Medicare: HCFA Can Improve Methods for Revising Physician Practice
Expense Payments (Letter Report, 02/27/98, GAO/HEHS-98-79).

Pursuant to a legislative requirement, GAO reviewed the Health Care
Financing Administration's (HCFA) proposed practice expense revisions
and its ongoing efforts to refine its data and methodologies, focusing
on: (1) HCFA's approach for estimating the practice expenses directly
associated with each medical service or procedure; (2) two methodologies
HCFA used to adjust the direct expense estimates; (3) practice expenses
excluded or limited by HCFA; (4) HCFA's method for assigning indirect
practice expenses to each medical service or procedure; and (5) the
potential impact of the new fee schedule allowances on beneficiary
access to care.

GAO noted that: (1) HCFA used expert panels--consisting of physicians,
administrators, and nonphysician clinicians--to estimate the direct
labor and other direct practice expenses associated with medical service
or procedures; (2) GAO explored alternative primary data gathering
methods, such as mailing out surveys, using existing survey data, and
gathering data on-site; (3) GAO concluded that each of those methods has
practical limitations that preclude its use as a reasonable alternative
to HCFA's use of expert panels; (4) gathering data directly from a
limited number of physician practices would, however, be a useful
external validity check on HCFA's practice expense rankings and would
also help HCFA identify refinements needed during phase-in of the fee
schedule revision; (5) the panels' estimates of direct practice expenses
needed several types of adjustment; (6) GAO found problems, however,
with one of HCFA's adjustment methods, which substantially altered the
practice expense rankings; (7) specifically, HCFA used a statistical
model to reconcile significant differences among various panels'
estimates for the same procedure; (8) GAO identified technical
weaknesses in the model that may have biased the estimates; (9) HCFA
excluded some physician practice expenses from the panels' estimates
because it believes that Medicare pays for those expenses through other
mechanisims; (10) physician groups, however, argue that shifts in
medical practices have resulted in physicians absorbing these expenses;
(11) HCFA also placed upper limits on the panels' administrative and
clinical labor estimates; and although these limits seem reasonable to
HCFA, they are not supported by any data or analysis; (12) HCFA's method
for assessing indirect expenses to medical procedures is
acceptable--there is no single best way on the basis of each procedure's
total relative value units (RVU) for physician work, direct practice
expenses, and malpractice expenses, factors that likely reflect some of
the variation on the ratio between direct and indirect expenses among
physician specialties; (13) however, the survey data collected by a
physician organization might provide more straightforward estimates of
specialty--specific indirect cost ratios, and that organization is
willing to expand its survey for HCFA's use; and (14) the 1992
implementation of the fee schedule resulted in lower Medicare payments
to some medical specialities, but subsequent studies found that Medicare
beneficiaries' access to care remained very good.

--------------------------- Indexing Terms -----------------------------

 REPORTNUM:  HEHS-98-79
     TITLE:  Medicare: HCFA Can Improve Methods for Revising Physician 
             Practice Expense Payments
      DATE:  02/27/98
   SUBJECT:  Health care programs
             Medical fees
             Malpractice (medical)
             Medical services rates
             Proposed legislation
             Health care costs
             Medical economic analysis
             Physicians
             Overhead costs
             Statistical methods
IDENTIFIER:  Medicare Program
             
******************************************************************
** This file contains an ASCII representation of the text of a  **
** GAO report.  Delineations within the text indicating chapter **
** titles, headings, and bullets are preserved.  Major          **
** divisions and subdivisions of the text, such as Chapters,    **
** Sections, and Appendixes, are identified by double and       **
** single lines.  The numbers on the right end of these lines   **
** indicate the position of each of the subsections in the      **
** document outline.  These numbers do NOT correspond with the  **
** page numbers of the printed product.                         **
**                                                              **
** No attempt has been made to display graphic images, although **
** figure captions are reproduced.  Tables are included, but    **
** may not resemble those in the printed version.               **
**                                                              **
** Please see the PDF (Portable Document Format) file, when     **
** available, for a complete electronic file of the printed     **
** document's contents.                                         **
**                                                              **
** A printed copy of this report may be obtained from the GAO   **
** Document Distribution Center.  For further details, please   **
** send an e-mail message to:                                   **
**                                                              **
**                                            **
**                                                              **
** with the message 'info' in the body.                         **
******************************************************************


Cover
================================================================ COVER


Report to Congressional Committees

February 1998

MEDICARE - HCFA CAN IMPROVE
METHODS FOR REVISING PHYSICIAN
PRACTICE EXPENSE PAYMENTS

GAO/HEHS-98-79

Medicare Physician Fee Schedule

(101590)


Abbreviations
=============================================================== ABBREV

  AMA - American Medical Association
  CPEP - clinical practice expense panel
  CPT - Current Procedural Terminology
  HCFA - Health Care Financing Administration
  HHS - Department of Health and Human Services
  MedPAC - Medicare Payment Advisory Commission
  MGMA - Medical Group Management Association
  PPRC - Physician Payment Review Commission
  RUC - Relative Value Update Committee
  RVU - relative value unit
  SMS - Socioeconomic Monitoring System

Letter
=============================================================== LETTER


B-278530

February 27, 1998

The Honorable William V.  Roth, Jr.
Chairman
The Honorable Daniel Patrick Moynihan
Ranking Minority Member
Committee on Finance
United States Senate

The Honorable Tom Bliley
Chairman
The Honorable John D.  Dingell
Ranking Minority Member
Committee on Commerce
House of Representatives

The Honorable Bill Archer
Chairman
The Honorable Charles B.  Rangel
Ranking Minority Member
Committee on Ways and Means
House of Representatives

The Medicare physician fee schedule specifies the payments to
physicians for more than 7,000 services and procedures, ranging from
a routine office visit to surgical removal of a brain tumor. 
Medicare's physician fee schedule payments totaled about $43 billion
in 1997.\1 Since many other insurers base their payments on
Medicare's, the fee schedule allowances also influence physicians'
non-Medicare income. 

Before implementation of the physician fee schedule in 1992, Medicare
based payments on each physician's charges and the charges of other
physicians in the same locality.  The fee schedule system was
instituted to relate payments to the resources physicians use to
provide a service, rather than to physicians' charges for a service. 
For each of three categories of resources--physician work,\2 practice
expenses, and malpractice expenses--each medical procedure is ranked
against all other procedures according to the amount of resources
used.  The fee schedule allowance for a procedure equals the sum of
the three rankings, expressed as relative value units (RVU),
multiplied by a conversion factor that translates RVUs into
dollars.\3 Currently, only the physician work RVUs, which account for
about half the total RVUs for each procedure, are resource based. 
The practice expense and malpractice expense RVUs, which account for
about 41 percent and 5 percent, respectively, of the fee schedule
allowances, are still based on historical charges for physician
services.  The reason for this is that a new system for calculating
resource-based values for these components was not available at the
time the fee schedule was implemented by the Health Care Financing
Administration (HCFA).  (App.  I provides additional information on
the Medicare fee schedule.)

The Social Security Amendments of 1994\4 required the Secretary of
Health and Human Services (HHS) to revise the fee schedule by 1998 so
that the practice expense RVUs would reflect the relative amount of
applicable resources physicians expend when they provide a service or
perform a procedure.  The Congress required that the revision be
budget neutral, so total Medicare payments to physicians for practice
expenses would not change.  However, Medicare payments could increase
for some services and procedures and decrease for others. 
Furthermore, depending upon their mix of services and procedures,
members of different physician specialties could experience gains or
losses in their Medicare payments. 

On June 18, 1997, HCFA published a notice of proposed rulemaking in
the Federal Register that described HCFA's proposed practice expense
revisions to the fee schedule.  HCFA estimated that the revisions
would generally increase Medicare payments to physician specialties
that provide more office-based services.  For example, HCFA estimated
that Medicare payments to family practice physicians would increase
by 12 percent, and payments to thoracic surgeons, who perform more
hospital-based procedures, would decrease by 28 percent.  In total,
HCFA estimated that its revisions, had they been in effect in fiscal
year 1997, would have reallocated $2 billion of the $18 billion
Medicare paid for physician practice expenses that year.  The
revisions could also affect physicians' non-Medicare income if other
insurers adopt the Medicare revisions. 

Some physician groups argued that HCFA based its proposed revisions
on invalid data and that the reallocations of Medicare payments would
be too severe.  Subsequently, the Congress included provisions in the
Balanced Budget Act of 1997 that delay the resource-based practice
expense revisions until 1999; establish a 3-year phase-in period for
the revisions; and require HCFA to publish a revised proposed rule by
May 1, 1998.\5 The act also required us to evaluate HCFA's June 1997
proposed rule, report to the Congress within 6 months, and consult
with representatives of physicians' organizations during our
evaluation.\6

This report provides our evaluation of HCFA's proposed practice
expense revisions and also includes information on HCFA's ongoing
efforts to refine its data and methodologies.  Specifically, this
report focuses on (1) HCFA's approach for estimating the practice
expenses directly associated with each medical service or procedure,
(2) two methodologies HCFA used to adjust the direct expense
estimates, (3) practice expenses excluded or limited by HCFA, (4)
HCFA's method for assigning indirect practice expenses to each
medical service or procedure, and (5) the potential impact of the new
fee schedule allowances on beneficiary access to care. 

We reviewed HCFA's proposed revisions, comments received on the
proposal, and selected documentation on the data and methodologies
HCFA and its contractor used.\7 We also held extensive meetings with
HCFA staff to gain an understanding of the methodologies they used
and the rationale behind some of their key decisions and assumptions. 
We did not gather new data on physician practice expenses, test the
reliability of HCFA's data, or independently verify HCFA's data
sources or calculations; but we did our own calculations, using
HCFA's data, to analyze some aspects of HCFA's methodology.  We also
met with researchers, representatives of physician organizations, and
others to obtain their views on HCFA's proposal and to discuss
potential alternative data sources and methodologies.  We performed
our evaluation from August through December 1997 in accordance with
generally accepted government auditing standards.  The physician
groups and others that we met with are listed in appendix II. 


--------------------
\1 For each service or procedure, Medicare pays 80 percent of the
allowed amount set by the fee schedule and Medicare patients are
responsible for the remaining 20 percent.  In this report, we refer
to the Medicare fee schedule allowance as the "Medicare payment."

\2 Physician work is based on the time the physician spends, the
intensity of effort and level of skill required, and stress as a
result of the risk of harm to the patient. 

\3 The fee schedule allowances are also adjusted for differences in
local costs using geographic practice cost indexes. 

\4 Sec.  121, P.L.  103-432, 108 Stat.  4398, 4408, Oct.  31, 1994. 

\5 Sec.  4505(a),(b), and (d), P.L.  105-33, 111 Stat.  251, 435, and
436, Aug.  5, 1997. 

\6 Sec.  4505(c), P.L.  105-33, 111 Stat.  251, 435, Aug.  5, 1997. 

\7 HCFA contracted with Abt Associates to help gather the data HCFA
used to develop new practice expense RVUs. 


   RESULTS IN BRIEF
------------------------------------------------------------ Letter :1

HCFA used expert panels--consisting of physicians, administrators,
and nonphysician clinicians--to estimate the direct labor and other
direct practice expenses associated with medical services or
procedures, and we found this to be an acceptable method.  We
explored alternative primary data gathering methods, such as mailing
out surveys, using existing survey data, and gathering data on-site;
we concluded that each of those methods has practical limitations
that preclude its use as a reasonable alternative to HCFA's use of
expert panels.  Gathering data directly from a limited number of
physician practices would, however, be a useful external validity
check on HCFA's practice expense rankings and would also help HCFA
identify refinements needed during phase-in of the fee schedule
revision. 

The panels' estimates of direct practice expenses needed several
types of adjustment.  We found problems, however, with one of HCFA's
adjustment methods, which substantially altered the practice expense
rankings.  Specifically, HCFA used a statistical model to reconcile
significant differences among various panels' estimates for the same
procedure (for example, hernia repair).  We identified technical
weaknesses in the model that may have biased the estimates. 

HCFA excluded some physician practice expenses from the panels'
estimates because it believes that Medicare pays for those expenses
through other mechanisms.  Physician groups, however, argue that
shifts in medical practices have resulted in physicians absorbing
these expenses.  HCFA also placed upper limits on the panels'
administrative and clinical labor estimates; and although these
limits seem reasonable to HCFA, they are not supported by any data or
analysis. 

HCFA's method for assigning indirect expenses, such as the cost of
general office supplies, to medical procedures is acceptable--there
is no single best way to make such assignments.  HCFA assigned
indirect expenses on the basis of each procedure's total RVUs for
physician work, direct practice expenses, and malpractice expenses,
factors that likely reflect some of the variation in the ratio
between direct and indirect expenses among physician specialties. 
However, the survey data collected by a physician organization might
provide more straightforward estimates of specialty-specific indirect
cost ratios, and that organization is willing to expand its survey
for HCFA's use. 

At this point, we do not know how the final revisions to the fee
schedule will alter Medicare payments to physicians, or whether those
changes will affect beneficiary access to physician services.  The
1992 implementation of the fee schedule resulted in lower Medicare
payments to some medical specialties, but subsequent studies found
that Medicare beneficiaries' access to care remained very good. 
However, the cumulative effect of the 1992 changes, the upcoming
practice expense revisions, and other changes to Medicare's fee
schedule could alter physicians' willingness to accept Medicare's fee
schedule payments for some procedures.  Therefore, there is a
continuing need to monitor indicators of beneficiary access to care,
focusing on services and procedures with the greatest reductions in
Medicare payments. 


   BACKGROUND
------------------------------------------------------------ Letter :2

A resource-based, relative-value payment system ranks procedures on a
common scale, according to the resources used for the procedure.  The
need to estimate and rank practice expenses for thousands of medical
procedures, coupled with the complex structure of the Medicare
program, presented HCFA with some enormous challenges. 

Cost data for physician practices are available in the aggregate and
for individual items, including the wages and salaries of
receptionists, nurses, and technicians employed by the physician; the
cost of office equipment such as examining tables, instruments, and
diagnostic equipment; the cost of supplies such as face masks and
wound dressings; and the cost of billing services and office space. 
For most physician practices, the total of each of these expenses is
probably readily available.  However, Medicare pays physicians by
procedure, such as a skin biopsy.  Therefore, HCFA had to develop a
way to estimate the portion of practice expenses associated with each
procedure--information that is not readily available. 

