Fee-for-service Medicare payments to physicians and certain other licensed clinical practitioners (including nurse practitioners, physician assistants, clinical nurse specialists, and occupational and physical therapists) are adjusted for geographic differences in market conditions and business costs. These geographic adjustments are intended to ensure that payment to providers reflects the local costs of providing care, so that the Medicare program does not overpay in certain areas and underpay in others.
Each of the three components of the Medicare Physician Fee Schedule (PFS)—physician work, practice expense (PE), and malpractice (MP) insurance—is adjusted for differences across geographic areas in the input prices related to each component. When they are combined, these three components are known as the geographic adjustment factor (GAF).1
This chapter describes the history, intent, and evolution of the geographic practice cost indexes (GPCIs) to provide background and context for the committee’s findings and recommendations about improving the accuracy of payment. The committee sought to develop a uniform and consistent approach to the GPCIs and the hospital wage index (HWI) (see Chapter 3) by employing comparable data sources and methods.
Throughout its deliberations about the GPCIs, the committee has made a distinction between geographic adjustments that are designed to adjust payments for input price differences that providers face, and those that might be made to help address perceived workforce shortages and achieve other policy goals. While the committee acknowledged that both cost and access are part of its charge, it took the position that preserving access to care in nonmetropolitan areas should be done explicitly rather than using the GPCIs to address both cost differences and access issues (Zuckerman and Maxwell, 2004). The committee viewed the combination
1 Unless otherwise specified, the term “practitioners” is used to describe both physicians and other eligible clinical practitioners who are permitted to furnish services and bill Medicare under the Physician Fee Schedule (CMS, 2009). Physician assistants must be supervised by a physician, but nurse practitioners and certain other practitioners may practice independently if their state laws allow it and may therefore bill Medicare directly. Their payment is a set percentage of the Physician Fee Schedule.
of the two sets of issues as conceptually problematic by making it difficult to distinguish the level of resources being allocated to each objective, which affected the determination of the accuracy of payment.
Accordingly, the committee’s conceptual distinction is reflected in the structure of its reports. The committee’s phase 1 report addresses geographic differences in input prices, focusing on improving accuracy by relying on the best possible input price measures from an independent source. Phase 2 of the committee’s work will address broader policy issues, including workforce supply and access to care in the context of geographic adjustment. For example, physician practices have an increasingly diverse mix of employment arrangements, and advanced practitioners such as nurse practitioners contribute to the work component as well as the practice component of physician work. Accordingly, the phase 2 report will also consider the impact of the committee’s phase 1 report recommendations on geographic adjustment to fee-for-service payment in the context of current market trends toward delivery system integration.
Fee-for-service Medicare payments to practitioners are based on the PFS. The PFS is based on a list of more than 7,000 distinct services defined according to the nomenclature of the Current Procedural Terminology (CPT®) codes developed by the American Medical Association (AMA) (2011a). CMS uses the CPT® codes to create an expanded coding system called the Healthcare Common Procedure Coding System (HCPCS) and assigns HCPCS codes to the 7,000+ procedures that Medicare recognizes in its fee-for-service payment system.
Medicare payment for physicians and other licensed health practitioners for each service is based on submission of a claim using one or more HCPCS codes (CMS, 2011a). Each HCPCS code has an assigned number of relative value units (RVUs) that represents the cost of resources required to provide a particular procedure or service relative to the resources associated with other procedures or services. For example, a follow-up office visit and a cataract removal require different amounts of resources than those needed to perform a colonoscopy, so all are assigned different RVUs (MedPAC, 2008). The total RVUs for a procedure are subdivided into the three components of the PFS: physician work, PE, and MP insurance:
- Physician work RVUs reflect the time, skill, effort, judgment, and stress associated with providing one service relative to other services.
- Practice expense RVUs address the direct costs of providing a service and the indirect costs of maintaining a clinical practice, including administrative and clinical staff compensation (salary and benefits), rent, and supplies and equipment (CMS, 2010a). For most services, there are different PE RVUs for services provided in facility settings and in office settings. Practice expenses associated with supplies and equipment are not adjusted geographically because they are typically purchased in a national market with practically uniform prices across areas.
- Malpractice premium RVUs represent payment for professional liability insurance (PLI), also known as MP insurance (CMS, 2010a). The mean MP premium for each payment area is weighted for state- and insurer-specific specialty mix and adjusted for each insurer’s market share (O’Brien-Strain et al., 2010).
Before Medicare pays for a service, the RVUs for that service are adjusted for geographic differences in input prices and for provider type (e.g., physician, nurse practitioner, podiatrist, and others who can bill Medicare independently). Policy adjustments are also made, such as for services furnished in a provider shortage area. Then, the sum of the three geographically adjusted total RVUs is multiplied by a conversion factor (CF) that determines Medicare payment in dollars (see Appendix B).
Physician services include office visits, surgical procedures, and a broad range of other services provided in offices, hospitals, clinics, post-acute care settings, and other clinical settings (MedPAC, 2007). For most physician services in most settings, Medicare pays the provider 80 percent of the fee schedule amount and the Medicare beneficiary is responsible for the remaining 20 percent2 (MedPAC, 2010) after meeting the $162 deductible (HHS, 2011). Medicare pays nurse practitioners, physician assistants, and clinical nurse specialists at 85 percent of the physicians’ fees, after the deductible is met (MedPAC, 2010). However, their services can be paid at 100 percent of the physicians’ fees if they are “incident to” services, or services that are rendered by a nurse, and billed by the supervising physician (MedPAC, 2010).
Medicare pays for physicians’ services under Section 1848 of the Social Security Act, which requires that payments be based on national uniform RVUs (CMS, 2010b; Hsiao et al., 1988). The basic concepts and methodology of the current Medicare physician payment approach, known as the Resource-Based Relative Value Scale (RBRVS), were enacted in the Omnibus Budget Reconciliation Act of 1989 (OBRA) and implemented by the Centers for Medicare and Medicaid Services (CMS) in 1992. The change was intended to make Medicare payments more equitable by basing them on relative input use rather than on historical prices, and to reflect local variation in input prices. Additional statutory changes that affect geographic adjustment have been made over the years (see Box 5-1).
CMS is required by law to update the GPCIs that adjust these RVU-based fees every 3 years. The CY 2011 final PFS rule implemented the following changes to the adjustment factors in response to new statutory requirements in the Patient Protection and Affordable Care Act (ACA):
- Extended the GPCI work floor of 1.0 through FY 2011, in accordance with a provision in the Medicaid and Medicare Extension Act of 2010;
- Kept the permanent 1.5 GPCI work floor for Alaska in effect; and
- Established a permanent, non-budget neutral floor of 1.0 for practice expense for “frontier” states (Montana, Nevada, North Dakota, South Dakota, and Wyoming).
By statute, any changes to the GPCIs that do not explicitly receive additional funding must be budget neutral. In practice, budget neutrality requires that the total amount of payment be unaffected by new adjustments, so that any adjustment upward for one payment area must
2 Participating providers receive the Medicare Part B allowed amount as payment in full for services and bill the beneficiary only for any coinsurance or deductible that may apply. Payment for nonparticipating physicians (those who have not signed a Participating Payment Agreement with the Part B enrollment department at CMS) is 5 percent below the Medicare Physician Fee Schedule amount (CMS, 2009), but these physicians are permitted to bill patients up to 15 percent in excess of the fee schedule amount (https://www.cms.gov/mlnproducts/downloads/physicianguide.pdf).
1989: The U.S. Congress requires that the U.S. Department of Health and Human Services (HHS) account for physician work, practice expenses, and malpractice expenses when calculating the geographic practice cost indexes (GPCIs) (Omnibus Budget Reconciliation Act of 1989. P.L. 101-239, December 19, 1989).
1992: Section 1848 of the Social Security Act establishes a fee schedule for physicians’ services.
1996: The Health Care Financing Administration reduces the number of payment areas from 210 to 89 (CMS, 1996).
1997: The U.S. Congress requires the Centers for Medicare and Medicaid Services (CMS) to implement resource-based malpractice relative value units (RVUs) for all services provided, effective in the year 2000 (The Balanced Budget Act of 1997. P.L. 105-33, August 5, 1997).
2003: The U.S. Congress mandates review of the practice expense GPCI (Medicare Prescription Drug, Improvement, and Modernization Act of 2003. P.L. 108-173, December 8, 2003).
2005: The Government Accountability Office (GAO) reports that the GPCIs are sound conceptually but that data and data collection methods could be improved, such as by collecting more data on physician assistant wages and using commercial rent data rather than residential rent rates (GAO, 2005).
2007: GAO recommends that CMS design a uniform approach to defining payment areas, so that there is consistency from state to state, and that CMS base its locality structure on the most recent data (GAO, 2007).
2007: The Medicare Payment Advisory Commission recommends that CMS exclude expenses that do not vary geographically (including supplies and medical equipment) from the GPCI formulas to improve their accuracy (MedPAC, 2007).
2008: Acumen report for CMS evaluates four smoothing techniques, and concludes that each method would significantly reduce large disparities between payment areas (O’Brien-Strain et al., 2008).
2010: On behalf of HHS Secretary Kathleen Sebelius, CMS commissions the Institute of Medicine to evaluate the accuracy of the geographic adjustment factors in a 2-year study.
2010: The U.S. Congress passes the Patient Protection and Affordable Care Act of 2010, which establishes a wage index floor of 1.0 for frontier states, sets a practice expense GPCI floor for frontier states, and extends the work GPCI floor through December 31, 2010 (P.L. 111-148).
