This is the accessible text file for GAO report number GAO-09-802 
entitled 'Medicare: Per Capita Method Can Be Used to Profile Physicians 
and Provide Feedback on Resource Use' which was released on 
October 26, 2009. 

This text file was formatted by the U.S. Government Accountability 
Office (GAO) to be accessible to users with visual impairments, as part 
of a longer term project to improve GAO products' accessibility. Every 
attempt has been made to maintain the structural and data integrity of 
the original printed product. Accessibility features, such as text 
descriptions of tables, consecutively numbered footnotes placed at the 
end of the file, and the text of agency comment letters, are provided 
but may not exactly duplicate the presentation or format of the printed 
version. The portable document format (PDF) file is an exact electronic 
replica of the printed version. We welcome your feedback. Please E-mail 
your comments regarding the contents or accessibility features of this 
document to Webmaster@gao.gov. 

This is a work of the U.S. government and is not subject to copyright 
protection in the United States. It may be reproduced and distributed 
in its entirety without further permission from GAO. Because this work 
may contain copyrighted images or other material, permission from the 
copyright holder may be necessary if you wish to reproduce this 
material separately. 

Report to the Chairman, Subcommittee on Health, Committee on Ways and 
Means, House of Representatives: 

United States Government Accountability Office: 
GAO: 

September 2009: 

Medicare: 

Per Capita Method Can Be Used to Profile Physicians and Provide 
Feedback on Resource Use: 

GAO-09-802: 

GAO Highlights: 

Highlights of GAO-09-802, a report to the Chairman, Subcommittee on 
Health, Committee on Ways and Means, House of Representatives. 

Why GAO Did This Study: 

The Medicare Improvements for Patients and Providers Act of 2008 
directed the Secretary of Health and Human Services to develop a 
program to give physicians confidential feedback on the Medicare 
resources used to provide care to Medicare beneficiaries. GAO was asked 
to evaluate the per capita methodology for profiling physicians—a 
method which measures a patient’s resource use over a fixed period of 
time and attributes that resource use to physicians—in order to assist 
the Centers for Medicare & Medicaid Services (CMS) with the development 
of a physician feedback approach. In response, this report examines (1) 
the extent to which physicians in selected specialties show stable 
practice patterns and how beneficiary utilization of services varies by 
physician resource use level; (2) factors to consider in developing 
feedback reports on physicians’ performance, including per capita 
resource use; and (3) the extent to which feedback reports may 
influence physician behavior. GAO focused on four medical specialties 
and four metropolitan areas chosen for their geographic diversity and 
range in average Medicare spending per beneficiary. To identify 
considerations for developing a physician feedback system, GAO reviewed 
the literature and interviewed officials from health plans and 
specialty societies. Further, GAO drew upon literature and interviews 
to develop an illustration of how per capita measures could be included 
in a physician feedback report. 

What GAO Found: 

Using 2005 and 2006 Medicare claims data and a per capita methodology, 
GAO found that specialist physicians showed considerable stability in 
resource use despite high patient turnover. This stability suggests 
that per capita resource use is a reasonable approach for profiling 
specialist physicians because it reflects distinct patterns of a 
physician’s resource use, not the particular population of 
beneficiaries seen by a physician in a given year. GAO also found that 
our per capita method can differentiate specialists’ patterns of 
resource use with respect to different types of services, such as 
institutional services, which were a major factor in beneficiaries’ 
resource use. In particular, patients of high resource use physicians 
used more institutional services than patients of low resource use 
physicians. 

GAO identified four key considerations in developing feedback reports 
on physician performance (see table). 

Table: Key Considerations in Developing Physician Feedback Reports: 

General considerations: Report content; Examples of specific 
considerations: Types of measures, comparative benchmarks. 

General considerations: Report design; Examples of specific 
considerations: Length, organization, graphics. 

General considerations: Report dissemination; Examples of specific 
considerations: Which physicians should receive reports, frequency of 
reporting, hardcopy versus electronic dissemination. 

General considerations: Transparency; Examples of specific 
considerations: Information about purpose, methods, data. 

Source: GAO. 

[End of table] 

To illustrate how per capita measures could be included in a physician 
feedback report, we developed a mock report containing three types of 
per capita measures. 

Although the literature suggested that feedback alone has no more than 
a moderate influence on physicians’ behavior, the potential influence 
of feedback from CMS on Medicare costs may be greater, in part because 
of the relatively large share of physicians’ practice revenues that 
Medicare typically represents. 

CMS reviewed a draft of this report and broadly agreed with our 
findings. 

View [hyperlink, http://www.gao.gov/products/GAO-09-802] or key 
components. For more information, contact A. Bruce Steinwald at (202) 
512-7114 or steinwalda@gao.gov. 

[End of section] 

Contents: 

Letter: 

Background: 

Per Capita Profiling Method Shows Specialist Physicians' Practice 
Patterns Relatively Stable Over 2 Years; Patients of High Resource Use 
Physicians Used More Institutional Services Than Other Patients: 

Research Literature, Health Insurers, and Specialists Identified 
Considerations in Developing Physician Feedback Reports on Resource 
Use: 

Potential Influence of Feedback Regarding Medicare Costs on Physician 
Behavior Is Uncertain: 

Concluding Observations: 

Agency and Professional Association Comments and Our Evaluation: 

CMS Comments: 

AAOS and ACC comments: 

Appendix I: Methodology: 

Appendix II: Comments from the Centers for Medicare & Medicaid 
Services: 

Appendix III: GAO Contact and Staff Acknowledgments: 

Bibliography: 

Tables: 

Table 1: Selected Events Preceding CMS Providing Physicians Feedback on 
Their Medicare Resource Use: 

Table 2: Average Stability of Physicians' Resource Use by Specialty-- 
Averaged Across Four Metropolitan Areas, 2005-2006: 

Table 3: Average Stability of Physicians' and Beneficiaries' Resource 
Use by Metropolitan Area--2005-2006: 

Table 4: Key Considerations in Developing Physician Feedback Reports: 

Table 5: Comments and Suggestions for Designing Physician Feedback 
Reports: 

Table 6: How the Nation's Five Largest Health Insurers Make Patient- 
Level Profiling Data Available to Physicians for Review and Appeal: 

Figures: 

Figure 1: Stability of Medicare Beneficiaries' and Specialist 
Physicians' Resource Use--Averaged Across Four Metropolitan Areas and 
Four Physician Specialties, 2005-2006: 

Figure 2: Share of Total Medicare Expenditures per Beneficiary for 
Services Provided by Physicians to their Patients, Institutional 
Services, and All Other Services--Practice Average Across Four 
Specialties in Four Metropolitan Areas, 2006: 

Figure 3: Mock Physician Feedback Report Illustrating Per Capita 
Measures: 

Figure 4: Example of a Hover in a Mock Physician Feedback Report: 

Figure 5: Beneficiary Resource Use by Health Category for Quintiles of 
Physician Resource Use--Four Specialties in Four Metropolitan Areas, 
2006: 

Abbreviations: 

CBO: Congressional Budget Office: 

CBSA: Core-Based Statistical Area: 

CMS: Centers for Medicare & Medicaid Services: 

FFS: fee-for-service: 

HCC: Hierarchical Condition Category: 

HHS: Department of Health and Human Services: 

MedPAC: Medicare Payment Advisory Commission: 

MIPPA: Medicare Improvements for Patients and Providers Act of 2008: 

[End of section] 

United States Government Accountability Office: 
Washington, DC 20548: 

September 25, 2009: 

The Honorable Pete Stark: 
Chairman: 
Subcommittee on Health: 
Committee on Ways and Means: 
House of Representatives: 

Dear Mr. Chairman: 

In recent years, evidence has mounted that the Medicare program is 
unsustainable in its present form.[Footnote 1] Because of rising health 
care costs and the aging of baby boomers into eligibility for Medicare, 
future program spending is projected to consume an increasing share of 
the government's resources. In their 2009 annual report, the Medicare 
Trustees projected that Medicare expenditures, which reached $468 
billion in 2008, will increase in future years at a faster pace than 
the overall economy, rising from 3.2 percent of gross domestic product 
in 2008 to 11.4 percent by 2083. 

Physicians play a central role in the generation of health care 
expenditures, through both the services they provide and the services 
they order, including hospital admissions, diagnostic tests, and 
referrals to other physicians. The evidence suggests that some of the 
spending for services provided and ordered by physicians may not be 
warranted. For example, the wide variation in Medicare spending for 
physician services--unrelated to beneficiary health status or outcomes--
indicates that health needs alone do not determine spending. 

Consistent with physicians' central role in providing and ordering 
services and their influence on the amount of spending for patient 
services, physician groups, insurers, and Medicare officials have 
turned to profiling as a possible tool to help identify and contain 
overuse of services and the resulting high expenditures. In profiling, 
the resource use of a physician's: 

patients is compared to a benchmark[Footnote 2]. In our previous report 
on profiling, Medicare: Focus on Physician Practice Patterns Can Lead 
to Greater Program Efficiency,[Footnote 3] we profiled generalist 
physicians and found that in each of the 12 metropolitan areas we 
studied there were physicians who, relative to their peers in the same 
area, treated a disproportionate share of overly expensive patients. In 
that report we used a profiling methodology known as per capita, which 
measures per patient resource use for a defined population over a fixed 
period of time and attributes that resource use to physicians. We 
recommended that the Administrator of the Centers for Medicare & 
Medicaid Services (CMS) develop a profiling system to identify 
individual physicians with inefficient practice patterns and provide 
incentives for physicians to improve the efficiency of care they 
provide.[Footnote 4] In our subsequent testimony on physician feedback 
to the Subcommittee on Health of the House Ways and Means Committee we 
stated that providing feedback to physicians on their practice patterns 
could be a promising step toward encouraging efficiency in 
Medicare.[Footnote 5] The Medicare Payment Advisory Commission (MedPAC) 
has also recommended providing feedback to physicians on their resource 
use.[Footnote 6] In its reports, MedPAC has explored an episode-based 
profiling methodology, which measures resource use for treating a 
particular episode of illness--for example, a stroke or heart attack--
and attributes that resource use to physicians. 

Following the issuance of our report and subsequent testimony, Congress 
passed the Medicare Improvements for Patients and Providers Act of 2008 
(MIPPA), which directed the Secretary of Health and Human Services 
(HHS) to develop a program to provide physicians confidential feedback 
on the Medicare resources used to provide care to Medicare 
beneficiaries.[Footnote 7] MIPPA gave HHS the flexibility to measure 
resource use on a per capita basis, an episode basis, or both. In 
response to this mandate, CMS is currently testing both profiling 
methodologies in its Physician Resource Use Measurement and Reporting 
Program. MIPPA also directed us to submit a report to Congress on CMS's 
physician feedback program by 2011. 

In your letter of August 22, 2007, you pointed out that both the per 
capita and episode-based methods could be used to identify inefficient 
physicians, but noted that less is known about the per capita method. 
At that time, you asked us to evaluate the per capita method for 
profiling physicians in order to assist CMS with the development of a 
physician feedback approach for Medicare. This report explores the use 
of a per capita method to profile physicians based on their patients' 
level of resource use, and discusses the development and influence of 
feedback reports. Specifically, this report examines (1) the extent to 
which physicians in selected specialties show stable practice patterns 
and how beneficiary utilization of services varies by physician 
resource use level; (2) factors to consider in developing feedback 
reports on physicians' performance, including per capita resource use; 
and (3) the extent to which feedback reports may influence physician 
behavior. 

We focused our analysis on four diverse specialties--a medical 
specialty (cardiology), a diagnostic specialty (diagnostic radiology), 
a primary care specialty (internal medicine), and a surgical specialty 
(orthopedic surgery); and four metropolitan areas--Miami, Fla.; 
Phoenix, Ariz.; Pittsburgh, Pa.; and Sacramento, Calif.[Footnote 8] We 
chose these areas for their geographic diversity, range in average 
Medicare spending per beneficiary, and number of physicians in each of 
the four specialties. We limited our study to physicians who 
participate in Medicare fee-for-service (FFS). Our results are not 
generalizable to other geographic areas and specialties. 

