas within each state for urban counties and rest-of-state nonmetropolitan areas for nonurban counties. The units of analysis for determining the percentile distribution are the 484 MedPAC geographic units, weighted by beneficiary count

Source:   MedPAC analysis of 2013 and 2014 beneficiary-level spending from the Medicare Beneficiary Summary Files and Medicare inpatient claims.

Removing PAC use from total use reduced the average difference from the mean from $56 to $34—that is, by 39 percent, even though PAC services were only 17 percent of the total. The variation in use of PAC services is particularly evident at the top and bottom of the distribution. At the top 10th percentile of areas ranked on total service use, use of PAC services was about $232 per capita compared with $100 per capita at the bottom 10th percentile. We also found a positive correlation between different categories of services, with a correlation of 0.31 between inpatient and ambulatory services, 0.37 between inpatient and PAC services, and 0.23 between ambulatory and PAC services, suggesting that areas with high use tended to have high use for all of these services.

What regional variation in drug spending and service use under Medicare Part D did MedPAC find?

Similar to our findings in the previous study, the use of drugs covered under Medicare Part D varied less than spending. Drug use in higher use areas (90th percentile) was about 21 percent higher than in lower use areas (10th percentile). By comparison, drug spending in high use areas was 38 percent higher than in low-use areas. Overall, variation in drug use was comparable to variation in medical services.

Medicare Part D drug use had less regional variation than Medicare Part D drug spending, but differences remained, 2013 and 2014

Note: “Drug use” is the average drug use per beneficiary per month among stand-alone prescription drug plan enrollees in each area. “Geographic areas” are the metropolitan statistical areas (MSAs) of the core-based statistical areas for urban counties and rest-of-state non-MSAs for nonurban counties. If an MSA crosses state borders, the MSA is divided into multiple areas based on state borders. The units of analysis for determining the percentile distribution are the 484 geographic units, with each unit weighted equally (i.e., not weighted by beneficiary count).

Source: MedPAC analysis of the 2013 and 2014 prescription drug event data.

For the subset of FFS beneficiaries with Part D drug coverage, we also evaluated the correlation between the use of inpatient, ambulatory, and post-acute services and use of Part D drugs. We did not find a significant relationship between the use of those services and the use of drugs. That is, areas with very low or very high use of medical services did not consistently have very low or very high use of drugs.

Do spending and service use vary systematically for urban and rural areas?

For the areas we looked at, we found that average service use was similar in metropolitan areas and nonmetropolitan areas. For the metropolitan areas, average service use was 99.9 percent of the national average, and for nonmetropolitan areas, average service use was 100.2 percent of the national average.