2010: In November 2010, CMS posts the final Physician Fee Schedule rule with comment period for the 2011 GPCI. The Final Rule describes updates to GPCI weights and includes new regulations in response to provisions in the Patient Protection and Affordable Care Act of 2010 (CMS, 2010b).
2011: On July 8, 2011, CMS issues the CY 2012 Physician Fee Schedule Proposed Rule, which proposes to change the GPCI cost share weights by decreasing the weight for work and increasing the practice expense (PE) weight; to add a new category for contract labor as a component of the PE; and to use American College of Surgeons (ACS) residential rent data for the office rent component of the GPCI.
be paid for by a downward adjustment for other areas. This requirement creates significant tensions among providers in high-versus low-cost areas.3
Another major source of disagreement is whether the geographic adjusters should be used as policy levers to help influence provider supply, particularly in nonmetropolitan areas. Some rural health policy experts and practitioners argue that because earning potential influences physicians’ decisions on where to practice, and because many private payers use Medicare prices as a basis for setting their own rates, the geographic adjustments should be used as policy tools to encourage physicians to practice in nonmetropolitan areas (Iowa Medical Society, 2010; MacKinney et al., 2003). Using the geographic price adjusters to raise payments in provider shortage areas has been called into question by others on the grounds that it is inconsistent with the underlying purpose of input price adjustments and reduces payment accuracy (Schwartz, 2010).
Another source of long-standing dissatisfaction over the geographic adjustment factors has been the use of proxy data from sources other than physician practices to measure geographic variation in the price of some inputs. Among practitioners, the complexity of the index construction and the lack of direct public access to some of the sources of data used for the index calculations have also been grounds for criticism.
The committee’s principles value transparency to stakeholders, but they also assign a high priority to the task of improving accuracy by relying on the best possible input price measures from an independent source. In the view of the committee members, proxy data for physician earnings are more accurate than are data on costs paid by providers because the proxy data are independent of local business decisions or other requirements, such as state laws on staffing ratios, which do not necessarily reflect input prices across labor markets. The committee also made a distinction between geographic payments that are intended to adjust payments for input prices and those adjustments that might be made to help reach policy goals, such as addressing shortages of clinical practitioners to maintain or improve access to care. Such policy adjustments will be addressed in the phase 2 report.
The GPCI payment adjustments are made for 89 different geographic areas in the United States, also known as payment areas (or localities). Some are defined according to metropolitan areas, but there are 34 statewide payment areas that include both metropolitan and nonmetropolitan areas (see Figure 1-5 in Chapter 1). Practice input prices may vary substantially within payment areas, particularly in the statewide areas. For example, although Texas has eight areas (Brazoria, Dallas, Galveston, Houston, Beaumont, Fort Worth, Austin, and the rest of Texas), San Antonio—the 25th largest metropolitan area in the country and the 3rd largest metropolitan area in Texas—is included within the “rest of Texas” payment area, despite the fact that practitioners there are unlikely to face prices equivalent to those in the nonmetropolitan areas of Texas.
Historically, CMS has relied on the advice of state medical associations when deciding whether to make changes to statewide payment areas. However, as the Texas example shows, statewide payment areas do not necessarily represent economically integrated areas with similar relative wages and rents, and they may not be the most accurate basis for adjustment. In
3 See, for example, statements to the IOM Committee on Geographic Adjustment Factors in Medicare Payment from Senator Grassley (2011), Eneida Roldan (2011), and Alice Tolbert Coombs (2010).
recent rules, CMS (2010a) noted that changes in demographics and local economic conditions have occurred since 1997, when the current payment area structure was developed and implemented. These changes may have led to inconsistencies between payment differences and input price differences that warrant reconsideration of the current configuration of payment areas.
The committee’s discussion and recommendations about revising payment area configurations are the subjects of Chapter 2. Because hospitals and physicians essentially draw from the same labor market, the committee recommends that the same set of payment areas be used for the HWI and the GPCIs, and that metropolitan statistical areas (MSAs) and statewide non-MSAs should serve as the basis for defining these labor markets. While the payment areas would stay the same for the HWI, implementing this recommendation would mean that the GPCI payment areas would expand from 89 to 441 areas, which would be a significant change. The impact of the change in payment areas will be assessed in the phase 2 report.
As described above, the GAF is a combination of three independent GPCIs, each used to adjust the fee schedule for geographic variation in input prices for a different component of the cost of physician care.
The relative contribution of these three components varies by type of service and the setting where it is provided. For example, the composition of the total RVU for the office visit code 99201 is roughly 40 percent work RVU, 57 percent PE RVU, and 3 percent malpractice RVU, while the composition for the emergency room visit code 99283 is roughly 74 percent work, 21 percent PE, and 5 percent malpractice. Because each CPT code is composed of a different mix of the three RVUs, and therefore the three GPCIs are combined in different proportions, each code has a different average GAF.
When it was introduced, the RBRVS was seen as a significant improvement over the previous system, which was based on the customary, prevailing, and reasonable (CPR) physician fees in each payment area. Payments based on the CPR method varied widely across areas but were only partially explained by differences in practice costs (Physician Payment Review Commission, 1991).
CMS updates the RBRVS to adjust values for new services and to reflect services that may be overvalued or undervalued after considering the recommendations of the AMA/Specialty Society Relative Value Scale Update Committee (RUC). The accuracy of the RUC’s valuation of services has been another source of discussion and debate for some time. According to the Medicare Payment Advisory Commission (MedPAC), the RUC process does not accurately identify services that are overvalued and tends to recommend higher values for specialty care (MedPAC, 2006).
In its discussions about accuracy and the work adjustment, the committee acknowledged the importance of the RVUs in the broader fee-for-service healthcare system, since most private insurers use the RVUs as the basis for negotiating fees with physicians in their networks. While the committee believes that further study of the accuracy of the RVUs is warranted in the near future, that effort is beyond the scope of this committee’s charge.
GPCI Cost Share Weights
To set the relative importance of each input category, CMS bases the GPCI cost share weights on those used in the Medicare Economic Index (MEI), which measures price differences (infla-
tion) from year to year (rather than across geographic areas) in the cost of providing services under the PFS (MaCurdy et al., 2011). The weight assigned to the GPCI for each component of the Medicare PFS is based on the sum of the MEI cost shares of the inputs that comprise that component. The MEI cost shares are updated annually to meet a statutory requirement, which states that any prevailing charge levels beginning after June 30, 1973 may not exceed the level from the previous year except to the extent that the Secretary find that a higher level is justified by year-to-year economic changes based on appropriate economic index data (P.L. 74-271).
In CY 2011, the GPCI cost share weights were based on the 2000 base-year MEI weights, reflecting physician expenses in 2000. In the PFS proposed rule for CY 2012, CMS announced plans to update the GPCI cost share weights with the 2006 base-year MEI cost share weights, which use more current practice expense data primarily from the 2006 AMA Physician Practice Information Survey (PPIS) (MaCurdy et al., 2011). This update would decrease the overall weight assigned to physician work, increase the overall weight assigned to practice expense, and disaggregate certain practice expense categories (see Table 5-1).
Within the practice expense component, the proposed rule for CY 2012 adds a new PE cost category for purchased services. The purchased services index reflects regional variation in input costs for contracted labor that would typically be outsourced, such as accounting, legal, and building maintenance services. These industries are included in the “all other services” category within the MEI office expense and the stand-alone “other professional expenses” category of the MEI (MaCurdy et al., 2011).
No geographic adjustment is applied to the portion of payment that corresponds to inputs, such as equipment and supplies that are generally purchased in national markets at prices that do not vary systematically by geographic area (GAO, 2005). Because the physician work GPCI is adjusted for only one-quarter of the geographic variation in the proxies used in the and no adjustment is applied to the equipment and supplies component of PE, only 48 percent of the GPCI cost share weights were adjusted for geographic input price variation in 2011. The changes in the proposed CY 2012 GAF would increase this percentage from 48 to 51 percent in CY 2012 (see Table 5-1) (MaCurdy et al., 2011).
|Cost Share Weights (%)||Geographically Adjusted Cost Share Weights (%)|
|Expense Category||Current Rule
|Equipment, Supplies, Other||12.81||9.97||0.0||0.0|
a Work cost share weight with the one-quarter work adjustment.
b Only 62 percent of the purchased services index is adjusted for geographic variation in contracted services. SOURCE: MaCurdy et al., 2011.
The physician work GPCI is designed to reflect geographic differences in the cost of physician labor across areas in comparison to the national average (CMS, 2010a). The committee discussed two key issues: (1) whether physician work should be adjusted for geographic differences in the price of physician labor and, if so, to what extent; and (2) what data should be used in computing the work adjustment.
The physician work GPCI has some unique characteristics compared to the practice expense GPCI. Practice costs such as office rent and wages of nonphysician personnel are determined in local real estate and labor markets, where geographic variation in input prices is well understood and reasonably well documented. Physician work costs are different, in that there is no broader market for this input beyond medical practices, making the physician labor market distinctive.
Moreover, many physicians are self-employed and have an ownership interest in their practice, and it is not uncommon for physicians in private practice to have a partial salary for administrative or clinical responsibilities. Earnings of self-employed physicians, including those in physician-owned groups, are therefore a combination of payment for their own labor and an entrepreneurial return on investment in their business (Gillis et al., 1993). There are so many variations in staffing arrangements in physician practices that physician income may not be accurately described by a measure that is based solely on the payments that physicians receive for providing services.
How Should Physician Work Be Geographically Adjusted?