To measure beneficiaries' resource use, we first adjusted for 
beneficiaries' health conditions, because sick beneficiaries are 
expected to use more resources than healthy beneficiaries. Using 
Hierarchical Condition Category (HCC) and expenditure data obtained 
from CMS, we estimated a risk adjustment model that uses the same 70 
HCCs as the model CMS uses to set managed care capitation rates. HCCs 
are a way of summarizing an individual's diagnoses into major medical 
conditions, such as vascular disease or severe head injury.[Footnote 9] 
Given beneficiaries' HCCs during the year, we used our model to 
estimate Medicare's expected annual expenditures for services provided 
to the beneficiaries in our study. Based on these expected 
expenditures, we placed beneficiaries into 25 discrete risk categories. 
Within each risk category and metropolitan area, we ranked 
beneficiaries from 1 to 100 by their total annual Medicare FFS 
expenditures such that the average beneficiary in a given risk category 
and metropolitan area had a rank of 50.[Footnote 10] We used this rank 
as our risk-adjusted measure of beneficiaries' resource use. 

Our measure of physicians' resource use is derived from the resource 
use of their patients.[Footnote 11] For all physicians in our study, we 
calculated the average rank of their patients. We then used this 
average to rank physicians within the same metropolitan area and 
specialty on a scale of 1 to 100. This measure reflects how expensive a 
physician's patients are compared to the patients of other physicians 
in the same specialty and area after adjusting for differences in 
patient health status. 

To examine the stability of physicians' resource use from a year-to- 
year perspective, we analyzed data for 2005 and 2006.[Footnote 12] We 
divided physicians' and beneficiaries' resource use into quintiles and 
examined which physicians and beneficiaries stayed in the same resource 
use quintile from 2005 to 2006 and which ones did not.[Footnote 13] We 
also examined the degree of turnover in the patients seen by physicians 
between 2005 and 2006. In addition, we used the physician quintiles to 
examine how beneficiary utilization of selected services in 2006 varied 
by physician resource use quintile[Footnote 14]. 

We concluded that the information on Medicare claims that we used in 
this report was sufficiently reliable for the purpose of our analysis, 
because it is a record of Medicare's payments to health care providers. 
We obtained beneficiaries' FFS expenditures from claims information, 
and we used data from CMS files containing enrollment and institutional 
status in order to determine whether beneficiaries were eligible for 
our study. CMS provided us with a file containing beneficiaries' HCCs, 
which we used to estimate their expected expenditures. We obtained 
physicians' specialties from Medicare physician files that CMS uses to 
administer the program and set payment rates. CMS and its contractors 
closely monitor these files, so they are generally considered reliable. 
In addition, we interviewed relevant CMS officials concerning the data 
and consulted data documentation maintained by CMS. We consider the 
data sufficiently reliable for our purposes. 

To determine factors to consider in developing reports to provide 
feedback to physicians on their performance, including their per capita 
resource use, and the extent to which feedback reports may influence 
physician behavior, we reviewed selected literature and interviewed 
experts.[Footnote 15] To identify relevant literature, we searched 31 
databases, including MEDLINE and Science Citation Index, using terms 
such as "physician performance feedback," for journal articles and 
other documents published between January 1, 2000, and February 13, 
2009. From reference lists in documents identified during that search, 
we identified additional documents that met our criteria. We selected 
for review three types of documents: (1) meta-analyses, reviews, or 
scans of the literature on the effectiveness of providing performance 
feedback to physicians; (2) evaluations of various efforts to provide 
performance feedback to physicians; and (3) documents that provided 
guidance from experts on methods for providing performance feedback to 
physicians. In addition to reviewing selected literature, we conducted 
interviews with officials of four specialty societies to identify 
specialty-specific perspectives and concerns, and to solicit officials' 
comments on a mock feedback report we designed. We also conducted 
interviews with officials of the five health insurers with the highest 
revenues in 2007 about their experiences with feedback reports. 
[Footnote 16] 

There are several limitations to our findings. Our findings cannot be 
generalized to other areas or specialties. We also restricted our scope 
to individual physicians and did not analyze group practices. Most 
importantly, we did not pilot our mock report, which illustrates how 
per capita measures could be included in a physician feedback report, 
or test it by giving physicians feedback based on actual resource use. 
Consequently, we are unable to evaluate how helpful it would be to 
physicians and, particularly, whether it has potential for increasing 
physicians' efficiency. 

We conducted our work from February 2008 to September 2009 in 
accordance with all sections of GAO's Quality Assurance Framework that 
are relevant to our objectives. The framework requires that we plan and 
perform the engagement to obtain sufficient and appropriate evidence to 
meet our stated objectives and to discuss any limitations in our work. 
We believe that the information and data obtained, and the analysis 
conducted, provide a reasonable basis for any findings and conclusions. 

Background: 

We, MedPAC, and the Congressional Budget Office (CBO) have all 
suggested that CMS profile physician resource use and provide feedback 
to physicians as a step toward improving the efficiency of care 
financed by Medicare. In July 2008, Congress passed MIPPA,[Footnote 17] 
which directed the Secretary of HHS to establish a program by January 
1, 2009, to provide physicians confidential feedback on the Medicare 
resources used to provide care to beneficiaries. MIPPA gave HHS the 
flexibility to measure resource use on a per capita basis, an episode 
basis, or both. In response to the MIPPA mandate, CMS is pursuing its 
Physician Resource Use Measurement and Reporting Program. (See table 
1.) 

Table 1: Selected Events Preceding CMS Providing Physicians Feedback on 
Their Medicare Resource Use: 

March 2005: 
MedPAC, in its report to Congress, recommended that the Secretary of 
HHS should use Medicare claims data to measure fee-for-service 
physicians' resource use and share results with physicians 
confidentially to educate them about how they compare with aggregated 
peer performance.[A] 

June 2006: 
MedPAC, in its report to Congress, stated that it is important to use a 
per capita profiling methodology in conjunction with an episode-based 
profiling methodology in order to get a complete picture of resource 
use.[B] 

March 2007: 
CBO, in its testimony to the Committee on Finance, United States 
Senate, stated that physicians participating in fee-for-service 
Medicare could be required or encouraged to participate in a program 
that would provide physicians feedback on how their practice patterns 
compared to their peers as a step toward encouraging more efficient 
care.[C] 

April 2007: 
GAO, in its report to Congress, recommended that CMS develop a 
physician profiling system that included feedback and incentives as 
part of a package of reforms to improve the efficiency of care financed 
by Medicare.[D] 

May 2007: 
GAO, in its testimony to the Subcommittee on Health, House Committee on 
Ways and Means, stated that providing feedback to physicians on their 
practice patterns could be a promising step toward encouraging 
efficiency in Medicare.[E] 

July 2008: 
Congress passed the Medicare Improvements for Patients and Providers 
Act of 2008, which mandated that the Secretary of HHS establish a 
program to provide physicians confidential feedback on the Medicare 
resources used to provide care to beneficiaries.[F] 

April 2008 to Present: 
CMS began a phased implementation of its Physician Resource Use 
Measurement and Reporting Program which, in Phase I, has disseminated 
approximately 310 Resource Use Reports to physicians in 13 areas. The 
program is exploring both per capita and episode-based 
methodologies.[G] 

Source: GAO. 

[A] MedPAC, Report to the Congress: Medicare Payment Policy 
(Washington, D.C.: March 2005), 142. 

[B] MedPAC, Report to the Congress: Increasing the Value of Medicare 
(Washington, D.C.: June 2006), xvi. 

[C] CBO, Medicare's Payments to Physicians: Options for Changing the 
Sustainable Growth Rate (Washington, D.C.: March 1, 2007), 16-17. 

[D] GAO, Medicare: Focus on Physician Practice Patterns Can Lead to 
Greater Program Efficiency, GAO-07-307 (Washington, D.C.: April 30, 
2007), 22. 

[E] GAO, Medicare: Providing Systematic Feedback to Physicians on their 
Practice Patterns Is a Promising Step Toward Encouraging Physician 
Efficiency, GAO-07-862T (Washington, D.C.: May 10, 2007). 

[F] Medicare Improvements for Patients and Providers Act of 2008 
(MIPPA), Pub. L. No. 110-275, §131(c), 122 Stat. 2494, 2520-25. 

[G] Federal Register, vol. 74, Physician Resource Use Measurement and 
Reporting Program, no. 132 (Washington, D.C.: July 13, 2009), 33589- 
33591. 

[End of table] 

Key Decisions in Physician Profiling: 

When profiling physicians on their resource use, five key decisions 
must be made: 

* Which resource use measurement methodology to use. There are two main 
profiling methodologies: per capita and episode-based. Using both types 
of measures of resource use may provide more meaningful results by more 
fully capturing the relevant characteristics of a physician's practice 
patterns. 

* How to account for differences in patient health status. Accounting 
for differences in patient health status, a process sometimes referred 
to as risk-adjustment, is an important and challenging aspect of 
physician profiling. Because sicker patients are expected to use more 
health care resources than healthier patients, we believe the health 
status of patients must be taken into account to make meaningful 
comparisons among physicians. There are various risk-adjustment methods 
and the suitability of a given method will depend on characteristics of 
the physicians to be profiled and their patients. 

* How to attribute resource use to physicians. Important attribution 
decisions include whether to assign a patient's resource use to the 
single physician who bears the greatest responsibility for the resource 
use, to all physicians who bore any responsibility, or to all 
physicians who met a given threshold of responsibility, such as 
providing a certain percentage of the expenditures or volume of 
services. A single attribution approach may not be applicable for all 
types of measures or for all types of physician specialties. 

* What benchmark(s) to use. Physician profiling involves comparing 
physicians' resource use to a benchmark. There are differing opinions 
on what are the most appropriate and meaningful comparative benchmarks. 

* How to determine what is a sufficient sample size to ensure 
meaningful comparisons. The feasibility of using resource use measures 
to compare physicians' performance depends, in part, on two factors: 
the availability of enough data on each physician to compute a resource 
use measure and a sufficient number of physicians to provide meaningful 
comparisons. It is important to calculate resource use measures only 
for physicians with sufficient sample sizes in order to address 
concerns that a physician's profile may be distorted by a few aberrant 
cases. There is no consensus on what sample size is adequate to ensure 
meaningful measures. 

CMS's Resource Use Measurement and Reporting Program: 

Responding to the MIPPA mandate to establish a physician feedback 
program by January 1, 2009, CMS began in April 2008 to develop its 
program for reporting to physicians on their resource use. In the first 
phase of the program, CMS identified eight priority conditions and 
disseminated approximately 310 Resource Use Reports to physicians in 
selected specialties who practiced in one of 13 geographic areas. The 
reports generally included both per capita and episode-based resource 
use measures that were calculated according to five different 
attribution rules. The reports also contained multiple cost benchmarks 
relative to physicians in the same specialty and geographic area. In 
Phase II, CMS is proposing to expand the program by adding quality 
measures and reporting on groups of physicians as a mechanism for 
addressing small sample size issues. 

Per Capita Profiling Method Shows Specialist Physicians' Practice 
Patterns Relatively Stable Over 2 Years; Patients of High Resource Use 
Physicians Used More Institutional Services Than Other Patients: 

Using a per capita profiling method, we found that from 2005 to 2006, 
specialist physicians showed considerable stability in their practice 
patterns, as measured by resource use--greater stability than their 
patients, despite high patient turnover. We also found that our per 
capita method can differentiate specialists' patterns of resource use 
with respect to different types of services, such as institutional 
services,[Footnote 18] which were a major factor in beneficiaries' 
resource use. In particular, patients of high resource use physicians 
used more institutional services than patients of low resource use 
physicians. 

Specialist Physicians' Resource Use More Stable Than Beneficiaries' 
Resource Use: 

Using a per capita method to profile specialist physicians, we found 
that their practice patterns, as measured by the level of their 
resource use, was relatively stable over 2005 and 2006 by comparison 
with individual beneficiaries' resource use (see figure 1).[Footnote 
19] This is true despite the fact that our measure of physicians' 
resource use is derived from their patients' resource use and that the 
specific patients whom physicians see are not always the same from year 
to year. Among the physicians we studied, less than one-third of 
patients seen by study physicians in 2005 were also seen by the same 
physician in 2006. This stability suggests that per capita resource use 
is a reasonable approach for profiling physicians, because it reflects 
distinct patterns of a physician's resource use, not the particular 
population of beneficiaries seen by a physician in a given year. 

