2

Empirical Analysis of Geographic Variation

As described in Chapter 1, the Centers for Medicare & Medicaid Services (CMS) charged the Institute of Medicine’s (IOM’s) Committee on Geographic Variation in Health Care Spending and Promotion of High-Value Care with examining “geographic variation in intensity, cost, and growth of health care services and in per capita health care spending among the Medicare, Medicaid, privately insured, and uninsured U.S. populations.” To this end, the committee commissioned new analyses to complement its evaluation of the existing literature. The purpose of these new analyses was to quantify the magnitude of geographic variation in spending, utilization, and quality across various populations, payers, and geographic units; to evaluate known (and measurable) factors that account for variation in the Medicare and commercial markets; and to identify types of health care services with disproportionately high rates of variation that drive total variation.

RESEARCH FRAMEWORK AND STATISTICAL MODELING APPROACH

The literature on geographic variation has focused largely on traditional, fee-for-service Medicare. Much less is known about variation in expenditures and outcomes in the private market and in other public programs, such as Medicaid and Medicare Advantage (also known as Medicare Part C). This gap in knowledge is significant. A recent study notes that in 2010, Medicare spending accounted for 23 percent of the $2.19 trillion spent on personal health care in the United States, while spending in



The National Academies | 500 Fifth St. N.W. | Washington, D.C. 20001
Copyright © National Academy of Sciences. All rights reserved.
Terms of Use and Privacy Statement



Below are the first 10 and last 10 pages of uncorrected machine-read text (when available) of this chapter, followed by the top 30 algorithmically extracted key phrases from the chapter as a whole.
Intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text on the opening pages of each chapter. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

Do not use for reproduction, copying, pasting, or reading; exclusively for search engines.

OCR for page 39
2 Empirical Analysis of Geographic Variation A s described in Chapter 1, the Centers for Medicare & Medicaid Ser- vices (CMS) charged the Institute of Medicine’s (IOM’s) Committee on Geographic Variation in Health Care Spending and Promotion of High-Value Care with examining “geographic variation in intensity, cost, and growth of health care services and in per capita health care spend- ing among the Medicare, Medicaid, privately insured, and uninsured U.S. populations.” To this end, the committee commissioned new analyses to complement its evaluation of the existing literature. The purpose of these new analyses was to quantify the magnitude of geographic variation in spending, utilization, and quality across various populations, payers, and geographic units; to evaluate known (and measurable) factors that account for variation in the Medicare and commercial markets; and to identify types of health care services with disproportionately high rates of variation that drive total variation. RESEARCH FRAMEWORK AND STATISTICAL MODELING APPROACH The literature on geographic variation has focused largely on tradi- tional, fee-for-service Medicare. Much less is known about variation in expenditures and outcomes in the private market and in other public pro- grams, such as Medicaid and Medicare Advantage (also known as Medicare Part C). This gap in knowledge is significant. A recent study notes that in 2010, Medicare spending accounted for 23 percent of the $2.19 tril- lion spent on personal health care in the United States, while spending in 39

OCR for page 39
40 VARIATION IN HEALTH CARE SPENDING the private sector and Medicaid made up 34 and 17 percent, respectively (MedPAC, 2012). Although Medicare beneficiaries represent just 15 percent of the total U.S. population, more than 60 percent of Americans are covered by private insurance (ASPE, 2011). Moreover, 28 percent of all Medicare beneficiaries are enrolled in the Medicare Advantage program, which allows private insurers to contract with CMS to provide Medicare-covered Part A, B, and D services (Gold et al., 2013). Medicare is the largest single payer for health care in the nation, and has for many years been the only available source of reliable national claims data (Bernstein et al., 2011; Reschovsky et al., 2011). Nonetheless, spending and utilization patterns in traditional Medicare should not be assumed to be representative of other payer mar- kets or of total U.S. health care spending and utilization. To better understand the causes of variation in the health care system, the committee commissioned original empirical analyses of the complete database of Medicare beneficiaries (by Acumen, LLC; Dartmouth Insti- tute of Health Policy and Clinical Practice; and the University of Pitts- burgh), as well as two nationwide commercial databases, OptumInsight (by The Lewin Group) and Thomson Reuters (TR) MarketScan (by Harvard University).1,2 The results of these analyses were synthesized and used to conduct a separate analysis of geographic variation in total health care spending (by Precision Health Economics, LLC [PHE]). The subcontractors conducted a series of regression and correlation analyses to examine geographic variation in spending, utilization, and quality among the overall Medicare and commercial populations (aggregate analyses), as well as among 15 subpopulations with acute and chronic clini- cal conditions (cohort analyses). As noted in Chapter 1, not all geographic variation is unacceptable. The analyses conducted for this study generally excluded acceptable variation, which occurs as a result of factors beyond the control of the health care system in a region. Specifically, the baseline regression model was used to examine geographic variation in spending and utilization, adjusted for input prices of areas, as well as the age, sex, and health status of patients. As detailed in later sections, except where noted, regression models were not adjusted for other factors beyond the 1  Two different commercial databases were used to improve the external validity of the analyses of variation in the private sector. Each database had unique advantages: While the TR MarketScan database is large and representative, the OptumInsight database provides rich, individual-level data on a number of demographic factors. Results for both commercial populations are presented throughout Chapter 2 and 3, alongside the Medicare findings. For details on each database, refer to Appendix C and the subcontractor reports. (All subcontrac- tor final reports and spreadsheets of results are publicly available on the IOM webpage at http://www.iom.edu/geovariationmaterials.) 2  Refer to Chapter 1, Box 1-2, for a complete list of subcontractors performing these analyses and corresponding data sources.