The goal of geographic adjustment is to pay physicians at a level that is equivalent across geographic areas, given cost of living differences and differences in amenities across geographic areas. Since the implementation of the PFS and the RBRVS in 1992, there have been differences of opinion about whether and how to make geographic adjustments to physician work payments and how much the adjustments should be. Committee members reflected the range of opinions when the deliberations began, and there was support for full, partial, and no work adjustment.
A full work adjustment would mean that variations in earnings would reflect the full extent of differences in cost of living, as attenuated by area amenities. The rationale for a full work adjustment is that compensation rates in the private sector, including the health care industry, vary across labor markets. Public sector wage rates for a variety of occupations, ranging from census workers to highly skilled professionals and managers, also vary geographically. Geographic variation in wages for nonphysician health care workers is recognized and reflected in the geographic adjustment of hospital and physician office labor expenses. Furthermore, a substantial and growing share of physicians (nearly 50 percent of new physicians, according to the Medical Group Management Association [MGMA] ), are employees who are paid at locally prevailing salary scales.
A partial physician work adjustment of 25 percent has been in place since the work adjuster was developed because there was such a wide variation in the earnings data used to calculate the adjustment and policy makers would not support a full adjustment (Zuckerman and Maxwell, 2004). Committee members supporting a partial adjustment took the position that an adjustment was needed, but the data used to calculate the adjustment might not adequately reflect the variation in compensation in different areas. Thus, the appropriate amount for the
adjustment might be overstated or understated, especially if the market for physician services was found to differ significantly from the market for other professional services.
There was also some support on the committee for no work adjustment. The argument against any physician work adjustment is based on the view that physicians providing an equivalent service for a federal program should receive the same reimbursement regardless of where they are located; “work is work.” According to this view, Medicare’s work RVU already takes into account physician work effort, and it takes no more or less effort to provide the same medical service in different geographic areas (Goertz, 2010).
Given the variety of opinions, the committee turned first to a consideration of economic theory and discussed the applicability of the labor economics theory of compensating wage differentials, which addresses the relationship between wage rates and various attributes of a particular job. The economic argument for adjusting Medicare physician payment across areas is that, in general, compensation varies inversely with the affordability and desirability of an area as a place to live and work. Thus, wages will tend to be lower if there is a lower cost of living and greater availability of amenities. (See Appendix I and the discussion of the theory of compensating wage differentials in Chapter 2.)
According to this theory, compensation for physician labor, like compensation for other labor, should reflect the cost of living in an area, along with amenities that might affect wage compensation, such as the quality of schools and housing, access to recreational facilities, and professional opportunities. The theory implies that workers will accept lower monetary compensation in return for amenities they value and will require higher compensation in return for giving up amenities they value (Borjas, 2010; Ehrenberg and Smith, 2009). The theory further holds that these differences not only reflect the requirements of the local labor market but also are fair in that workers—especially relatively mobile professionals such as physicians—can move between areas if they perceive their salaries are misaligned with amenities and costs of living. The committee recognized that there may be substantial differences in preferences for amenities among individuals in the labor market. The committee also recognized that preferences for amenities may differ among persons in professional occupations from those in other occupations and also may differ between health professionals and those in other professions. The extent to which such differences exist and are related to differences in compensation by occupation in general and by profession in particular, however, has not yet been adequately measured.
Another perspective on geographic differences in the cost of providing services was provided in testimony from clinical practitioners about geographic differences in the requirements for support services that are not adequately accounted for in the national average RVUs by CPT code. For example, in rural areas, physicians can be isolated in solo or small practices with few available professional resources to assist with discharge planning or family counseling. In these circumstances, primary care providers take on many different roles that may not be reimbursed (Iowa Medical Society, 2011). Providers in medically underserved urban areas may also lack necessary supports—translators, for example (Flores, 2005), which increases the time required to communicate with patients. While the committee acknowledges the potential for such resource and payment gaps, its position is that payment for these support services is more appropriately provided through a different targeted mechanism than through a geographic adjuster focused on variation in input prices. These other issues will be examined further in the committee’s phase 2 report.
The committee next sought to reconcile its differences by pursuing an evidence-based approach to determining the level of desired adjustment, and whether it should be no adjust-
ment, partial, or full adjustment. A study by the Center for Studying Health System Change (HSC) found that mean physician incomes in metropolitan and nonmetropolitan areas were not statistically significantly different (Reschovsky and Staiti, 2005).4 However, a finding of no difference on average does not necessarily mean that there are no important differences among individual metropolitan and nonmetropolitan areas that should be reflected in Medicare payments to providers. Another study found that primary care physicians (general practitioners, family physicians, internists, and pediatricians) in nonmetropolitan areas earned about 5 percent less than their urban counterparts did, after making similar adjustments to those made in the HSC study (Weeks and Wallace, 2008). Neither study assessed possible differences among individual metropolitan and nonmetropolitan areas. Data for both studies were more than 10 years old and do not reflect the most recent trends in provider payment. The committee therefore concluded that new empirical evidence will be needed to confirm the full extent of differences in compensation across geographic areas.
After extensive discussion, the committee came to agreement that geographic areas vary in terms of prices of goods and services and desirability in terms of places to live and work, even if there are individual and professional differences in the ways that desirability is perceived by health professionals. The committee was also in agreement about addressing in its phase 2 report differences in resource use and the ways that services are provided in medically underserved areas.
Given the inconclusive empirical evidence on geographic variation in compensation, the committee concluded that new empirical evidence will be needed to confirm the full extent of differences in fee-for-service compensation of physicians and other clinicians across geographic areas. The committee therefore recommended that the work adjustment should be based on a set of principles involving accuracy, consistency, and transparency, as described in Chapter 1, and a systematic empirical process to generate new empirical evidence about geographic variation in compensation.
To generate this new empirical evidence, the committee recommended a multiple regression model using the incomes of proxy or reference occupations to predict physician incomes region by region. The approach is based on the logic of compensating wage differentials, which suggests that anything less than a full cost of living adjustment should be offset by the region’s desirable amenities. The proposed approach assumes that the preferences for amenities among the individuals in the proxy occupations—and thus the offsets from a full cost of living adjustment—are similar to those of physicians. If that were found not to be the case using proxy data, the statistical model could systematically compare physician salary data from different sources to improve the model’s explanatory power. The committee’s recommended approach to testing various statistical models for predicting physician compensation is discussed in more detail in the following section and in Appendix I.
How Much of the Variation in Physician Work Should Be Adjusted?
When the geographic adjuster for physician work was originally developed, it was based on nonphysician professional earnings that ranged from 28 percent above the national average, in Manhattan, New York, to 16 percent below the national average, in rural Missouri (Zuckerman and Maxwell, 2004). Policy makers concluded that the range appeared too large, and Congress
4 The study used self-reported data on net income from the 2000–2001 HSC Community Tracking Study Physician Survey, adjusting for hours worked, specialty, practice ownership, and payer mix, factors that also affect physician income.
required that the physician work GPCI reflect only one-quarter of the variation observed in professional earnings. This reduced the range to 9 percent above average for Manhattan and 5 percent below average for rural Missouri (Zuckerman and Maxwell, 2004).
Over time, Congress further limited the extent of geographic adjustments to physician work. In addition to the one-quarter work adjustment, two additional statutory provisions limited downward adjustments to the work component of physician fees. First, section 1848(e)(1)(G) of the Social Security Act requires that the state of Alaska receive a permanent 1.5 work GPCI floor for services furnished beginning January 2009, meaning that physician payment will remain above the national average of 1.0. Second, a provision in the Medicaid and Medicare Extension Act of 2010 extended the 1.0 temporary work GPCI floor, enacted in the Medicare Modernization Act (MMA) through December 31, 2011. These provisions raised Medicare fees to physicians in low-cost areas and narrowed urban-rural fee differences (GAO, 2005).
The congressional decision to adjust for one-quarter of the variation in physician work was the result of political compromise rather than empirical evidence. One subsequent study in the early 1990s found that the one-quarter work adjustment was a better fit than was the full adjustment or no adjustment in a statistical model relating the work GPCI and physician net hourly earnings as measured by the AMA’s Socioeconomic Monitoring System survey in 1990 and 1991 (Gillis et al., 1993). After adjustment with the one-quarter work GPCI, physician earnings still varied, though less so than for the other levels of work adjustment. However, this study did not attempt to estimate the optimal fraction for the adjustment or assess the proxy occupations selected, and the committee was reluctant to draw firm conclusions from one study with data that are now more than 20 years old.
The committee therefore concluded that the one-quarter work adjustment lacks empirical foundation and sought to develop an alternative using statistical modeling based on multiple regression, a standard statistical technique that allows testing and modeling of independent or explanatory variables to predict a dependent or outcome variable. The inputs to the analysis would be indexes representing the ratio of median compensation for an occupation in each payment area to the national mean of these median compensation levels, both for physicians and for the proxy occupations. Preferably, if appropriate data can be found, these income indexes should be calculated based on employed professionals. The statistical analysis would then be a linear regression5 to determine which occupations’ earnings best track physician earnings, then creating an adjustment index based on geographic variation in earnings in the other occupations. (The analysis is summarized in this section and described in detail in Appendix I.)
After fitting this linear statistical model, there are at least two ways to use the fitted regression model to calculate the work adjustment. One approach is to calculate an index to represent the predicted value for physician compensation from the regression model. This resembles the committee’s approach for nonphysician labor expense in the PE GPCI, but with an difference. For nonphysician labor expense, the geographic adjustment is based on the weighted average hourly wage of health care workers in each geographic area relative to the weighted average national wage for those same health care workers, where the weights used for the averaging are national employment for all occupations in all physician offices.