We divided both physician and beneficiary resource use into five groups 
of approximately equal size (quintiles) and found that, on average 
across the four metropolitan areas and four specialties, 58 percent of 
physicians and 30 percent of beneficiaries were in the same quintile of 
resource use in 2005 and 2006. The pattern was even more pronounced for 
the top resource use quintile: 72 percent of physicians and 35 percent 
of beneficiaries remained in that quintile. If the level of physicians' 
and beneficiaries' resource use was purely random, only 20 percent 
would be expected to have remained in the same quintile. 

Figure 1: Stability of Medicare Beneficiaries' and Specialist 
Physicians' Resource Use--Averaged Across Four Metropolitan Areas and 
Four Physician Specialties, 2005-2006: 

[Refer to PDF for image: vertical bar graph] 

Resource use remained in same quintile[A] both years: 
Physicians: 58%; 
Beneficiaries: 30%. 

Resource use remained in top quintile[A] both years: 
Physicians: 72%; 
Beneficiaries: 35%. 

Percentage expected by chance: 20%. 

Source: GAO analysis of Medicare claims data. 

Note: The specialist physicians include cardiologists, diagnostic 
radiologists, internists, and orthopedic surgeons in Miami, Fla.; 
Phoenix, Ariz.; Pittsburgh, Pa.; and Sacramento, Calif. 

[A] Beneficiaries and physicians are divided into five ascending groups 
of nearly equal size based on the level of their resource use. 

[End of figure] 

We also examined the stability of physicians' resource use by specialty 
and found a similar pattern, although not to the same extent in all 
specialties. The average percentage of physicians who were in the same 
resource use quintile in 2005 and 2006 ranged from 48 percent for 
orthopedic surgeons to 60 percent for internists. Resource use in the 
top quintile was more stable and ranged from 69 percent for diagnostic 
radiologists to 74 percent for internists. (See table 2.) 

Table 2: Average Stability of Physicians' Resource Use by Specialty-- 
Averaged Across Four Metropolitan Areas, 2005-2006: 

Physician Specialty: Cardiology; 
Average percentage remaining in same quintile[A]: 59; 
Average percentage remaining in the top quintile[A]: 71. 

Physician Specialty: Diagnostic radiology; 
Average percentage remaining in same quintile[A]: 58; 
Average percentage remaining in the top quintile[A]: 69. 

Physician Specialty: Internal medicine; 
Average percentage remaining in same quintile[A]: 60; 
Average percentage remaining in the top quintile[A]: 74. 

Physician Specialty: Orthopedic surgery; 
Average percentage remaining in same quintile[A]: 48; 
Average percentage remaining in the top quintile[A]: 70. 

Source: GAO analysis of Medicare claims data. 

Note: The four metropolitan areas are Miami, Fla.; Phoenix, Ariz.; 
Pittsburgh, Pa.; and Sacramento, Calif. 

[A] Physicians are divided into five ascending groups of nearly equal 
size based on the level of their resource use. 

[End of table] 

In each of the four metropolitan areas, physicians showed greater 
stability in their resource use than individual beneficiaries, although 
the percentages varied. For example, the percentage of physicians 
remaining in the top quintile ranged from 68 percent in Phoenix to 76 
percent in Miami. For beneficiaries, the percentage in the top quintile 
ranged from 31 percent in Phoenix to 39 percent in Miami. (See table 
3.) 

Table 3: Average Stability of Physicians' and Beneficiaries' Resource 
Use by Metropolitan Area--2005-2006: 

Metropolitan Area: Miami; 
Physicians[A]: Average percentage remaining in same quintile[B]: 62; 
Physicians[A]: Average percentage remaining in same quintile[B]: 76; 
Beneficiaries: Average percentage remaining in same quintile[B]: 31; 
Beneficiaries: Average percentage remaining in the top quintile[B]: 39. 

Metropolitan Area: Phoenix; 
Physicians[A]: Average percentage remaining in same quintile[B]: 56; 
Physicians[A]: Average percentage remaining in same quintile[B]: 68; 
Beneficiaries: Average percentage remaining in same quintile[B]: 29; 
Beneficiaries: Average percentage remaining in the top quintile[B]: 31. 

Metropolitan Area: Pittsburgh; 
Physicians[A]: Average percentage remaining in same quintile[B]: 52; 
Physicians[A]: Average percentage remaining in same quintile[B]: 70; 
Beneficiaries: Average percentage remaining in same quintile[B]: 30; 
Beneficiaries: Average percentage remaining in the top quintile[B]: 32. 

Metropolitan Area: Sacramento; 
Physicians[A]: Average percentage remaining in same quintile[B]: 58; 
Physicians[A]: Average percentage remaining in same quintile[B]: 71; 
Beneficiaries: Average percentage remaining in same quintile[B]: 30; 
Beneficiaries: Average percentage remaining in the top quintile[B]: 32. 

Source: GAO analysis of Medicare claims data. 

[A] Cardiologists, diagnostic radiologists, internists, and orthopedic 
surgeons. 

[B] Beneficiaries and physicians are divided into five ascending groups 
of nearly equal size based on the level of their resource use. 

[End of table] 

The greater stability of physicians' resource use compared to 
beneficiaries' resource use could be due to their individual practice 
styles, as well as to a range of other factors, such as participation 
in formal or informal referral networks. These networks have a range of 
providers, including other physicians, who treat their patients and 
refer them for treatment, testing, and admissions to hospitals. 

Beneficiary Use of Institutional Services Varies by Physician Resource 
Level: 

Beneficiaries seen by high resource use physicians generally were 
heavier users of institutional services than those seen by lower 
resource use physicians, and institutional services accounted for more 
than one-half of total patient expenditures. This pattern was 
consistent across three of the four specialties we studied, with 
orthopedic surgery being the exception. 

Institutional services were the major driver of Medicare expenditures 
for beneficiaries in physicians' practices, accounting on average for 
54 percent of expenditures. Services provided by a particular physician 
in our study directly to that physician's patients accounted for only 2 
percent of total expenditures or about $350 for each beneficiary in a 
physician's practice. All other services--those provided by other 
physicians, home health care, hospice care, outpatient services, and 
durable medical equipment--accounted for the remaining 44 percent of 
expenditures. (See figure 2.) 

Figure 2: Share of Total Medicare Expenditures per Beneficiary for 
Services Provided by Physicians to their Patients, Institutional 
Services, and All Other Services--Practice Average Across Four 
Specialties in Four Metropolitan Areas, 2006: 

[Refer to PDF for image: pie-chart] 

Services provided by a particular physician to his or her patients: 2%; 
Institutional services[A]: 54%; 
All other services[B}: 44%. 

Source: GAO analysis of CMS claims data. 

Note: The percentages shown are the average share of Medicare 
expenditures for the beneficiaries in the practices of cardiologists, 
diagnostic radiologists, internists, and orthopedic surgeons in Miami, 
Fla.; Phoenix, Ariz.; Pittsburgh, Pa.; and Sacramento, Calif., with a 
minimum of 100 Medicare patients in their practice. 

[A] Institutional services include inpatient hospital and skilled 
nursing care. 

[B] Part B physician and supplier services (other than those provided 
directly by the physician), home health care, hospice care, outpatient 
hospital services, and durable medical equipment. 

[End of figure] 

Expenditures for institutional services for a physician's patients grew 
as the level of physician resource use increased. Dividing the level of 
physician resource use into quintiles, we examined the relationship of 
physicians' resource use and expenditures for services provided to 
their patients. Average expenditures for institutional services 
increased more steeply by physician resource quintile than expenditures 
for all other services.[Footnote 20] 

The four specialties all exhibited this pattern of increasing 
beneficiary expenditures for institutional services accompanying 
increasing physician resource use, although for orthopedic surgery the 
increase was small. The increase in average beneficiary expenditures 
for all other services that accompanied increasing physician resource 
use was similar for three of the four specialties and was steeper for 
internal medicine. 

We also examined the average number of physicians seen by the Medicare 
beneficiaries we studied and found that it was positively associated 
with increasing physician resource use. Overall, the number of 
physicians seen increased from an average of about 13 physicians per 
beneficiary in the lowest quintile of resource use to more than 23 in 
the highest. The increase in the number of physicians seen was 
accompanied by an increase in average beneficiary expenditures for 
institutional services that was steeper than the rise in other 
services. 

Research Literature, Health Insurers, and Specialists Identified 
Considerations in Developing Physician Feedback Reports on Resource 
Use: 

Through our review of selected literature and interviews with officials 
of health insurance companies, specialty societies, and profiling 
experts, we identified several key considerations in developing reports 
to provide feedback to physicians on their performance, including their 
per capita resource use. We also drew on information from these sources 
to develop an example of how per capita measures could be presented in 
a physician feedback report. 

Key Considerations in Developing Physician Feedback Reports Include 
Content, Design, Dissemination Strategy, and Transparency: 

We identified four key considerations in developing reports to provide 
feedback to physicians (see table 4). 

Table 4: Key Considerations in Developing Physician Feedback Reports: 

General considerations: Report content; 
Examples of specific considerations: Types of measures, comparative 
benchmarks. 

General considerations: Report design; 
Examples of specific considerations: Length, organization, graphics. 

General considerations: Report dissemination; 
Examples of specific considerations: Which physicians should receive 
reports, frequency of reporting, hardcopy versus electronic 
dissemination. 

General considerations: Transparency; 
Examples of specific considerations: Information about purpose, 
methods, data. 

Source: GAO analysis. 

[End of table] 

Report Content: 

Our review of selected literature suggested that a physician feedback 
report should contain three basic elements: an explanation of the 
information contained in the report (which we will discuss in the 
context of transparency), measures describing the performance of the 
physician or physicians to whom the report is directed, and comparative 
benchmarks. 

Measures. Both the selected literature we reviewed and the officials we 
interviewed supported including measures of quality along with measures 
of cost, and ensuring that measures are actionable by providing 
information that can help physicians improve their performance. The 
officials we interviewed were divided as to whether these measures 
should reflect physicians' performance at the individual level or the 
group level. 

* Quality measures. All five of the insurers we contacted were 
profiling physicians in terms of quality and cost, and four of the five 
had adopted a model code for physician ranking programs that called for 
rankings to be based on quality as well as cost.[Footnote 21] Most of 
the specialty society officials we interviewed also called for the 
inclusion of quality measures in physician feedback reports, and some 
cautioned that focusing solely on costs could create perverse 
incentives--for example, encouraging physicians to reduce 
inappropriately the level of care provided to patients. The lack of 
widely accepted, claims-based quality measures for some specialties has 
limited the number of specialties some insurers profile. For example, 
at the time of our interview, one insurer was profiling physicians in 
only one specialty (cardiology) while planning to begin profiling other 
specialties within a year. 

* Actionable measures. According to one research report we reviewed, 
little research has been done to determine how the reporting of global 
scores--such as an overall per capita cost rank--influences physician 
behavior,[Footnote 22] but experts on physician profiling and a broad 
array of stakeholders, including physicians and insurance company 
officials, agreed that performance data should be disaggregated into 
enough categories to enable physicians to identify practice patterns to 
change. According to some profiling experts, resource use reports must 
pinpoint physicians' overuse and misuse of resources, and identify 
practices that add costs but do not improve desired outcomes. 
Similarly, specialty society officials we interviewed emphasized the 
importance of including measures that focus on areas in which the 
physician has control. 

* Individual versus group measures. Another measurement consideration 
is whether physicians in group practices should be profiled as 
individuals or as a group. The insurers we contacted took varying 
approaches. In some cases, the approach was driven by contracting 
arrangements, with insurers constructing group profiles for physicians 
with whom they had group contracts. One insurance company official 
pointed out that profiling at the group level allows more physicians to 
be profiled, as it increases the data available to construct a profile. 
Another official advocated profiling at the individual level because he 
believes physicians are more interested in assessments of individual 
performance. Officials of the four specialty societies generally saw 
some merit to both approaches, but some underscored the difficulty of 
identifying group affiliations or noted that groups are not necessarily 
homogeneous enough for a group assessment to be appropriate. 