OCR for page 39
EMPIRICAL ANALYSIS OF GEOGRAPHIC VARIATION 41 control of the health care system that often are associated with poor health status or higher spending, such as beneficiaries’ race and income, or factors that cannot be measured using claims data, such as patient or physician preferences. The specific research methodologies of these analyses are sum- marized in Appendix C and further detailed in the individual subcontractor reports. The results of the empirical analyses are presented in this chapter and Chapter 3. The committee also commissioned analyses of the Medicaid database (by Acumen, LLC). As noted in Chapter 1, however, those findings are not presented in this report because of concerns about their reliability and va- lidity due to incomplete data. The Acumen report notes that in 2007, more than 64 percent of all Medicaid beneficiaries were at least partially covered by a managed care program. Data on these beneficiaries had large gaps, as CMS did not begin collecting encounter claims data on managed care en- rollees until 2012 (Acumen, LLC, 2013a). In addition, studies of Medicaid generally are restricted to populations enrolled in fee-for-service programs, thus limiting the reliability and generalizability of results based on Medicaid data (Autor et al., 2011). Unlike Medicare, moreover, Medicaid can vary considerably in programming and policies because states can request waiv- ers from CMS to operate outside of federal guidelines. Medicaid program- ming and policies also have changed over time, so the data available for individual states vary widely. As described below, however, PHE imputed values for a hospital referral region’s (HRR’s) Medicaid population in its calculation of total U.S. health care spending. The committee’s analysis of geographic variation for the uninsured population also was restricted. According to the 2011 Current Popula- tion Survey, the uninsured population, one in seven of whom is an un- documented immigrant, made up approximately 16 percent of the total U.S. population in 2010 (ASPE, 2011; Zuckerman et al., 2011). Although uninsurance rates are known to vary greatly among states, a comprehen- sive analysis of geographic variation among the uninsured could not be conducted because of the lack of a coordinated database on the financing and delivery of care for this population. As discussed in later sections of this chapter, however, PHE adjusted its calculation of total U.S. health care spending, and associated analyses of variation for spending incurred by the uninsured were conducted using census data and Medical Expenditure Panel Survey (MEPS) data (PHE, 2013). This chapter presents findings from the committee’s empirical evalu- ation of geographic variation, with support from the existing literature. After briefly addressing the methodological issue of the unit of analysis, the chapter confirms the robust presence of regional variation in both Medi- care and commercial health care spending and utilization across multiple geographic units as well as over time. It then explores the sources of this

OCR for page 39
42 VARIATION IN HEALTH CARE SPENDING geographic variation, evaluating the role of price; other patient-, provider-, and market-level factors also are examined. Next, the influence of high variation in post-acute care services on total variation in Medicare spending and utilization is discussed. Finally, the chapter briefly assesses the limita- tions of efforts to analyze variation in quality and presents the committee’s recommendation for future research. GEOGRAPHIC VARIATION AND THE UNIT OF ANALYSIS The performance of the health care system varies across different units of analysis, including physician, practice, health care system, and geo- graphic unit. Geographic units can in turn be defined by economic markets (e.g., HRRs), political boundaries (e.g., county, state), administrative areas (e.g., zip codes, census tracts), or where people live (e.g., metropolitan statistical areas). In accordance with its statement of task, the committee examined variation within “areas of different sizes” to determine how variation is affected by different levels of aggregation. Box 2-1 defines tech- nical geographic units referenced throughout the literature on geographic variation and this report. To the extent possible, the committee considered variation across and within individual providers in an area, although in practice, concerns about patient privacy, proprietary information, and small sample sizes precluded public release of analyses at the individual physician level (and even results pertaining to small geographic areas). CONFIRMING REGIONAL VARIATION IN SPENDING AND UTILIZATION Health care spending is a measure of expenditures for care, and re- flects the effects of both the utilization of health services and their prices. “Utilization” captures the total number of units or intensity of health care services, as well as the mix of services provided. Recent reports by the Dart- mouth Institute for Health Policy and Clinical Practice and the Medicare Payment Advisory Commission (MedPAC) estimate that unadjusted Medi- care spending per beneficiary is 50-55 percent higher in HRRs in the highest quintile of spending relative to those in the lowest quintile. Medicare service use (adjusted for demographics and beneficiary health) is approximately 30 percent greater in the highest quintile compared with the lowest (MedPAC, 2011; Zuckerman et al., 2010). These findings are corroborated by a large body of literature that highlights the robust presence of variation in health care spending and utilization across regions in the United States (CBO, 2008; Fisher and Wennberg, 2003; Fisher et al., 2003a,b; GAO, 2009; MedPAC, 2003, 2009; Wennberg et al., 2002, 2008).

OCR for page 39
EMPIRICAL ANALYSIS OF GEOGRAPHIC VARIATION 43 BOX 2-1 Definitions of Geographic Units Frequently Used in Health Services Research ·  Hospital service areas (HSAs)—Created by Dartmouth and defined by assigning to an HSA the zip codes from which a hospital or several hospitals draw the greatest proportion of their Medicare patients. There are 3,426 HSAs (Dartmouth Institute for Health Policy and Clini- cal Practice, 2013). ·  Hospital referral regions (HRRs)—Created by Dartmouth to repre- sent regional health care markets for tertiary (complex) medical care. Dartmouth defined 306 HRRs by assigning HSAs to regions where the greatest proportion of major cardiovascular procedures were per- formed, “with minor modifications to achieve geographic contiguity, a minimum total population size of 120,000, and a high localization index” (Dartmouth Institute for Health Policy and Clinical Practice, 2013). · etropolitan statistical areas (MSAs, or metropolitan core-based sta- M tistical areas [CBSAs])—Created by the Office of Management and Budget using counties. Each of 388 MSAs (OMB, 2013) includes one or more counties with one core urban area of 50,000 individuals or more, as well as “adjacent counties exhibiting a high degree of social and economic integration” (as measured by such factors as com- muting patterns) with an urban core (OMB, 2010). Areas that do not qualify as MSAs are often classified as “outside” MSAs (OMB, 2010) or non-MSAs. The Centers for Medicare & Medicaid Services (CMS) ad- justs hospital payments according to a hospital wage index calculated for MSAs and non-MSAs* (CMS, 2012). *CBSAs are geographic entities that the Office of Management and Budget implemented in 2003 (OMB, 2010). The committee’s commissioned analyses used MSAs (a subcomponent of CBSAs also referred to as metropolitan CBSAs), as well as non-MSA “rest of state” regions. For simplicity, and in accordance with expert practice in this area (Acumen, LLC, 2009; MedPAC, 2012; OMB, 2010), the commit- tee uses the term “metropolitan CBSA” throughout this report. Medicare and Commercial Spending Varies Across All Levels of Geography For the present study, variation was examined at three geographic units of measurement: hospital service area (HSA), HRR, and metropolitan core-based statistical area (metropolitan CBSA). In the Acumen Medicare analysis, total spending, measured per capita, includes all costs incurred (by beneficiary and insurer) in traditional fee-for-service Medicare (Parts A, B,