The committee also discussed a second approach to the work adjustment, in which the relative weights for each of the seven reference (proxy) occupations would be derived from
5 A linear regression model is used to explain the relationship between two or more variables by using a straight line to plot the strength of the relationship. For example, linear regression can be used to fit a predictive model to an observed data set of independent and dependent variables.
the regression equation. Under the current method of GPCI calculations, none of the reference (proxy) occupations are parts of the physician workforce; in fact, only two—nurses and pharmacists—are part of the health care workforce. By using the proposed regression equation to determine relative weights of the proxy occupations, occupations with a higher regression coefficient would receive a higher weight in the predicted value used to compute physician work.
For example, if monetary compensation in Occupation A tracks physician pay more closely than does Occupation B, Occupation A compensation would receive a higher weight in determining values of physician work. If the wages of the alternate occupations used as predictor variables were found to be highly correlated with each other, the choice of occupations would need to be reevaluated, perhaps by testing alternative choices of reference occupations and replacing the less predictive occupations (those with smaller coefficients) with more predictive ones (with larger coefficients). Furthermore, the total weight given to all occupations would also be determined empirically through the magnitude of the coefficients; thus, the choice of a one-quarter work GPCI or something larger or smaller would be determined through an objective empirical procedure.
There are many possible variations to developing a statistical model to set the level of the work adjustment, in terms of the data sources, specific variable definitions, and the possible of influence of high or low outlier values. The committee did not perform a full evaluation of each of the alternatives in the limited time available, but it recommended that CMS consider statistical modeling as a general approach. The committee concluded that an empirical alternative using statistical modeling would be an important improvement over the way the work adjustment is currently calculated. Appendix I presents a detailed discussion of some possible ways in which the modeling might be accomplished.
Which Data Should Be Used for Adjusting Physician Work?
In CY 2011, CMS computed the work GPCI using the relative median hourly earnings from 2006–2008 Bureau of Labor Statistics (BLS) Occupational Employment Statistics (OES) data of seven nonphysician occupation categories:
- architecture and engineering;
- computer, mathematical, life and physical sciences;
- social science, community and social service, and legal;
- education, training and library;
- registered nurse;
- pharmacists; and
- art, design, entertainment, sports and media (MaCurdy et al., 2011).
The use of the relative median earnings of these seven nonphysician (proxy) occupations to compute the work GPCI has been a source of disagreement among stakeholders since the GPCIs were introduced. Some believe that only actual physician wage data should be used in the work GPCI calculations. They question how accurately the relative median earnings of the seven nonphysician occupations reflect actual relative differences in physician compensation (e.g., Reding, 2010), and the committee addressed this concern in its recommendations.
When the geographic adjuster for physician work was originally developed in the early 1990s, it was based on the median hourly earnings of workers in professional occupations with 5 or more years of college education. This group of highly educated workers was assumed to be similar to physicians in the types of goods and services they purchase and in their preferences for area amenities (Zuckerman and Maxwell, 2004). Physician wages were available from the U.S. Census Bureau, but the GPCI developers considered those data inappropriate for several reasons. Primary among them was the concern that the data captured existing patterns of the very fee-based reimbursement system which the PFS would replace. This would result in endogeneity, or circularity: that is, existing fees could influence the value of the adjuster that would be applied to the new fee schedule.
Alternate Data Sources for the Work Adjustment
There is strong support in the provider community for continuing to use provider-generated data, such as those from surveys of physicians by the AMA and MGMA, for the work adjustment (e.g., Reding, 2010). The committee’s position was that the best approach to the work adjustment is to use variations in compensation for other professions as proxies, but it also decided that it was important to determine whether any existing data sources generated by providers might be adequate. The committee therefore considered several alternative data sources that might provide information on geographic variation in physician earnings for purposes of comparison with the physician proxy data that are used for the work adjustment. These sources included two nonphysician surveys: the BLS OES and the U.S. Census American Community Survey (ACS), and two surveys of physicians: the PPIS conducted by the AMA and the MGMA Physician Compensation and Production Survey (see Table 5-2).
The committee’s evaluation of the appropriateness of these four data sources was based on the following key characteristics:
- Sample size. The data source should have an adequate sample size to ensure that the variables described above are available at the level of MSAs and statewide non-MSAs, which define the labor markets recommended by the committee (see Chapter 2 for a discussion of labor markets). Labor markets are the payment areas by which physician payment rates vary. A large sample size is necessary so that each labor market contains sufficient data for reliable computation of the work GPCI.
- Response rate. The data source should have an adequate response rate to ensure that the intended population is represented by the sample. A low response rate increases the likelihood of a sample that is unrepresentative of the entire population of physicians.
- Representativeness. The sample should reflect a broad geographic distribution, and the survey respondents should represent the range of physician practice characteristics, including single and multi-specialty practices, self-employed and salaried physicians, and solo and group practices. There should be a representative balance of these types of physicians to avoid the potential for biases. Additionally, residents should be excluded, since their wages substantially underestimate physician earnings and would introduce geographic distortions.
- Timeliness of data. The wage data should be current and collected regularly.
|The BLS Occupational Employment Survey (Department of Labor)||2009 American Community Survey (Bureau of Census, DOC)||MGMA Physician Compensation and Production Survey||2006 AMA Physician Practice Information Survey|
|Sample size||1.2 million
|57,791 clinicians||5,825 physicians|
|Sampling unit||Employers||Households||Physicians and nonphysicians that bill independently||Physicians only (MD/OD)|
|Can be aggregated to level of MSA/statewide non-MSA||Yes||Yes||Yes||No|
|Representativeness of data||Random sample and high response rate; excludes self-employed physicians; resident salaries skew mean wage downward||Random sample and high response rate||Convenience sample and poor response rate||Random sample but poor response rate|
|Geographic representativeness||Broad representation (all U.S. physicians are represented at any level at which wage data are published)||Broad representation (all U.S. physicians are represented at any level at which data are published)||Regional composition of respondents:
|Regional composition of respondents:
|Frequency of data collection||Semiannually (200,000 establishments in each panel)||Continuously||Annually||One-time survey|
|Frequency of data reporting||Annually, based on a 3-year rolling average||Annually, based on 1, 3, and 5 years of data; first 5-year report expected in 2012||Annually||n/a|
|Requirements for reporting data in each cell||A minimum of 2 responses from at least 3 establishments||A minimum of 2 responses from at least 3 establishments||A minimum of 10 respondents from 3 practices||Varies by question type|
|Physician benefits||No||No||Retirement benefits only||Yes|
|Physician specialty||Anesthesiologists, family/general practitioners, internists, obstetricians/gynecologists, pediatricians; psychiatrists; surgeons||Yes, but varies: wages are reported only by specialties with sufficient sample size||Yes|
|Physician hours worked||Yes||Yes||No||Yes|
|Staff wages by SOC code||Yes||Yes, but varies: wages are reported only by occupations with sufficient sample size||No||No|
|Characteristics of Surveyed Physicians|
|Salaried v. self-employed||Salaried only||Both||Both; entrepreneurial return may be included in reported wages||Both|
|Solo v. group practice||Group only (plus hospitals and other health care employers)||Both||Both||Both|
|Single- v. multi-specialty practice||Both||Both||Both; 73% multi-specialty, 27% single-specialty||Both|
Bureau of Labor Statistics Occupational Employment Statistics (BLS OES)
The BLS OES data provide estimates of wages and employment rates for 800 occupations in 450 industries in the United States (BLS, 2011b), excluding self-employed individuals. The data are collected through a voluntary mail survey distributed to about 200,000 establishments nationally every 6 months (BLS, 2011b). Wages and employment rates are published twice yearly on the basis of a rolling 3-year average, based on a sample size of 1.2 million.
The BLS OES data include data on wages and hours, by several specialties, at the labor market level.6 Included specialties are anesthesiologists; family and general practitioners; internists; obstetricians and gynecologists; pediatricians; psychiatrists; and surgeons. The BLS OES has a large sample size (1.2 million establishments) and a relatively high survey response rate, 78.2 percent.
The limitations of the BLS data are the following:
- The survey does not include data on benefits;
- Precise salary information is not available at the higher levels because wages above $187,200 are collected in a single category of “$187,200 or higher,” and OES then assigns a mean wage to workers in that interval that is above $187,200;
- The data include wages for medical residents, which may result in underestimates of median hourly wages for physicians in areas with teaching programs; and
- Data at the MSA/statewide non-MSA level are available only for a limited number of specialties.
American Community Survey (ACS)
The American Community Survey, launched in 2005 by the U.S. Census Bureau, is a nationwide continuous survey of households that collects demographic, housing, social, and economic data, including wages and hours worked by occupation (U.S. Census Bureau, 2008). The ACS replaced what would have been the decennial census long form in 2010, and most of the questions are identical or nearly identical to the decennial census long form. CMS used long-form Census data for the physician work GPCI until CY 2011. The ACS surveys approximately 2.9 million households annually, with a response rate of 98 percent.
The ACS currently publishes 1-, 3-, and 5-year rolling estimates. One-year estimates of economic characteristics, such as wages, are provided for geographic areas that have a population of least 65,000 (U.S. Census Bureau, 2008). The ACS publishes period estimates7 of wages that represent data collected over 3 and 5 years for less-populated geographic areas such as micropolitan statistical areas and statewide non-MSAs. The U.S. Census released the ACS 5-year public use data in December 2010. The 5-year data include wage estimates for the less-populated areas, for which 1- or 3-year wage estimates were not published.