Comparative benchmarks. One consideration addressed by multiple 
publications we reviewed was the kind of benchmark to which physicians' 
performance should be compared. For example, a physician's performance 
may be compared to (1) an evidence-based standard, (2) a standard based 
on professional judgment, such as the consensus standards endorsed by 
the National Quality Forum, or (3) to a statistical norm, such as the 
average for a physician's peers locally or nationally. 

Although studies we reviewed offered conflicting evidence as to whether 
including peer comparisons in physician feedback reports increases 
their effectiveness, some profiling experts and specialty society 
officials believe comparative information is useful and of interest to 
physicians. In the literature we reviewed, for example, one profiling 
expert suggested that such comparisons can motivate behavior change by 
taking advantage of physicians' desire to perform at least as well as 
their peers; another stated that performance statistics are not 
meaningful to physicians without peer comparisons. 

A physician's peer group can be defined in various ways. According to 
one study, some organizations that provide performance feedback to 
physicians have found comparisons within specialty and locality most 
useful to and most frequently requested by physicians.[Footnote 23] 
Representatives of some of these organizations said physicians find 
local information more relevant because it reflects the practice 
patterns of their geographic area. All five insurers we contacted 
compare physicians to others in the same market and specialty; one of 
the five also compares physicians to peers nationwide on some measures. 
In contrast, officials of all four specialty societies recommended 
comparisons at the national level, with officials of one society 
stating that there is no scientific basis for regional variations in 
practice patterns.[Footnote 24] There was less agreement about whether 
physicians should be compared to others in their specialty or to a more 
narrowly defined group. Officials of one specialty society advocated 
comparisons at the subspecialty level in recognition of the variation 
in resource use patterns among subspecialists. Another official pointed 
out that such comparison groups could be difficult to define because 
physicians in some specialties tend to have multiple subspecialties. 
Because views differ on appropriate comparison groups, one hospital- 
owned healthcare alliance plans to incorporate in its physician reports 
a customizable feature that will allow users to select the peer 
comparison they wish to see. 

Comparisons to physicians' own past performance (trend data) are 
commonly presented in feedback reports, and the majority of physicians 
surveyed in one study found these comparisons useful. 

Report Design: 

The selected literature we reviewed offered little hard evidence on how 
feedback reports should be designed to engage physicians' interest or 
to prove their comprehension of the material. However, researchers and 
profiling experts offered some comments and suggestions based either on 
their experience with clinical performance measurement or on an 
analysis of the literature on consumer behavior and its possible 
implications for physician reporting (see table 5). 

Table 5: Comments and Suggestions for Designing Physician Feedback 
Reports: 

Topic: Amount of material and report length; 
Comments: 
* Effective reports do not necessarily provide a high level of detail; 
* Detailed supporting data can be made available in a separate 
drilldown section; 
* Physician feedback reports can vary greatly in length depending on 
the number of topics covered and the level of detail. 

Topic: Organization; 
Comments: 
* The organization of the report may be more important than its length; 
* All high-level summary information should be in one place so that 
it's easy to absorb; 
* Reports should move from gross measures to more refined; 
* Spatial organization, through the use of headings and lists, is 
critical for helping readers find information. 

Topic: Graphics; 
Comments: 
* Visual formats provide the best methods for data interpretation and 
are useful for highlighting the most important measures; 
* Information can be conveyed visually in tables, graphs, and score 
cards; 
* Tables may be better to show specific numeric values, while graphs 
may be better to display information for comparative purposes, because 
they facilitate the organization of material into meaningful groups; 
* A score card or summary-rating format consolidates data even further 
than tables or graphs, using colors or symbols to help readers easily 
identify successes as well as areas for improvement. 

Source: GAO analysis of selected literature. 

[End of table] 

The amount and combination of material that should be included in a 
single report is an important consideration. According to one 
publication that summarized a review of multiple feedback reports, some 
organizations issue separate reports on efficiency/cost and 
effectiveness/clinical quality, in part to avoid diluting the impact of 
either set of measures. Others believe a single report gives physicians 
a more complete picture of their performance. 

Officials of the three insurers we contacted that routinely issued 
feedback reports to physicians said that their companies produced 
summary reports, typically one to two pages in length, containing high- 
level information, but also made more detailed information, such as 
patient-level data, available to physicians. One insurer's summary 
report consisted of one page of cost efficiency measures and one page 
of effectiveness measures. The cost efficiency page presented average 
cost per episode of care by service category for the physician and the 
physician's peer group, as well as the ratio of the two, in both 
tabular and graphic form. The effectiveness page presented process-of- 
care measures for selected conditions, including cardiovascular disease 
and asthma. Company officials said summary reports were limited to two 
pages to accommodate physicians' attention spans and that the two sets 
of measures were presented separately to discourage attempts to link 
the two. Specialty society officials agreed reports should be short-- 
most proposed one to two pages--and strongly recommended that 
information be presented graphically to the extent possible. One 
official, noting that physicians are very visually oriented, 
recommended feedback reports consisting mainly of easily understood 
graphics. 

The selected literature we reviewed, our interviews with specialty 
society officials, and existing physician feedback reports suggested 
reports can be kept short by segmenting some information into separate 
documents--for example, a cover letter that explains the report's 
purpose, a description of the profiling methodology, a set of 
frequently asked questions, and a list of definitions. 

Report Dissemination: 

Some key considerations with respect to report dissemination are which 
physicians should receive reports, how frequently to issue reports, and 
whether to issue reports in hardcopy or electronically. 

Which physicians should receive feedback reports. One major decision is 
whether to issue reports to all physicians for whom performance 
measures can be calculated or only to a subset who fail to meet certain 
performance standards--a decision that may involve weighing reporting 
costs against potential impacts. None of the studies we reviewed 
directly addressed this issue, but all of the specialty society 
officials we interviewed advised sending reports to all or nearly all 
physicians, rather than just to poor performers. They gave several 
reasons: to provide positive recognition to physicians who are 
performing well; to avoid singling out certain physicians as poor 
performers, especially on the basis of excess costs over which they 
have little control; and to create opportunities for voluntary peer-to- 
peer learning among physicians who are at different points along the 
performance spectrum. Similarly, all three of the insurers that 
routinely issued feedback reports sent them to all physicians for whom 
they had performance measures. 

Frequency of reporting. According to one book we reviewed, 
organizations that provide feedback to physicians should do so more 
than once a year to give physicians an opportunity to improve their 
performance in a timely manner.[Footnote 25] However, because of the 
time needed to gather sufficient data to identify trends and patterns 
of performance, many organizations provide feedback no more than twice 
a year. Of the two insurers that told us how frequently they issued 
feedback reports, one did so annually and the other at least every 6 
months. Officials of the latter company said the frequency of their 
reporting was limited by the number of claims in their dataset and 
suggested that CMS would not face the same limitations. 

Hardcopy versus electronic dissemination. Reports can be disseminated 
in hardcopy through various channels, such as the mail, or 
electronically, through e-mail or a Web site. One literature scan we 
reviewed cited certain advantages of electronic formats such as Web- 
based applications. Specifically, they allow users to organize 
information as they choose and are well suited to presenting data from 
the general to the specific, which facilitates information processing. 
Although this report noted some concerns about physicians' access to 
the Internet, according to a report based on a national survey of 
physicians in December 2002 and January 2003, almost all respondents 
said they had Internet access, and most said they considered it 
important for patient care.[Footnote 26] 

Of the three insurers that routinely issued feedback reports, two 
issued them electronically and one issued them in hardcopy. Officials 
of the latter company said that staff typically hand-delivered the 
reports to physicians during on-site visits in order to discuss the 
results.[Footnote 27] Officials of most of the specialty societies we 
contacted did not advocate one dissemination mode over the other, but 
some noted that organizations that issue reports electronically must 
confront certain challenges, such as ensuring that security features do 
not make access difficult, addressing the lack of high-speed Internet 
service in some areas, and determining whether to send reports by e- 
mail or to instruct physicians to access them on the Internet.[Footnote 
28] One specialty society official recommended using both modes of 
dissemination to accommodate different preferences. 

Transparency: 

Both the selected literature we reviewed and our interviews with 
officials from insurance companies and specialty societies underscored 
the importance of ensuring transparency regarding the purpose of the 
report and the methodology and data used to construct performance 
measures. 

Purpose. According to one literature scan, feedback reports should 
explicitly state their purpose--for example, to reduce costs, improve 
quality, or simply to provide information--and should highlight any 
items for which the physician will be held accountable.[Footnote 29] 

Methodology. Two important considerations are where to provide 
information about methodology--whether in the report itself or through 
some other mechanism, such as a Web page--and how much technical detail 
to provide. Some of the insurers we contacted provide information on- 
line about their profiling methodologies, including details about 
measures, attribution of care to physicians, risk adjustment, and 
statistical issues. In addition, some of the officials we interviewed 
said that company staff will meet with physicians to explain the 
profiling methodology, if requested. For example, officials of one 
company said that it has on staff four profiling experts, mostly 
nurses, in addition to about 20 medical directors who can answer 
physicians' questions. 

Specialty society officials we interviewed highlighted a potential 
trade-off between providing enough information in the report to 
persuade physicians of the validity of the measures and keeping the 
report concise enough to maintain physicians' interest. All of the 
officials we interviewed agreed that physicians should have access to 
details about the methodology; some suggested this information might 
best be disseminated through a Web site. Explaining how the data are 
risk-adjusted to account for differences in physicians' patient 
populations was cited by specialty society officials as particularly 
important. 

Data. Another consideration is ensuring transparency with regard to the 
data used in profiling--making patient-level detail available so 
physicians can reconcile performance measures with their own 
information about their practices. All five of the health insurers we 
contacted provided opportunities for physicians to examine patient- 
level data and file appeals before results are made public, although 
their processes or policies for doing so varied (see table 6). 

Table 6: How the Nation's Five Largest Health Insurers Make Patient- 
Level Profiling Data Available to Physicians for Review and Appeal: 

Availability of patient-level data: 
Insurer A: Generally e-mailed upon request; 
Insurer B: Generally mailed upon request; 
Insurer C: Accessible on-line to each physician; 
Insurer D: Accessible on-line to each physician; 
Insurer E: Hand-delivered to each physician group during site visits. 

Window for review and appeal: 
Insurer A: 90 days; 
Insurer B: 45 days; 
Insurer C: 60 days; 
Insurer D: 45 days; 
Insurer E: 45 days. 

Source: GAO analysis of information provided by insurers. 

[End of table] 

Officials of one of the two insurers that made detailed data available 
on-line said their company previously sent hardcopy reports to 
physicians, but learned from medical office managers that they would 
prefer an on-line format that could be manipulated to facilitate 
physician comparisons. Officials of the other insurer said that their 
company planned to make the data available in a manipulatable format 
soon. Most of the specialty society officials we interviewed agreed 
that patient-level data should be made available to physicians, but 
some predicted that few physicians would access them. Two interviewees 
suggested practice size would probably be a factor; one added that 
physicians in smaller groups would likely lack the resources and skills 
to analyze the data. 

Per Capita Measures Can Be Presented in a Physician Feedback Report: 

Drawing upon lessons culled from the literature and our interviews, we 
developed a mock report that illustrates how per capita measures could 
be included in a physician feedback report. Such a report could also 
include other measures such as quality measures and episode-based 
resource use measures. We included two types of per capita measures-- 
risk-adjusted cost ranks and risk-adjusted utilization rates--each 
presented with local and national comparative benchmarks. To provide 
further context, we also included per capita measures showing how the 
average Medicare costs of patients the physician treated at least once 
were distributed among service categories, and the percentage of those 
costs that were for services directly provided by the physician to whom 
the report is directed. We kept the mock report under two pages and 
included minimal text, while ensuring transparency by indicating the 
availability of methodology details and supporting data. To accommodate 
physicians' differing dissemination preferences, we designed the mock 
report to be available in both electronic and hardcopy formats. (See 
figure 3.) 