OCR for page 39
44 VARIATION IN HEALTH CARE SPENDING and D). Medicare Advantage (Part C) was evaluated in a separate analysis discussed below. Similarly, in the OptumInsight (Lewin) and MarketScan (Harvard) analyses of commercial data, total spending includes all facil- ity, provider, and prescription drug costs incurred by the beneficiary, the insurer, and any additional (secondary) payers. To keep the presentation manageable, the analysis results are shown as a summary measure of varia- tion—the 90th percentile of spending compared with the 10th percentile (for aggregated years 2007-2009). This value is approximately the ratio of average spending in the highest-spending quintile of geographic units to average spending in the lowest-spending quintile. The analysis results, displayed in Table 2-1, show that without adjust- ments for any differences among regions, the HRR in the 90th percentile spent 42 percent more per Medicare beneficiary each month than the HRR in the 10th percentile. At the metropolitan CBSA level, the 90th percentile spent 38 percent more than the 10th percentile per beneficiary each month. Similar analyses of commercial insurance data confirm the presence of spending variation for all geographic units. Table 2-1 also shows that considerably greater variation exists at the smaller, HSA level. The policy implications of increasing levels of variation for smaller geographic units are discussed in Chapter 4. The committee, however, has chosen to present analysis results at the HRR level in the re- mainder of this report, as the corresponding area served by a major tertiary care hospital is the most widely established unit of analysis in the literature on geographic variation. TABLE 2-1 Ratio of the 90th to the 10th Percentiles of Unadjusteda Per-Member-Per Month (PMPM) Medicare and Commercial Spending Across Geographic Units HSA HRR Metropolitan CBSA Medicare 1.47 1.42 1.38 Commercial 1 (OptumInsight) 1.71b 1.42 1.50 Commercial 2 (MarketScan) 1.43 1.36 1.36 NOTE: Metropolitan CBSA = metropolitan core-based statistical area (also referred to as metropolitan statistical area [MSA]); HRR = hospital referral region; HSA = hospital service area. a“Unadjusted spending” refers to all-cause spending that has not been adjusted for any factors other than year of analysis and length of beneficiary enrollment. bThe OptumInsight results in this table are based on 2,896 HSAs with at least 500 observations. The analysis was conducted using only “large” HSAs to mitigate the effect of outliers. The Medicare and MarketScan databases were much larger; the data generally had normal distribution and were less affected by outliers. SOURCE: Committee analysis of subcontractor data.

OCR for page 39
EMPIRICAL ANALYSIS OF GEOGRAPHIC VARIATION 45 While variation occurs for all levels of geography, public- and private- sector spending per beneficiary (adjusted for age, sex, and health status) are only weakly correlated at the HRR level (PHE, 2013). In other words, areas that are high spenders in Medicare are not necessarily high spenders in the commercial market and vice versa. As shown in Table 2-2, the two commercial databases correlate well with each other, but are weakly cor- related with Medicare. As described in later sections, spending variation in Medicare is driven by variation in utilization of post-acute services, whereas in the commercial population, price has a greater influence than utilization on overall spending variation. As noted previously, most of the literature on geographic variation has of necessity relied on data from Medicare Parts A and B. Any measure of total spending by HRR is necessarily incomplete because of data limita- tions but is useful as a measure of the total resources potentially available to medical decision makers in an HRR. It is surprising that the correlation of Medicare spending with total spending is not higher (see Table 2-3), as Medicare accounts for a substantial fraction of total health care spend- ing. Moreover, it is unclear why a phenomenon responsible for variation TABLE 2-2 Correlation of Spending Measures Between Medicare and Commercial Payers Medicare & Medicare & MarketScan & MarketScan OptumInsight OptumInsight “Raw”a Baseline 0.112 0.081 0.663 Spending Input Price Adjusted -0.094 -0.032 0.632 Baseline Spending NOTE: All hospital referral region means are from the “Baseline” regression model, and thereby adjusted for partial year enrollment, age, sex, age*sex, and health status. a“Raw” spending refers to all-cause spending not adjusted for input-prices. SOURCE: PHE, 2013. TABLE 2-3 Correlation Between Total Spending and Payer-Specific Spending MarketScan Medicare Total Spending (Input Price Adjusted) 0.21 0.30 SOURCE: PHE, 2013.

OCR for page 39
46 VARIATION IN HEALTH CARE SPENDING in Medicare expenditures, such as practice patterns, would not matter for total spending as well. A map showing PHE’s estimate of quintiles of total spending across HRRs is included in Appendix G. The committee commissioned a separate analysis of variation in Medi- care Advantage (Part C) spending. That analysis was limited in scope as individual-level claims data were not available for the 2007-2009 study period; therefore, the analysis examined spending variation based on to- tal monthly Medicare reimbursement paid to Medicare Advantage plans (Acumen, LLC, 2013a).3 In part because of a policy decision to raise reimbursements in HRRs with lower traditional Medicare spending, the analysis found somewhat less variation in Medicare Advantage spending compared with traditional Medicare: HRRs in the 90th percentile spent 36 percent more per Medicare Advantage beneficiary than HRRs in the 10th percentile, while Table 2-1 shows a slightly higher differential ratio for fee- for-service beneficiaries. The distribution of the 90th to the 10th HRR cost percentile is narrower, as Medicare Advantage monthly spending is based on benchmarks set by the Congress. Although average per capita spending is higher for Medicare Advantage ($986) than for traditional Medicare ($958), HRR-level expenditures are correlated between the two programs (0.66). A complementary analysis by PHE examined geographic variation in total health care spending at the HRR level. This measure accounts for the total population in the United States by synthesizing estimates from Acu- men’s population-specific study of Medicare and Medicaid fee-for-service (Acumen, LLC, 2013a,b) and Harvard’s analysis of the MarketScan da- tabase as a proxy for commercial spending (Harvard University, 2012). Spending for the uninsured population was imputed by estimating a fac- tor price-adjusted national average based on census and MEPS data. The spending estimate for Medicaid Managed Care was generated using enroll- ment and total dollars paid for Medicaid health maintenance organization (HMO) beneficiaries by state, using data from the Medicaid Statistical Information System (MSIS). To create the total spending measure, payer- specific weights were applied.4 The analysis of total spending found that HRRs in the 90th percentile spend 50 percent more per beneficiary each month than HRRs in the 10th percentile, a larger variation than that shown in Table 2-1 for Medicare or commercial insurers (PHE, 2013). 3  CMS began collecting individual encounter claims data for Medicare Advantage benefi- ciaries in April 2012. 4  See PHE (2013) for detailed methodology. That report is publicly available on the IOM webpage at the following link: http://www.iom.edu/geovariationmaterials.