The limitations of the ACS data are similar to, but not the same as, those of the BLS data. Resident wages are also included. Unlike BLS data, which represent employed physicians only, ACS data include both employed physicians and self-employed physicians, whose reported wages may also reflect profitability from practice ownership or the degree to which a physician
6 Occupations are defined by the Standard Occupational Classification system. Industries are defined by the North American Industry Classification System. The Office of Management and Budget coordinated the development of both systems on the basis of the work of interagency and intergovernmental committees of statistical experts.
7 Period estimates are defined by the U.S. Census (2008) as estimates “based on information collected over a period of time.”
may draw a partial salary for clinical or administrative work (e.g., medical director of a clinical service area) related to a local medical group or hospital. In addition, the availability of annual specialty wage data in the ACS varies depending on the sample size of the specialties of the physician reporting wage data.
AMA Physician Practice Information Survey (PPIS)
The Physician Practice Information Survey is a national survey sponsored by the AMA for the purpose of updating the practice cost data used to develop the PE RVUs and to set the cost share weights for the MEI (Kane, 2009). The survey collected physician wage data in 2006–2007 by specialty, including employed and self-employed physicians and excluding residents.
In its review of the survey data, the committee was concerned about the survey’s small sample size (5,825 physicians) and low response rate (11.7 percent) (see Table 5-2). The PPIS threshold for presenting data is 20 observations, indicating possible small sample sizes in some of the cost data metropolitan categories. In addition, the AMA has indicated that it does not plan to conduct the PPIS again, so these data would not be available for future adjustments.
MGMA Physician Compensation and Production Survey
The 2009 Physician Compensation and Production Survey is a national survey conducted annually by MGMA. This survey collects physician wage data, including time worked, by specialty. Employed and self-employed physicians are included and residents are excluded. In comparison to the AMA PPIS, the MGMA Physician Compensation and Production Survey has a larger sample size (57,791 clinicians) and higher response rate (18.72 percent).
While the MGMA survey is not a reliable data source for computing the work GPCI, the committee considered whether it might be useful as a source of physician data for a statistical model to ascertain how physician wage variation compares to the wages of other professional wages. An advantage of the MGMA data for this purpose is that the data include information on the number of RVUs performed by each physician respondent, which would provide a way to control statistically for service mix, incorporating time, intensity, and skill per unit of physician work. However, as mentioned elsewhere in the report, the committee finds independent sources of data to be more accurate for calculating geographic adjustments to payment. The ACS data, when they become available, might be appropriate for such a model.
The practice expense GPCI adjusts for geographic variation in the direct costs of providing services and the indirect costs of maintaining a clinical practice, including administrative and clinical staff compensation (salary and benefits), rent, and supplies and equipment (CMS, 2010a). Practice expenses associated with supplies and equipment are not adjusted geographically because they are assumed to be purchased in a national market in which prices are similar across the country. As of 2011, the PE GPCI accounted for 43.7 percent of the geographic adjustment, on average.
Geographic adjustments to wages for clinical and administrative office staff are based on median wage data from the BLS OES for four occupations: registered nurses (RNs), licensed practical nurses (LPNs), health technicians, and administrative staff (CMS, 2010a).8 This selection of occupations dates back to the first iteration of the GPCIs and is based on a 1983 survey of physician expenditure data. At that time, those four occupations were the top earnings categories for employees in physician practices, although it was noted that the employee occupational mix varied by specialty. For example, radiologists were more likely to employ technicians, whereas psychiatrists tended to have only administrative staff support (Zuckerman et al., 1987).
Since 1983, the health care system and its workforce has evolved, and the current 4 occupations used for the employee compensation component of the PE GPCI may not accurately reflect the current practice costs of office staff. Physician practices have an increasingly diverse mix of employment arrangements and staffing configurations, many of which vary by specialty and subspecialty, as well as by local workforce supply and other factors that physicians do not control. Therefore, the committee considered the use of a PE employee compensation index using a broader range of occupations, which would better reflect the current workforce, thus improving the accuracy of the adjustment.
BLS collects wage data at three different levels: all-industry, health care sector only, and physician offices. Having decided to broaden the number of occupations included in the adjustment, the committee discussed which of these levels of BLS wage data should be used to compute the PE employee compensation index. All industry wage data have the largest sample size, but the committee is concerned that the large sample does not represent physician offices. Physician office industry level wage data are most representative of physician offices, but the sample size is smaller and the data do not address the problem of endogeneity. Health care industry-level data have a sufficient sample size that is more representative of physician offices than the all industry-level data and addresses the endogeneity problem. Therefore, the committee concluded that BLS health sector-level wage data are a more acceptable data source for computing the employee compensation PE GPCI. The committee found no compelling reason to restrict the number of occupations in the PE adjustment, as long as the weights used in the adjustments are specific to employment in physicians’ offices.
Because employment data are not available by practice type and are thought to be highly variable for reasons other than geographic variation, the committee will consider other ways to address occupation mix in the second phase of the study. Their considerations will be subject to the availability of data.
The committee also explored the degree of geographic differences in the mix of employees in clinical practices. On reviewing the data presented in Table 5-3, the committee considered whether the adjusters should reflect those geographic differences, or should be held instead to a national standard occupational mix. Variability in staffing patterns in MSAs and rest-of-state areas will be considered further in the second year of the study, subject to the availability of data.
8 In the CY 2012 PFS, CMS proposed to expand the four occupations used to compute the employee compensation index to 33 health sector occupations, which account for 90 percent of the total wage share in physician offices (CMS, 2011b).
|Occupations||National Employment Shares||Northeast||Midwest||South||West|
|Receptionists and Information Clerks||9||13||10||8||10||10||12||6||11|
|Billing and Posting Clerks||4||6||4||5||5||4||6||5||4|
|Licensed Practical and Vocational Nurses||4||4||7||5||11||5||10||2||5|
|Supervisors and Admin. Managers||4||5||4||2||3||4||5||4||3|
|Office Clerks, General||3||3||4||4||4||5||6||3||3|
|Radiologic Technologists and Technicians||2||3||2||2||2||2||2||2||2|
Alternate Employee Compensation Data Sources
An alternative to using the BLS data that CMS is using might be to use wage data from the ACS, as proposed in 2005 by the Government Accountability Office (GAO). Because ACS reports wage data annually, the GAO (2005) report suggests that the use of ACS wage data would make the PE GPCI more current and will allow the PE GPCI to be updated annually. For the sixth PFS update, CMS chose not to use wage data from the ACS because the 3-year public-use microsample reflected only 3 percent of households, which resulted in small sample sizes in certain geographic areas. For example, the pharmacist occupational category had fewer than 10 observations in the Manhattan, Kansas; Beaumont, Texas; and southern Maine areas.
However, in late 2010, additional ACS data became publicly available, offering certain advantages over BLS data, including a higher response rate, larger sample size (including wage data at the zip code level), and more frequent data collection (see Table 5-2). CMS (2010a) indicated that it will review the ACS data, and it has proposed to use them in the construction of the practice expense adjustment factor in the future. In the proposed revisions to the sixth update released in July 2011, CMS proposed to use ACS data to estimate regional variation in the cost of office space (MaCurdy et al., 2011).
Geographic differences in office rents are calculated on the basis of the median rent for a two-bedroom apartment, using data from the U.S. Department of Housing and Urban Development (HUD) (Pope et al., 1989). Even though physicians’ offices are located in commercial as well as residential areas, HUD price information is the only source publicly available for all metropolitan and nonmetropolitan areas. The appropriateness of using these data rests on the assumption that residential rents and commercial office rents are influenced by similar factors—for example, land scarcity and population density—although the lack of publicly available commercial data makes it difficult to fully test this assumption.
Data on median rents for a two-bedroom apartment are used to minimize the effect of outliers, which is reasonable for estimation of relative rental costs. The rent adjuster is based on the “fair market rent” under HUD’s Section 8 Housing Program, which has been criticized as not reflecting commercial space or actual cost differences in metropolitan and nonmetropolitan areas (Grassley, 2011).
Alternate Sources of Office Rent Data
To assess the accuracy of HUD data and respond to stakeholders’ concerns, the committee identified alternative public and commercially available sources of commercial rent data and compared the data available, the frequency and methods of data collection, sample sizes, and demographic information with the characteristics of HUD data currently being used. Table 5-4 presents these comparisons.