Figure 3: Mock Physician Feedback Report Illustrating Per Capita 
Measures: 

[Refer to PDF for image: illustration] 

FFS Medicare: Physicians Report: 

Physician Information: 
Name: Dr. John Doe: 
NPI: •••••••2487: 
Area: Cityville. 

Report Information: 
Reporting Period: 01/01/2007 - 12/31/2007: 
Specialty: Cardiology. 

Overall Medicare Resource Use Measures: 

How the average risk-adjusted Medicare costs of patients you treated at 
least once compared to those of other cardiologists: 

Average patient risk-adjusted cost rank: 

U = your rank; 
A = Area rank; 
N = National rank. 

All services: 
U: 62; 
A: 47; 
N: 49. 

Physician visits: 
U: 59; 
A: 45; 
N: 51. 

Physician procedures: 
U: 57; 
A: 47; 
N: 51. 

Imaging: 
U: 56; 
A: 45; 
N: 47. 

Laboratory: 
U: 43; 
A: 46; 
N: 50. 

Hospital inpatient: 
U: 62; 
A: 47; 
N: 49. 

Hospital outpatient: 
U: 48; 
A: 41; 
N: 50. 

Skilled nursing and home health: 
U: 43; 
A: 47; 
N: 53. 

How the average Medicare costs of patients you treated at least once 
were distributed: 

Payments to all providers and you: 

All services: 
All providers: 100% ($13,422) of total; 
You: 11% ($1,449) of category. 

Physician visits: 
All providers: 10% ($1,342) of total; 
You: 45% ($604) of category. 

Physician procedures: 
All providers: 7% ($939) of total; 
You: 65% ($610) of category. 

Imaging: 
All providers: 6% ($805) of total; 
You: 20% ($161) of category. 

Laboratory: 
All providers: 3% ($403) of total; 
You: 18% ($73) of category. 

Hospital inpatient: 
All providers: 44% ($5,906) of total; 
You: 0% ($0) of category. 

Hospital outpatient: 
All providers: 8% ($1,074) of total; 
You: 0% ($0) of category. 

Skilled nursing and home health: 
All providers: 14% ($1,879) of total; 
You: 0% ($0) of category. 

Risk-adjusted utilization rates: 

Hospitalizations (per 100 patients):
You: 58.3; 
Cardiologists In Your Area: 41.3; 
Cardiologists Nationwide: 46.0 . 

Rehospitalizations (per 100 patients): 
You: 8.6; 
Cardiologists In Your Area: 4.5; 
Cardiologists Nationwide: 5.1. 

Evaluation & Management Visits (per patient): 
You: 3.7; 
Cardiologists In Your Area: 3.1; 
Cardiologists Nationwide: 3.5. 

Note: This figure is an illustration of how per capita measures could 
be included as part of a physician feedback report, which could include 
a cover letter, quality measures, and other resource use measures. All 
of the data presented in the figure are hypothetical. 

Source: GAO. 

[End of figure] 

Specialty society officials who vetted a draft of the mock report made 
several recommendations. Some recommendations centered on taking 
advantage of electronic capabilities, such as adding hovers to define 
key terms (see figure 4), creating interactive features to let 
physicians explore "what if" scenarios, and including links to 
educational materials and specialty guidelines. Officials also 
recommended adding information on pharmaceutical costs, a category we 
did not include because not all beneficiaries are enrolled in a 
Medicare Part D prescription drug plan. 

Figure 5: Example of a Hover in a Mock Physician Feedback Report: 

[Refer to PDF for image: illustration] 

The electronic version of the feedback report could make use of 
interactive features. 

For example, the report could use “hovers” to display a short 
definition of key terms in the document and a reference to a page with 
more information. 

To meet requirements for federal agencies to make electronic 
information accessible to disabled individuals, the feedback report 
could show all hover definitions on the last page of the document, if 
it were printed. 

Average patient risk-adjusted cost rank: 
Hover definition: 
A patient’s risk adjusted cost rank is calculated by comparing the 
patient’s Medicare costs to all other Cityville patients with similar 
risk scores and represents how unexpectedly expensive or inexpensive 
the patient’s Medicare-covered care was. Your rank is the average rank 
of all patients you treated at least once. See Glossary for more 
details. 

Source: GAO. 

[End of figure] 

More generally, specialty society officials said that they particularly 
liked the graphs and charts in our mock report. One official added that 
our report was easier to understand than other reports he had seen and 
that he thought it would get physicians' attention. Another official 
commented how the presented per capita measures could give physicians 
insight on the care their patients are receiving that they were not 
previously aware of--a perspective other cost measures could not 
provide. However, multiple officials said the measures as presented 
were too broad to be actionable and might not seem relevant to 
physicians, as most physicians feel responsible only for the costs of 
services they directly order or provide, not for the total cost of 
patients' care. Two officials suggested that these per capita measures 
would have more value in health care systems that emphasized 
coordination of care. 

Potential Influence of Feedback Regarding Medicare Costs on Physician 
Behavior Is Uncertain: 

Our review of available literature on the effectiveness of physician 
feedback suggests that feedback alone generally has no more than a 
moderate influence on physician behavior. However, the potential 
influence of feedback from CMS regarding Medicare costs is uncertain, 
and may be greater than that of feedback from other sources, because 
Medicare reimbursement typically represents a larger share of 
physicians' practice revenues than that from other insurers. 

In general, studies examining the effect of feedback on physicians' 
behavior have found it to have a small to moderate effect.[Footnote 30] 
Factors that appear to influence the effectiveness of feedback include 
its source, frequency, and intensity. For example, one review of the 
literature concluded that physicians were more likely to be influenced 
by reports from a source they expected to continue monitoring their 
performance. This review also found that repeated feedback over a 
period of several years may be more likely to get physicians' 
attention.[Footnote 31] Another review reported that the intensity of 
the feedback appeared to influence its effectiveness. The review cited 
individual, written feedback containing information about costs or 
numbers of tests, but no personal incentives, as among the least 
intensive, and therefore likely to be among the least effective 
approaches.[Footnote 32] 

Consistent with the literature we reviewed, most of the insurance 
company officials we interviewed questioned whether providing 
performance feedback to physicians would have a significant impact on 
the physicians' behavior in the absence of other incentives. While all 
five insurers profiled physicians, none used the results solely to 
provide feedback.[Footnote 33] Officials of four of the five insurance 
companies said that to affect physicians' behavior, profiling results 
must be made public, thus influencing patients' choice of physicians, 
or linked to monetary incentives, as in pay-for-performance 
arrangements. However, officials of one company disagreed, stating that 
feedback alone can affect physicians' behavior if the reports show how 
they rank against their peers and make clear what behavior they need to 
change to improve their efficiency. These officials also said that the 
impact of feedback could depend on the size of physicians' practices 
and whether they have the resources to review the reports and the 
management structure to affect changes. 

Whether the experiences of private insurers or the lessons from the 
literature on the influence of feedback will hold in the case of the 
Medicare program is uncertain. A survey conducted in 2004-2005 found 
that, for most physicians, Medicare represented more than one-quarter 
of practice revenue, and for 17 percent of physicians, the proportion 
was more than one-half.[Footnote 34] Because physicians typically 
contract with a dozen or more health insurance plans, few, if any, of 
these plans are likely to represent as large a share of physicians' 
practice revenue as Medicare. Hence, the impact of feedback from CMS 
might be greater than that from other sources. In addition, one 
profiling expert suggested that physicians might expect feedback from 
CMS to be only the first step in efforts to influence physicians' 
behavior--to be followed, for example, by public reporting of profiling 
results. This perspective comports with recommendations in our earlier 
report.[Footnote 35] Two interviewees said that providing feedback on a 
confidential basis would be an appropriate first step. One said it 
would allow time to test the profiling methodology and gauge 
physicians' reactions; the other said it would provide an opportunity 
for physicians to vet the measures and identify any errors. 

Most of the specialty society officials predicted that feedback from 
CMS would have a small to moderate effect on physician behavior, 
similar to that described in the literature we reviewed, but some 
officials offered suggestions for enhancing its effectiveness. Other 
suggestions can be drawn from the literature we reviewed. These 
suggestions included: 

* providing advance notice of feedback reports (through presentations, 
letters, or other communications) to help ensure that physicians open 
and read the reports; 

* working through credible intermediaries, such as medical societies or 
locally prominent physicians, to assure physicians that the feedback 
process is reasonable and legitimate; 

* providing opportunities for physicians to discuss the reports through 
videoconferences, teleconferences, or on-line discussion groups; and: 

* offering in-person follow up, possibly drawing on the resources of 
the Medicare Quality Improvement Organizations.[Footnote 36] 

Involving physicians in the development of a feedback system may also 
enhance its effectiveness. One literature scan concluded that physician 
involvement in system design was vital for obtaining physician buy-in. 
[Footnote 37] Information from insurers suggested that, although 
physicians may not always be involved in initial development of 
feedback systems, their feedback can prompt modifications. Some 
insurance officials we interviewed described an iterative process 
involving ongoing communication with physicians and continuous 
modification of reports and systems. For example, officials of one 
insurance company said that the company did not seek initial input from 
physicians--in the belief that they would not have been able to provide 
much input without a complete understanding of the data and 
methodology--but took into account physicians' responses to earlier, 
less formal systems. Officials of other companies described various 
mechanisms for obtaining physicians' perspectives, including formal 
physician advisory councils, regular meetings with officials of 
national medical societies, and town hall meetings with physicians at 
the local level. 

Concluding Observations: 

Profiling physicians to improve efficiency is used by some private 
insurance companies and, at the direction of Congress, is being adopted 
by the Medicare program. We believe that a per capita methodology is a 
useful approach to profiling physicians on their practice efficiency 
and could be part of a feedback program that could also include quality 
measures and episode-based resource use measures. 

Our findings are consistent with those of our previous report on 
physician profiling in which, through analysis of physician practice 
patterns, we determined that CMS could use profiling to improve the 
efficiency of Medicare. Despite a more diverse mix of physician 
specialties in our present analysis, and with certain exceptions noted 
in our findings, we found substantial consistency in certain patterns 
we observed across metropolitan areas and specialties. We also found 
consistency across time in that physicians who showed high resource use 
in one year tended to stay high in the subsequent year. 

Agency and Professional Association Comments and Our Evaluation: 

We provided a draft of this report to the HHS for comment and received 
written comments from CMS, which are reprinted in appendix II. We also 
solicited comments on the draft report from representatives of the 
American Academy of Orthopaedic Surgeons (AAOS), the American College 
of Cardiology (ACC), the American College of Physicians, and the 
American College of Radiology. We received oral comments from the first 
two. 

CMS Comments: 

Our draft report did not include any recommendations for CMS to respond 
to. CMS broadly agreed with each of our three findings: 

* CMS agreed that the per capita methodology is a useful approach to 
measuring physicians' resource use and noted that per capita 
measurement is one of the cost of care measures included in CMS's 
Physician Resource Use Management and Reporting Program. CMS also 
agreed that the consistency of our per capita measure across years is 
an important finding and stated that the agency intends to examine 
measure consistency in the ongoing administration of its program. 

* CMS found the attention in our report to considerations for 
developing a physician feedback system to be particularly helpful. CMS 
listed several examples of how its program already addresses many of 
these considerations and is in the process of addressing others. We 
agree with CMS that some of the approaches described in our report 
would require significant resources and recognize that CMS will need to 
investigate how to balance the trade-offs between different approaches 
in order to best leverage its resources. 

* CMS agreed that physician feedback may have a moderate influence on 
physician behavior. CMS further stated its commitment to developing 
meaningful, actionable, and fair measurement tools for physician 
resource use that, along with quality measures, will provide a 
comprehensive assessment of performance. We continue to believe that 
providing physicians feedback on their performance could be a promising 
step toward encouraging greater efficiency in Medicare; however, we are 
still concerned that efforts to achieve greater efficiency that rely 
solely on physician feedback without financial or other incentives will 
be suboptimal. 

CMS also provided technical comments, which we incorporated as 
appropriate. 