Another difficulty inherent in developing a resource-based fee
schedule is that significant variations in practice expenses exist
among physicians and their practice settings.  For example, a general
practice physician in a solo practice likely has expenses different
from those of a physician in a group practice. 

Both these problems--the difficulty in allocating expenses to
procedures and the variation in expenses among physician
practices--are mitigated somewhat by basing the fee schedule
allowances on relative rankings of practice expenses.  Even though
the absolute expenses associated with a procedure cannot be precisely
measured and will vary among physicians, the expense of one procedure
relative to another is easier to estimate and less likely to vary
across physician practices. 

Because of these issues, revising the practice expense component of
the fee schedule has been a difficult task for HCFA, and the
revisions HCFA proposed in June 1997 have been the subject of
widespread controversy among physician specialty groups.  This
controversy is not unexpected, since the legislative requirement that
fee schedule changes be budget neutral means that some physician
specialty groups would benefit from the changes at the expense of
other groups.  Among the physician specialty groups that we met with,
groups representing physicians whose Medicare payments were projected
to increase generally supported HCFA's proposal, while groups whose
members' payments were projected to decrease were far more critical
of HCFA's approach and methodology. 


      OVERVIEW OF HCFA'S
      METHODOLOGIES
---------------------------------------------------------- Letter :2.1

In the spring of 1994, HCFA staff and leading researchers discussed
potential approaches for developing detailed estimates of the
practice expenses associated with each medical procedure.  On the
basis of these discussions, HCFA decided to use a variety of data
gathering and mathematical methods to estimate the direct and
indirect practice expenses at the procedure level.  HCFA used those
estimates as the basis for its June 1997 proposed revision to the
physician fee schedule.  The key aspects of HCFA's approach were the
use of (1) expert panels to estimate direct costs, (2) a series of
HCFA's own methods to adjust the direct cost estimates, and (3)
another HCFA method to allocate indirect expenses to procedures. 
Since then, HCFA has continued to refine both its data and its
methodologies. 


      DEVELOPING DIRECT EXPENSE
      ESTIMATES
---------------------------------------------------------- Letter :2.2

To estimate the direct expenses for individual procedures, HCFA
convened 15 clinical practice expense panels (CPEP), organized by
specialty.\8 Each panel included 12 to 15 members.  About half the
members of each panel were physicians, and the remaining members were
practice administrators and nonphysician clinicians such as nurses. 
HCFA provided national medical specialty societies an opportunity to
nominate the panelists, and over 60 specialties and subspecialties
were represented on the panels. 

HCFA and contractor staff identified more than 6,000 procedure codes
for which direct expense data would be developed.\9 The contractor
arranged these codes into families of clinically related codes. 
Within each family of codes, the contractor also selected "reference
codes" that formed a basis for ranking the other codes within the
family.  HCFA and its consultants reviewed the contractor's work and
made the family grouping and reference code information available to
medical specialty societies for their review and comment.  Some
codes, called "redundant codes," were assigned to two or more CPEPs
so that HCFA and its contractor could analyze differences in the
estimates developed by the various panels.  For example, HCFA
included the repair of a disk in the lower back among the procedures
reviewed by both the orthopedic and neurosurgery panels.\10

Within each panel, the members attempted to reach consensus on the
type and quantity of nonphysician labor, medical equipment, and
medical supplies required to perform each procedure.  The CPEP
members were instructed to base their estimates on the typical
patient--the patient who most frequently undergoes a particular
procedure--not necessarily a Medicare patient.  For example, most
women receiving hysterectomies are in their 40s and 50s and are not
Medicare patients. 

Once the CPEPs completed their work, HCFA's contractor calculated the
dollar costs of the labor, medical equipment, and medical supplies
that the CPEPs had estimated for each procedure.  Nonphysician labor
costs were calculated using about 100 occupational categories of
clinical and administrative staff.  For example, if a CPEP estimated
that a procedure required 10 minutes of a registered nurse's time,
the associated cost was calculated from the salaries and benefits
paid to registered nurses.  For equipment, the contractor estimated
costs using the price of the item, an applicable finance rate, and a
depreciation schedule for that item based on an assumed utilization
rate.  For medical supplies, the contractor identified a
representative cost for each item on the basis of information from
catalogues, suppliers, and CPEP members. 


--------------------
\8 For example, one panel reviewed general surgery codes, while
another reviewed orthopedic codes. 

\9 The Current Procedural Terminology (CPT), compiled by the American
Medical Association, is used by the Medicare program and most other
payers to identify, classify, and bill for medical procedures.  It
consists of procedure codes, descriptions, and modifiers to
facilitate billing and payment for medical services and procedures
performed by physicians.  When the terms "code" and "procedure code"
are used in this report, they refer to CPT codes. 

\10 This was procedure code 63030. 


      ADJUSTING DIRECT EXPENSE
      ESTIMATES
---------------------------------------------------------- Letter :2.3

For several reasons, HCFA applied a series of adjustments to the
direct expenses estimated by the CPEPs.  First, HCFA reviewed the
data to ensure that the costs arrived at were allowable under
Medicare policy and revised the costs where necessary.  Next, HCFA
used a process called "linking" to convert the direct expense
estimates of the different CPEPs to a common scale.  HCFA also
adjusted the revised estimates for labor, equipment, and supply costs
to make them consistent with national practice expense data collected
by the American Medical Association (AMA), a process that HCFA called
"scaling." Lastly, HCFA adjusted estimates that appeared to be
unreasonable.  After making all these adjustments, HCFA used a
mathematical process to rank the procedures by direct expenses and
convert the rankings into direct expense RVUs. 


      DEVELOPING INDIRECT EXPENSE
      ESTIMATES
---------------------------------------------------------- Letter :2.4

Indirect expenses, such as the cost of general office supplies and
utilities, cannot be associated with specific medical procedures;
therefore, HCFA had to develop a method to identify and assign
indirect expenses to the procedures.  HCFA originally intended to
survey a random sample of approximately 5,000 physician practices,
obtain data on their direct and indirect practice expenses, and use
those data to develop a method for allocating indirect expenses. 
However, HCFA abandoned this effort because few practices responded
with the detailed information HCFA requested. 

After exploring several alternatives, HCFA decided to allocate
indirect expenses to procedures on the basis of the physician work,
direct practice expense, and malpractice expense RVUs associated with
the procedure.  Thus, procedures that ranked high in each of these
categories were allocated proportionately more indirect expenses. 


      ONGOING REFINEMENT OF HCFA'S
      DATA AND METHODOLOGIES
---------------------------------------------------------- Letter :2.5

Since publication of the proposed fee schedule revisions, HCFA has
been reviewing the comments on its proposal and reexamining its data
and methodologies.  In October 1997, HCFA convened "validation"
panels, composed primarily of physicians, to review the direct costs
estimated by the original CPEPs for several hundred procedures and
revise those estimates as they believed necessary.  In November 1997,
HCFA asked representatives of physician groups to review and comment
on its methodology for estimating and assigning indirect expenses. 
Then, in December 1997, HCFA convened a "cross-specialty" panel
comprising 38 members nominated by various medical specialty
societies.  HCFA asked this panel to develop standard time estimates
for selected administrative tasks.  HCFA officials said they may
change the methods they used to convert the direct expense estimates
to a common scale as well as the method they used to compute and
assign indirect expenses. 


   HCFA USED AN ACCEPTABLE
   METHODOLOGY TO DEVELOP DIRECT
   EXPENSE ESTIMATES
------------------------------------------------------------ Letter :3

Ideally, estimates of the relative resources associated with each
medical procedure would be based on resource data obtained from a
broad, representative sample of physician practices.  However, the
feasibility of completing such an enormous data collection task
within reasonable time and cost constraints is doubtful, as evidenced
by HCFA's unsuccessful attempt to survey 5,000 practices.  After
considering this option and the limitations of survey data already
gathered by other organizations, HCFA decided to use expert panels to
estimate the relative resources associated with medical procedures. 
Various medical specialty and physician groups, however, have
criticized this methodology.  They advocate using other methodologies
as the basis for the fee schedule revisions--methodologies that have
their own limitations.  At this time, HCFA is considering what, if
any, changes it will make to the CPEP data before incorporating these
data into its next proposed rule. 


      SOME SUPPORT THE CPEP
      METHODOLOGY
---------------------------------------------------------- Letter :3.1

Researchers we contacted who specialize in physician reimbursement
issues support HCFA's use of CPEPs to estimate direct practice
expenses.  Generally, they believe that bringing together
knowledgeable physicians, practice administrators, and clinical staff
to identify the direct inputs used in providing a service or
procedure is an acceptable, cost-effective approach.  The Physician
Payment Review Commission (PPRC),\11 which advises the Congress on
health care policy issues, also supports HCFA's approach and asserts
that the CPEP process is an adequate way to collect credible direct
expense data for use in developing practice expense RVUs. 

Some of the national medical societies we met with also support the
CPEP process.  For example, AMA representatives told us that they
believe the use of expert panels is an acceptable method for
estimating the direct expenses associated with a procedure.  Also,
American Society of Internal Medicine representatives said that the
feedback they received indicated that their members thought the CPEP
meetings were an effective way to develop direct expense data for
procedures.  The American Academy of Family Physicians has also been
supportive of HCFA's use of practice expense panels. 


--------------------
\11 In Oct.  1997, the Medicare Payment Advisory Commission (MedPAC)
replaced PPRC.  Throughout this report, we generally refer to PPRC,
rather than to MedPAC, because we relied on materials previously
issued by PPRC and PPRC was in existence at the time we began our
review. 


      OTHER GROUPS CRITICIZE THE
      CPEP METHODOLOGY
---------------------------------------------------------- Letter :3.2

Some medical specialty groups have criticized HCFA's design and
implementation of the CPEP process.  As discussed below, these
criticisms focus on three issues. 

First, some physician groups stated that the panels were not
representative of the different practice settings or types of
physicians who provide a particular service.  That is, the panels did
not contain a broad spectrum of small and large practices or those
from rural and urban areas, nor did they include representatives from
all of the medical specialties.  This criticism ignores the efforts
HCFA made, working closely with the major medical societies, to
constitute representative panels.  HCFA asked the societies to
nominate physicians and others who the medical societies believed
could appropriately represent their membership.  Although the number
of individuals representing each medical specialty was small,
collectively the panels included about 180 physicians from 61 medical
specialties and subspecialties, and those physicians worked in
different types and sizes of practices.  The benefits to be gained by
further expanding the panels could be outweighed by the problems that
would be encountered in structuring and moderating much larger
groups. 

Second, some groups believe the CPEP data are invalid because they
represent the "best guesses" of physicians rather than actual
practice expense data.  This criticism implies that the panelists
lacked the knowledge to make informed judgments about the
nonphysician labor, medical supplies, and equipment associated with
individual procedures and that they were not prepared to participate
in the panel discussions.  However, panelists based their judgments
on a wide variety of factors, such as their knowledge of their own
practices and results of surveys conducted by medical specialty
societies; this demonstrates that the panelists' collective knowledge
about practice expenses was broader than their individual knowledge
about their own practices. 

Third, the CPEP process has been criticized because some of the
panels used different assumptions and definitions than other panels,
leading to differences in the resources identified by different
panels for the same procedures.  For example, one panel's estimates
included time for staff to resubmit denied claims, while another
panel did not include this activity in its estimates.  In another
case, two panels differed on the type of patient on which to base
their estimates; one considered "typical" patients, while the other
considered "problem" patients.  HCFA officials acknowledge these
differences but note that each panel was consistent internally when
it identified the resources associated with individual procedures. 
Because different panels identified different resources for the
redundant procedures, and because HCFA believed that the panels used
different scales to rank their procedures, HCFA made a number of
adjustments to the data.  The adjustments HCFA made are addressed
separately in this report. 


      OTHER METHODS FOR ESTIMATING
      DIRECT EXPENSES HAVE
      LIMITATIONS
---------------------------------------------------------- Letter :3.3

Several medical specialty groups have recommended that HCFA use
actual practice expense data, rather than the CPEP process, as the
basis for estimating direct expenses and calculating RVUs.  They
propose that HCFA obtain these data by either asking physician
practices to complete a survey instrument mailed to them or
collecting data on-site.  These approaches have their own
limitations, some of which they share with the CPEP process. 
Starting over and using one of these approaches as the primary means
for developing direct expense estimates, we believe, would needlessly
increase costs and further delay implementation of the fee schedule
revisions. 

Surveys to gather procedure-level information require either the
physician or other practice staff to use their best judgment to
identify the expenses associated with procedures, since cost
accounting systems do not allocate expenses in this manner.  Also,
once the results of a survey are received, analysts must adjust the
data in order to make valid comparisons among respondents, since
responses will likely vary by the type and size of responding
practice.  The CPEP process shares both these problems--CPEP members
were asked for their best judgment on the resources associated with
specific procedures, and HCFA staff adjusted the CPEP data to ensure
consistency in the data reported by different panels. 

A greater problem with surveys of physicians and physician practices
is low response rates.  HCFA and its contractor developed a lengthy
survey that asked physician practices to provide detailed data on
their direct and indirect expenses for the procedures they billed to
Medicare.  For example, the survey asked practices to provide
information on the number and types of clinical and administrative
staff they employed, the frequency with which they provided each
service in a year, and the square footage of office space that they
leased or purchased.  After an initial test involving approximately
1,700 practices, HCFA canceled the survey because only about a
quarter of the practices surveyed responded. 

Low response rates have also been encountered in national surveys
conducted by the AMA and the Medical Group Management Association
(MGMA).  The AMA's annual Socioeconomic Monitoring System (SMS)
survey polls about 4,000 physicians and asks them for information on
a number of topics; the survey includes eight general questions about
practice expenses.  For example, the survey asks respondents for
their share of the practice's total nonphysician payroll expenses and
total expenses.  About one-third of the respondents are resurveyed
the following year.  Generally, about 40 percent of the physicians
surveyed respond and provide complete information on the practice
expense questions.  MGMA collects practice expense data through a
membership survey.  Its 12-page survey instrument asks for
information on a practice's current assets and liabilities; operating
costs; total number of patients treated in a year; and percentage of
income from Medicare, Medicaid, and managed care plans.  The response
rate for this survey is less than 30 percent. 

Moreover, estimates derived from existing or specially designed
surveys may be biased because the respondents may not be
representative of the broader population of physician practices.  For
example, the practice expense portion of the AMA's SMS survey is
addressed only to self-employed physicians--about 60 percent of the
total physician population, according to AMA officials.  Also, this
survey is not sufficiently large to statistically project estimates
to smaller medical specialties.  Similarly, MGMA's survey does not
cover all types of physician practices and is sent only to MGMA
members.  Regarding HCFA's own attempt to survey practices, HCFA
officials told us they were concerned that the respondents may be
primarily from larger practices.  As a result, any estimates derived
from the survey would likely not have reflected the expenses incurred
by smaller practices. 

Gathering expense data on-site at physician practices also has
limitations.  In the early 1990s, PPRC contracted with three
multispecialty group practices to collect detailed data on the
resources they used in delivering selected services.  Staff at the
practices filled out schedules showing the activities they performed
and the time spent on these activities.  The practices also developed
information on the medical equipment and supplies used for particular
procedures.  The problem with this approach is its cost.  PPRC spent
about $135,000 collecting data at just one multispecialty group
practice.  It is unlikely that HCFA could fund this type of data
collection effort in sufficient magnitude to use the information as
its primary data source.  And basing new practice expense RVUs on
data gathered at only a few practices would raise legitimate concerns
about adequate representation of the diverse range of physician
practices. 

Several medical specialty societies have urged HCFA to gather
practice expense data using activity-based costing, a method that was
developed for use by manufacturing companies in the 1980s and has
subsequently been applied in some health care organizations.  Using
this approach to identify costs for individual procedures involves
several steps.  First, all of a practice's costs are obtained from
information such as financial statements and tax returns.  Next, the
major processes within a practice, such as serving patients in the
office, maintaining medical records, and billing, are identified by
conducting on-site interviews and having staff complete worksheets
that capture the time staff spend on their daily activities.  Costs
are then identified with, or allocated to, each process and, in turn,
to the procedure codes associated with the process.  Activity-based
costing generally does not distinguish between individual procedure
codes; rather, it groups codes together and then assigns costs to the
group of codes.  As a result, it does not provide the specificity
needed to adjust the Medicare fee schedule. 


      HCFA'S PLAN FOR REFINING THE
      CPEP DATA IS UNCLEAR
---------------------------------------------------------- Letter :3.4

After the June 1997 publication of HCFA's proposed fee schedule
revisions, HCFA used additional expert panels to review some of the
data obtained from the 15 original panels.  In October 1997, HCFA
convened 17 "validation" panels composed primarily of physicians and
again organized by specialty.  Each panel was assigned between 14 and
36 procedures for which to review the original nonphysician labor,
equipment, and supply estimates.  The validation panels generally
increased the estimates of nonphysician labor identified by the
original CPEPs.  Then, in December 1997, HCFA convened a
cross-specialty panel that included representatives from all the
major medical specialty societies and medical directors from HCFA's
claims processing contractors.  HCFA asked this panel to reach
consensus on the direct labor estimates for 57 high-cost, high-volume
procedures previously reviewed by the CPEP and validation panels. 
The cross-specialty panel failed to reach consensus, but it did
provide HCFA with some insight into the reasons for differences among
the estimates of previous panels. 

At this time, it is unclear what approach HCFA will take in preparing
its next proposed role, which is due in May 1998.  HCFA has not yet
decided whether to rely on the direct cost estimates developed by the
original CPEPs, the refinements made by the validation panels, or its
own adjustments to the data (which have come under strong criticism
from some specialty groups).  Various physician groups have advocated
that HCFA validate its CPEP data through means other than the expert
panel process.  Given HCFA's time and resource constraints, using
surveys and on-site data gathering methodologies as the primary means
to estimate the direct expenses associated with procedures is not
practicable.  However, these methods could be used on a limited basis
to check the basic accuracy of the CPEP data.  HCFA could conduct a
small number of on-site reviews, similar to PPRC's approach, to test
the validity of the CPEP data.  Gathering some direct expense data
through either surveys or on-site reviews would enable HCFA to
identify any egregious problems with the direct expense relative
rankings and help focus its efforts on correcting those problems. 
HCFA officials told us that they are considering such a check of the
CPEP data but have not yet finalized their decision. 


   CPEP ESTIMATES NEED ADJUSTMENT,
   BUT ONE OF HCFA'S ADJUSTMENTS
   RAISES QUESTIONS
------------------------------------------------------------ Letter :4

HCFA staff believed that each of the CPEPs developed reasonable
relative rankings of its assigned procedures but found that for some
procedures the labor estimates often varied considerably across CPEPs
for the same procedures.  These observed variations in estimates
suggest that adjustments to the CPEP data are necessary.  To correct
for these variations, HCFA used an adjustment process referred to as
"linking" to place the estimates on a common scale.  Although we
consider linking to be desirable, we found that certain features of
the CPEP data cast doubt on HCFA's particular linking model.  HCFA
also adjusted the CPEP data so that they were consistent in the
aggregate with national practice expense data developed from the
AMA's SMS survey--a process HCFA calls "scaling." It is not clear
whether HCFA will use linking in its next proposed rule. 


      EFFECTS OF LINKING
      METHODOLOGY
---------------------------------------------------------- Letter :4.1

HCFA staff adjusted the CPEPs' administrative and clinical labor
estimates because different panels developed very different estimates
for the same procedures.  For example, two CPEPs reviewed procedure
code 43117--partial removal of the esophagus.  One CPEP estimated the
administrative labor associated with this code at 375 minutes and the
clinical labor at 697 minutes.  In contrast, the other CPEP estimated
administrative labor at 465 minutes and clinical labor at 1,647
minutes.  Such variations support the need for adjusting the CPEP
estimates.  HCFA staff believe that some CPEPs had higher labor time
estimates than others primarily because the CPEPs included activities
performed by physicians or because they double-counted some
activities that staff may do simultaneously.\12 \13

While recognizing these differences in labor estimates, HCFA staff
concluded that the CPEPs ranked the procedures similarly.  That is,
while two CPEPs may have developed different labor estimates for the
same codes, HCFA staff believe that the ratios between the estimates
were generally constant.  For example, one CPEP's estimates were
generally twice as high as those of another.  To correct for these
differences between CPEPs, HCFA used a statistical regression
methodology to standardize the different CPEP labor estimates and
"link" them, that is, place them on a common scale.  HCFA's
methodology separately adjusted the administrative labor estimates
and the clinical labor estimates developed by the panels. 

HCFA's linking significantly reduced some of the original CPEP
estimates.  For instance, linking reduced the administrative
estimates for one CPEP by 80 percent and the clinical estimates of
another CPEP by 50 percent.  As a result of these changes, linking
also affected the ranking of codes among CPEPs.  For example, before
linking, code A might have had a higher ranking than code B; but
after linking, the two might have been ranked equally or code B might
have been ranked higher.  HCFA officials told us that linking's
impact on the rankings was appropriate because it adjusted for some
of the incorrect assumptions used by the panels, such as the
assumption that a staff person always performs tasks sequentially. 

Representatives from several medical societies believe that HCFA's
linking methodology is seriously flawed and is unwarranted.  The
American College of Surgeons, for example, believes that the higher
labor time estimates developed by the panels reviewing surgical
procedure codes are not necessarily inflated and therefore do not
need to be adjusted.  Rather, the American College of Surgeons
believes that these estimates reflect higher practice expenses
incurred by surgeons, such as expenses associated with the need to
obtain prior authorization for surgeries and with updating referring
physicians on a patient's status.  On the other hand, the American
Society of Internal Medicine believes that linking is necessary. 
Representatives told us that some CPEP labor time estimates are
overstated because panelists on some CPEPs uniformly assigned higher
labor time estimates to the codes they reviewed than did other CPEPs
and that HCFA therefore needed to adjust these estimates downward to
make them comparable across all panels. 


--------------------
\12 Physician time is included in the physician work RVUs and
therefore should not be included in the practice expense RVUs. 

\13 HCFA did not see the same kind of variation in the CPEPs' supply
and equipment estimates and so did not make similar adjustments for
these items. 


      HCFA'S LINKING METHODOLOGY
      RAISES SOME QUESTIONS
---------------------------------------------------------- Letter :4.2

In developing its linking methodology, HCFA wanted to generate
separate adjustment factors for a panel's clinical labor estimates
and its administrative labor estimates.  For example, if a panel's
clinical labor estimates were too high by a factor of two, they would
all be cut in half.  According to HCFA, these linking adjustments are
most appropriate when the actual relationships between CPEPs conform
to certain patterns. 

HCFA staff told us that their review suggested that the data
generally followed these patterns, but we found that in a number of
cases the CPEP data departed considerably from these patterns.  For
example, the ratios between any two panels' estimates for redundant
codes generally have to be similar.  An intuitive test of this
assumption is to examine the related assumption:  that the CPEPs rank
the redundant codes generally in the same order.  If this were true,
all the CPEPs would, for example, rank codes A, B, and C in the order
first, fifth, and tenth.  We found that, on the contrary, some CPEPs
ranked redundant codes in very different orders. 

As a second example, HCFA asserted that its linking methodology is
more appropriate when the actual CPEP data conform to a second
pattern--that the ratios of estimates between CPEPs are generally
constant.  If two CPEPs evaluated codes A and B, the first CPEP's
labor estimates might generally be twice those of the second CPEP--70
minutes versus 35 minutes for code A and 120 minutes versus 60
minutes for code B.  If roughly constant ratios were found for all
pairs of CPEPs, then all the labor estimates by a particular CPEP
could be adjusted by a constant percentage without significantly
affecting their relationship or ratios.  However, we found a number
of cases that did not display generally constant ratios. 

HCFA believes that such discrepancies do not compromise its model's
reliability.  Rather, it believes these discrepancies reflect
real-world deviations from its model as a result of random
differences between the CPEPs.  However, our review of an analysis of
HCFA's statistical model identified potentially significant problems
that signal omission of a systematic factor.  This suggests the
regression estimates may be statistically biased--too high or too
low.  This same analysis, though, points to modifications of the HCFA
regression that might yield a satisfactory linking method.  HCFA
staff acknowledge that their regression model has some anomalies, but
they do not believe the anomalies are serious enough to negate the
overall validity of the model. 

In addition, HCFA's linking methodology relies on CPEP estimates for
redundant codes, but critics have questioned the characteristics of
the redundant codes selected and HCFA's process for selecting them. 
HCFA's linking methodology assumes that the practice expenses
associated with a redundant code are the same, no matter which
medical specialty provides the service.  For example, the methodology
assumes that, for an office visit, a cardiologist and a primary care
physician incur the same labor expenses.  Several medical specialties
criticized this assumption, noting that the higher labor expense
estimates of some CPEPs may reflect that different tasks are
performed or that more time is needed for similar tasks.  For
example, administrative labor expenses for an office visit to a
cardiologist might be higher than those for an office visit to a
primary care physician because the cardiologist's staff may have to
spend more time obtaining precertification approval and handling a
higher percentage of denied claims. 

To select redundant codes, HCFA and its contractor identified codes
that were frequently billed by two different specialties.  This
approach differed from the one HCFA used to select redundant codes
while developing physician work RVUs in the late 1980s.  As part of
that process, clinicians representing different medical specialties
reviewed potential redundant codes to ensure that they involved
equivalent physician work.  In some cases, the clinicians determined
that, because a code did not involve equivalent physician work
between two specialties, its use as a linking code in the regression
was not appropriate.  HCFA staff told us that they did not use the
same process for the practice expense RVUs.  They believed that any
code evaluated by two or more CPEPs was an appropriate redundant code
because Medicare pays the same amount regardless of which specialty
performs the procedure. 

Representatives from both PPRC and the AMA agree that HCFA needs to
adjust the labor estimates from the different CPEPs to make them
comparable.  However, representatives from both organizations
question the process HCFA used to select the redundant codes and
believe that physician participation in the selection of redundant
codes would have improved the linking adjustments.  Nevertheless,
PPRC staff do not believe that HCFA should select new redundant
codes, assemble a new set of CPEPs, and estimate the linking
regression on new data.  Instead, PPRC staff agreed that it might be
useful for HCFA to have physicians review the redundant codes used,
eliminate any questionable codes, and rerun its regression model on
this subset of the original CPEP data.  (App.  III provides more
detailed information regarding HCFA's linking methodology and its
limitations.)


      EFFECTS OF SCALING
      METHODOLOGY
---------------------------------------------------------- Letter :4.3

Following linking, HCFA compared the aggregate CPEP data with data
from the AMA's 1996 SMS survey.  HCFA found that the aggregate CPEP
estimates for labor, supplies, and equipment each accounted for a
different portion of total direct expenses than the AMA data did. 
For instance, labor accounted for 73 percent of total direct expenses
in the SMS survey data but only 60 percent of the total direct
expenses in the CPEP data.  To make the CPEP percentages mirror the
SMS survey percentages, HCFA inflated the CPEPs' labor expenses for
each code by 21 percent and the medical supply expenses by 6 percent
and deflated the CPEPs' medical equipment expenses by 61 percent. 
HCFA staff told us that they believe this scaling was necessary to
ensure that the proportions of practice expense RVUs devoted to
labor, supplies, and equipment were consistent with an external
benchmark.  For example, without scaling, HCFA would have no means to
ensure that the labor expense estimates, as adjusted by linking and
other steps in HCFA's methodology, represented the appropriate labor
expenses.  In addition, inaccuracies in HCFA's estimates of supply
and equipment prices, as well as HCFA's assumed equipment utilization
rate, might have distorted the expenses among the three components. 
As a result, expenses associated with supplies and equipment might be
overrepresented in total practice expense RVUs.  (App.  IV contains
more details on HCFA's scaling methodology and its effects on the
CPEP data.)

The AMA believes that scaling was appropriate because the CPEPs,
given their limited size, were not necessarily representative of all
medical practices.  Therefore, the CPEP data needed to be adjusted to
reflect national averages.  PPRC staff, too, believe that scaling is
warranted.  However, they said that modifications to scaling are
needed to ensure that the CPEP data are consistent with the SMS
survey data, because HCFA eliminated certain labor expenses developed
by the CPEPs that are contained in the SMS survey data. 

Some physician groups believe that if HCFA utilizes a scaling
methodology in the future, it should develop different scaling
factors for each medical specialty, since the percentages of labor,
medical supplies, and medical equipment to total practice expenses
vary among medical specialties.  For example, physicians who provide
equipment-dependent procedures, such as echocardiography, have a
higher percentage of equipment expenses compared with other
specialties, such as family practitioners, that are not as dependent
upon medical equipment. 


      HCFA IS CONSIDERING OTHER
      METHODS TO ADJUST CPEP DATA
---------------------------------------------------------- Letter :4.4

HCFA officials told us that they may eliminate linking in their May
1998 proposed rule because it is a complex and confusing methodology
that has caused considerable controversy in the medical community. 
Instead, they may make other adjustments to the CPEP data so that
comparisons can be made among the different CPEPs.  For example, HCFA
officials told us that they may use standard administrative labor
estimates, such as the time it takes a receptionist to schedule a
patient's next appointment, across broad categories of codes.  This
may reduce much of the variation in administrative labor estimates
developed by the CPEPs and eliminate the need for linking these
estimates.  HCFA may also shift administrative labor devoted to
billing and other administrative activities from the direct expense
category to the indirect expense category.  This change too may
eliminate the need for linking the administrative labor estimates. 
At this time, however, HCFA has made no decisions on whether it will
continue to rely on its linking methodology as part of its May 1998
proposed rule. 


   SOME OTHER ADJUSTMENTS TO THE
   CPEP ESTIMATES LACK SUPPORTING
   DATA
------------------------------------------------------------ Letter :5

HCFA disallowed some direct expenses identified by the CPEPs because
it believes Medicare pays for these expenses outside the physician
fee schedule.  HCFA also limited some administrative and clinical
labor estimates:  HCFA believes these estimates were too high but did
not test the basis for its reductions.  Various physician groups have
suggested that HCFA reclassify certain administrative labor
activities as indirect expenses, a move that could eliminate the need
for limitations on estimates developed by the CPEPs.  Further, HCFA
made certain assumptions regarding equipment utilization
rates--assumptions that it has not tested and that some physician
groups believe have negative effects on RVUs. 


      HCFA DISALLOWED SOME DIRECT
      EXPENSE ESTIMATES
---------------------------------------------------------- Letter :5.1

HCFA edited the CPEP data for both policy considerations and
reasonableness.  The most controversial policy edit concerned HCFA's
elimination of nearly all expenses related to physicians' staff who
accompany them in the hospital--primarily nurses.  Such staff
reportedly (1) assist physicians at surgery, (2) serve as scrub
nurses at surgery or perform other nursing functions, (3) assess
patients following surgery and provide patient education, or (4)
communicate with hospital staff to arrange for patient discharge and
posthospital care.  HCFA officials said that they disallowed the
expenses for these services primarily because Medicare pays for them
through other mechanisms.  For example, Medicare's policy is to pay
for assistants at surgery only if they are either physicians or
physician assistants; Medicare does not pay for other medical
professionals serving in this role.  According to HCFA, hospitals are
responsible for providing the nurses who work in the hospital
setting, and Medicare's payments to the hospital for surgical
procedures already cover the expense of scrub nurses who participate
in surgeries.  Medicare pays for postoperative patient assessment and
education through the physician work component of the Medicare fee
schedule; paying again for these expenses through the practice
expense component of the fee schedule would represent double payment,
according to HCFA officials.  Regarding physician staff who
communicate with hospital staff, HCFA allowed 15 minutes of a nurse's
time as a direct expense for surgical codes. 

The American College of Surgeons and several other physician groups
argue that surgeons are not separately reimbursed for their
hospital-related practice expenses and that Medicare should therefore
recognize them as a legitimate practice expense.  Representatives
from these organizations said that hospitals have been cutting back
on their nursing staff, prompting physicians to bring their own
nurses to the hospital to assist them with their work.  According to
HCFA officials, however, neither the American Hospital Association
nor any physician group has been able to provide HCFA with
information on the extent to which this practice occurs or how often
physicians absorb these expenses.  In an October 1997 Federal
Register notice, HCFA asked for specific data from physicians,
hospitals, and others on the extent to which staff accompany
physicians to hospitals, ambulatory surgical centers, and other
facilities and are not otherwise reimbursed by Medicare.  Subsequent
to completion of our fieldwork, HCFA received some limited
information on this issue, but we did not review or evaluate it. 
HCFA officials said that they will review this information before
deciding whether to change their decision in HCFA's next proposed
rule. 


      HCFA ALSO CAPPED SOME DIRECT
      EXPENSE ESTIMATES
---------------------------------------------------------- Letter :5.2

HCFA staff also conducted reasonableness edits of the CPEP data that
resulted in reducing the allowed expenses for certain codes. 
Physicians and clinical staff within HCFA, in consultation with other
government physicians and Medicare claims processing contractor
staff, reviewed the CPEP data and identified two problems.  They
found that (1) the administrative labor time estimates developed by
the CPEPs for many diagnostic tests and minor procedures appeared to
be excessive when compared with the administrative labor time
estimates for a mid-level office visit and (2) the nonphysician
clinical labor time estimates for many procedures were excessive when
compared with the time physicians spend performing the procedures. 
Therefore, HCFA capped the administrative labor time for several
categories of services at the level of a mid-level office visit. 
With certain exceptions, HCFA also capped nonphysician clinical labor
time at 1-1/2 times the minutes used by a physician to perform a
procedure.  HCFA has not, however, conducted tests or studies that
validate these changes and thus cannot be assured that they are
necessary or reasonable. 

It is not surprising that HCFA staff believed an administrative labor
time cap was needed, given the variation in administrative labor time
estimates developed by the CPEPs and the controversy surrounding
estimates of administrative billing activities.  The AMA reported
that the CPEP administrative billing estimates seemed unreasonable,
as they were based primarily on guesses.  We observed that estimates
of billing times were frequently the most contentious issue within
the validation and cross-specialty panels.  Representatives from
different physician groups told us that physicians are more familiar
with clinical tasks, and the time needed to complete them, than they
are with administrative tasks.  Consequently, physician estimates for
administrative tasks would be less accurate than those for clinical
tasks, they said.  Others, however, told us that physicians deferred
in some cases to practice administrators on the panels, resulting in
administrative labor estimates that were more reliable. 

Both the AMA and most participants in the cross-specialty panel
recommended that HCFA treat billing activities as indirect expenses
rather than as direct expenses--a shift that would be consistent with
accounting standards in the federal sector.  HCFA officials told us
that they are considering this recommendation but have made no final
decision.  Treating billing and other administrative expenses as
indirect expenses could make a cap on administrative labor estimates
unnecessary. 


      HCFA LACKS DATA TO SUPPORT
      ITS ASSUMPTIONS ON EQUIPMENT
      USE
---------------------------------------------------------- Letter :5.3

HCFA assumed that equipment associated with specific procedures, such
as a treadmill used for a cardiology stress test, is used 50 percent
of the time that a practice was operating, while equipment that
supports all or nearly all services provided by a practice, such as
an examination table, is used 100 percent of the time.  HCFA
officials told us that actual data on equipment utilization rates
were not available from the medical community.  Therefore, HCFA had
to make assumptions about the rate at which equipment is used.  HCFA
officials also told us that they could eliminate all equipment
expenses from their direct expense RVU calculations without
significantly altering the final RVUs for most procedures because
equipment typically represents a small fraction of a procedure's
direct expenses.  They acknowledged, however, that the equipment
utilization rate affects each medical specialty differently and that
they have not conducted a sensitivity analysis to determine the
effect of different equipment utilization rates on the different
specialties. 

The AMA and other physician groups that we contacted said that HCFA's
estimates greatly overstate the use of most equipment, resulting in
an underestimation of the equipment expenses used in calculating
RVUs.  The American Academy of Ophthalmology, for example, surveyed
its members and found that argon lasers used in eye surgery are used
no more than 10 percent of the time that offices are open.  These
physician groups believe that HCFA should seek input from large group
practices as well as data from MGMA on equipment utilization rates. 

In its October 1997 Federal Register notice, HCFA asked for copies of
any studies or other data showing the actual use of equipment, by
procedure code, that it could use to adjust its equipment utilization
rate assumptions.  This is consistent with the Balanced Budget Act of
1997 requirement that HCFA use actual data in setting equipment
utilization rates. 


   OPTIONS FOR ASSIGNING INDIRECT
   EXPENSES WARRANT FURTHER
   CONSIDERATION
------------------------------------------------------------ Letter :6

Direct expenses can be specifically identified for a service or
procedure, whereas indirect expenses, by definition, cannot. 
Therefore, total indirect expenses must be identified and assigned in
some manner.  Recognizing that there is no one right answer, HCFA
considered several methodologies for assigning indirect expenses to
individual procedure codes and selected a method that is based on the
three components of the Medicare fee schedule.  Physician groups we
contacted criticized this methodology because they believe that it
fails to recognize that indirect expenses, as a percentage of all
practice expenses, differ among medical specialties. 


      OPTIONS HCFA CONSIDERED FOR
      ASSIGNING INDIRECT EXPENSES
---------------------------------------------------------- Letter :6.1

According to HCFA's proposed rule of June 1997, HCFA considered four
assignment methodologies.  HCFA's selected methodology assigns
indirect expenses on the basis of (1) physician work RVUs, (2) direct
practice expense RVUs, and (3) malpractice expense RVUs for each
code.  To calculate the indirect expense RVUs for a procedure, HCFA
adds the values of the three RVU components and then multiples the
total by a factor of .219.  This factor is constant for all codes and
ensures that the total pool of indirect expense RVUs does not exceed
45 percent of all practice expense RVUs.  PPRC supports this
approach, which is generally consistent with the method PPRC proposed
for assigning indirect expenses to procedure codes. 

HCFA officials said that they selected this methodology for several
reasons.  First, it assigns indirect expenses on the basis of the
variables that HCFA believes are the primary drivers of indirect
expenses.  For example, higher physician work RVUs generally reflect
greater complexity of a procedure and more time required to carry it
out--meaning that a physician can perform fewer of these procedures
in a day.  This, in turn, means that a physician's indirect expenses
associated with operating a practice, such as rent and utilities,
must be allocated over a smaller pool of procedures.  Second, HCFA's
methodology reduced the redistribution effects of the proposed rule
on various physician groups.  For example, surgeons and other
physicians who provide hospital-based services benefited because
their commonly performed procedures typically have higher physician
work RVUs than the procedures performed by physicians who provide
office-based services.  Higher physician work RVUs result in greater
indirect expense RVUs under HCFA's assignment methodology. 


      PHYSICIANS' GROUPS HAVE
      CRITICIZED HCFA'S METHOD FOR
      ASSIGNING INDIRECT EXPENSES
---------------------------------------------------------- Letter :6.2

Physician groups have criticized HCFA's methodology, saying that it
fails to recognize that different medical specialties have different
direct-to-indirect expense ratios.  The American College of Surgeons,
for example, reports data that indicate indirect expenses, as a
percentage of total practice expenses, range from 54 percent for
urologists to 71 percent for neurosurgeons.  These physician groups
believe that it is inappropriate for HCFA to ignore these differences
and assume a constant direct-to-indirect expense ratio across all
medical specialties. 

Our review of HCFA's methodology shows that the ratio of direct to
indirect expenses differs by procedure.  For example, procedure code
13100--repair of a wound--consists of 1.39 direct expense RVUs and
1.01 indirect expense RVUs, resulting in an indirect expense ratio of
42 percent (relative to total expenses).  In contrast, procedure code
24587--repair of an elbow fracture--consists of 1.51 direct expense
RVUs and 4.12 indirect expense RVUs, resulting in an indirect expense
ratio of 73 percent.  Depending upon the procedures performed, the
indirect expense ratios will vary from physician to physician and
will reflect their medical specialty.  What is not clear from HCFA's
methodology is whether the indirect expense ratio for each procedure,
and therefore each medical specialty, is correct. 

The American Society of Internal Medicine and other physician groups
that we met with believe that HCFA should develop separate indirect
expense ratios for each medical specialty and use these ratios when
calculating indirect expense RVUs.  HCFA could develop these ratios,
they say, on the basis of data contained in the AMA's SMS survey and
would not have to rely upon an assumption.\14 HCFA has already used
the SMS survey data in its proposed fee schedule revisions to
determine that indirect expense RVUs constitute 45 percent of the
total pool of practice expense RVUs.  HCFA officials told us that
they will evaluate this alternative indirect expense assignment
methodology before issuing HCFA's next proposed rule. 


--------------------
\14 AMA representatives noted that the SMS survey does not include
data on all specialties or subspecialties, but they have expressed
their willingness to work with HCFA to expand the sample size so that
additional data would be collected. 


   ACCESS TO CARE NEEDS CONTINUED
   MONITORING
------------------------------------------------------------ Letter :7

It is not clear if beneficiary access to care will be adversely
affected by Medicare's new fee schedule allowances for physician
practice expenses.  This will depend upon such factors as the
magnitude of the Medicare payment reductions experienced by different
medical specialties, other health care insurers' use of the fee
schedule, and fees paid by other purchasers of physician services. 
While beneficiary access to care has remained very good since
implementation of the fee schedule in 1992, the cumulative effect of
prior and proposed changes to the fee schedule will need to be
monitored to ensure that Medicare beneficiaries are not denied access
to needed care because of lower payment levels. 


      IMPACT ON PHYSICIAN INCOME
      WILL VARY BY SPECIALTY
---------------------------------------------------------- Letter :7.1

As part of its June 1997 proposed rule, HCFA prepared an impact
analysis showing the rule's potential effect on physicians' income
from Medicare, by medical specialty.  Generally, Medicare payments to
surgeons and some specialists would decrease, while payments to
generalists would increase.  Whether HCFA's final rule will result in
similar effects is not known. 



                                Table 1
                
                Estimated Changes in Physicians' Income
                      From Medicare as a Result of
                  Implementing HCFA's Proposed Rule of
                               June 1997

                                                                Percen
                                                                  tage
                                                                change
                                                                    in
                                                                physic
                                                                   ian
                                                                income
                                                                  from
                                                                Medica
Medical specialty                                                   re
--------------------------------------------------------------  ------
Family practice                                                     12
General internal medicine                                            3
Urology                                                              1
General surgery                                                    (9)
Cardiac surgery                                                   (32)
----------------------------------------------------------------------
Source:  HCFA. 

Because a large number of public and private health care insurers
base their payments on Medicare's fee schedule, the total impact on
physicians' incomes resulting from changes in Medicare's fee schedule
allowances could be greater than shown in table 1.  Yet, if other
health care purchasers pay physicians about the same fees as Medicare
for the same services, physicians have little or no incentive to
provide more care to privately insured patients and less care to
Medicare patients. 


      CUMULATIVE EFFECTS OF FEE
      SCHEDULE CHANGES COULD
      AFFECT ACCESS TO CARE
---------------------------------------------------------- Letter :7.2

Since Medicare began paying physicians for their services on the
basis of a national fee schedule in 1992, both HCFA and PPRC have
monitored indicators of beneficiary access to care to determine if
there have been adverse consequences.  HCFA surveys approximately
12,000 beneficiaries annually to gather information on such issues as
beneficiary satisfaction with care, difficulties obtaining care, and
whether beneficiaries ever had a medical problem but did not seek
physician treatment.  PPRC's analysis of these data, along with data
from other sources, indicates that access for most beneficiaries
remains very good and that indicators of access remain essentially
unchanged since implementation of the fee schedule.  Any decreases
observed in selected services do not appear to be related to the fee
schedule but rather to other factors, such as changes in medical
practices. 

Some medical specialties that experienced reduced Medicare payments
after implementation of the fee schedule in 1992 would experience
further reductions under HCFA's proposed rule.  For example, between
1992 and 1996, cardiologists experienced a 9-percent reduction in
their Medicare payments; gastroenterologists, an 8-percent reduction;
and ophthalmologists, a 12-percent reduction.  HCFA's proposed rule
would result in further reductions of 17 percent, 20 percent, and 11
percent, respectively, once the new practice expense component of the
fee schedule is fully implemented in 2002.  Total potential
reductions of approximately 25 percent are significant and could
affect physician decisions regarding their care of Medicare
beneficiaries. 


      OTHER CHANGES TO MEDICARE'S
      FEE SCHEDULE COULD ALSO
      AFFECT ACCESS
---------------------------------------------------------- Letter :7.3

To convert RVUs into a dollar amount, HCFA uses a conversion factor. 
Between 1994 and 1997, there were separate conversion factors for
surgical services, primary care services, and other nonsurgical
services.  The Balanced Budget Act of 1997 established a single
conversion factor for all physician services beginning in January
1998.  As a result of this change, surgical services experienced a
10.4-percent reduction in Medicare payments that are in addition to
the proposed practice expense fee schedule changes.  The American
College of Surgeons reported that this change to a single conversion
factor has already resulted in some surgeons concluding that they can
no longer treat Medicare patients.  However, there is no evidence on
the extent to which this is occurring.  It also estimated that
between 1997 and 2002 the combined effects of fully implementing the
changes in Medicare's practice expense payments as proposed by the
June 1997 rule and projected declines in the single conversion factor
would severely reduce Medicare payments for some surgical procedures
and could further reduce beneficiary access to care.  For example,
Medicare payments for a total hip replacement could decrease by 45
percent and by 34 percent for a laparoscopic removal of the gall
bladder. 


   CONCLUSIONS
------------------------------------------------------------ Letter :8

HCFA has made considerable progress in developing new practice
expense RVUs, but much remains to be done before the new fee schedule
payments are implemented in 1999.  Although HCFA worked closely with
the medical specialty societies both before and after issuing its
proposed rule, considerable controversy remains within the medical
community over HCFA's methods for developing direct and indirect
expense data.  However, there is no need for HCFA to start over and
utilize different methodologies for creating new practice expense
RVUs; doing so would needlessly increase costs and further delay
implementation of the fee schedule revisions.  HCFA will need to
continue working with these societies as it refines its data and
methodologies. 

HCFA's use of expert panels is an acceptable method to develop direct
cost estimates.  Not only is this method supported by medical
researchers and PPRC, but other options for developing these
estimates have practical limitations that preclude their use as
reasonable alternatives.  However, data generated by the panels
represent a starting point, not an end point, for developing the
direct expense RVUs.  Collecting actual data on key procedures from a
limited number of physician practices through surveys or on-site
reviews during the 3-year phase-in period would enable HCFA to check
the reliability of the CPEP data and test the assumptions HCFA used
for its adjustments.  Medical specialty and physician groups need
assurances that a process exists for periodically updating the
practice expense RVUs and identifying and correcting significant
problems.  Yet, HCFA does not have a plan to refine the practice
expense data during or after phase-in of the new fee schedule
revision. 

While some adjustments to the CPEP estimates are necessary to correct
for differences in the estimates between panels, HCFA's adjustments
to link the estimates of the expert panels raise some questions.  If
HCFA plans to rely on a regression-based linking methodology, its
regression model will need to be reevaluated as we found significant
discrepancies in some cases between the CPEP data and the assumptions
underlying HCFA's particular model.  In addition, an analysis of the
regression suggests a possible bias in the linking factors.  Other
nonregression approaches HCFA is considering may also be appropriate
to deal with variations in the panels' estimates, but we cannot
evaluate them until they are more fully defined.  Scaling seems to be
necessary because of various steps in HCFA's methodology that
affected the proportion of RVUs allocated to labor, medical supplies,
and medical equipment.  Without scaling, HCFA would not have an
external benchmark to ensure that labor, supplies, and equipment were
appropriately apportioned among the total direct practice expense
RVUs. 

At the time of its proposed rule, HCFA appropriately disallowed
nearly all of the expenses related to staff who accompany physicians
to the hospital since there was no available evidence that these
expenses are not already reimbursed by Medicare or that this is a
widespread, common physician practice.  Information supplied by some
physician groups indicates, however, that there may have been a shift
in hospital and physician practices that Medicare has not recognized
in its methods for reimbursing nonphysician clinical labor expenses. 
Hospitals may no longer be providing the same level of nursing
support that they did at the time Medicare established its current
method for paying hospitals for their expenses.  Additionally,
physicians may now be relying on their own staff to perform work in
the hospital, work that Medicare recognized as a physician
responsibility when establishing the physician work RVUs. 

HCFA has not examined its assumptions regarding its capping of
administrative and clinical labor time estimates to ensure that they
are necessary and reasonable.  By including billing and other
administrative labor as direct expenses--expenses that accounting
standards in the federal sector typically include as indirect
expenses--HCFA needlessly made it more difficult for the panels to
develop reliable, consistent estimates. 

HCFA's use of physician work, direct practice expense, and
malpractice expense RVUs is an acceptable option for assigning
indirect expenses to procedures since these factors likely reflect
the drivers of indirect expenses.  However, there are other
alternatives.  For example, HCFA could use specialty-specific
indirect expense ratios, based on the SMS survey data.  This would be
more clearly consistent with the Balanced Budget Act of 1997
requirement that HCFA utilize actual data for its key assumptions. 

Despite its having made significant adjustments to the CPEP estimates
and its use of various methodologies to develop the new practice
expense RVUs, HCFA has done little in the way of performing
sensitivity analyses to determine which of its data adjustments and
methodologies have the greatest effects on rankings and the RVUs. 
Having such information would enable HCFA to target its refinement
efforts on those areas most susceptible to weaknesses in the data or
methodologies.  While it is unreasonable to expect HCFA to conduct
such analyses before it issues its May 1998 proposed rule, HCFA has
time between the rule's issuance and implementation of the fee
schedule revisions in 1999, as well as early in the initial phase-in
period, to conduct sensitivity analyses, gather needed data, and make
necessary adjustments to the RVUs. 

The potential impact of the proposed new fee schedule allowances for
physician practice expenses on beneficiary access to care is unknown
at this time.  However, the combined impact of the proposed and prior
fee schedule changes on physicians' incomes will affect some medical
specialties more than others.  Therefore, indicators of beneficiary
access to care that focus on the medical specialties most adversely
affected by the cumulative changes in Medicare's fee schedule
allowances will require continued monitoring. 


   RECOMMENDATIONS
------------------------------------------------------------ Letter :9

We recommend that the Administrator of HCFA take the following
actions: 

  -- Use sensitivity analyses to test the effects of (1) the limits
     HCFA placed on the panels' estimates of clinical and
     administrative labor and (2) HCFA's assumptions about equipment
     utilization.  Where HCFA's adjustments or assumptions
     substantially alter the rankings and RVUs of specific
     procedures, HCFA should collect additional data to assess the
     validity of its adjustments and assumptions, focusing on the
     procedures most affected. 

  -- Evaluate (1) classifying the administrative labor associated
     with billing and other administrative expenses as indirect
     expenses, (2) alternative methods for assigning indirect
     expenses, and (3) alternative specifications of the regression
     model used to link the panels' estimates.  Since these three
     aspects of HCFA's methodology are interrelated, HCFA should
     determine how changes in one aspect of the methodology, such as
     reclassifying some labor from direct to indirect expenses,
     affect other aspects of the methodology, such as the
     specification of the regression model to link the panels'
     estimates of administrative labor and the method used to
     allocate indirect expenses. 

  -- Determine whether changes in hospital staffing patterns and
     physicians' use of their clinical staff in hospital settings
     warrant adjustments between Medicare reimbursements to hospitals
     and physicians.  Similarly, HCFA should determine whether
     physicians have shifted tasks to nonphysician clinical staff in
     a way that warrants reexamining the physician work RVUs. 

  -- Work with physician groups and the AMA to develop a process for
     collecting data from physician practices as a cross-check on the
     calculated practice expense RVUs, and to periodically refine and
     update the RVUs. 

  -- Monitor indicators of beneficiary access to care, focusing on
     those services with the greatest cumulative reductions in
     Medicare fee schedule allowances, and consider any access
     problems when making refinements to the practice expense RVUs. 


   AGENCY COMMENTS AND OUR
   EVALUATION
----------------------------------------------------------- Letter :10

We provided a draft of this report for comment to HCFA officials.  We
also gave copies of the draft to representatives of medical
societies, physician groups, a medical group we contacted during our
work, and MedPAC.  The following summarizes the comments and our
responses. 


      HCFA
--------------------------------------------------------- Letter :10.1

HCFA officials agreed with our analysis regarding the use of CPEPs
for developing direct practice expense data and the use of an
allocation formula to assign indirect expenses to individual
procedures.  Regarding our recommendations that HCFA collect and
analyze additional data to test the validity for its adjustments and
assumptions, they asked that we clarify the time frames in which we
expect HCFA to conduct this work.  We provided this clarification in
the report.  HCFA officials explained that they plan to analyze
options on how to treat billing and other administrative expenses and
that it would be premature to make a decision on this issue before
they have analyzed their options.  We agree that HCFA should evaluate
available options before making a final decision, and we therefore
modified our recommendation on this issue. 

HCFA officials disagreed with some of our discussions regarding their
linking and scaling methodologies.  Regarding HCFA's linking
methodology, HCFA officials were concerned that the report's
discussion was overly negative.  They believed that the report
ignored the distinction between ideal data and real-world data, which
typically deviate from the ideal.  Specifically, HCFA officials
stated that the linking process works best when the redundant codes
have similar ranks across CPEPs and when the relative spacing among
the redundant codes across CPEPs is the same.  They also emphasized,
however, that not all CPEP data will fit neatly into the assumptions
or patterns that underlie their regression specification. 
Nevertheless, the officials said they recognize that their model
would appear more appropriate if the CPEP data conformed more closely
with these patterns. 

We agree that there is randomness in sample data.  Random variation
will cause predictions from even the best-specified regression model
to deviate from the data on which the model is estimated. 
Nonetheless, we are not convinced that HCFA's linking model is free
of statistical problems.  We did not expect to see such substantial
and often striking deviations from the assumptions or patterns that
HCFA staff had told us were the basis for their model.  This point
holds, especially with regard to deviations of the ordinal ranks from
the expected pattern.  Although it is possible that these deviations
exclusively reflect the randomness in the estimates of the various
CPEP panels, they may also reflect systematic factors that, omitted
from HCFA's regression analysis, could make its linking factors too
high or too low.  In any case, the observed discrepancies between the
CPEP data and the expected patterns suggest potential limitations of
linking as well as the need for further analysis of the implications
of these discrepancies for HCFA's statistical model. 

Moreover, analysis of the HCFA regression's residuals suggests that
the linking factors estimated by the regression may be statistically
biased; that is, even if different samples of CPEP data were used,
the estimated linking factors drawn from the same model would deviate
from the true coefficients.\15 Specifically, the analysis of the
residuals indicates that the broad types of procedure codes (for
example, invasive procedures versus lab tests) reviewed by a CPEP
affect the size and sign (positive or negative) of the residual
associated with a procedure code.  Consequently, the mix of procedure
codes reviewed by a CPEP appears to affect the CPEP's estimates. 
Since HCFA's model does not account for this effect, the estimated
coefficients drawn from HCFA's model may be statistically biased. 
Despite these apparent difficulties, a statistical model is in
principle an acceptable approach to linking.  However, we believe
that HCFA should evaluate the issues highlighted by the residual
analysis and revise the regression model as necessary. 

HCFA officials did not believe that the draft provided a balanced
discussion of its scaling methodology and provided new information to
support their use of this methodology.  We included this information
to clarify our discussion of scaling and concluded that scaling seems
to be appropriate. 


--------------------
\15 Such bias is potentially important.  By contrast, any statistical
estimate, biased or not, deviates from the true value because of
variability inherent in estimating coefficients on the basis of a
sample of data instead of the entire population. 


      COMMENTS FROM
      REPRESENTATIVES OF
      PHYSICIANS' ORGANIZATIONS
      AND OTHERS AND OUR RESPONSE
--------------------------------------------------------- Letter :10.2

Comments from these groups and our responses are provided below;
separate discussions are presented for each major section of our
report. 

Regarding HCFA's approach to developing direct expense estimates,
many of the representatives, including those from MedPAC and the
American Society of Internal Medicine, supported our conclusion that
the expert panel process represents an acceptable method.  The AMA
agreed with us that starting data collection over would needlessly
increase costs and further delay implementation.  Both the AMA and
MGMA support the need for HCFA to collect limited, additional data as
a cross-check on the CPEP data.  But MGMA believed that, for
developing direct expense estimates, surveys would be better than
HCFA's informal, subjective method of convening panels.  MGMA and the
American College of Surgeons questioned both the way HCFA convened
and conducted its panels and the validity of the data they generated. 

We continue to consider the expert panel process to be an acceptable
method and believe that the limitations of surveys of physicians and
their practices--low response rates, potential response biases, and
answers based on the judgments of the respondents--preclude their use
as HCFA's primary data gathering approach.  As we note in the report,
however, data from surveys could help HCFA evaluate and, if
necessary, modify its panel data. 

We cannot comment on how well the panels were conducted, because HCFA
convened them months before we began our work.  Nonetheless, the
panel-generated data represent only the first phase of HCFA's
development of practice expense RVUs.  That is why we recommend that
HCFA validate its data by, for example, collecting actual data from
physician practices and testing the sensitivity of the results to
each of its key adjustments to the panel data.  The phase-in period
authorized in law gives HCFA significant time to validate the data
before the fee schedule revisions are fully effective. 

Regarding HCFA's linking and scaling adjustments, the representatives
generally agreed that the CPEP data need to be adjusted for
differences between panels' cost estimates.  For example, a
representative from the American Academy of Family Physicians
emphasized that the Academy would consider HCFA's use of "raw" CPEP
data to be unacceptable.  The AMA also favored adjustment, but not
necessarily by using HCFA's linking method.  The AMA preferred more
targeted adjustments to the CPEP data to improve their consistency. 
However, representatives from the American College of Surgeons
disagreed with our conclusion that a linking adjustment is needed as
they believe that the data are so flawed that after-the-fact
manipulations, such as linking, cannot correct them. 

We continue to believe in the necessity of adjusting the CPEP data,
given the substantial, often striking disparities in the estimates
made by two or more panels for the same codes.  A linking regression
is one way to do this.  HCFA is exploring alternative methods, but
without more specifics, we cannot comment on them at this time. 

Representatives, other than those from MedPAC, had few comments on
HCFA's scaling methodology.  MedPAC staff support scaling if done
correctly, but they were concerned about HCFA's implementation of
this methodology because they believe the raw SMS data are not
directly comparable with the CPEP data.  HCFA adjusted the CPEP data
by removing the labor time estimates associated with staff that
accompany physicians to the hospital, and MedPAC staff believe HCFA
should remove such data from the SMS data so that SMS and CPEP data
are comparable.  HCFA believes, however, that physicians' bringing
staff to the hospital is a relatively infrequent practice and has
only a minor impact on the SMS data.  We cannot comment on this
issue, since information on it was unavailable when we conducted our
work. 

With respect to HCFA's other adjustments to the CPEP data, the AMA
said that it would be particularly useful for HCFA to collect data on
administrative and equipment costs from group practices, firms that
provide billing and other administrative services to physician
practices, and associations such as MGMA.  We support HCFA's
collecting additional data about its labor time caps and equipment
utilization rates if its sensitivity analyses show that these
adjustments and assumptions substantially alter the rankings and RVUs
of particular procedures. 

Representatives of the Practice Expense Coalition and the American
College of Surgeons disagreed with our view that HCFA appropriately
disallowed nearly all the expenses associated with staff that
accompany physicians to the hospital.  The Practice Expense Coalition
contends that this represents a real, unreimbursed cost to physicians
and that hospitals, in an effort to cut costs, are not paying for
these services.  It also said that quality of patient care could
suffer if hospitals are not forced to change their behavior or if
other parts of the Medicare program fail to reimburse physicians for
these costs. 

We continue to believe that, according to Medicare policy, HCFA
appropriately disallowed these expenses at the time of its proposed
rule.  However, the information supplied to HCFA by some physician
groups indicates that there may have been a shift in hospital and
physician practices affecting Medicare reimbursement policy for these
expenses.  Therefore, we modified our report to recommend that HCFA
determine whether Medicare needs to revise how it pays for these
expenses. 

Regarding assigning indirect expenses to procedures, representatives
of the different groups and medical societies generally believed that
HCFA should evaluate using specialty-specific indirect cost ratios as
opposed to its current assignment method.  AMA representatives said
that rather than using indirect expense ratios, they would prefer
that HCFA assign indirect expenses to procedures on the basis of
specialty-specific data derived from the SMS survey.  We agree that
HCFA should evaluate alternative methods for assigning indirect
expenses, because there is no one best way to do so.  But, as noted
in our report, we also believe that the method contained in HCFA's
proposed rule is acceptable. 

Some representatives believed that HCFA should consider treating
billing and other administrative labor time as indirect expenses
rather than as direct expenses.  They believed that this is one of
several possible options HCFA should study before making a final
decision.  For example, MGMA proposed that HCFA convene a separate
expert panel composed of medical managerial and billing personnel to
consider what administrative expenses can be defined as direct as
opposed to indirect expenses.  The Practice Expense Coalition
cautioned that shifting billing and other administrative expenses to
the indirect expense category could have serious implications for
physician reimbursement unless accompanied by other corrections. 
Additionally, one representative noted that the cross-specialty panel
convened by HCFA did not vote unanimously to treat these expenses as
indirect expenses. 

While representatitves of the American Society of Internal Medicine
supported the inclusion of billing expenses as indirect expenses,
they believed these expenses should be assigned using a methodology
that differs from how the other indirect expenses are assigned under
HCFA's formula.  Specifically, they believed that the billing
expenses associated with each procedure are not reflective of
physician work values, which are a primary determinant of other
indirect expenses in HCFA's current formula. 

On the basis of these comments and those expressed by HCFA, we
modified slightly our recommendation on this issue. 

Concerning changes in beneficiary access to care, representatives of
the different groups and medical societies supported our
recommendation that HCFA monitor indicators of beneficiary access to
care following implementation of the fee schedule revisions. 
Representatives from the American Academy of Family Physicians said
that other changes in Medicare's fee schedule may also affect access
and questioned whether HCFA can isolate changes in practice patterns
that are attributable only to the practice expense fee revisions. 
The American College of Surgeons also commented that other changes in
Medicare's fee schedule payments will have especially serious
consequences on Medicare's payments for surgical procedures that
could reduce beneficiary access to care and adversely affect faculty
practice plans of teaching institutions. 

The AMA representatives said that evaluating access to care is only
one part of analyzing the impact of changes in the fee schedule. 
They said that while doctors may continue to treat patients, they may
cut costs in other areas, such as salaries, equipment purchases, and
satellite offices.  Thus, quality of care may be adversely affected
even if access remains generally good.  The Practice Expense
Coalition said that the fee schedule revisions may cause other
changes in medical practice.  For example, specialists may no longer
choose to perform certain medical procedures, resulting in only
generalists performing such procedures and potentially affecting
quality of care. 

In response to the comments on access to care, we added additional
information in our report on the potential effects of the proposed
fee schedule revisions. 


--------------------------------------------------------- Letter :10.3

As agreed with your offices, we are sending copies of this report to
the Secretary of HHS, the Administrator of HCFA, interested
congressional committees, physicians' organizations, and other
interested parties.  We will also make copies available to others
upon request. 

This report was prepared by Robert Dee, Frank Putallaz, Suzanne
Rubins, and Michelle St.  Pierre, with assistance from Jonathan
Ratner.  Please call me at (202) 512-7114 or William Reis, Assistant
Director, at (617) 565-7488 if you have any questions. 

William J.  Scanlon
Director, Health Financing
 and Systems Issues


OVERVIEW OF MEDICARE'S FEE
SCHEDULE
=========================================================== Appendix I

Efforts to reform Medicare physician payments began in the 1980s,
prompted by concerns that the existing methods of physician
reimbursement were flawed, that program costs were increasing, and
that beneficiary access to care required monitoring.  Medicare
spending for physician expenses per beneficiary had been growing at
almost twice the rate of the gross national product.  At the time,
Medicare reimbursed physicians through the "customary, prevailing,
and reasonable charge" system.  This payment method had been
criticized as inflationary and inequitable because it resulted in
widely varying fees for the same service.  Concerns were also raised
that the payment levels favored surgical services at the expense of
primary care services, resulting in distorted financial incentives. 
Limits on actual charges and a series of freezes and reductions in
payment levels for particular services made the system increasingly
complex. 

The Consolidated Omnibus Budget Reconciliation Act of 1985\16
required the Secretary of the Department of Health and Human Services
(HHS) to study and report to the Congress on a resource-based,
relative-value scale system for reimbursing physicians for their
services.  Such a system, as opposed to a charge- or cost-based
payment system, ranks services on a common scale according to the
resources expended in providing them.  Payment for a service is
dependent upon its ranking; services with a high ranking receive
greater payment than those with a low ranking.  In its 1989 report to
the Congress, the Physician Payment Review Commission (PPRC)
recommended that a resource-based, relative-value scale be adopted. 

The Omnibus Buget Reconciliation Act of 1989\17 established a uniform
national fee schedule with three relative-value components--physician
work, practice expense, and malpractice expense--and required that
the schedule be phased in over 5 years beginning in 1992. 
Implementation was to be accomplished in a budget-neutral manner. 
Also included in the legislation were geographic adjustment factors
for each component of the fee schedule, elimination of
specialty-specific payment differentials for providing the same
service, a process for calculating the annual update for the
conversion factor that converts relative values into payment rates,
and establishment of volume performance standards to track changes in
the volume or intensity of Medicare services.  The 1989 legislation
relied upon the extensive work done by Health Care Financing
Administration (HCFA) contractors at the Harvard School of Public
Health that responded to earlier legislation requiring development of
a resource-based physician work component.  Methods for calculating
resource-based relative values for practice and malpractice expenses
were not available at the time. 


--------------------
\16 Sec.  9305(b), P.L.  99-272, 100 Stat.  82, 192, Apr.  7, 1986. 

\17 Sec.  6102, P.L.  101-239, 103 Stat.  2106, 2171, Dec.  19, 1989. 


   DEVELOPMENT OF PHYSICIAN WORK
   RVUS
--------------------------------------------------------- Appendix I:1

The development of resource-based relative value units (RVU) for the
physician work component of the fee schedule took about 7 years to
complete.  Building on preliminary studies conducted earlier that
decade, Harvard researchers undertook a complex, multiphased process
with the cooperation of the American Medical Association (AMA) and
the assistance of about 100 physicians organized into technical
consulting groups.  These groups developed vignettes to describe
standard scenarios for delivering services that were included in the
AMA's physicians' Current Procedural Terminology (CPT).  A national
survey was conducted in which physicians were asked to rank services
on the basis of four standard elements:  (1) physician time, (2)
mental effort and judgment, (3) technical skill and physical effort,
and (4) stress due to risk of harm to the patient.  The researchers
reported a high level of consistency in how physicians in the same
specialty ranked the relative work required for services they
performed.  Cross-specialty panels drawn from the physician
consulting groups chose procedure codes that represented equivalent
or similar work within different specialties.  Those codes then
served as the basis for a statistical process to link all the codes
ranked by each specialty along a common scale. 

Physician work RVUs for about 800 procedure codes were developed
through the survey process.  RVUs for the remaining codes were
extrapolated from these 800 codes.  For extrapolation, codes were
assigned to families of codes and the relative work values were
determined by small groups of physicians who had participated in the
previous development stages. 


   PROCESS FOR REFINING THE RVUS
   AND CREATING NEW RVUS
--------------------------------------------------------- Appendix I:2

Before phase-in of the physician work RVUs could begin in 1992, HCFA
had to create a process to both refine the existing values and create
values for new procedure codes in the future.  HCFA's early
refinement process involved using carrier medical directors to revise
some of the newly created work RVUs and to assign RVUs to some
low-volume codes and other codes not included in the Harvard study. 
Today, a different refinement process is in place that includes a
multispecialty committee known as the AMA/Specialty Society Relative
Value Update Committee (RUC).  RUC, created in 1991, makes
recommendations to HCFA on the relative values to be assigned to new
or revised procedure codes.  HCFA then convenes a meeting of selected
carrier medical directors to review RUC's recommendations. 
Currently, HCFA accepts most of these recommendations.  According to
PPRC and AMA representatives, the RUC process is supported by most
physicians and has increased the medical community's confidence in
the physician work RVUs. 


   HCFA'S CURRENT FEE SCHEDULE
   PAYMENTS FOR PHYSICIAN PRACTICE
   EXPENSES
--------------------------------------------------------- Appendix I:3

Unlike physician work, the practice expense component of the fee
schedule is still calculated according to a charge-based system set
up in 1989.  Two main data sources are used:  Medicare claims and
allowed charge data from 1991, and information on the percentage of
revenue expended on practice expenses from national surveys of
physicians, specialists, and nonphysician practitioners reimbursed
under the Medicare fee schedule.  The RVUs for practice expenses are
computed as follows: 

1.  Using national survey data, determine the average proportion of
revenue devoted to practice expenses for physicians overall, for
various specialties, and for the nonphysician practitioners paid
under the Medicare fee schedule. 

2.  Using 1991 Medicare allowed charges, multiply the allowed charge
for each procedure code by the average percentage of revenue devoted
to practice costs for the specialty that performs that procedure. 

Example:  For a service with a 1991 allowed charge of $100 performed
only by family practitioners (whose practice expense-to-revenue
proportion is 52.2 percent), the calculation would be as follows: 

$100 x 0.522 = 52 (initial dollar) RVUs\18

3.  For procedures performed by more than one specialty, multiply the
practice expense proportion by the frequency each specialty performs
that service, then add the product and multiply by the 1991 allowed
amount. 

Example:  For a service with a 1991 allowed charge of $100 performed
70 percent of the time by family practitioners and 30 percent of the
time by internists (whose practice expense-to-revenue proportion is
46.4 percent), the calculation would be as follows: 

((0.522 x .70) + (0.464 x .30)) x $100 = 50.5 (initial dollar) RVUs

Malpractice RVUs are computed under a similar statutory formula. 


--------------------
\18 This and the following example are found in the AMA's Medicare
RBRVS:  The Physicians' Guide, 1994. 


   FEE SCHEDULE ADJUSTMENTS AND
   CONVERSION OF RVUS TO DOLLARS
--------------------------------------------------------- Appendix I:4

Before the physician work, practice expense, and malpractice expense
RVUs can be converted to dollars, they are adjusted by HCFA. 
Specifically, HCFA computes a geographic adjustment factor for each
of the three types of RVUs; each factor is designed to reflect
variation in value or cost of the relevant component from the
national average within fee schedule areas established by HCFA. 

After the three RVU components for each service are multiplied by
their respective geographic adjustment factors and combined, the
uniform national conversion factor is applied.\19 This factor
converts each total RVU into a dollar amount representing Medicare's
allowed charge for each service, including the 80 percent reimbursed
to physicians and the 20 percent beneficiary coinsurance.  HCFA must
compute the conversion factor in a manner that ensures budget
neutrality:  That is, the total Medicare expenditures for physicians'
services must not differ by more than $20 million from what the
expenditures would have been if the current fee schedule had not been
adopted.  The conversion factor is determined annually so that total
expected Medicare expenditures for physician services meet the
performance standard (the target rate of increase in expenditures)
established by the Congress or by formulas in the original fee
schedule legislation. 


--------------------
\19 The Omnibus Budget Reconciliation Act of 1989 allowed for
different conversion factors for different categories or groups of
services.  The Balanced Budget Act of 1997 required a single
conversion factor starting in 1998 and allowed for targeted increases
in physician payments based on a sustainable growth rate in the
Medicare program. 


MEDICAL SOCIETIES, PHYSICIAN
GROUPS, RESEARCHERS, AND OTHERS
GAO CONTACTED
========================================================== Appendix II


   MEDICAL SOCIETIES AND OTHER
   RELATED PHYSICIAN GROUPS
-------------------------------------------------------- Appendix II:1

American Academy of Family Physicians
American College of Emergency Physicians
American College of Rheumatology
American College of Surgeons
American Hospital Association
American Medical Association
American Osteopathic Association
American Society of Internal Medicine
The Cleveland Clinic Foundation
The Mayo Foundation
Practice Expense Coalition (which represents 43 medical specialty
organizations, including the American College of Cardiology, American
Academy of Ophthalmology, and American Society of General Surgeons)


   RESEARCHERS AND OTHERS
-------------------------------------------------------- Appendix II:2

Abt Associates
Coopers & Lybrand
Gary Siegel Organization, Inc.
Integrated Healthcare Information Systems, Inc. 
The Lewin Group
Medical Group Management Association


DESCRIPTION AND ANALYSIS OF HCFA'S
LINKING METHODOLOGY
========================================================= Appendix III

The development of an RVU system requires that all services be
directly comparable on a common scale.  With different panels of
physicians evaluating different codes, HCFA was concerned about
whether the labor time estimates developed by its clinical practice
expense panels (CPEP) were, in fact, directly comparable.  In other
words, if some CPEPs overestimated labor times while others
underestimated labor times, it might be necessary to normalize those
estimates to make them comparable and ensure that the relative
rankings among CPEPs would be correct. 

To assess the consistency of data from different CPEPs, HCFA assigned
several hundred codes to more than one CPEP and referred to these as
redundant codes.  HCFA found that the panel estimates for a redundant
code often differed.  For example, for removal of the thyroid (CPT
code 60270), the general surgery and otolaryngology CPEPs differed in
their estimates of total administrative labor time by 116
minutes--375 minutes versus 259 minutes.  Similarly, their estimates
of the total clinical labor time differed by 109 minutes--537 minutes
versus 646 minutes.  In another example--the partial removal of an
esophagus (CPT code 43117)--the general surgery and cardiothoracic
CPEPs differed in their estimates of total administrative labor time
by 90 minutes--375 minutes versus 465 minutes.  The two CPEPs'
estimates of total clinical labor time required for this procedure
differed by 950 minutes--697 minutes versus 1,647 minutes.  HCFA
staff believe that some CPEPs had higher labor estimates than others
because they included physician work in their estimates for practice
expenses or because they double counted some activities that staff
may do simultaneously. 

While recognizing these differences in the labor estimates, HCFA
concluded that the relationship among codes was generally constant
from panel to panel.  Specifically, after observing the CPEPs and
reviewing different CPEPs' labor estimates for redundant codes, HCFA
decided that, despite differences in the absolute labor time
estimates for a given code, the relative rankings of redundant codes
among CPEPs were generally similar.  That is, the relationships or
ratios between the estimates for pairs of codes were generally
similar.  For instance, if two CPEPs evaluated codes A and B, the
first CPEP's labor estimates might always be twice those of the
second CPEP; for example, 70 minutes versus 35 minutes for code A and
120 minutes versus 60 minutes for code B.  This means that any given
CPEP's estimates for a set of codes would differ from another CPEP's
estimates by a generally constant percentage. 

HCFA utilized a statistical approach called regression that used the
redundant codes to normalize--or "link"--the labor estimates of all
the CPEPs to make them comparable.  A linking regression was also
used during development of the physician work RVUs.  Researchers at
the Harvard School of Public Health obtained physician work estimates
for different codes from panels of physicians in different medical
specialties and used a similar linking regression to normalize the
physician work estimates. 

In the practice expense linking methodology, the regression analysis
produced two adjustment factors for each CPEP--one for clinical labor
estimates and a second for administrative labor estimates.  For
example, all of the clinical labor estimates for the ophthalmology
CPEP were reduced by multiplying them by an adjustment factor of
0.73, while all the administrative labor estimates were reduced by
multiplying them by a factor of 0.46.  Similarly, the clinical and
administrative labor estimates for the obstetrics and gynecology CPEP
were reduced by multiplying them by factors of 0.88 and 0.51,
respectively.  Generally, the adjustment factors produced larger
reductions in the administrative labor estimates than in the clinical
labor estimates.  For instance, the administrative labor estimates
for CPEPs 8 and 15 were reduced by 76 and 80 percent, respectively. 
See table III.1 for a listing of the linking adjustors. 



                        Table III.1
          
            Linking Adjustment Factors, by CPEP

                                    Clinical  Administrati
                                  labor time      ve labor
                                     linking  time linking
                                  adjustment    adjustment
CPEP                                  factor        factor
------------------------------  ------------  ------------
CPEP 1: Integumentary and               0.76          0.52
 physical medicine
CPEP 2: Male genital and                0.42          0.38
 urinary
CPEP 3: Orthopedics                     0.43          0.31
CPEP 4: Obstetrics and                  0.88          0.51
 gynecology
CPEP 5: Ophthalmology                   0.73          0.46
CPEP 6: Radiology                       0.78          0.48
CPEP 7: Evaluation and                  1.00          1.00
 management\a
CPEP 8: General surgery                 0.45          0.24
CPEP 9: Otolaryngology                  0.46          0.34
CPEP 10: Miscellaneous                  0.85          0.72
 internal medicine
CPEP 11: Gastroenterology               0.77          0.39
CPEP 12: Cardiothoracic and             0.50          0.24
 vascular surgery
CPEP 13: Cardiology                     0.74          0.44
CPEP 14: Pathology and                  1.00          1.00
 anesthesia
CPEP 15: Neurosurgery                   0.84          0.20
----------------------------------------------------------
\a CPEP 7, as the reference panel, was assigned a value of 1.00.  The
linking coefficients for the other CPEPs were developed in
relationship to CPEP 7. 

The fact that the linking adjustments reduced the estimates for
almost all CPEPs is not inherent to this methodology but rather
results from HCFA's choice of CPEP 7 as the "reference panel." Had a
different panel, say, CPEP 12, been chosen, the regression analysis
would have produced factors that raised some CPEPs' estimates and
lowered others.  Nonetheless, the relative relationships between
these adjusted estimates would be the same as with the factors in
table III.1.  Additionally, using a different CPEP as a reference
panel would not have resulted in different RVUs for any procedure
under this methodology. 


   HCFA'S LINKING METHODOLOGY
   CHANGED THE RANKINGS OF CODES
------------------------------------------------------- Appendix III:1

HCFA's linking adjustments significantly altered the relative ranking
of codes among CPEPs.  This follows from the large reductions in cost
estimates for some CPEPs after the linking adjustment was applied. 
This change in ranking is illustrated in the following example.  On
the basis of original CPEP data, HCFA calculated the nonphysician
labor expenses for application of a body cast (code 29035) performed
in the office at $71.52 and the nonphysician labor expenses for a
vaginal hysterectomy (code 58260) performed in a hospital at
$56.21.\20 Without any adjustments, this means that Medicare would
pay about 27 percent more in nonphysician labor expenses for a
full-body cast than it would for a vaginal hysterectomy.  However, as
shown in table III.2, the labor expenses for these two procedures are
about equal after HCFA applied its linking adjustment factors. 



                              Table III.2
                
                    Comparison of Direct Labor Cost
                   Estimates for Two Procedure Codes,
                        Before and After Linking

                                                                 Labor
                                                Initia            cost
                                                     l  Linkin  estima
                                                 labor       g      te
                                                  cost  adjust   after
                                                estima    ment  linkin
                                                    te  factor       g
----------------------------------------------  ------  ------  ------
Application of a body cast
----------------------------------------------------------------------
Clinical labor cost                             $49.37    0.43  $21.23
Administrative labor cost                        22.15    0.31    6.87
Total labor cost                                $71.52          $28.10

Vaginal hysterectomy
----------------------------------------------------------------------
Clinical labor cost                              27.38    0.88   24.10
Administrative labor cost                        28.83    0.51   14.70
======================================================================
Total labor cost                                $56.21          $28.80
----------------------------------------------------------------------

--------------------
\20 This estimate excludes the expenses of labor employed by the
hospital. 


   SOME FEATURES OF THE LINKING
   REGRESSION RAISE QUESTIONS
------------------------------------------------------- Appendix III:2

Given some of the large differences in labor estimates for redundant
codes observed among different CPEPs, some adjustments to the CPEP
data are warranted to ensure they are comparable on a common scale. 
A linking methodology based on a regression analysis for developing
physician work RVUs has been accepted by independent researchers,
albeit with suggestions for improvements and alternative
methodologies.  Such a linking regression may be appropriate to use
in developing adjustments to the CPEP data for practice expense RVUs. 
However, the appropriateness of HCFA's selected linking methodology
is related to a number of features about the data, for example, that
CPEPs generally ranked redundant codes in the same order.  Our
preliminary review of HCFA's methodology indicates that some of these
assumptions are not true for all CPEPs. 

The remainder of this section compares the actual CPEP data with the
assumptions underlying HCFA's regression model.  The more the pattern
of the CPEP data correspond to the actual pattern, the more
appropriate HCFA's regression model is likely to be.  Some
discrepancies are to be expected because of random variability in
sample data.  However, our analysis of the CPEP data indicates that
portions of the panels' data differ considerably from the assumptions
HCFA used in developing its regression model and that HCFA needs to
evaluate alternative specifications to its regression model. 

1.  HCFA's linking methodology relies on the CPEPs' ranking the
redundant codes in generally the same order.  A preliminary review of
the labor estimates for redundant codes indicates that this
consistent ranking of the codes may not be true for all of the CPEPs. 
For some pairs of CPEPs, we calculated correlation coefficients,
which measure how strongly two CPEPs' rankings of redundant codes are
correlated.  The closer the correlation coefficient is to 1, the more
highly correlated the CPEP rankings are.\21 Table III.3 shows the
correlation coefficients for 18 pairs of CPEPs. 



                              Table III.3
                
                Ranking Correlations for Selected CPEPs

                                Spearm
                                    an
                                correl
                                 ation
                                coeffi
                                cient\
CPEP pairs\a                         b    Number of redundant codes
------------------------------  ------  ------------------------------
8 + 10 (AI)                      -0.86                27
1 + 3 (AI)                       -0.32                64
12 + 13 (CO)                     -0.05                29
5 + 9 (CO)                        0.18                26
8 + 10 (AO)                       0.22                29
3 + 15 (CO)                       0.29                81
3 + 15 (AO)                       0.32                88
7 + 10 (AI)                       0.46                37
12 + 13 (AO)                      0.63                78
8 + 12 (AO)                       0.76                91
2 + 8 (AO)                        0.79                48
1 + 8 (AI)                        0.79                38
8 + 12 (CO)                       0.79                80
2 + 8 (CO)                        0.81                39
7 + 13 (AI)                       0.84                25
10 + 12 (AO)                      0.93                39
3 + 7 (AI)                        0.98                50
6 + 13 (AO)                       1.00                31
----------------------------------------------------------------------
\a AI refers to administrative, in-office labor estimates; AO refers
to administrative, out-of-office labor estimates; and CO refers to
clinical, out-of-office labor estimates. 

\b The observed significance levels for the panel pairs were less
than .01, except for panel pairs 12 + 13 (CO), 5 + 9 (CO), and 8 + 10
(AO). 

The results of the correlation analysis for these 18 pairs of CPEPs
indicate that some panels ranked redundant codes very differently
from other panels and that HCFA's assumption about consistent ranking
is questionable.  Overall, few of the 18 CPEP pairs' rankings were
strongly and positively correlated.  Three of the CPEP pairs had
negative correlations, which means that codes ranked high by one CPEP
were ranked low by the other CPEP--exactly the opposite of HCFA's
assumption.  Of the remaining CPEP pairs, we judged the four with
coefficients of 0.32 or less to be weakly correlated.  Six other CPEP
pairs with coefficients of between 0.46 and 0.79 could be considered
modestly correlated.  Only five CPEP pairs had coefficients of over
0.80, which we would consider moderately to highly correlated.  While
we did not develop correlation coefficients for all of the possible
CPEP pairs, the results from these 18 pairs contradict HCFA's
assumption that the CPEPs generally ranked redundant codes in the
same order. 

2.  In developing the linking regression, HCFA also relies on the
CPEPs' having generally similar relative ranks for the redundant
codes.  In other words, the relationships or ratios between the codes
were assumed to be similar.  For example, if two CPEPs ranked codes A
and B, the estimates of the absolute time for each code might be
different.  Nonetheless, the two codes would have the same relative
rank if the labor estimate for code A was about twice as high as that
for code B for both CPEPs.  If this similarity in ranking between
CPEPs were to hold for most codes, then the labor estimates of the
two CPEPs would generally differ by a constant percentage.  HCFA
staff told us that they did a quick review of the redundant codes and
observed, for example, that the estimates for redundant codes from
one CPEP were usually about 16 percent greater than the estimates
from a second CPEP.  However, because of time constraints and other
factors, HCFA staff did not conduct a formal, comprehensive analysis
to confirm that this relationship was true across all CPEPs and all
redundant codes. 

If labor time estimates for redundant codes generally differ between
CPEPs by a constant percentage, then the ratios of the redundant
estimates should be similar.  However, our comparison of the ratios
for redundant codes for selected CPEPs shows considerable variation. 
For instance, CPEPs 3 and 15 examined 81 redundant codes for their
clinical, out-of-office estimates.  The ratios of CPEP 3 to CPEP 15's
clinical labor time estimates were not constant over the 81 codes. 
Instead, these ratios ranged from 0.71 to 2.48 (see table III.4). 
Not only is this a threefold difference in the ratios, but in some
cases the labor estimates from CPEP 3 are higher than those from CPEP
15, and in some cases they are lower.  This shift from ratios
exceeding 1 to ratios of less than 1 indicates a lack of consistency
in the relationship between the estimates of these two CPEPs.  If the
differences between CPEPs are not consistent, then adjusting all
estimates within a CPEP by a fixed amount, as HCFA's linking
regression does, may not be appropriate. 



                        Table III.4
          
           Comparison of Selected Clinical, Out-
           of-Office Labor Estimates for CPEPs 3
                           and 15

                              CPEP 3--   CPEP 15--
                            orthopedic  neurosurge
                                     s          ry
CPT code                     (minutes)   (minutes)   Ratio
--------------------------  ----------  ----------  ------
Incise spinal column and           449         633    0.71
 cord--63199
Incise spinal column/              429         527    0.81
 nerves--63185
Open bone biopsy--20250            228         264    0.86
Revise spinal cord                 516         576    0.90
 ligaments--63182
Neck spine disk surgery--          419         438    0.96
 63075
Fusion of spine--22810             536         518    1.03
Low back disk surgery--            419         368    1.14
 63030
Revision of lumbar spine-          623         515    1.21
 -22224
Spine and skull spinal             613         491    1.25
 fusion--22590
Remove part, thorax                419         308    1.36
 vertebra--22101
Application of head brace-         480         329    1.46
 -20661
Repair thorax spine                633         408    1.55
 fracture--22327
Biopsy soft tissue of              409         229    1.79
 back--21925
Decompression of tibia             456         184    2.48
 nerve--28035
----------------------------------------------------------
We note that HCFA's linking regressions are expressed in terms of the
natural logarithm of the CPEP estimates, while our analysis is based
on the actual CPEP estimates.  A natural logarithm, or log, is a way
of expressing a number as an exponent of a common base.  For our
purposes, this difference has no effect.  The logarithmic
transformation simply compresses the range of variation, when the
ratio of the estimates is compared to the ratio of the log of those
estimates.  The properties we focus on--ranking, ratios being greater
than or less than one, and so on--do not depend on whether an
estimate is expressed as 449 minutes or, in natural logs, as 6.11. 


--------------------
\21 A correlation coefficient of 1 means that the estimates of two
CPEPs are perfectly correlated.  That is, if CPEP A ranks three codes
in the order 1,2,3, then CPEP B ranks the codes in the same order.  A
correlation coefficient of -1 means that the estimates have a perfect
negative correlation.  For example, if CPEP A ranks three codes in
the order 1, 2, 3, then CPEP B ranks the same codes 3, 2, 1. 


      RESIDUAL ANALYSIS SUGGESTS
      POSSIBLE BIAS IN REGRESSION
      ESTIMATES
----------------------------------------------------- Appendix III:2.1

In HCFA's regression equation, the coefficient on the CPEP variable
is critical, because HCFA uses that estimated coefficient as a
linking adjustment factor.  Under certain circumstances, statistical
estimates can be inaccurate in a way statisticians term "biased." A
correctly specified regression equation can yield unbiased estimates
of its coefficients.  These estimates then accurately reflect the
average effect of an explanatory variable (in this case, the variable
denoting the CPEP making the cost estimate) on the dependent variable
(in this case, the CPEP labor cost estimates).\22 While the predicted
values (of the dependent variable) from any regression will differ
from the actual values because of random variation caused by sampling
error, differences between predicted and actual values should not be
systematic--that is, correlated with factors not included in the
regression.  However, when a factor omitted from the regression model
is correlated with an explanatory factor that the regression does
include, the estimated regression coefficients are biased.  The
omission of a factor from HCFA's linking regression that is
correlated with both the dependent variable and the CPEP variable
would mean that the estimated linking factors were biased. 

In HCFA's linking regression, the CPEP labor estimates should differ
only by some constant percentage that reflects differences in CPEPs'
implicit scales, and other factors should not contribute to
differences.  HCFA recently contracted with an external researcher to
conduct a preliminary analysis of the residuals associated with the
regression equation.\23 The researcher found that other factors may
influence differences among CPEP estimates.  Because regression
equations are based on a sample of data, their predictions are
generally less than completely accurate, and their residuals signal
the degree of accuracy and other properties of the regression
equation.  For the regression estimates to be unbiased and have other
desirable properties, a plot of the residuals against the dependent
variable (in this case, the natural log of the CPEP labor time
estimates multiplied by wage rates) should be randomly distributed
about the horizontal axis, which represents a residual equal to zero. 
Such a random pattern indicates that the actual data do not deviate
from the regression model in any systematic way. 

Our review of the residual plots for HCFA's linking regressions
indicates that, for some CPEPs, the residuals corresponding to the
services evaluated by the CPEP are systematically related to the
broad category within which a specific procedure code is found:  A
hernia repair belongs to the "invasive procedure" category, while a
mid-level office visit belongs to the "evaluation and management"
category.  For example, with respect to the clinical labor
regression, when the residuals for CPEP 8 (general surgery) are
analyzed, the residuals for evaluation and management codes tend to
be positive but those for invasive procedures tend to be negative. 
This suggests that having fewer evaluation and management codes among
the redundant codes would probably increase CPEP 8's linking adjustor
for clinical labor.  By contrast, the residuals for CPEP 8's
administrative labor estimates display the opposite pattern--more
evaluation and management redundant codes would likely increase this
CPEP's linking factor for administrative labor.  The residuals for
all CPEPs did not exhibit these differences.  However, as the
researcher points out, it appears that for some CPEPs, the mix of
redundant codes by category of service influenced the estimated
linking factors.  Consequently, selection of a different set of
redundant codes would likely lead to different values for the linking
factors.  The extent to which these values would differ from those
that HCFA has published is unknown. 

In addition to differences in CPEP estimates related to the category
of service, another factor may be the rating scales the CPEPs
explicitly used.  According to the researcher who conducted the
residuals analysis, some differences among CPEPs' estimates might
indicate that some CPEPs' rating scales were, in effect, more
compressed, while other CPEPs evaluated codes using a scale with a
wider spread.  For example, in the regression for clinical labor
costs, the residuals for invasive redundant codes for CPEP 12
(cardiothoracic and vascular surgery) increase with the size of the
dependent variable (natural log of clinical labor costs).  This
association between residuals and labor costs suggests that, for
invasive procedures, this CPEP underestimated costs for redundant
codes whose estimated levels of clinical labor are small but
overestimated costs for codes whose estimates of clinical labor are
large (relative to other CPEPs).  That is, this CPEP's scale is
stretched out compared with those of other CPEPs. 

Such associations between residuals and the dependent variable
constitute a problem in the linking regression as currently
specified.  In estimating and using its regression model, HCFA
assumed that the difference between CPEPs' estimates is a constant
percentage for any CPEP.  If, however, some CPEPs rated the same
services differently, depending on whether the service had high labor
input or low labor input, then the CPEPs' relative rankings for
redundant codes may not be similar.  Consequently, a linking process
that adjusts all estimates within a CPEP by a constant factor may not
be appropriate. 

Although further analysis to replicate these preliminary findings
would be desirable, alternate specifications of the regression
approach to linking could mitigate the problems discussed above and
could improve the accuracy and credibility of the linkage adjustment
factors.  Specifically, a more appropriate linking regression model
would take into account potential differences in CPEPs' ratings
related to the category of service (for example, invasive procedures)
and the level of estimated labor input (low versus high).  In
addition, it might account for some of the deviations we noted
between the CPEP data and the patterns identified by HCFA. 


--------------------
\22 In HCFA's linking regression model, the dependent variable is
(the natural logarithm of) a particular CPEP's estimate of the labor
cost for a procedure.  The dependent variable is related to two
explanatory variables, which reflect measurable, systematic
influences, and a random term, which reflects the idiosyncratic,
unmeasurable factors that affected each CPEP's decisions.  In effect,
HCFA assumes that the cost estimate for a code depends on (1) which
panel made the estimate and (2) which code was evaluated--the two
explanatory variables.  The first explanatory variable is intended to
capture the systematic differences in scale between the CPEP
estimating the code and the reference CPEP.  The regressions'
coefficient estimates for the second variable are used as linking
adjustors for the CPEPs. 

\23 A residual is the difference between the predicted value of the
variable being analyzed and its observed value.  In this case, it is
the difference between the actual CPEP estimate and the value
predicted by the linking regression.  For an equation estimated on 20
observations of the dependent variable (for example, labor cost) and
independent variables (procedure code and CPEP), 20 residuals would
be generated (1 for each observation). 


DESCRIPTION AND ANALYSIS OF HCFA'S
SCALING METHODOLOGY
========================================================== Appendix IV

After it linked the CPEP estimates, HCFA conducted a second series of
data adjustments referred to as scaling.  In the aggregate, the
CPEPs' estimates imply that labor, medical supplies, and medical
equipment constitute 60 percent, 17 percent, and 23 percent,
respectively, of all direct expenses.  HCFA compared these estimates
with the AMA's Socioeconomic Monitoring System (SMS) survey data--one
of the few sources of national data on practice expenses--and found
that they differed.  The SMS data attributed significantly higher
proportions of practice expense to labor and less to equipment. 
According to the 1996 SMS data, labor, medical supplies, and medical
equipment represented 73 percent, 18 percent, and 9 percent,
respectively, of total direct expenses. 

To match the CPEP percentages with the SMS percentages, HCFA inflated
CPEP labor expenses for each code by 21 percent, inflated CPEP
medical supply expenses by 6 percent, and deflated CPEP equipment
expenses by 61 percent.\24

HCFA staff believed that scaling was necessary to ensure that the
proportions of practice expense RVUs devoted to labor, supplies, and
equipment were consistent with an external benchmark.  For example,
without scaling, HCFA would have no means to ensure that its total
labor expense estimates, as adjusted by linking and other steps in
HCFA's methodology, were appropriate.  The amount of the total labor
expenses depends upon the CPEP chosen as the reference panel because
the labor expense estimates from all CPEPs are linked to the
reference panel.  If a different CPEP had been chosen as the
reference panel, total labor expenses might have been much larger. 
Scaling thus enabled HCFA to use any CPEP as the reference panel and
still arrive at the appropriate amount of total labor expenses. 

HCFA officials also told us that scaling was necessary because of
their concerns regarding pricing of labor, supplies, and equipment. 
For example, HCFA used list prices as its basis for pricing both
supplies and equipment, but officials believe that physicians and
physician practices typically pay less than list price for these
items.  As a result, total practice expenses for supplies and
equipment were likely overstated.  Scaling, however, eliminated the
effects of using inflated pricing estimates. 

HCFA officials also said that they needed to use scaling because the
CPEPs did not provide them with data on equipment utilization rates,
thus requiring HCFA to make assumptions regarding how often an item
of equipment is used within physician practices.  Utilization rates,
in turn, affect how much of Medicare's practice expense payments
relate to equipment expenses.  Scaling provided HCFA with a cap on
the total amount of practice expenses devoted to equipment that was
not dependent upon the equipment utilization rate assumptions HCFA
used. 

Table IV.1 illustrates the impact of the scaling adjustments by using
a hypothetical example.  Before scaling, both codes A and B have $100
of direct expenses, so they would receive the same number of direct
expense RVUs.  After scaling, however, code A's direct expenses
increase to $109, while code B's decrease to $85.  These changes
reflect code A's higher labor expenses and lower equipment expenses
compared with those of code B.  Because of scaling, code A now
receives 1.3 times as many direct practice expense RVUs as code B.\25



                               Table IV.1
                
                 Illustration of Scaling Adjustments on
                               Two Codes

                                                        Scalin   Costs
                                                             g  (after
                                                  CPEP  adjust  scalin
                                                 costs      or      g)
----------------------------------------------  ------  ------  ------
Code A
----------------------------------------------------------------------
Labor costs                                        $60    1.21     $73
Supply costs                                        30    1.06      32
Equipment costs                                     10    0.39       4
======================================================================
Total                                             $100            $109

Code B
----------------------------------------------------------------------
Labor costs                                        $40    1.21     $48
Supply costs                                        20    1.06      21
Equipment costs                                     40    0.39      16
======================================================================
Total                                             $100             $85
----------------------------------------------------------------------

--------------------
\24 These scaling adjusters are simply the ratio of the SMS aggregate
percentages divided by the CPEP aggregate percentages.  HCFA computed
these ratios as follows:  73/60 = 1.21; 18/17 = 1.06; 9/23 - 1 =
-0.61. 

\25 $109/$85 equals 1.3. 


*** End of document. ***