Each of the sources reviewed has strengths and weaknesses. For example, both the American Housing Survey and the Basic Housing Allowance collect only residential rental data. The General Services Administration (GSA) collects data on commercial rent for federal office space only and has limited geographic coverage. The U.S. Postal Service (USPS) collects rental data for commercial properties it leases or owns, but the reported lease costs may reflect a number of factors, including the date that the lease was signed and the type of building. REIS, Inc. collects
|U.S. Department of Housing and Urban Development (HUD)||American Housing Survey (U.S. Census Bureau and HUD)||General Services Administration (GSA)||Basic Allowance for Housing (U.S. Department of Defense)||U.S. Postal Service (USPS)||MGMA Physician Cost Survey for Single-Specialty Practice||REIS, Inc.|
|Data available||Residential rental rates for 0–5+-bedroom apartments at 40th or 50th percentile of a distribution of standard-quality housing units||Average price of residential properties by region, according to type of house (focus on structure, utilities, and amenities, rather than geography)||Commercial rent for federal government properties only||Residential rent rates for 1–4-bedroom apartments/detached houses, utilities, and renters’ insurance rates||Commercial properties leased or owned by USPS||Data on building and occupancy, reported as percentage of total revenue||Commercial rent rates for properties larger than 10,000 sq. ft, at zip code, county, and MSA levels|
|Collection methods||2000 census long-form survey, updated with ACS 1-year survey, BLS Consumer Price Index, and trending/random digit dialing data on market trends||Census employees call or visit to conduct personal interviews||GSA subscribes to various commercial data sources and hires independent appraisers to estimate the values of their properties; in particularly small markets, GSA uses a return on investment process to establish rent rates||A contractor collects data from multiple sources, including newspapers, real estate listings, and apartment management companies; utilities data are from the American College of Surgeons (ACS)||Distribute cost survey questionnaires to both medical group practices and others involved in physician practice management (participants included MGMA members and nonmembers)||Reis, Inc. conducts its own surveys|
|Frequency||Every 10 years, and updated annually||Biannually (odd years)||Appraisals every 1–5 years||Annually||Annually||Quarterly|
|Demographics||Only collects data on 2-bedroom residential units; excludes new units (<2 years old), units below the public housing rent threshold, and units with renters who have occupied the unit longer than 15 months||Federal government buildings only; does not reflect traditional market behavior or all geographic regions||Excludes “undesirable” neighborhoods||All USPS properties (leased and owned)||Nonmetropolitan (<50,000): 21.15%
Metropolitan (50,000–250,000): 29.29%
Metropolitan (250,000–1,000,000): 32.67%
Metropolitan (>1,000,000): 16.88%
|Sample size||530 metropolitan areas and 2,045 nonmetropolitan county areas||National: 55,000 housing units; Metropolitan: 4,100 units||All federal government buildings||400 military housing areas (in the United States), defined by zip code||25,300+ leased properties, 8,500+ owned properties||1,871 practices||169 MSAs total; Reis, Inc. samples 40% of each region each quarter|
|Available to public||Yes, free of charge||Yes, free of charge||Limited data are available||Yes, free of charge||Yes, free of charge||Yes, for a fee: rent as a percentage of physician operational expense|
commercial rental rates for properties larger than 10,000 square feet in metropolitan areas, but has limited information for nonmetropolitan areas.
Data from MGMA’s Cost Survey for Single-Specialty Practices do reflect physicians’ actual rental costs. However, the MGMA (2010) cost survey has a low response rate (19.06 percent), and the 2009 data are limited in sample size (n = 1,871) and representativeness. Specifically, sample sizes by state appear to be uneven, with 10 states having fewer than 10 observations each. In addition, as discussed elsewhere in the phase 1 report, the committee preferred an independent source of data that would accurately reflect input prices faced by providers, not the costs incurred by providers.
In addition to reviewing the limitations of the individual data sources, the committee also compared HUD’s data with the REIS, Inc. and USPS data for a select number of metropolitan areas. The REIS, Inc. and USPS data on commercial rents were expressed in price per square foot, while HUD’s data were expressed as price per entire residential unit. In order to compare the data, the committee standardized the different units by converting the data into index values (see Table 5-5). The analysis shows substantial variation across the three sources, with HUD data providing higher index values in metropolitan markets in California, but lower values in other locations, such as Chicago and Raleigh-Durham.
In the CY 2012 PFS proposed rule, CMS proposed replacing HUD data with ACS residential rent data on the grounds that ACS data provide more detailed geographic information, rely on more current survey data, and will serve as a more standardized data source in the event that ACS wage data are adapted to compute the employee wage index and work GPCI (CMS, 2011b). It was estimated that 26 percent of localities would experience a change in their office rent index that would be greater than 5 percent if ACS data were used (MaCurdy et al., 2011). The proposal was in response to an Affordable Care Act mandate for CMS to explore using ACS data for portions of the PE GPCI (CMS, 2011b).
On the basis of its analyses for this study, the committee concluded that all of these sources had significant limitations. Most of them are not geographically complete, as they do not reflect market prices in both metropolitan and nonmetropolitan areas. Each source of data also yields a substantially different wage index, which indicates that they may not be representative of the market in which physicians rent space. Small sample sizes, low response rates, and sample biases also led the committee to conclude that these surveys do not accurately represent the physician population.
A variety of possible alternative sources of data were discussed. The committee favored adding a question on commercial rent prices to an existing federal survey, but no current survey was found that would be appropriate. The committee also considered the CMS proposal to mount a physician cost survey, but found data on costs incurred by providers to be less accurate than an independent source of data on prices faced by providers in the commercial market. Another problem with using practice data as a basis for market rent is that many physician practices pay rent to properties in which they have a partial ownership interest, and additional income produced through these arrangements may not have been excluded from self-reported data.
The committee also discussed whether the use of residential or commercial rent data would be more accurate conceptually, given that empirical comparisons of the available sources would be problematic for a variety of reasons. The committee concluded that the cost of space is not adequately addressed with residential data only. Therefore, the committee recommends that a new source of commercial rent data be developed for the PE GPCI.
|Metro Area, Using Reis Description||“Best match” CBSA code||REIS, Inc. “Effective rent”/sq ft, commercial||Un-weighted index||USPS median lease cost/sq ft||Un-weighted index||HUD median residential 2-bdr rent||Un-weighted index|
|District of Columbia||47894||$41.13||2.123||$3.65||0.706||$1,885.00||1.285|
|Salt Lake City||41620||$14.18||0.732||$6.06||1.172||$1,264.00||0.862|
Physicians purchase professional liability or MP insurance to protect themselves from possible financial losses due to MP lawsuits. The majority of physicians’ MP insurance policies provide coverage for $1 million per incident and $3 million per year (GAO, 2003). This is the standard for comparing costs from place to place.
The MP premiums that physicians pay are likely to vary depending both on their specialties and on the location of their medical practices (Jena et al., 2011). For example, specialists who conduct medical interventions that are more likely to result in medical malpractice claims, such as obstetricians, neurosurgeons, and orthopedic surgeons, pay higher premiums than do primary care physicians who do more clinical evaluation and management and fewer claim-prone procedures. MP premiums vary greatly from region to region. In 2010, on average, a general surgeon practicing in Miami–Dade County, Florida, might have faced an annual premium of $192,982 for liability insurance, whereas a general surgeon practicing in Nebraska paid $10,928 for the same liability coverage (Lowes, 2010).
The level of physicians’ concerns about the risk of malpractice litigation has been found to be high across a range of specialties, practice settings, and geographic areas at the state level, with wide state-to-state variation in the liability environments (Carrier et al., 2010). One reason for the geographic differences in MP premiums is that states have different tort laws governing medical malpractice and medical malpractice insurance. Medical liability and medical malpractice insurance are subject to state laws and regulations. Ultimately, the degree to which states monitor MP insurance carriers, control premium prices, and interpret liability can substantially affect MP premiums (Sloan and Chepke, 2008). The concentration of specialists and claims experience in a given location could also affect premiums. If an area has a high concentration of specialists with high liability risk, then the insurance carrier may charge them higher premiums to cover higher anticipated losses.
As described earlier in this chapter, the Omnibus Budget Reconciliation Act of 1989 (OBRA) (1989) required CMS to establish a Medicare PFS that used GPCIs to measure cost differences in physician work, practice expenses, and MP insurance and to adjust Medicare fees accordingly. If geographic differences in MP premiums were not taken into account, physicians working in areas with higher MP premiums would be subject to an additional practice cost not within their control (GAO, 2005). The current MP insurance portion of the Medicare payment formula consists of MP RVUs and the MP GPCIs, as discussed in the next section.
Malpractice GPCI Methodology and Data Collection
As of CY 2011, the MP cost share weight is 3.9 percent, which means that on average across all procedures, MP costs represent 3.9 percent of the total RVUs. The MP GPCI is based on MP premium data for 25 physician specialties collected from state insurance commissioners and private insurers that are averaged for each payment area. When CMS calculates the mean MP premium for each physician payment area, it is weighted for state- and insurer-specific specialty mix and adjusted for each insurer’s market share (O’Brien-Strain et al., 2010).
In 2003, the U.S. Congress directed GAO to evaluate the Medicare GPCIs, including the MP GPCI. The mandated review included an evaluation of the methods used to determine MP
costs, a review of the increases in MP insurance premiums and the variation of premium costs across states and physician specialty, and an evaluation of the MP GPCI and its relative weights.9
GAO recommended that CMS collect MP premium data more frequently from all states and from insurers that account for at least 50 percent of the MP insurance business in a state (GAO, 2005). In addition, GAO advised that CMS should collect data on each insurer’s market share by physician specialty, so that it could adjust average premiums for differences in specialty mix (GAO, 2005). GAO also recommended that CMS standardize the procedures used to collect data from insurers to improve the comparability of premiums within and between payment areas (GAO, 2005).
In response, CMS increased the number of states from which it was able to collect premium data from 33 in 2004 to 49 in the 201210 GPCI update (O’Brien-Strain et al., 2010a). Premium data were also collected from insurance carriers that represented 50 percent of the market share, or from at least two operating MP insurers per state. In addition, CMS increased the depth of the MP premium data from 20 specialties in 2009 to 25 specialties in 2012.
The primary sources used to collect market share data were the state departments of insurance; an alternative source was the National Association of Insurance Commissioners’ market share data. The primary source used to collect premium data was state rate filings, and the alternative source for filling in any gaps was the 2005 to 2008 Medical Liability Monitor survey.