AAOS and ACC comments: 

The representatives of AAOS and ACC raised no major issues with regard 
to the substance of the report. The AAOS representative said that the 
report captured well the key aspects of physician profiling and the key 
considerations in developing physician feedback reports. The ACC 
representatives endorsed the overall approach of a feedback report 
consisting of a high-level summary accompanied by additional sections 
with greater detail and a separate document that explains the 
methodology in detail. The representatives of both groups said that 
physicians should be provided feedback on both quality and resource 
use, but differed on whether they should be presented in the same 
report. Both groups also stressed that physicians should only be 
compared to physicians within their specialty or subspecialty. 

Both the AAOS and the ACC representatives commented on the design of 
our mock report. Both said that the measures of physician resource use 
by type of service and the benchmark comparisons were easy to 
understand. They had difficulty, however, in understanding a related 
measure that shows the physician's share of payments by service 
category. We did not alter our mock report in response to these 
comments, but believe that the concerns they expressed should be taken 
into account by organizations designing physician feedback reports. 

The representatives of both groups stressed the importance of risk 
adjustment in the measurement of physician resource use and suggested 
that we include a fuller explanation of risk adjustment techniques in 
our report. We did not expand our explanation of such techniques 
because they are not the focus of this report; however, we acknowledge 
the important role played by risk adjustment techniques in constructing 
physician feedback reports on resource use. 

As agreed with your office, unless you publicly announce the contents 
of this report earlier, we plan no further distribution until 30 days 
from the report date. At that time, we will send copies of this report 
to the Acting Administrator of CMS, committees, and others. The report 
will also be available at no charge on the GAO Web site at [hyperlink, 
http://www.gao.gov. 

If you or your staff have any questions, please contact me at (202) 512-
7114 or steinwalda@gao.gov. Contact points for our Offices of 
Congressional Relations and Public Affairs may be found on the last 
page of the report. GAO staff who made major contributions to this 
report are listed in appendix III. 

Sincerely yours, 

Signed by: 

A. Bruce Steinwald: 
Director, Health Care: 

[End of section] 

Appendix I: Methodology: 

This appendix describes the per capita methodology that we used to 
measure beneficiaries' and physicians' Medicare fee-for-service (FFS) 
resource use. We focused our analysis on four diverse specialties: a 
medical specialty (cardiology), a diagnostic specialty (diagnostic 
radiology), a primary care specialty (internal medicine), and a 
surgical specialty (orthopedic surgery). We included diagnostic 
radiologists in our study because they are less amenable to episode 
grouping, the major alternative to per capita profiling of physicians. 
We limited our analysis to physicians in these specialties who 
practiced in one of four areas: Miami, Fla.; Phoenix, Ariz.; 
Pittsburgh, Pa.; and Sacramento, Calif.[Footnote 38] We chose these 
areas for their geographic diversity, range in average Medicare 
spending per beneficiary, and number of physicians in each of the four 
specialties. Our results apply only to the four specialties in the four 
metropolitan areas we studied. 

To conduct our analysis, we obtained 2005 and 2006 Centers for Medicare 
& Medicaid Services (CMS) data from the following sources: (1) Medicare 
claims files that include data on physician, durable medical equipment, 
skilled nursing, home health, hospice, and hospital inpatient and 
outpatient services; (2) Denominator File, a database that contains 
enrollment and entitlement status information for all Medicare 
beneficiaries in a given year; (3) Hierarchical Condition Category 
(HCC) files that summarize Medicare beneficiaries' diagnoses; (4) files 
summarizing the institutional status of beneficiaries; and (5) Unique 
Physician Identification Number Directory, which contains information 
on physicians' specialties. 

Adjustment for Differences in Patient Health Status: 

In order to develop a resource use measure that accounts for 
differences in health status between beneficiaries, we developed a risk 
adjustment model that uses an individual's diagnoses during the year to 
estimate the total Medicare FFS expenditures expected for the 
individual in that year. As our inputs to the model, we used the same 
70 HCCs as those in the model CMS uses to set managed care capitation 
rates.[Footnote 39] HCCs are a way of summarizing an individual's 
diagnoses into major medical conditions, such as vascular disease or 
severe head injury.[Footnote 40] To estimate our model, we used HCC and 
expenditure data for 2005 and 2006 five percent national samples of 
Medicare FFS beneficiaries.[Footnote 41],[Footnote 42] 

Methodology Used to Determine Beneficiaries' Resource Use: 

For all Medicare FFS beneficiaries who received at least one service in 
2005 or 2006 from a physician located in any of our four metropolitan 
areas and who also did not meet our exclusion criteria (see footnote 
5), we used our risk adjustment model to estimate their total expected 
Medicare FFS expenditures. Based on their expected expenditures, we 
placed beneficiaries into 1 of 25 discrete risk categories.[Footnote 
43] The categories were ordered in terms of health status from 
healthiest (category 1) to sickest (category 25). Next, within each 
risk category and metropolitan area, we ranked beneficiaries from 1 to 
100 by their total actual annual Medicare expenditures, such that the 
average beneficiary in a given risk category and metropolitan area had 
a rank of 50.[Footnote 44] We used this rank as our risk-adjusted 
measure of beneficiary resource use. 

To examine the stability of beneficiaries' resource use, we divided the 
2005 and 2006 beneficiary populations into five ascending groups of 
nearly equal size (quintiles) based on the level of their resource use. 
[Footnote 45] We then identified beneficiaries in each of the four 
metropolitan areas who saw a physician in their area in 2005 and again 
in 2006. We measured the stability of beneficiaries' resource use as 
the percentage of beneficiaries who remained in the same quintile in 
2006 that they were in during 2005. In addition, we determined the 
percentage of beneficiaries who remained in the highest resource 
quintile. 

Methodology Used to Determine and Compare Physicians' Resource Use: 

For the purposes of this study, we defined a physician's practice as 
all Medicare FFS beneficiaries who did not meet our exclusion criteria 
and who had at least one evaluation and management visit with the 
physician during the calendar year for cardiologists, internists, and 
orthopedic surgeons, or who received any service from the physician for 
diagnostic radiologists.[Footnote 46],[Footnote 47] To ensure that a 
physician's resource use measure would not be overly influenced by a 
few patients with unusually high or low Medicare expenditures, we 
excluded physicians with small practices--those who treated fewer than 
100 of the Medicare patients in our study during the year.[Footnote 48] 
For all physicians, we calculated the average beneficiary resource use 
rank of the patients in their practices,[Footnote 49] which ranged from 
a low of 26.0 to a high of 91.8 in 2006. Next, within each metropolitan 
area and specialty, we ranked physicians on the basis of this average 
from 1 to 100 such that the average measure of physician resource use 
was 50. We used this rank as our measure of physician resource use. 
This measure reflects how expensive a physician's patients are compared 
to the patients of other physicians in the same specialty and area 
after adjusting for differences in patient health status. For example, 
a cardiologist in Miami is only compared to other cardiologists in 
Miami. 

To examine physicians' resource use, we divided the physicians into 
five ascending groups (quintiles) of nearly equal size based on the 
measure of their resource use described above.[Footnote 50] In the same 
manner as we measured the stability of beneficiaries' resource use, we 
measured the stability of physicians' resource use by determining the 
percentage of them who remained in the same physician resource use 
quintile from 2005 to 2006. We also measured the degree of turnover in 
the patients seen by physicians by computing the percentage of patients 
seen in 2005 by each physician that were also seen by the same 
physician in 2006. 

We examined utilization patterns by physician resource use quintile by 
decomposing the 2006 Medicare expenditures of physicians' patients into 
those for institutional services (inpatient hospital and skilled 
nursing care), those for services provided directly by the physician to 
his or her patients, and those for all other services--outpatient 
hospital, home health care, hospice care, durable medical equipment, 
and all other Part B services of Part B providers and suppliers. We 
also measured the number of physicians seen by a physicians' patients 
by physician resource use quintile. 

Although our measure of a beneficiary's resource use is independent of 
the beneficiary's health status, there was an association between 
physician resource use and the mix of healthy and sick patients in 
physicians' practices--physicians who ranked high in terms of resource 
use also treated a larger proportion of beneficiaries who were in poor 
health than did physicians who ranked low in resource use. However, the 
resource use of all their patients was also consistently higher than 
that of low resource use physicians' patients regardless of patient 
health status. Figure 5 shows the average resource use of beneficiaries 
in five health status categories across the five physician resource use 
quintiles.[Footnote 51] For example, patients in the healthiest 
category who were treated by physicians in the highest resource use 
quintile had an average resource use rank of 74, whereas similarly 
healthy patients treated by physicians in the lowest quintile had 
average resource use rank of 53. This ordering of the differences in 
patient resource use by the level of physician resource use is repeated 
across all health categories. It indicates that physicians have 
consistent patterns of resource use with respect to all of their 
patients, regardless of their patients' health status. 

Figure 6: Beneficiary Resource Use by Health Category for Quintiles of 
Physician Resource Use--Four Specialties in Four Metropolitan Areas, 
2006: 

[Refer to PDF for image: line graph] 

Beneficiary health category[A]: 1 (healthiest); 	
Beneficiary resource use: 1 (low resource use): 53; 
Beneficiary resource use: 2: 59; 
Beneficiary resource use: 3: 63; 
Beneficiary resource use: 4: 68; 
Beneficiary resource use: 5 (high resource use): 74. 

Beneficiary health category[A]: 2; 
Beneficiary resource use: 1 (low resource use): 50; 
Beneficiary resource use: 2: 56; 
Beneficiary resource use: 3: 61; 
Beneficiary resource use: 4: 66; 
Beneficiary resource use: 5 (high resource use): 73. 

Beneficiary health category[A]: 3; 
Beneficiary resource use: 1 (low resource use): 49; 
Beneficiary resource use: 2: 55; 
Beneficiary resource use: 3: 59; 
Beneficiary resource use: 4: 64; 
Beneficiary resource use: 5 (high resource use): 72. 

Beneficiary health category[A]: 4; 
Beneficiary resource use: 1 (low resource use): 49; 
Beneficiary resource use: 2: 53; 
Beneficiary resource use: 3: 56; 
Beneficiary resource use: 4: 59; 
Beneficiary resource use: 5 (high resource use): 66. 

Beneficiary health category[A]: 5 (sickest); 
Beneficiary resource use: 1 (low resource use): 50; 
Beneficiary resource use: 2: 52; 
Beneficiary resource use: 3: 54; 
Beneficiary resource use: 4: 57; 
Beneficiary resource use: 5 (high resource use): 63. 

Source: GAO analysis of CMS claims data. 

Note: Beneficiary resource use is averaged across the cardiologists, 
diagnostic radiologists, internists, and orthopedic surgeons in Miami, 
Fla.; Phoenix, Ariz.; Pittsburgh, Pa.; and Sacramento, Calif. who met 
our requirement for a minimum of 100 Medicare patients in their 
practice. 

[A] Each health category consists of 5 risk categories, which span the 
following ranges of risk scores (r): r £ .5, .5 < r £ 1.0, 1.0 < r £ 
2.0, 2.0 < r £ 5.0, 5.0 < r £ 18.0. The first health category includes 
the healthiest beneficiaries and comprises 43.8 percent of the study 
population; the fifth includes the sickest beneficiaries and comprises 
5.5 percent of the study population. The second, third, and fourth 
health categories comprise, respectively, 17.3, 17.4, and 16.1 percent 
of the study population. 

[B] Physicians are divided into five ascending groups of nearly equal 
size based on the level of their resource use, which is based on the 
average level of resource use of their patients. 

[End of figure] 

The mix of healthy and sick patients in physicians' practices did not 
affect the positive relationship we found between average institutional 
expenditures per beneficiary and physician resource use level. Within 
each beneficiary health category, the patients of high resource use 
physicians had average institutional expenditures that exceeded those 
of the patients of physicians with lower resource use. Similar analyses 
showed that patient mix did not affect (1) the positive relationship 
between physicians' resource use and the average number of physicians 
seen by their patients, (2) the positive relationship between 
physicians' resource use and expenditures for all other services 
provided their patients, and (3) the steeper rise in the use of 
institutional services by physicians' patients with increasing 
physician resource use as compared to the rise in the use of all other 
services. 