OCR for page 39
EMPIRICAL ANALYSIS OF GEOGRAPHIC VARIATION 47 Medicare and Commercial Utilization of Health Care Varies Across Service Categories The committee also commissioned analyses of geographic variation in utilization, which is measured in two ways. The first entails measuring utilization as “counts” of specific medical services, such as the number of emergency department and office visits per beneficiary per month. Because there are many different types of services, any measure of total utiliza- tion must be a weighted sum of those services (for example, a hospital day should count more than a single laboratory test). The subcontractors weighted each service by a standard national price for that service to re- move the effect of different prices across geographic locations and derive a measure of the aggregate quantity of services. This is the second measure of total utilization.5 The ratios of the 90th to the 10th percentile of risk-adjusted utili- zation (measured as counts) point to the presence of regional variation across different types of health care services within both the Medicare and commercial payer populations (see Table 2-4). The high use of emer- gency department services among the MarketScan commercial population is particularly striking, as utilization among HRRs in the 90th percentile (measured as counts of visits per beneficiary per month) is more than twice as high as that among HRRs in the 10th percentile. A recent MedPAC analysis also reveals substantial regional variation in service-specific utilization. Metropolitan CBSAs in the 90th percentile utilized approximately 2.01 times as much post-acute care per beneficiary as metropolitan CBSAs in the 10th percentile (MedPAC, 2011).6 After post- acute care services, the ambulatory care (outpatient visit) and inpatient visit categories varied the most, with 90th to 10th percentile ratios of 1.24 and 1.22, respectively (MedPAC, 2011). The impact of post-acute care services on variation in total Medicare spending and utilization is discussed in greater detail later in this chapter. The wide variation in inpatient hospitalization spending has been a key focus in the literature. Findings have shown that these regional differences may result from variation in the per capita rates of admission and readmis- sion (Fisher et al., 1994; Wennberg and Cooper, 1998), average lengths of hospital stay (Yuan et al., 2000), and mix of patient diagnosis-related groups (DRGs) (Frick et al., 1985; Steinwald, 2003), and indirectly from 5  Refer to Appendix E for detail on the methodology. 6  This MedPAC analysis adjusts for input price and health status. The MedPAC report notes that this regression model used a “service sector-specific health status adjustor. For example, metastatic cancer would have a much greater coefficient for total service utilization than it would for post acute care. This is different than using a beneficiary’s HCC [hierarchical condi- tion category] score to adjust for health status” (MedPAC, 2011, p. 7).

OCR for page 39
48 VARIATION IN HEALTH CARE SPENDING TABLE 2-4 Ratio of the 90th to the 10th Percentile of Per-Member-Per Month (PMPM) Risk-Adjusted Utilization (measured as counts) of Selecteda Service Categories Among Medicare and Commercial Populations at the Hospital Referral Region Level Emergency Inpatient Outpatient Prescription Department Imaging Admission Visits Drug Fills Visits Procedures Medicare 1.29 1.33 1.19 1.33 1.23 Commercial 1 1.48 1.46 1.36 1.54 1.76 (OptumInsight) Commercial 2 1.45 1.30 1.34 2.04 1.33 (MarketScan) NOTE: Utilization figures have been adjusted for age, sex, and health status. aThe committee was limited in the number of utilization measures it could investigate across Medicare and commercial databases due to time and budget constraints. Hence, post-acute care was not included in the committee’s main analysis. However, upon receiving preliminary results of the analysis, the committee asked Acumen to investigate post-acute care. Commercial payers did not conduct a similar analysis due to their younger populations, who receive very little post-acute care as a population. This table only presents utiliza- tion measures that are common across payer populations. SOURCE: Committee analysis of subcontractor data. a host of inpatient-care and efficiency-related factors (nurse staffing, tests, procedures, drugs) that may differ across and within geographic regions (Franzini et al., 2010). The empirical analyses of Medicare and commercial data demonstrate considerable variation in emergency department and outpatient visits. This variation may be due to underlying differences in the regional distribution of socioeconomic factors shown to influence emergency department use (Cunningham, 2006). Significant regional variation also may exist in the organizational structure of emergency response staff, as well as the techni- cal capacity of emergency department facilities (Cummins, 1993) and/or the supply and availability of primary care services (Cunningham, 2006). The University of Pittsburgh analysis of Medicare Part D found HRR- level variation in prescription drug spending and use,7 with 90th to 10th percentile ratios of 1.24 and 1.17, respectively (University of Pittsburgh, 2013). Although a multitude of studies have assessed variation in total uti- lization of health care, the literature on variation in the use of prescription drugs is limited. 7  The Pittsburgh analysis adjusted drug spending for patient demographics (age, sex, race, income), insurance status, and clinical characteristics (CMS-HCC risk scores, prescription drug HCC [RxHCC] risk scores, and institutionalization status) (University of Pittsburgh, 2013, p. 9).

OCR for page 39
EMPIRICAL ANALYSIS OF GEOGRAPHIC VARIATION 49 Geographic Variation Persists Over Time Although total health care spending has grown steadily over time (Fisher et al., 2009), Figure 2-1 demonstrates that in the past four decades, per capita health care spending has been growing more rapidly in the com- mercial sector than in Medicare. The committee commissioned “growth” analyses to assess whether geographic variation persists over time, and thus can be considered a true signal rather than a result of random noise. At any fixed point in time, area-level expenditures reflect the underlying spending habits of individual beneficiaries while also reflecting some degree of random noise arising from the uncertainty of individual health episodes. The former factor persists over time, while the latter does not. Trends in prices paid by commercial insurers and the demographics of those whom they insure also may change over time in ways that do not mimic Medicare. Acumen’s analysis of Medicare data found that spending and utilization growth rates have not differed much over time between high- and low- cost regions of the country; regions that were high- (or low-) cost in 1992 remained so in 2010 (Acumen, LLC, 2013b). This finding is illustrated in Figure 2-2, which classifies HRRs into cost quintiles based on their expen- diture levels in 1992; quintiles 1 and 5 represent the lowest- and highest- cost regions, respectively. After regional differences due to input price, age, sex, and health status are removed, expenditure growth patterns are highly similar in each quintile (with the exception of some regression toward the mean for quintile 5). In short, spending differences between low- and high- cost geographic regions persist over time and thus do not simply reflect random variation at a point in time. Utilization growth rates mirror the spending patterns presented in Figure 2-2. These results are consistent with the existing literature, which reports that variation in Medicare spending persists across areas over time (Cutler and Sheiner, 1999). Another demonstration of the stability of the HRR cost quintiles over time is shown in Tables 2-5 and 2-6, which display the change in quintile rank between 1992 and 2010 for spending and utilization, respectively (Acumen, LLC, 2013b). As shown in Table 2-5, among all HRRs in the lowest-cost quintile of spending in 1992, 61 percent remained in the lowest- cost quintile in 2010, 28 percent moved to the second-lowest-cost quintile, and so on. Stability in utilization, displayed in Table 2-6, is weaker, with only 46 percent of HRRs in the lowest-cost quintile remaining in that quintile in 2010. As a final way of distinguishing the portion of geographic variation that arises from systematic differences and the portion that is random, the subcontractors examined correlations of year-to-year spending and year-to- year utilization during 1992 and 2010 (Acumen, LLC, 2013b). Area-level Medicare spending is highly correlated from one year to the next, with an

OCR for page 39
68 $500 $500 $400 $400 $300 $300 $200 $200 $100 $100 $0 $0 -$100 -$100 -$200 -$200 (a) Total Monthly Adjusted Differences from (b) Post-Acute Care Monthly Adjusted Differences from the National Mean of Spending Across HRR the National Mean of Spending Across HRR $500 $500 $400 $400 $300 $300 $200 $200 $100 $100 $0 $0 -$100 -$100 -$200 -$200 (c) Acute Care Monthly Adjusted Differences from (d) Diagnostic Monthly Adjusted Differences from the National Mean of Spending Across HRR the National Mean of Spending Across HRR