The MP component of the Medicare PFS has received little specific criticism lately. This may reflect the small percentage of total RVU cost attributed to MP prices, or the perception that the adjuster is accurately based on real data on insurance prices that physicians actually face. Given the very short time frame of this study and the number of other issues under consideration, the committee determined that it would make no recommendations about potential improvements to the accuracy of the MP GPCI.
The committee’s charge is to evaluate the sources of data and methods used to calculate the GPCIs and to make recommendations about how to improve the accuracy of the geographic adjusters. In order to validate the use of geographic adjustment for the work and practice expense GPCIs, the committee in its analyses first sought to confirm the degree of metropolitan-nonmetropolitan and regional differences in physician compensation and in clinical and administrative staff compensation.
The committee then considered the accuracy of a variety of data sources that had been used or proposed for use in the GPCIs. The shortcomings of the available data on physician compensation, staffing patterns, contract labor, and occupational mix for different types of physician practices made it difficult to conduct thorough quantitative assessments.
The recommendations presented in this chapter relied on many of the same data sources that were used for analyses presented in other chapters. As indicated in the discussion of the committee’s principles in Chapter 1, these recommendations are intended to improve the accuracy of the GPCIs and also reflect the committee’s preferences for consistency in data sources
9 P.L. 108-173, § 403(c), 117 Stat. 2055, 2277-78.
10 Premium data from Mississippi and Puerto Rico were not collected.
whenever possible. If the use of new data sources were to change the total payments, CMS would need to make a budget neutrality adjustment to recalibrate payment levels, as required by law.
In phase 2 of the study, the committee will consider the role of advanced practitioners in different employment arrangements in physician practices. These analyses will be subject to the availability of data and may include simulations and modeling with different types of practitioners and practice settings. The committee will also consider recruitment and retention issues across areas and review available data on how specialty and geographic location decisions are made by the workforce, including contract labor. In addition, the committee will review the impact of previous policy adjustments to address workforce shortages and other strategies to address access to needed care in medically underserved areas.
Recommendation 5-1: The Geographic Practice Cost Index (GPCI) cost-share weights for adjusting fee-for-service payments to practitioners should continue to be national, including the three GPCIs (work, practice expense, and liability insurance) and the categories within the practice expense (office rent and personnel).
Geographic adjustments should be made for the prices of inputs that are purchased and/or produced locally and that vary from the national average. Inputs that are purchased in a national market without systematic variation in prices across geographic areas should not be adjusted geographically. In future Physician Fee Schedule (PFS) updates, the Centers for Medicare and Medicaid Services (CMS) should take steps to ensure accuracy in distinguishing between national and local market input prices. The statutory requirement to use the Medicare Economic Index (MEI) cost-share weights as the source of GPCI cost-share weights is reasonable and should be continued.
Recommendation 5-2: Proxies should continue to be used to measure geographic variation in the physician work adjustment, but Centers for Medicare and Medicaid Services (CMS) should determine whether the seven proxies currently in use should be modified.
Geographic variations in the price of physician time can be measured in two ways: by directly measuring variation in physician income, or by using income data from proxy occupations as indicators of variations in physician income. In keeping with its principles about accuracy and independence of data sources, the committee prefers an independent source of data that reflects geographic variation in compensation levels for comparable professions rather than using physician compensation data that are affected by Medicare’s payment adjustments.
Therefore, the continued use of proxy data for rate-setting to avoid the circularity of using physician income data is appropriate. However, in keeping with its principles of accuracy, consistency, and transparency of data sources, the committee recommends that CMS empirically reevaluate the accuracy of the seven proxies it currently employs using the most current Bureau of Labor Statistics (BLS) Occupational Employment Statistics (OES) data. The statistical process for this assessment is described in detail in Appendix I.
The committee recognizes that this empirical approach is conceptually challenging because there is no obvious “gold standard” against which the proxy-based estimates can be judged. Although the committee does not favor basing the geographic adjuster on actual physician
incomes in each area, it would be useful to assess the extent to which the proxy-based estimates are related to variation in physician compensation among geographic areas on a national basis. This process would validate their status as proxies. If the proxy data were not found to have predictive value for physician compensation, CMS might compare the predictive value of physician salary data from several different sources, such as the Medical Group Management Association (MGMA) and the American Community Survey (ACS). A proposed methodology for such a reevaluation using statistical modeling is discussed in the section on the physician work adjustment and is described in Recommendation 5-3 and Appendix I.
Recommendation 5-3: The Centers for Medicare and Medicaid Services (CMS) should consider an alternative method for setting the percentage of the work adjustment based on a systematic empirical process.
The committee recommends that the work adjustment should be based on a systematic empirical process that generates new evidence to confirm the extent of differences in compensation across geographic areas. There is clearly a policy precedent for the current one-quarter adjustment, given that the Geographic Practice Cost Indexes (GPCIs) have been updated six times since the Physician Fee Schedule (PFS) was implemented, and the “quarter work” adjustment has been in place by law throughout all of the updates. Many will view that precedent as adequate justification for continuing the same approach.
The committee members did not think there is an adequate conceptual justification for choosing that level of adjustment. However, based on the available empirical evidence, the committee was unable to determine a more appropriate level for the adjustment.
The committee therefore advises CMS to test various statistical models using multiple regression, a versatile technique that allows testing and modeling of multiple independent or explanatory variables to predict a dependent or outcome variable (see Appendix I for more detail). Once the necessary data are assembled, CMS has reviewed the data to ensure that they are credible, and the model is estimated, CMS would determine the empirically derived percentage for the work adjustment by using the model that provides maximum explanatory power.
Several alternative data sets could be used for the modeling, each with different strengths, weaknesses, and predictive power. At a minimum, the wage index data used in the modeling would have to be adjusted to control for specialty mix, RVUs, and residency training status to ensure that the variability in wages attributable to these non-geographical factors would not affect the geographic adjuster based on the models.
While the committee strongly supports an empirical approach to determining the work adjustment, it also acknowledges that it is impossible to determine in advance how much predictive power the most appropriate statistical model may attain. If the correlations between the proxy occupation wages and the physician wages were found to be low or not statistically significant, for example, that might indicate that the factors determining physician wages are too distinctive to be adequately captured by this methodology. The committee has considered the possibility that geographical variations in the market for physician services or in amenities (including professional amenities) valued by physicians might not parallel the corresponding variations for other professionals. If that were found to be the case, CMS would need to re-evaluate the use of the current proxies, as indicated in Recommendation 5-2. For purposes of modeling (but not rate-setting), CMS might also compare the predictive power of different sources of provider-generated data, such as Medical Group Management Association (MGMA) survey data and American Community Survey (ACS) data, when they become available.
Recommendation 5-4: The practice expense Geographic Practice Cost Index (GPCI) should be constructed with the full range of occupations employed in physicians’ offices, each with a fixed national weight based on the hours of each occupation employed in physicians’ offices nationwide.
The committee finds that independent, health-care-specific data from the Bureau of Labor Statistics (BLS) provide the most conceptually appropriate measure of differences in wages for health professional labor and clinical and administrative office staff. Although acknowledging that there are some regional differences in occupational mix of employees in the limited data available, the committee prefers a consistent set of national weights applied to wage data from the full range of health sector occupations so that hourly wage comparisons can be made. The exceptions are those health professionals who bill independently under Medicare Part B, whose compensation should be captured through the work geographic practice cost index.
The expanded set of occupations will be a better reflection of the current workforce and a broader range of health professions, which will help to improve accuracy of the adjustment. In addition, the expansion will anticipate future changes in the workforce brought by changes in the labor market, including the increasing demand for expertise in the adoption and use of health information technology. Further study of the mix of occupations by specialties will be valuable to determine whether geographic differences in approaches to clinical service integration and care teams should be addressed in future assessments of the geographic adjustment factors.
Recommendation 5-5: The Centers for Medicare and Medicaid Services (CMS) and the Bureau of Labor Statistics (BLS) should develop a data use agreement allowing BLS to analyze confidential BLS data for CMS.
The committee recommends that the data source for office staff wages should be all health sector employers’ wages and benefits data from the BLS. Comparable to the analyses and recommendations about the Hospital Wage Index (HWI), the committee concluded that independent data that reflect market prices faced by providers are more appropriate than provider data on costs paid, because actual costs also reflect business decisions that are not necessarily an accurate reflection of input prices. Further, the committee concluded that independent data on health sector wages would be a closer proxy to physicians’ office staff wages than all-industry data from BLS.
The committee recognizes that there is a need to increase coverage in areas where current data are not made available in public data files by BLS because of the need to protect confidentiality. Some areas have a very small number of providers; thus, increased sampling to improve accuracy may not be possible. A data use or other formal agreement between CMS and BLS would allow additional analyses to be conducted in the interest of improving transparency. Using all occupations instead of a limited number would be new, but BLS could compute an index that includes all data, including those data that are suppressed due to confidentiality.
Recommendation 5-6: A new source of data should be developed to determine the variation in the price of commercial office rent per square foot.
The committee reviewed several available sources of data to determine whether an accurate alternative is available to replace the U.S. Department of Housing and Urban Development
(HUD) residential data that are currently used in the practice expense geographic practice cost index. These included rental data from the American Housing Survey (U.S. Census Bureau and HUD), the General Services Administration (GSA), The Basic Allowance for Housing (U.S. Department of Defense [DOD]), the U.S. Postal Service (USPS), the Medical Group Management Association (MGMA) Physician Cost Survey, and REIS, Inc.
Each of these sources yielded a substantially different index, which indicates that they may not be representative of the market in which physicians rent space. They also collected and reported data differently (e.g., monthly rent v. price per square foot), which made comparisons difficult. Based on the limitations associated with each data source, such as low response rates, small sample sizes, and sample bias, the committee concluded that all of these sources would be imperfect or geographically incomplete proxies for variation in physician office rental costs. Because the committee also concluded that the cost of space is not adequately measured with residential data, the committee recommends the development of a new data source.
Recommendation 5-7: Nonclinical labor-related expenses currently included under practice expense (PE) office expenses should be geographically adjusted as part of the wage component of the PE.
The update for the physician payment rule proposed for comment in July 2011 included setting several labor-related expenses to a national index. These included occupations in the “All Other, Labor-Related” category (e.g., security guard and janitor) and the “Other Professional Expenses” category (e.g., accountants and attorneys). The Centers for Medicare and Medicaid Services (CMS) proposed to create a new category for contracted/outsourced services for these labor categories and to create a new purchased services index. Including professional and other labor expenses in labor categories would promote consistency between labor-related hospital and physician payment adjustments, and it would also take into account geographic variations in wages for the services reflected in Bureau of Labor Statistics (BLS) data.
AMA (American Medical Association). 2010. IOM staff correspondence with AMA about the AMA PPIS. November 23, 2010.
______. 2011. About CPT. Chicago, Il. http://www.ama-assn.org/ama/pub/physician-resources/solutions-managing-your-practice/coding-billing-insurance/cpt/about-cpt.shtml (accessed January 7, 2011). BLS (Bureau of Labor Statistics). 2008. Occupational employment statistics. Washington, DC.
______. 2011b. Occupational employment statistics query system. Washington, DC: Bureau of Labor Statistics. http://data.bls.gov/oes/search.jsp?data_tool=OES (accessed July 22, 2011).
Borjas, G. J. 2010. Labor economics. Fifth edition. Boston, MA: McGraw-Hill.
Carrier, E. R., J. D. Reschovsky, M. M. Mello, R. C. Mayrell, and D. Katz. Physicians’ fears of malpractice lawsuits are not assuaged by tort reforms. 2010. Health Affairs 29(9):1585–1591.
CMS (Centers for Medicare and Medicaid Services). 1996. Physician Fee Schedule (1997 CY): Payment policies: Revisions. Federal Register 61(128):34614-34622.
______. 2009. Medicare physician guide: A resource for residents, practicing physicians, and other health care professionals. Medicare Learning Network. Washington, DC: Centers for Medicare and Medicaid Services.
______. 2010a. Medicare program; payment policies under the Physician Fee Schedule and other revisions for Part B for CY 2010. Federal Register 75(228):73170–73860.
______. 2010b. Physician Fee Schedule—PFS federal regulation notices. Washington, DC: Centers for Medicare and Medicaid Services. http://www.cms.gov/PhysicianFeeSched/PFSFRN/list.asp#TopOfPage (accessed March 7, 2011).
______. 2011a. HCPCS: General information. Washington, DC: CMS. http://www.cms.gov/medhcpcsgeninfo/01_overview.asp (accessed February 14, 2011).
______. 2011b. Medicare program; payment policies under the physician fee schedule, five-year review of work relative value units, clinical laboratory fee schedule: Signature on requisition and other revisions to part B for CY 2012. Federal Register (76)228:73026–73474.
Department of Defense. 2009. A primer on basic allowance for housing (BAH).
Department of Health and Human Services. 2011. Medicare premiums and co-insurance rates for 2011.Washington, DC: HHS. http://questions.medicare.gov/app/answers/detail/a_id/2305/~/medicare-premiums-and-coinsurance-rates-for-2011 (accessed July 14, 2011).
Ehrenberg, R. G., and R. S. Smith. 2009. Modern labor economics: Theory and public policy. Tenth edition. Boston, MA: Pearson-Addison-Wesley.
Flores, G. 2005. The impact of medical interpreter services on the quality of health care: A systematic review. Medical Care Research Review 62(3):255–299.
GAO (Government Accountability Office). 2003. Medical malpractice insurance: Multiple factors have contributed to increased premium rates. Washington, DC: GAO.
______. 2005. Physician fees: Geographic adjustment indices are valid in design, but data and methods need refinement. Washington, DC: GAO.
______. 2007. Geographic areas used to adjust physician payments for variation in practice costs should be revised. Washington, DC: GAO.
Gillis, K. D., R. J. Willke, and R. A. Reynolds. 1993. Assessing the Validity of the Geographic Practice Cost Indexes. Inquiry 30:265-280. Blue Cross and Blue Shield Assocation and Blue Cross and Blue Shield of the Rochester Area.
Goertz, R. 2011. Testimony presented to the Institute of Medicine Committee on Geographic Adjustment Factors in Medicare Payment. Washington, DC. September 17, 2010.
Grassley, C. 2011. Statement of Senator Chuck Grassley to the Institute of Medicine Committee on Geographic Adjustment Factors in Medicare Payment. Washington, DC. January 5, 2011.
GSA (General Services Administration). 2010. IOM correspondence with Nicholas S. Hufford, Chief Appraiser and Director, Asset Management and Valuations Division Office of Portfolio Management. November 18, 2010.
Hsiao, W. C., P. Braun, D. Yntema, and E. R. Becker. 1988. Estimating physicians’ work for a resource-based relative-value scale. New England Journal of Medicine 319(13):835–841.
HUD (Housing and Urban Development). 2011a. Fair market rents: Overview. Washington, DC: HUD. http://www.huduser.org/portal/datasets/fmr.html (accessed December 10, 2010).
______. 2011b. 2010 Fair market rents. County level data file. http://www.huduser.org/portal/datasets/fmr.html (accessed January 15, 2011).
IMS (Iowa Medical Society). 2010. Geographic differences in Medicare payment to physicians—GPCIs. West Des Moines: Iowa Medical Society.
______. 2011. Supplemental Report on the Work GPCI and GPCI Weights. West Des Moines: Iowa Medical Society.
Jena, A. B., et al. 2011. Malpractice risk according to physician specialty. New England Journal of Medicine 365:629–636.
Kane, C. 2009. The practice arrangements of patient care physicians, 2007–2008: An analysis by age cohort and gender. Chicago, IL: American Medical Association.
Lowes, R. 2010. Regional variation in malpractice premiums defies tort reform. Medscape Today News. November 2, 2010.
MacKinney, A. C., T. D. McBride, M. D. Shambaugh-Miller, and K. J. Mueller. 2003. Medicare physician payment: Practice expense. RUPRI Rural Policy Brief 8(9).
MaCurdy, T., J. Shafrin, and M. Bounds. 2011. Proposed revisions to the sixth update of the Geographic Practice Cost Index. Burlingame, CA: Acumen, LLC.
MedPAC (Medicare Payment Advisory Commission). 2006. Report to Congress: Medicare payment policy. Washington, DC: MedPAC.
______. 2007. Report to Congress: Promoting greater efficiency in Medicare. Washington, DC: MedPAC.
______. 2008. Payment basics: Physician services payment system. Washington, DC: MedPAC.
______. 2010. Report to Congress: Aligning incentives in Medicare. Washington, DC: MedPAC.
Mello, M. M., D. M. Studdert, and T. A. Brennan. 2003. The new medical malpractice crisis. New England Journal of Medicine 348(23):2281–2284.
MGMA (Medical Group Management Association). 2010. Physician Compensation and Production Survey: 2010 Report (Based on 2009 Data).
O’Brien-Strain, M. A., W. Addison, E. Coombs, N. Hinnebusch, M. Johansson, and S. McClellan. 2008. Review of alternative GPCI payment locality structures. Burlingame, CA: Acumen, LLC.
O’Brien-Strain, M., S. McClellan, S. Frances, and N. Theobald. 2010. Final report on GPCI malpractice RVUs for the CY 2010 Medicare physician fee schedule rule. Burlingame, CA: Acumen, LLC.
Physician Payment Review Commission. 1991. Annual report to Congress. Washington, DC.
Pope, G. C., W. P. Welch, S. Zuckerman, and M. G. Henderson. 1989. Cost of practice and geographic variation in Medicare fees. Health Affairs (Millwood) 8(3):117–128.
Reding, D. J. 2010. Testimony presented to the Institute of Medicine Committee on Geographic Adjustment Factors in Medicare Payment. September 16, 2010.
Reschovsky, J. D., and A. B. Staiti. 2005. Physicians incomes in rural and urban America. Washington, DC: Center for Studying Health System Change.
Schwartz, A. 2010. Testimony presented to the Institute of Medicine Committee on Geographic Adjustment Factors in Medicare. Washington, DC. September 17, 2010.
Sloan, F. A., and L. M. Chepke. 2008. Medical malpractice. Cambridge, MA: MIT Press.
U.S. Census Bureau. 2008. A compass for using and understanding the American community survey data.
Washington, DC: U.S. Census Bureau.
USPS (U.S. Postal Service). 2011. USPS leased facilities report. Washington, DC: USPS. http://www.usps.com/foia/readroom/leasedfacilitiesreport.htm (accessed March 8, 2011).
Weeks, W.B., and A. E. Wallace. 2008. Rural-urban differences in primary care physicians’ practice patterns, characteristics, and incomes. Kansas City, MO: National Rural Health Association.
Zuckerman, S., and S. Maxwell. 2004. Reconsidering geographic adjustments to Medicare physician fees. Washington, DC: The Urban Institute.
Zuckerman, S., W. P. Welch, and G. C. Pope. 1987. The development of an interim geographic Medicare economic index. Washington, DC: The Urban Institute and Center for Health Economics Research.