[End of section] 

Appendix II: Comments from the Centers for Medicare & Medicaid 
Services: 

Department Of Health & Human Services: 
Office Of The Secretary: 
Assistant Secretary for Legislation: 
Washington, DC 20201: 

September 11 2009: 

A. Bruce Steinwald: 
Director, Health Care: 
U.S. Government Accountability Office: 
441 G Street N.W. 
Washington, DC 20548: 

Dear Mr. Steinwald: 

Enclosed are comments on the U.S. Government Accountability Office's 
(GAO) report entitled: "Medicare: Per Capita Method Can Be Used to 
Profile Physicians and Provide Feedback on Resource Use" (GAO-09-802). 

The Department appreciates the opportunity to review this report before 
its publication. 

Sincerely, 

Signed by: 

Andrea Palm: 
Acting Assistant Secretary for Legislation: 

Enclosure: 

[End of letter] 

Department Of Health & Human Services: 
Centers for Medicare and Medicaid Services: 
Administrator: 
Washington, DC 20201: 

September 10, 2009: 

To: A. Bruce Steinwald: 
Director, Health Care: 

From: [Signed by] Charlene Frizerra: 
Acting Administrator: 

Subject: Government Accountability Office's Draft Report: "Per Capita 
Method Can Be Used to Profile Physicians and Provide Feedback on 
Resource Use" (GAO-09-802): 

The Centers for Medicare & Medicaid Services (CMS) appreciates the 
opportunity to review and comment on the Government Accountability 
Office's (GAO) draft report entitled "Per Capita Method Can Be Used to 
Profile Physicians and Provide Feedback on Resource Use." We agree 
that, given the role of physicians in total Medicare spending, there 
are opportunities to increase the efficiency of the Medicare program by 
measuring and reporting on physician resource use. In addition, we 
found the attention in your report to considerations for developing a 
physician feedback system to be particularly helpful. 

As GAO notes, CMS was given the authority to administer a Physician 
Resource Use Measurement and Reporting Program (the "program") by the 
Medicare Improvements for Patients and Providers Act of 2008 (MIPPA). 
CMS has implemented a fully operational program in 13 selected 
geographic sites and has made many of the same conclusions that GAO 
includes in this report. 

The GAO concludes that the "per capita profiling method shows 
specialist physicians' practice patterns are relatively stable over 2 
years." Per capita measurement is one cost of care measure included in 
CMS' program for physicians who met a minimum threshold of at least 20 
patients. Measurement consistency across years is an important finding, 
although CMS notes that the single touch attribution rule GAO used may 
inadvertently inflate the measure's consistency. CMS intends to also 
examine measure consistency in the ongoing administration of the 
program. GAO further concludes that feedback reports should include 
quality measures. CMS has included a regulatory proposal in the 
Calendar Year 2010 Physician Fee Schedule Proposed Rule (74 FR 33591) 
to include quality measures in the program. GAO also recommends that
performance data should be disaggregated into categories. We have 
disaggregated cost data into several categories including inpatient, 
outpatient, home health, and skilled nursing facility services, among 
others. In the report, GAO acknowledges that feedback reports are 
useful at both the individual physician level and the physician group 
level. CMS has included a regulatory proposal (74 FR 33591) to include 
both individual and group level feedback in the program. 

Regarding report dissemination, GAO's environmental scanning advised 
sending feedback reports to all physicians rather than just poor 
performers. To date, we have disseminated reports to all physicians in 
the 13 selected geographic-sites that meet a minimum threshold of 
patients/episodes. Further, GAO recognized the need for both 
dissemination of reports in hard copy and electronically. To date, we 
have only disseminated hard copy reports, but CMS is actively pursuing 
electronic dissemination of reports. GAO concluded that the methodology 
used to compile the reports should be disseminated in a transparent 
fashion, such as a public posting on a web site. GAO also recognized 
that some private insurers have made significant investments in the 
operation of feedback programs; for example one company is utilizing a 
staff of 4 profiling experts and 20 medical directors to support 
physicians' questions about the feedback. We note that these insurers 
are dedicating many more resources per profiled physician than CMS 
currently has available to it. CMS is investigating the feasibility of 
these investments while noting that significant resources would likely 
be needed to fund some approaches. 

Beyond general conclusions, GAO also identified the following key 
points: 

1. Using both per capita and episode of care measures of resource use 
may provide meaningful results by capturing the relevant 
characteristics of a physician's practice pattern; 

2. There are various risk adjustment methodologies, and the suitability 
of a given method will depend on characteristics of the physicians to 
be profiled and their patterns; 

3. A single attribution approach may not be applicable for all types of 
measures or for all types of physician specialties; 

4. There are differing opinions as to what are the most appropriate and 
meaningful comparative benchmarks; and; 

5. There is no consensus on what sample size is adequate to ensure 
meaningful measurement. 

In summary, we applaud GAO's recognition of these five key areas in the 
ongoing management of physician resource use feedback programs. We 
agree with GAO's conclusion that physician resource use feedback 
reports have a moderate influence on physician behavior and the per 
capita methodology is a useful approach to measuring physicians in a 
feedback program that could also include episode-based resource use and 
quality measures. CMS is committed to developing meaningful, 
actionable, and fair measurement tools for physician resource use that, 
along with quality measures, will provide a comprehensive assessment of 
performance under a physician value-based purchasing program. 

[End of section] 

Appendix III: GAO Contact and Staff Acknowledgments: 

GAO Contact: 

A. Bruce Steinwald, (202) 512-7114, or steinwalda@gao.gov. 

Staff Acknowledgments: 

In addition to the contact named above, Phyllis Thorburn, Assistant 
Director; Alison Binkowski; Nancy Fasciano; Richard Lipinski; Drew 
Long; Jessica Smith; Maya Tholandi; and Eric Wedum made key 
contributions to this report. 

[End of section] 

Bibliography: 

Balas, E. Andrew, Suzanne Austin Boren, Gordon D. Brown, Bernard G. 
Ewigman, Joyce A. Mitchell, and Gerald T. Perkoff. "Effect of Physician 
Profiling on Utilization: Meta-analysis of Randomized Clinical Trials." 
Journal of General Internal Medicine, vol. 11, no. 10 (1996): 584-590. 

Beckman, Howard B., Anthony L. Suchman, Kathleen Curtin, and Robert A. 
Greene. "Physician Reactions to Quantitative Individual Performance 
Reports." American Journal of Medical Quality, vol. 21, no. 3 (2006): 
192-199. 

Beckman, Howard B., Thomas Mahoney, and Robert A. Greene. Current 
Approaches to Improving the Value of Care: A Physician's Perspective. 
The Commonwealth Fund. November 2007. 

Hartig, J.R., and Jeroan J. Allison. "Physician Performance 
Improvement: An Overview of Methodologies." Clinical and Experimental 
Rheumatology, vol. 25, supplement 47 (2007): S50-S54. 

Jamtvedt, Gro, Jane M. Young, Doris Tove Kristoffersen, Mary Ann 
O'Brien, and Andrew D. Oxman. "Audit and Feedback: Effects on 
Professional Practice and Health Care Outcomes." Cochrane Database of 
Systematic Reviews, no. 2 (2006). 

Jamtvedt, Gro, Jane M. Young, Doris Tove Kristoffersen, Mary Ann 
O'Brien, and Andrew D. Oxman. "Does Telling People What They Have Been 
Doing Change What They Do? A Systematic Review of the Effects of Audit 
and Feedback." Quality & Safety in Health Care, vol. 15, no. 6 (2006): 
433-436. 

Kiefe, Catarina I., Jeroan J. Allison, O. Dale Williams, Sharina D. 
Person, Michael T. Weaver, and Norman W. Weissman. "Improving Quality 
Improvement Using Achievable Benchmarks for Physician Feedback: A 
Randomized Controlled Trial." JAMA, vol. 285, no. 22 (2001): 2871-2879. 

Marder, Robert J., Mark A. Smith, and Richard A. Sheff. "Changing 
Physician Practice: Providing Physicians Useful Feedback." In Effective 
Peer Review: A Practical Guide to Contemporary Design, 2nd ed., 153- 
164. Marblehead, Mass.: HCPro, Inc., 2007. 

Micklitsch, Christine N. and Theresa A. Ryan-Mitlyng. Physician 
Performance Management: Tool for Survival and Success. Englewood, 
Colo.: Medical Group Management Association, 1996. 

Mold, James W., Cheryl A. Aspy, and Zsolt Nagykaldi. "Implementation of 
Evidence-Based Preventive Services Delivery Processes in Primary Care: 
An Oklahoma Physicians Resource/Research Network (OKPRN) Study." 
Journal of the American Board of Family Medicine, vol. 21, no. 4 
(2008): 334-344. 

Nathanson, Philip. "Influencing Physician Practice Patterns," Topics in 
Health Care Financing, vol. 20, no. 4 (1994): 16-25. 

Pacific Business Group on Health. Advancing Physician Performance 
Measurement: Using Administrative Data to Assess Physician Quality and 
Efficiency. September 2005. 

Paxton, E. Scott, Barton H. Hamilton, Vivian R. Boyd, and Bruce L. 
Hall. "Impact of Isolated Clinical Performance Feedback on Clinical 
Productivity of an Academic Surgical Faculty." Journal of American 
College of Surgeons, vol. 202, no. 5 (2006): 737-745. 

Teleki, Stephanie S., Rebecca Shaw, Cheryl L. Damberg, and Elizabeth A. 
McGlynn. Providing Performance Feedback to Individual Physicians: 
Current Practice and Emerging Lessons. RAND Health Working Paper 
Series. July 2006. 

Van Hoof, Thomas J., David A. Pearson, Tierney E. Giannotti, Janet P. 
Tate, Anne Elwell, Judith K. Barr, and Thomas P. Meehan. "Lessons 
Learned from Performance Feedback by a Quality Improvement 
Organization." Journal for Healthcare Quality, vol. 28, no. 3 (2006): 
20-31. 

Veloski, Jon, James R. Boex, Margaret J. Grasberger, Adam Evans, and 
Daniel B. Wolfson. "Systematic Review of the Literature on Assessment, 
Feedback and Physicians' Clinical Performance: BEME Guide No. 7." 
Medical Teacher, vol. 28, no. 2 (2006): 117-128. 

[End of section] 

Footnotes: 

[1] Medicare is the federally financed health insurance program for 
persons aged 65 and over, certain individuals with disabilities, and 
individuals with end-stage renal disease. Medicare Part A covers 
hospital and other inpatient stays. Medicare Part B covers physician, 
outpatient hospital, home health, and other services. Medicare Parts A 
and B are known as original Medicare or Medicare fee-for-service (FFS). 

[2] Physicians can be profiled both in terms of the resources used in 
providing care to their patients and in terms of the quality of that 
care. In this report, we focus on profiling physicians on their 
resource use, which can be measured in terms of utilization or 
expenditures. 

[3] See GAO, Medicare: Focus on Physician Practice Patterns Can Lead to 
Greater Program Efficiency, [hyperlink, 
http://www.gao.gov/products/GAO-07-307] (Washington, D.C.: April 30, 
2007), 22. 

[4] The Centers for Medicare & Medicaid Services (CMS) is the agency 
within the Department of Health and Human Services (HHS) that oversees 
Medicare. 

[5] See GAO, Medicare: Providing Systematic Feedback to Physicians on 
their Practice Patterns Is a Promising Step Toward Encouraging Program 
Efficiency, [hyperlink, http://www.gao.gov/products/GAO-07-862T] (May 
10, 2007). 

[6] The Medicare Payment Advisory Commission (MedPAC) is an independent 
congressional agency established by the Balanced Budget Act of 1997 to 
advise Congress on issues affecting the Medicare program. 

[7] See Medicare Improvements for Patients and Providers Act of 2008 
(MIPPA), Pub. L. No. 110-275, §131(c), 122 Stat. 2494, 2520-27. 