OCR for page 39
$500 $500 $400 $400 $300 $300 $200 $200 $100 $100 $0 $0 -$100 -$100 -$200 -$200 (e) Prescription Drug Monthly Adjusted Differences from (f) Other Monthly Adjusted Differences from the National Mean of Spending Across HRR the National Mean of Spending Across HRR $500 $500 $400 $400 $300 $300 $200 $200 $100 $100 $0 $0 -$100 -$100 -$200 -$200 (g) ER/Ambulance Monthly Adjusted Differences from (h) Procedures Monthly Adjusted Differences from the National Mean of Spending Across HRR the National Mean of Spending Across HRR FIGURE 2-5a–h Medicare service category utilization (monthly cost residual) by hospital referral region (HRR). NOTE: In this analysis, utilization is measured as the the total per-member-per-month input-price-adjusted cost (the dollar amounts shown at the left of each figure). The predictor variables include beneficiary age; sex; age-sex interaction; race; health status coded by hierarchical condition category; partial- year enrollment; and indicators for supplemental Medicare insurance, institutionalization status, new enrollee status (prior-year diagnoses are not available for them), dual-enrollment status, and year of analysis (2007, 2008, 2009). Selected results displaying the residual total post-acute and acute care costs for all 306 HRRs are available in Appendix G. 69 SOURCE: Acumen, LLC, 2013a.

OCR for page 39
70 VARIATION IN HEALTH CARE SPENDING TABLE 2-10 Proportion of Variance Attributable to Each Medicare Service Category Adjusted Total Medicare Spending Remaining Reduction in Variance Variance (%)* Variation in Total Medicare Spending 6,974 — If No Variation in Post-Acute Care Only 1,864 73 If No Variation in Acute Care Only 5,085 27 If No Variation in Either Post-Acute or Acute 780 89 If No Variation in Prescription Drugs 6,374 9 If No Variation in Diagnostic Tests 5,986 14 If No Variation in Procedures 6,020 14 If No Variation in Emergency Department 6,972 0 Visits/Ambulance Use If No Variation in Other 6,882 1 NOTE: Total Medicare spending and each component are input-price- and risk-adjusted. Each row shows the reduction in variance from eliminating only the variation in that service, with the exception of the acute and post-acute care rows. * he individual reductions sum to more than 100 percent because of covariance terms. T SOURCE: Committee analysis of Medicare data. counties were penalized for being low-value, all legitimate providers in those counties would bear the consequences. Conclusion 2.4. Variation in total Medicare spending across geographic areas is driven largely by variation in the utilization of post-acute care services, and to a lesser extent by variation in the utilization of acute care services. LIMITATIONS OF EFFORTS TO MEASURE VARIATION IN QUALITY Health care quality “composite measures” allow measurement of mul- tiple aspects of quality by collapsing individual measures to create a single score (NQF, 2009). Composite quality indicators, developed and main- tained by the Agency for Healthcare Research and Quality (AHRQ), are based largely on administrative data (billing- or claims-related informa- tion), with inclusion and exclusion criteria being based on diagnosis or procedure codes. The committee’s commissioned analyses evaluated quality

OCR for page 39
EMPIRICAL ANALYSIS OF GEOGRAPHIC VARIATION 71 TABLE 2-11 Wide Variation in Spending for Durable Medical Equipment and Home Health Care in Contiguous Florida Counties DME Spending Home Health Care per Capita ($) Spending per Capita ($) Area 2006 2008 2006 2008 South Florida County Broward 394 321 1,002 1,390 Collier 207 202 305 395 Miami-Dade 2,043 828 2,591 5,318 Monroe 237 210 237 334 National 263 282 382 488 NOTES: DME = durable medical equipment. Spending data are annualized for beneficiaries with either Part A or Part B coverage for at least 1 month during 2006. The results are not adjusted for differences in beneficia- ries’ health status or prices. In March 2007, the U.S. Department of Justice, the U.S. Attorney’s Office for the Southern District of Florida, the Department of Health and Human Services (HHS), OIG, and state and local law enforcement launched the Medicare Fraud Strike Force in South Florida. In its early stages, the task force targeted fraud in HIV infusion therapy and DME (DOJ, 2013; Katz, 2012), which may explain the significant drop in DME spending observed between 2006 and 2008 in Miami-Dade County. SOURCE: MedPAC, 2011, p. 11. of care using both individual measures and the following nationally estab- lished composite quality indicators: · Prevention Quality Indicator (PQI) #90: Reflects the quality of am- bulatory care in preventing medical complications for both acute and chronic illness. · Inpatient Quality Composite Indicator (IQI) #91: Reflects the quality of care delivered in an inpatient hospital setting, and in- cludes mortality indicators, as well as procedures for which there is a question of inefficient use. · Patient Safety Indicator (PSI) #90: Reflects the quality of care within a hospital, particularly as it relates to potentially prevent- able surgical complications or iatrogenic events. · Pediatric Quality Indicator (PDI) #19: Reflects the quality of care among the pediatric population. Although these national quality indicators represent the “current state of the art in assessing the health care system as a whole,” performance mea- sures based on administrative data have a number of limitations (Farquhar, 2008). For instance, the complex association between preventive care in an outpatient setting (PQI) and beneficiary socioeconomic status often

OCR for page 39
72 VARIATION IN HEALTH CARE SPENDING complicates an assessment of regional variation because such factors as patient access to care, patient preferences, and other unmeasured barriers in traditionally underserved populations cannot be accounted for (AHRQ, 2007a). Measurement of preventable complications (PSI) may be limited by inaccuracies within the underlying data, as providers who fear negative consequences are unlikely to maintain a thorough record of preventable complications in their patients (AHRQ, 2007b). As discussed previously, studies have established differences in coding practices among physicians, as well as among hospitals, making a fair comparison of hospitals (based on IQI) difficult. Moreover, administrative data are “clinically vague,” as the same diagnosis code may be applied to a heterogeneous pool of clinical states. As a result, risk adjustment of administrative claims is likely to be imperfect, and this may affect the measurement of quality. As discussed earlier, although the subcontractors’ regression analyses risk adjust for certain known predictors (including age, sex, and health status), a number of unmeasured factors may account for variation in qual- ity across areas.16 This report does not quantify the amount of geographic variation in health care quality as it does for spending and utilization because of limitations in the measurement of quality composites and the underlying data. All of the commissioned analyses report on two measures of health care quality—PSI and PQI. The Harvard and Lewin analyses each include quality measures that are “rare” among commercially insured populations. Although previous research has noted some differences in quality patterns across the United States (MedPAC, 2003), greater emphasis has been placed on studying relationships among quality, overall spend- ing, and “high value.” In Chapter 4, the committee evaluates the use of a geographically based value index and further explores the empirical inter- relationships among quality of care and health care spending and utilization across Medicare and private payers. RESEARCH AGENDA This study represents the largest-scale analysis of geographic variation in health care spending in the United States, covering Medicare and repre- sentative private payer populations. The availability of and access to CMS’s complete Medicare (Parts A, B, C, and D) claims database were instrumen- tal to the successful completion of this research. As discussed previously, however, the lack of access to encounter claims information for Medicare 16  The subcontractors all followed AHRQ guidelines but used varying methodologies to calculate quality composites. Refer to Appendix C for a brief summary and to the Acumen, Lewin, and Harvard reports for complete details. (Subcontractor reports can be accessed through the Institute of Medicine website at http://www.iom.edu/geovariationmaterials.)

OCR for page 39
EMPIRICAL ANALYSIS OF GEOGRAPHIC VARIATION 73 and Medicaid managed care enrollees during the 2007-2009 period limits the generalizability of the study findings. Although CMS has in recent years made an effort to improve and simplify the process of obtaining historical data, significant operational, procedural, and financial barriers continue to exist. Congress could help remove these barriers by supporting CMS’s ef- forts to expand the availability of Medicare and Medicaid data for research purposes, with particular emphasis on releasing previously unavailable or limited Medicare Part C and D data. For its part, CMS could use its existing authority to broaden data access for the purposes of primary research and evaluation while safeguarding patient privacy and maintaining standards established by the Social Security Act, the Health Insurance Portability and Accountability Act (HIPAA) Privacy Rule, the Privacy Act of 1974, and the Federal Information Security and Management Act of 2002 (FISMA). As noted in earlier sections of this chapter, current understanding of health care utilization and quality is limited as it depends on information available in administrative (claims and billing) data, which do not capture the extent or severity of a patient’s illness. CMS could enrich claims data- bases by better adjusting for population health and could assist in the cre- ation of more accurate quality measures by creating a platform for clinical and behavioral information (e.g., electronic medical records). More research on health care outcomes and quality is needed, par- ticularly for commercially insured populations. To date, many nationally established quality composite measures have been designed to measure process and outcomes in the Medicare population and are not necessarily applicable to privately insured beneficiaries. Although in its estimate of total health care spending in the United States, PHE attempted to include estimates from Medicare, Medicaid, and commercial payers, as well as the uninsured, the generalizability of this analysis is limited. Further research on this topic is needed and would benefit from the availabilty of a national- level all-payer database. Moreover, combined use of Medicare and private administrative or claims data would allow for more accurate measurement of provider performance and quality of care. Collaboration between CMS and private payers would be an important first step toward creating an enriched national data source. RECOMMENDATION 1: Congress should encourage the Centers for Medicare & Medicaid Services (CMS), and provide the necessary resources, to make accessing Medicare and Medicaid data easier for research purposes. CMS should collaborate with private insurers to collect, integrate, and analyze standardized data on spending, as well as clinical and behavioral health outcomes, to enable more extensive comparisons of payments and quality and evaluation of value-based payment models across payers.

OCR for page 39
74 VARIATION IN HEALTH CARE SPENDING REFERENCES Acumen, LLC. 2009. Revision of Medicare Wage Index—Final Report: Part I. Burlingame, CA. ———. 2013a. Geographic variation in spending, utilization and quality: Medicare and Med- icaid beneficiaries. Washington, DC: Institute of Medicine. ———. 2013b. IOM study of geographic variation: Growth analysis. Washington, DC: In- stitute of Medicine. Adler, N. E., and K. Newman. 2002. Socioeconomic disparities in health: Pathways and poli- cies. Health Affairs 21(2):60-76. AHRQ (Agency for Healthcare Research and Quality). 2007a. AHRQ guide to prevention quality indicators: Hospital admission for ambulatory sensitive conditions. Edited by U.S. Department of Health and Human Services. Rockville, MD. ———. 2007b. AHRQ guide to patient safety indicators. Edited by U.S. Department of Health and Human Services. Rockville, MD, March 12. American Hospital Association. 2009. Geographic variation in health care spending: A closer look. In Trend Watch. Washington, DC: American Hospital Association. ———. 2011. Report of the Task Force on Variation in Health Care Spending. Washington, DC: American Hospital Association. Anthony, D. L., M. B. Herndon, P. M. Gallagher, A. E. Barnato, J. P. Bynum, D. J. Gottlieb, E. S. Fisher, and J. S. Skinner. 2009. How much do patients’ preferences contribute to resource use? Health Affairs 28(3):864-873. ASPE (Assistant Secretary for Planning and Evaluation). 2011. Overview of the uninsured in the United States: A summary of the 2011 current population survey. ASPE Issue Brief. http://aspe.hhs.gov/health/reports/2011/cpshealthins2011/ib.shtml (accessed July 18, 2013). Autor, D., A. Chandra, and M. Duggan. 2011. Public health expenditures on the working age disabled: Assessing Medicare: Public health expenditures on the working age disabled: Assessing Medicare. Cambridge, MA: National Bureau of Economic Research. Baicker, K., and A. Chandra. 2004. Medicare spending, the physician workforce, and benefi- ciaries’ quality of care. Health Affairs Supplemental Web Exclusives W4-184-97. Baicker, K., A. Chandra, J. S. Skinner, and J. E. Wennberg. 2004. Who you are and where you live: How race and geography affect the treatment of Medicare beneficiaries. Health Affairs Supplemental Variation 33-44. Baicker, K., A. Chandra, and J. S. Skinner. 2005. Geographic variation in health care and the problem of measuring racial disparities. Perspectives in Biology and Medicine 48(Suppl. 1):S42-S53. Baker, L. C., M. K. Bundorf, and D. P. Kessler. 2010. HMO coverage reduces variations in the use of health care among patients under age sixty-five. Health Affairs 29(11):2068-2074. Barnato, A. E., M. B. Herndon, D. L. Anthony, P. M. Gallagher, J. S. Skinner, J. P. Bynum, and E. S. Fisher. 2007. Are regional variations in end-of-life care intensity explained by pa- tient preferences? A study of the US Medicare population. Medical Care 45(5):386-393. Bernstein, J., J. D. Reschovsky, and C. White. 2011. Geographic variation in health care: Changing policy directions. Washington, DC: National Institute for Health Care Reform. Berwick, D. M., and A. D. Hackbarth. 2012. Eliminating waste in U.S. health care. Journal of the American Medical Association 307(14):1513-1516. Buntin, M. B., A. D. Garten, S. Paddock, D. Saliba, M. Totten, and J. J. Escarce. 2005. How much is postacute care use affected by its availability? Health Services Research 40(2):413-434.

OCR for page 39
EMPIRICAL ANALYSIS OF GEOGRAPHIC VARIATION 75 Case, A., and A. S. Deaton. 2005. Broken down by work and sex: How our health declines. In Analyses in the economics of aging. Chicago, IL: University of Chicago Press. Pp. 185-212. CBO (Congressional Budget Office). 2008. Geographic variation in health care spending. Washington, DC: Congress of the United States, CBO. Chernew, M. E., L. Sabik, A. Chandra, and J. P. Newhouse. 2009. Would having more primary care doctors cut health spending growth? Health Affairs (Millwood) 28(5):1327-1335. Chernew, M. E., L. M. Sabik, A. Chandra, T. B. Gibson, and J. P. Newhouse. 2010. Geo- graphic correlation between large-firm commercial spending and Medicare spending. American Journal of Managed Care 16(2):131-138. CMS (Centers for Medicare & Medicaid Services). 2012. Wage index. http://www.cms.gov/ Medicare/Medicare-Fee-for-Service-Payment/AcuteInpatientPPS/wageindex.html (ac- cessed March 29, 2013). Cummins, R. O. 1993. Emergency medical services and sudden cardiac arrest: The “chain of survival” concept. Annual Review of Public Health 14(1):313-333. Cunningham, P. J. 2006. What accounts for differences in the use of hospital emergency departments across U.S. communities? Health Affairs (Millwood) 25(5):w324-w336. Cutler, D. M., and L. Sheiner. 1999. The geography of Medicare. American Economic Review 89(2):228-233. Cutler, D., J. Skinner, D. Stern, and D. Wennberg. 2013. Physician beliefs and patient prefer- ences: A new look at regional variation in spending. https://www.lafollette.wisc.edu/ research/health_economics/Stern.pdf (accessed July 18, 2013). Dartmouth Institute for Health Policy and Clinical Practice. 2013. The Dartmouth Atlas of Health Care. http://www.dartmouthatlas.org (accessed July 18, 2013). DeNavas-Walt, C., B. D. Proctor, and J. C. Smith. 2009. Income, poverty, and health insurance coverage in the United States: 2009. Washington, DC: U.S. Census Bureau. DOJ (Department of Justice). 2013. Health care fraud and abuse control program: Annual report for fiscal year 2012. Washington, DC. Donohue, J. M., N. E. Morden, W. F. Gellad, J. P. Bynum, W. Zhou, J. T. Hanlon, and J. Skin- ner. 2012. Sources of regional variation in Medicare Part D drug spending. New England Journal of Medicine 366(6):530-538. Dunn, A., A. H. Shapiro, and E. Liebman. 2012. Geographic variation in commercial medi- cal (care expenditures: A framework for decomposing price and utilization. http://www. bea.gov/papers/pdf/RegionalPriceVariation_7_17_11Final.pdf (accessed July 18, 2013). Farley, P. J. 1985. Who are the underinsured? Milbank Memorial Fund Quarterly. Health and Society 476-503. Farquhar, M. 2008. AHRQ quality indicators. Edited by R. Hughes, Patient safety and qual- ity: An evidence-based handbook for nurses. Rockville, MD: Agency for Healthcare Research and Quality. Finkelstein, A., S. Taubman, B. Wright, M. Bernstein, J. Gruber, J. P. Newhouse, H. Allen, K. Baicker, and the Oregon Health Study Group. 2012. The Oregon health insurance experi- ment: Evidence from the first year. The Quarterly Journal of Economics 127:1057-1106. Fiscella, K., P. Franks, M. R. Gold, and C. M. Clancy. 2000. Inequality in quality. Journal of the American Medical Association 283(19):2579-2584. Fisher, E. S., and J. E. Wennberg. 2003. Health care quality, geographic variations, and the challenge of supply-sensitive care. Perspectives in Biology and Medicine 46(1):69-79. Fisher, E. S., J. E. Wennberg, T. A. Stukel, and S. M. Sharp. 1994. Hospital readmission rates for cohorts of Medicare beneficiaries in Boston and New Haven. New England Journal of Medicine 331(15):989-995.

OCR for page 39
76 VARIATION IN HEALTH CARE SPENDING Fisher, E. S., D. E. Wennberg, T. A. Stukel, D. J. Gottlieb, F. L. Lucas, and E. L. Pinder. 2003a. The implications of regional variations in Medicare spending. Part 1: The content, qual- ity, and accessibility of care. Annals of Internal Medicine 138(4):273-287. ———. 2003b. The implications of regional variations in Medicare spending. Part 2: Health outcomes and satisfaction with care. Annals of Internal Medicine 138(4):288-298. Fisher, E. S., J. P. Bynum, and J. S. Skinner. 2009. Slowing the growth of health care costs— Lessons from regional variation. New England Journal of Medicine 360(9):849-852. Fowler, F. J., P. M. Gallagher, D. L. Anthony, K. Larsen, and J. S. Skinner. 2008. Relationship between regional per capita Medicare expenditures and patient perceptions of quality of care. Journal of the American Medical Association 299(20):2406-2412. Franzini, L., O. I. Mikhail, and J. S. Skinner. 2010. McAllen and El Paso revisited: Medi- care variations not always reflected in the under-sixty-five population. Health Affairs 29(12):2302-2309. Frick, A. P., S. G. Martin, and M. Shwartz. 1985. Case-mix and cost differences between teaching and nonteaching hospitals. Medical Care 283-295. Fuchs, V. R., M. B. McClellan, and J. Skinner. 2001. Area differences in utilization of medical care and mortality among U.S. elderly. NBER Working Paper 8628. Cambridge, MA: National Bureau of Economic Research. GAO (Government Accountability Office). 2009. Medicare physician services: Utilization trends indicate sustained beneficiary acess with high and growing levels of service in some areas of the nation. Washington, DC: Government Printing Office. Gold, M. 2004. Geographic variation in Medicare per capita spending: Should policy-makers be concerned? In The synthesis project: New insights from research results. Princeton, NJ: Robert Wood Johnson Foundation. Gold, M., G. Jacobson, A. Damico, and T. Neuman. 2013. Medicare Advantage 2013 spot- light: Enrollment market update. Kaiser Family Foundation 8448:2. Goldman, T. R. 2012. Health policy brief: Eliminating fraud and abuse. Health Affairs, July 31. Gottlieb, D. J., W. Zhou, Y. Song, K. G. Andrews, J. S. Skinner, and J. M. Sutherland. 2010. Prices don’t drive regional Medicare spending variations. Health Affairs 29(3):537-543. Harvard University. 2012. Geographic variation in health care spending, utilization, and qual- ity among the privately insured. Washington, DC: Institute of Medicine. Hsu, J., J. Huang, V. Fung, M. Price, R. Brand, R. Hui, B. Fireman, W. Dow, J. Bertko, and J. P. Newhouse. 2009. Distributing $800 billion: An early assessment of Medicare Part D risk adjustment. Health Affairs 28(1):215-225. Kane, R. L., W. C. Lin, and L. A. Blewett. 2002. Geographic variation in the use of post-acute care. Health Services Research 37(3):667-682. Katz, P. 2012. Medicare fraud strike force: Past, present, and future. Fraud and Abuse 1(1):1-4. Manning, W. G., E. C. Norton, and A. S. Wilk. 2012. Explaining geographic variation in health care spending, use and quality, and associated methodological challenges. Paper commissioned by the Committee on Variation in Health Care Spending and Promotion of High-Value Care, Institute of Medicine, Washington, DC. http://www. iom.edu/Reports/2013/Variation-in-Health-Care-Spending-Target-Decision-Making-Not- Geography/~/media/Files/Report%20Files/2013/Geographic-Variation2/Commissioned- Papers/Manning_Norton_Wilk.pdf (accessed August 29, 2013). MedPAC (Medicare Payment Advisory Commission). 2003. Variation and innovation in Medi- care. In Report to the Congress: June 2003. Washington, DC: MedPAC. ———. 2008. Reforming the delivery system. In Report to the Congress: June 2008. Wash- ington, DC: MedPAC.

OCR for page 39
EMPIRICAL ANALYSIS OF GEOGRAPHIC VARIATION 77 ———. 2009. Measuring regional variation in service use. In Report to the Congress: Decem- ber 2009. Washington, DC: MedPAC. ———. 2011. Regional variation in Medicare service use. In Report to the Congress: January 2011. Washington, DC: MedPAC. ———. 2012. Health care spending and the Medicare program. Washington, DC: MedPAC. Newhouse, J., and Insurance Experiment Group. 1993. Free for all? Lessons from the RAND health insurance experiment, commercial books. Cambridge, MA: Harvard University Press. NQF (National Quality Forum). 2009. National Quality Forum approves criteria for evaluat- ing composite measures. http://www.qualityforum.org/Publications/2013/04/Composite_ Performance_Measure_Evaluation_Guidance.aspx (accessed December 30, 2011). OIG (Office of Inspector General). 2012. CMS and contractor oversight of home health agen- cies. Washington, DC: OIG. OMB (Office of Management and Budget). 2010. 2010 standards for delineating metropolitan and micropolitan statistical areas; notice. Federal Register 75(123):37246-37252. ———. 2013. Revised delineations of metropolitan statistical areas, micropolitan statistical areas, and combined statistical areas, and guidance on uses of the delineations of these areas. http://www.whitehouse.gov/sites/default/files/omb/bulletins/2013/b13-01.pdf (ac- cessed March 31, 2013). PHE (Precision Health Economics, LLC). 2013. Geographic variation in health care spending and the promotion of high-value care. Washington, DC: Institute of Medicine. Pine, M., H. S. Jordan, A. Elixhauser, D. E. Fry, D. C. Hoaglin, B. Jones, R. Meimban, D. Warner, and J. Gonzales. 2007. Enhancement of claims data to improve risk adjustment of hospital mortality. Journal of the American Medical Association 297(1):71-76. Radley, D. 2012. Geographic variation in access to care—The relationship with quality. New England Journal of Medicine 367(1):3-6. Reschovsky, J. D., J. Hadley, C. B. Saiontz-Martinez, and E. R. Boukus. 2011. Following the money: Factors associated with the cost of treating high-cost Medicare beneficiaries. Health Services Research 46(4):997-1021. Sirovich, B. E., D. J. Gottlieb, H. G. Welch, and E. S. Fisher. 2005. Variation in the tendency of primary care physicians to intervene. Archives of Internal Medicine 165(19):2252-2256. Sirovich, B., P. M. Gallagher, D. E. Wennberg, and E. S. Fisher. 2008. Discretionary deci- sion making by primary care physicians and the cost of U.S. health care. Health Affairs 27(3):813-823. Song, Y., J. Skinner, J. Bynum, J. Sutherland, J. E. Wennberg, and E. S. Fisher. 2010. Regional variations in diagnostic practices. New England Journal of Medicine 363(1):45-53. Steinwald, A. B. 2003. Specialty hospitals: Information on national market share, physician ownership, and patients served. Letter to the Honorable Bill Thomas, Chairman, Com- mittee on Ways and Means, House of Representatives and the Honorable Jerry Kleczka, House of Representatives. April 18, 2003. Sutherland, J. M., E. S. Fisher, and J. S. Skinner. 2009. Getting past denial—The high cost of health care in the United States. New England Journal of Medicine 361(13):1227-1230. The Lewin Group. 2013. Geographic variation in health care spending and promotion of high-value care. Washington, DC: Institute of Medicine. University of Pittsburgh. 2013. Geographic variation in health care spending and promotion of high-value health care Medicare Part D. Washington, DC: Institute of Medicine. Welch, W. P., M. E. Miller, H. G. Welch, E. S. Fisher, and J. E. Wennberg. 1993. Geographic variation in expenditures for physicians’ services in the United States. New England Journal of Medicine 328(9):621-627.

OCR for page 39
78 VARIATION IN HEALTH CARE SPENDING Wennberg, J. E., and M. M. Cooper. 1998. The surgical treatment of common diseases. In Dartmouth Atlas of Health Care 1998. Chicago, IL: American Hospital Publishing. Pp. 108-111. ———. 1999. The quality of medical care in the United States: A report on the Medicare program. The Dartmouth Atlas of Health Care 1999. Hanover, NH and Chicago, IL: Dartmouth Medical School and American Hospital Association. Wennberg, J. E., E. S. Fisher, and J. S. Skinner. 2002. Geography and the debate over Medicare reform. Health Affairs Supplemental Web Exclusives W96-W114. Wennberg, J. E., E. S. Fisher, T. A. Stukel, J. S. Skinner, S. M. Sharp, and K. K. Bronner. 2004. Use of hospitals, physician visits, and hospice care during last six months of life among cohorts loyal to highly respected hospitals in the United States. British Medical Journal 328(7440):607. Wennberg, J. E., S. Brownlee, E. S. Fisher, J. S. Skinner, and J. N. Weinstein. 2008. An agenda for change: Improving quality and curbing health care spending: Opportunities for the Congress and the Obama administration. Hanover, NH: Dartmouth Institute for Health Policy and Clinical Practice. Yuan, Z., G. S. Cooper, D. Einstadter, R. D. Cebul, and A. A. Rimm. 2000. The association between hospital type and mortality and length of stay: A study of 16.9 million hospital- ized Medicare beneficiaries. Medical Care 38(2):231. Zuckerman, S., T. Waidmann, R. Berenson, and J. Hadley. 2010. Clarifying sources of geo- graphic differences in Medicare spending. New England Journal of Medicine 363(1): 54-62. Zuckerman, S., T. A. Waidmann, and E. Lawton. 2011. Undocumented immigrants, left out of health reform, likely to continue to grow as share of the uninsured. Health Affairs 30(10):1997-2004.