[8] These areas refer to the following Core-Based Statistical Areas 
(CBSA), an umbrella term for micropolitan and metropolitan statistical 
areas: Miami-Fort Lauderdale-Pompano Beach, Fla.; Phoenix-Mesa- 
Scottsdale, Ariz.; Pittsburgh, Pa.; and Sacramento--Arden-Arcade-- 
Roseville, Calif. For CBSA definitions, see [hyperlink, 
http://www.gao.gov/products/http://www.census.gov/population/www/metroar
eas/metroarea.html]. 

[9] Hierarchical Condition Categories (HCC) collapse the over 15,000 
diagnosis codes into 189 clinically-meaningful condition categories 
which are additionally grouped into hierarchies of increasing severity. 
See appendix I. 

[10] We did not include Part D (drug) expenditures because not all 
beneficiaries are enrolled in a Medicare Part D prescription drug plan. 

[11] Our measure of physicians' resource use therefore includes all 
resources used by their patients, including those ordered by other 
providers. Patients were assigned to a physician if they had at least 
one evaluation and management visit with the physician during the 
calendar year for cardiologists, internists, and orthopedic surgeons, 
or if they received any service from the physician for diagnostic 
radiologists. According to our definition of a physician's practice, a 
beneficiary could belong to the practice of multiple specialists in our 
study. 

[12] These were the most recent data available when we began our study. 

[13] We divided the physicians into five ascending groups (quintiles) 
of nearly equal size based on the measure of their resource use. 

[14] See appendix I for further discussion of our methodology. 

[15] See the bibliography. 

[16] We interviewed officials of four specialty societies: the American 
Academy of Orthopaedic Surgeons, the American College of Cardiology, 
the American College of Physicians, and the American College of 
Radiology. We also interviewed officials of Aetna, Inc.; Cigna 
Corporation; Humana, Inc.; UnitedHealthGroup, Inc.; and WellPoint, Inc. 

[17] See Pub. L. No. 110-275, §131(c). 

[18] For the purposes of this report, we defined institutional services 
as hospital inpatient and skilled nursing facility services. 

[19] We defined beneficiaries' resource use in terms of their resource 
use compared to that of other beneficiaries with similar health 
conditions. Physicians' resource use is derived from beneficiaries' 
resource use. It is defined as the average resource use of those 
Medicare beneficiaries in our study population whom the physician saw 
compared to the average resource use of other physicians' Medicare 
beneficiaries. To determine stability of beneficiaries' resource use, 
we identified beneficiaries who were in our study population in both 
2005 and 2006. To determine stability of physicians' resource use, we 
identified physicians in the four specialties we studied who saw at 
least one of the Medicare beneficiaries in our 2005 study population 
and at least one beneficiary in our 2006 study population. We divided 
physicians and beneficiaries into quintiles according to their resource 
use. See appendix I. 

[20] Increases in per beneficiary expenditures across the physician 
resource use quintiles were accompanied by an increase in the average 
risk score of beneficiaries for all the specialties. However, this 
tendency did not significantly affect our physician resource use 
measure, because the resource use of physicians in higher quintiles was 
higher than that of physicians in lower quintiles for all of the 
patients they saw, regardless of health status. See appendix I. 

[21] The model code was developed in 2007 by the New York State 
Attorney General's office in consultation with medical societies, 
including the American Medical Association, and consumer groups. The 
model code was developed during the course of an investigation by the 
Attorney General's office into insurers' potentially deceptive steering 
of patients to the least expensive physicians under the guise of 
physician ranking programs. As of February 2009, the Attorney General's 
office had settled with eight insurers, instituting reforms designed to 
ensure that ranking programs are based on accurate and transparent 
measures. 

[22] Beckman, et al., "Current Approaches to Improve the Value of 
Care," p. 9. 

[23] Teleki, et al., p.7. 

[24] Officials of two specialty societies also recommended state or 
local comparisons. 

[25] Marder et al., p. 162. 

[26] Bennett, Nancy L., Linda L. Casebeer, Robert E. Kristofco, Sheryl 
M. Strasser. "Physicians' Internet Information-Seeking Behavior." The 
Journal of Continuing Education in the Health Professions, Vol. 24 
(2004), pp. 31-38. 

[27] At the time of our interview, this insurer was profiling and 
providing feedback to physicians in only one specialty. However, 
officials said that the company would continue to hand-deliver results 
to all physicians even after it begins reporting to physicians in 
additional specialties. 

[28] When disseminating information electronically, federal agencies, 
including CMS, must comply with requirements under Section 508 of the 
Rehabilitation Act of 1973 (29 U.S.C. §794d), which requires that 
federal employees and members of the public who are individuals with 
disabilities have access to and use of the information that is 
comparable to the access to and use of the information by federal 
employees and members of the public who are not individuals with 
disabilities. 

[29] Teleki, et al., pp. 5-6. 

[30] These studies varied in terms of the type of feedback provided 
(verbal and/or written; directed to individuals or groups; delivered by 
senior colleagues, professional standards review organizations, or 
other sources), types of clinicians to whom the feedback was delivered 
(physicians, dentists, nurses, or other providers), frequency and 
duration of the feedback, the content, whether the feedback was 
combined with other interventions, and the outcomes studied. 

[31] Veloski, p. 125. 

[32] Jamtvedt, "Audit and Feedback," pp. 5 and 24. 

[33] Two used the information to assign ratings to physicians in the 
provider directories made available to members, two used it to select 
physicians for high-performance networks, and one used it for both 
purposes. 

[34] These data are weighted national estimates from the Community 
Tracking Study Physician Survey conducted by the Center for Studying 
Health System Change. 

[35] See [hyperlink, http://www.gao.gov/products/GAO-07-307]. 

[36] Medicare Quality Improvement Organizations are private 
organizations that contract with CMS to monitor and improve the care 
delivered to Medicare beneficiaries in the 50 states, the territories, 
and the District of Columbia. 

[37] Although another review of the literature concluded that physician 
involvement had little or no impact on the effectiveness of a system in 
changing physician behavior, the researchers acknowledged that this 
finding was unexpected and could be related to a lack of detail in the 
studies they reviewed about the level of physicians' involvement. 

[38] These areas refer to the following Core-Based Statistical Areas 
(CBSA), an umbrella term for micropolitan and metropolitan statistical 
areas: Miami-Fort Lauderdale-Pompano Beach, Fla.; Phoenix-Mesa- 
Scottsdale, Ariz.; Pittsburgh, Pa.; and Sacramento--Arden-Arcade-- 
Roseville, Calif. For CBSA definitions, see [hyperlink, 
http://www.census.gov/population/www/metroareas/metroarea.html]. 

[39] We also included one additional variable to represent 
beneficiaries who did not have any of the included 70 HCCs. 

[40] Hierarchical Condition Categories (HCCs) collapse the over 15,000 
diagnosis codes into 189 clinically meaningful condition categories 
which are additionally grouped into hierarchies of increasing severity. 
If a beneficiary's diagnoses correspond to more than one condition 
within a hierarchy, he or she is assigned only the most severe one. 

[41] We derived our expenditure data from beneficiaries' Part A and 
Part B Medicare FFS claims. We did not include Part D claims because 
not all Medicare beneficiaries are enrolled in a Part D prescription 
drug plan. We made two adjustments to the data used to estimate the 
model: (1) we annualized the expenditures of beneficiaries who died 
during the year and (2) we capped total annual expenditures for all 
beneficiaries at $100,000 in order to reduce the effect of 
beneficiaries with extreme values in the model's estimation. 

[42] We excluded several types of beneficiaries: (1) those who were 
institutionalized for more than 3 consecutive months during the year, 
(2) those who were enrolled in a Medicare Advantage plan for any part 
of the year, (3) those who were newly enrolled in Medicare, and (4) 
those enrolled on the basis of having end-stage renal disease. 

[43] We chose the break points for the risk categories based on 
beneficiaries' risk scores--the ratio of their predicted cost to the 
sample mean. The first 10 risk categories had intervals of 0.1, while 
the subsequent 15 had intervals ranging from 0.2 to 4. We initially 
specified 26 risk categories, but dropped the final one containing 
beneficiaries with risk scores exceeding 18.0 because it contained less 
than 120 beneficiaries in each year. 

[44] We included expenditures from all claims submitted on the 
beneficiary's behalf, including claims from locations outside the four 
selected metropolitan areas and claims from all provider types 
(hospital inpatient, outpatient, physician, durable medical equipment, 
skilled nursing facility, home health, and hospice). We did not include 
Part D prescription drug costs because not all Medicare beneficiaries 
are enrolled in a Medicare Part D prescription drug plan. 

[45] Each beneficiary resource use quintile includes 20 ranks such that 
the first quintile consists of beneficiaries with ranks 1-20 and the 
last quintile consists of beneficiaries with ranks 81-100. 

[46] According to our definition of a physician's practice, a 
beneficiary could belong to the practice of multiple specialists in our 
study. 

[47] We applied this criterion for diagnostic radiologists because they 
typically do not have evaluation and management visits. 

[48] We excluded 28 percent of the physicians in the four specialties 
in 2005 and 29 percent in 2006 because they treated less than 100 
Medicare patients a year. Our analyses included 5,890 physicians in 
2005 and 5,828 in 2006. 

[49] Our measure of physicians' resource use therefore includes all 
resources used by their patients, including those ordered by other 
providers. 

[50] Each physician resource use quintile includes 20 ranks such that 
the first quintile consists of physicians with ranks 1-20 and the last 
quintile consists of physicians with ranks 81-100. 

[51] The five health status categories collapse the 25 risk categories 
into five broader health status categories. Each health status category 
consists of 5 risk categories, which span the following ranges of risk 
scores (r): r £ .5, .5 < r £ 1.0, 1.0 < r £ 2.0, 2.0 < r £ 5.0, 5.0 < r 
£ 18.0. The first health category includes the healthiest beneficiaries 
and comprises, on average, 27 percent of the Medicare patients seen by 
the physicians in our study in 2006; the fifth includes the sickest 
beneficiaries and comprises 16 percent of their Medicare patients. The 
second, third, and fourth health categories comprise, respectively, 14, 
18, and 25 percent of the physicians' Medicare patients. 

[End of section] 

GAO's Mission: 

The Government Accountability Office, the audit, evaluation and 
investigative arm of Congress, exists to support Congress in meeting 
its constitutional responsibilities and to help improve the performance 
and accountability of the federal government for the American people. 
GAO examines the use of public funds; evaluates federal programs and 
policies; and provides analyses, recommendations, and other assistance 
to help Congress make informed oversight, policy, and funding 
decisions. GAO's commitment to good government is reflected in its core 
values of accountability, integrity, and reliability. 

Obtaining Copies of GAO Reports and Testimony: 

The fastest and easiest way to obtain copies of GAO documents at no 
cost is through GAO's Web site [hyperlink, http://www.gao.gov]. Each 
weekday, GAO posts newly released reports, testimony, and 
correspondence on its Web site. To have GAO e-mail you a list of newly 
posted products every afternoon, go to [hyperlink, http://www.gao.gov] 
and select "E-mail Updates." 

Order by Phone: 

The price of each GAO publication reflects GAO’s actual cost of
production and distribution and depends on the number of pages in the
publication and whether the publication is printed in color or black and
white. Pricing and ordering information is posted on GAO’s Web site, 
[hyperlink, http://www.gao.gov/ordering.htm]. 

Place orders by calling (202) 512-6000, toll free (866) 801-7077, or
TDD (202) 512-2537. 

Orders may be paid for using American Express, Discover Card,
MasterCard, Visa, check, or money order. Call for additional 
information. 

To Report Fraud, Waste, and Abuse in Federal Programs: 

Contact: 

Web site: [hyperlink, http://www.gao.gov/fraudnet/fraudnet.htm]: 
E-mail: fraudnet@gao.gov: 
Automated answering system: (800) 424-5454 or (202) 512-7470: 

Congressional Relations: 

Ralph Dawn, Managing Director, dawnr@gao.gov: 
(202) 512-4400: 
U.S. Government Accountability Office: 
441 G Street NW, Room 7125: 
Washington, D.C. 20548: 

Public Affairs: 

Chuck Young, Managing Director, youngc1@gao.gov: 
(202) 512-4800: 
U.S. Government Accountability Office: 
441 G Street NW, Room 7149: 
Washington, D.C. 20548: