Medicare: Graduate Medical Education Payment Policy Needs to Be
Reexamined (Letter Report, 05/04/94, GAO/HEHS-94-33).

It is widely held that the United States is not training enough primary
care physicians relative to types of physicians.  In 1961, about half of
all doctors were in primary care practice; if current trends continue,
that number could drop to about 26 percent by 2020.  At the same time,
if health care reform establishes a delivery system that incorporates
managed care, the need for primary care physicians will increase.  The
Medicare program is the primary vehicle through which the federal
government helps finance physician training and education.  Although
data are limited, some researchers argue that hospitals are using
Medicare funds to disproportionately underwrite the training of
nonprimary care physicians at a time when more primary care physicians
are needed. This report (1) describes how Medicare compensates hospitals
for the costs of graduate medical education and (2) determines the
extent of Medicare support for the graduate medical education of primary
and nonprimary care physicians.

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

 REPORTNUM:  HEHS-94-33
     TITLE:  Medicare: Graduate Medical Education Payment Policy Needs 
             to Be Reexamined
      DATE:  05/04/94
   SUBJECT:  Health care cost control
             Hospitals
             Medical services rates
             Hospital care services
             Health maintenance organizations
             Physicians
             Medical education
             Medical economic analysis
             Education or training costs
             Health insurance cost control
IDENTIFIER:  Health Security Act
             National Resident Matching Program
             HCFA Intern and Resident Information System
             HCFA Hospital Cost Report Information System
             HCFA Second National Graduate Medical Education Data 
             Collection System
             Medicare Program
             National Health Care Reform Initiative
             Clinton Health Care Plan
             
**************************************************************************
* 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.                                          *
*                                                                        *
* A printed copy of this report may be obtained from the GAO Document    *
* Distribution Facility by calling (202) 512-6000, by faxing your        *
* request to (301) 258-4066, or by writing to P.O. Box 6015,             *
* Gaithersburg, MD 20884-6015. We are unable to accept electronic orders *
* for printed documents at this time.                                    *
**************************************************************************


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


Report to Congressional Requesters

May 1994

MEDICARE - GRADUATE MEDICAL
EDUCATION PAYMENT POLICY NEEDS TO
BE REEXAMINED

GAO/HEHS-94-33

Medicare GME Payment Policy


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

  DGME - direct graduate medical education
  DRG - diagnosis-related group
  FTE - full-time equivalent
  GME - graduate medical education
  HCFA - Health Care Financing Administration
  HCRIS - Hospital Cost Report Information System
  HHS - Department of Health and Human Services
  IME - indirect medical education
  IOM - Institute of Medicine
  IRB - intern and resident to bed
  IRIS - Intern and Resident Information System
  NGMEDC - National Graduate Medical Education Data Collection
  OB-GYN - obstetrics/gynecology

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


B-254528

May 5, 1994

The Honorable David H.  Pryor
Chairman
The Honorable William S.  Cohen
Ranking Minority Member
Special Committee on Aging
United States Senate

The Honorable John Conyers, Jr.
Chairman
Committee on Government Operations
House of Representatives

It is widely stated that the United States is not training a
sufficient number of primary care physicians relative to nonprimary
care physicians.\1 In 1961, about 50 percent of physicians were in
primary care practice.  In 1990, about 33 percent of physicians were
in primary care practice, and it is estimated that if current trends
continue, the number will decrease to about 26 percent by 2020.  In
contrast, to the extent that health care reform may bring a delivery
system that incorporates managed care, the need for primary care
physicians will increase given the significant role of primary care
physicians in managed care organizations.\2

The Medicare program is the primary vehicle through which the federal
government contributes to the financing of physician training and
education, also referred to as graduate medical education (GME). 
Medicare financing of physician training and education began with the
enactment of the program in 1965; at that time, the Congress was
concerned about a shortage of physicians to serve newly insured
individuals, including those under Medicare.  In 1992, Medicare total
payments for GME amounted to $5.2 billion.  Although data are
limited, some researchers assert that Medicare funds are used by
hospitals to disproportionately support the training of nonprimary
care physicians at a time when more primary care physicians are
needed. 

Concerned about the declining ratio of primary care to nonprimary
care physicians, you asked us to assess the role of medical education
in physician specialty choice and how federal financing may influence
such choices.  In this report we will (1) describe how Medicare
compensates hospitals for the costs of GME and (2) determine the
extent of Medicare's support for the GME of primary care and
nonprimary care physicians.\3 In a separate report, we will address
the larger concerns of factors beyond Medicare GME financing that
play a role in determining the types of physicians produced in the
United States. 


--------------------
\1 Primary care generally refers to family medicine, general internal
medicine, and general pediatrics. 

\2 In managed care organizations, primary care physicians serve as
the patient's initial contact for medical referrals and comprise as
much as 50 percent of physician staff. 

\3 While this report focuses on Medicare financing of GME, we
recognize that Medicare has a limited role in the overall financing
of GME and that many physicians choose their specialty during their
undergraduate medical education (medical and osteopathic school). 


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

The Medicare program pays for about 29 percent of the total direct
costs of GME.  These payments, which amounted to $1.46 billion in
1992, are intended to compensate hospitals for Medicare's share of
the costs associated with training physicians.\4 Historically, the
Medicare program has based these payments on distributions of interns
and residents determined by hospitals.  In 1985 and 1986, the
Congress modified Medicare's payment methodology for GME in an
attempt to promote primary care training programs.  Under the
American Health Security Act of 1993, the administration has proposed
several changes to further promote primary care training.  However,
the extent to which Medicare pays for the training of primary care
and nonprimary care physicians has never been analyzed. 

For the 1989-91 period, our analysis showed that about 60 percent of
interns and residents were training in nonprimary care specialties
versus about 40 percent in primary care specialties.  About 55
percent of Medicare direct graduate medical education (DGME) payments
were associated with the training of nonprimary care interns and
residents while about 45 percent were associated with the training of
primary care interns and residents. 

However, some interns and residents in primary care training will
ultimately complete their residency training in nonprimary care and
enter practice as nonprimary care physicians.  This phenomenon is
often referred to as "branching." When branching is considered, our
analysis showed that the proportion of interns and residents
categorized as nonprimary care physicians changed from about 60
percent to about 75 percent.  The proportion of interns and residents
categorized as primary care changed from about 40 percent to about 25
percent.  (See fig.  1.)

   Figure 1:  Distribution of
   Interns and Residents in
   Training Before and After
   "Branching" Is Considered

   (See figure in printed
   edition.)

Also, when branching is considered, the proportion of DGME payments
associated with the training of nonprimary care interns and residents
changed from about 55 percent to about 72 percent, while the
proportion associated with the training of primary care interns and
residents changed from about 45 percent to about 28 percent.  (See
fig.  2.)

   Figure 2:  Distribution of
   Direct Graduate Medical
   Education Payments Before and
   After "Branching" Is Considered

   (See figure in printed
   edition.)

These distributions were primarily driven by hospitals' decisions
regarding their residency programs.  However, there is reason to
question whether hospitals should be the primary decisionmakers in
determining such distributions. 


--------------------
\4 Another $180 million went toward nursing and allied health
training programs.  Direct costs include salaries and fringe
benefits, the costs of classroom space, equipment, and overhead. 
Medicare also provides payment for indirect costs, which are the
portion of higher patient care costs due to the presence of GME
activities.  In 1992, Medicare provided about $3.56 billion in
payments for indirect costs; estimates of total indirect costs of GME
are unavailable. 


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

The Medicare program, authorized by title XVIII of the Social
Security Act, helps pay medical costs for about 32.3 million people
aged 65 years and older, as well as for about 3.8 million individuals
with disabilities.  Medicare is administered by the Health Care
Financing Administration (HCFA), within the Department of Health and
Human Services (HHS).  As part of paying for individuals'
hospitalization costs, Medicare also pays for the costs associated
with providing GME. 


      PHYSICIANS RECEIVE SPECIALTY
      TRAINING THROUGH RESIDENCY
      PROGRAMS
---------------------------------------------------------- Letter :2.1

During the fourth year of medical school, students formally elect the
medical specialty area they intend to pursue.\5 Students typically
are then matched, through the National Resident Matching Program,
with a residency training program that will prepare them for practice
in the chosen specialty area.\6 This period of training is referred
to as GME and generally takes 3 to 7 years after graduation from
medical school, depending on the specialty or subspecialty.\7 During
this time, physicians are generally called "interns" or
"residents."\8

In the primary care specialties, which include family medicine,
general internal medicine, and general pediatrics, residency training
takes 3 years.\9 After completing the training for internal medicine
and pediatrics, these physicians may choose to enter practice (as
general internists or as general pediatricians) or continue with
additional training.  Internal medicine and pediatric graduates who
pursue additional training become subspecialists.  For example, they
may become cardiologists or gastroenterologists.  Subspecialists are
required to maintain their competency in general internal medicine or
general pediatrics, as it pertains to the subspecialty area.  Thus,
internists and pediatricians who become subspecialists are generally
no longer classified as primary care physicians.\10 In contrast,
family physicians who pursue additional training do not necessarily
become subspecialists as a result of additional training.  Family
physicians may pursue additional training for added qualifications in
geriatrics, which is not a subspecialty but rather provides new
expertise.  Consequently, family physicians with added qualifications
remain primary care physicians.  Although some physicians elect to
pursue additional subspecialty training after several years of
practice, many physicians elect to pursue this training immediately
after their initial residency training. 

Residents primarily receive their training in teaching hospitals.\11
About 1,250 of the nation's more than 5,000 hospitals are categorized
as teaching hospitals.  In 1992, 89,368 interns and residents were
training in 7,065 residency programs in such hospitals throughout the
United States. 


--------------------
\5 Factors thought to influence a student's specialty choice include
the type of training experiences he or she has during medical school,
role models, and other factors such as working hours, loan
indebtedness, income, and prestige afforded by the specialty area
chosen. 

\6 This applies to allopathic (M.D.) and osteopathic (D.O.)
physicians.  Allopathic medicine is the most common form of medical
practice.  Osteopathic medicine is a form of medical practice similar
to allopathic medicine that also incorporates manual manipulation of
the body as therapy. 

\7 Some disciplines require a preliminary year that may be done in
internal medicine, surgery, or a transitional internship; e.g.,
certification to practice ophthalmology requires 1 preliminary year
of training in internal medicine and then 3 years in ophthalmology. 

\8 The term "intern" refers to osteopathic physicians in their first
year of graduate medical education, after which they are referred to
as residents. 

\9 Some studies include obstetrics/gynecology (OB-GYN) in their
definition of primary care.  Several studies have shown that other
specialists provide some primary care to their patients.  In
addition, nurse practitioners, physician's assistants, and others
make significant contributions to the provision of primary care. 
While these other primary care practitioners are important to any
discussion of physician supply, they are beyond the scope of this
study. 

\10 In some cases, nonprimary care physicians also provide some
primary care services to patients. 

\11 "Teaching hospitals" refer to hospitals with one or more graduate
medical education programs approved by the Accreditation Council for
Graduate Medical Education or the American Osteopathic Association. 


      MEDICARE PAYS FOR GME TO
      MEET COMMUNITY NEEDS
---------------------------------------------------------- Letter :2.2

GME is funded primarily through revenue generated by hospital patient
care services and, to a lesser extent, by payments from the Medicare
program.  Hospital charges are generally set at levels high enough to
cover a portion of the facilities' GME costs; private payers who pay
charges contribute toward GME costs in this way.  The Medicare
program makes separate payments to hospitals for GME using
methodologies to calculate payments for Medicare's portion of GME
costs.\12

When it established the Medicare program, the Congress acknowledged a
need for Medicare to support the financing of GME.\13 According to
the committee reports accompanying the original Medicare legislation,
Medicare support for residency training programs was viewed as
necessary to help meet the needs of the community for trained health
personnel.  At that time, increased availability of private health
insurance had stimulated public demand for health services and there
was a public perception of a shortage of health professionals. 
Efforts to provide health insurance to the elderly through Medicare
contributed to growing public and congressional concerns that this
increased demand for health services could not be met due to a
shortage of health professionals.  Because of the perceived overall
physician shortage, Medicare's original payment methodology paid the
portion of costs associated with training residents regardless of
their specialty or the length of training.\14

While the committee reports did not define "community need," Medicare
historically has based GME payments on distributions of interns and
residents determined by hospitals.\15 In effect, Medicare has relied
on hospitals to determine the specialty distribution of physicians to
be trained. 


--------------------
\12 The federal government also contributes to the financing of
graduate medical education through programs administered by the
Department of Veterans Affairs, the Department of Defense, the Public
Health Service within HHS, and through federal sharing in states'
costs of the Medicaid program. 

\13 The committee reports indicated that these educational activities
enhance the quality of care in an institution and that Medicare
should recognize these costs for reimbursement purposes until
communities undertake to bear such costs in another manner. 

\14 However, legislative changes in 1985, 1986, and 1993 had the
effect of limiting payments by specialty and based on length of
training.  (See pp.  10-11 for a complete discussion.)

\15 Hospitals determine the number and types of residency training
programs they offer within parameters set by the Accreditation
Council for Graduate Medical Education or the American Osteopathic
Association. 


      SCOPE AND METHODOLOGY
---------------------------------------------------------- Letter :2.3

To describe how Medicare compensates hospitals for the costs of
graduate medical education, we reviewed documents from HCFA and
interviewed agency officials.  To determine (1) the number and
specialty distribution of physicians in training and (2) Medicare
expenditures for GME, we analyzed data from HCFA's Intern and
Resident Information System (IRIS), Hospital Cost Report Information
System Minimum Data Set, and the Second National Graduate Medical
Education Data Collection.  Because the IRIS dataset was incomplete
at the time of our analysis, our average annual Medicare payment for
1989-91 was less than the $1.07 billion average reported by HCFA. 
However, our payment estimate represented about 78 percent of total
Medicare payments.  (See app.  I for objectives and additional
information on our sources and methodology.)

We conducted our work from March 1993 to January 1994 in accordance
with generally accepted government auditing standards. 


   MEDICARE PAYS HOSPITALS FOR A
   PORTION OF GRADUATE MEDICAL
   EDUCATION COSTS
------------------------------------------------------------ Letter :3

Medicare's payment methodology for the costs of graduate medical
education has two components:  Medicare reimburses teaching hospitals
for both the direct and indirect costs of medical education.\16 These
payments are intended to compensate hospitals for Medicare's "share"
of the costs associated with providing graduate medical education.\17

The direct costs of providing medical education include salaries and
fringe benefits for residents and teaching physicians, the cost of
conference and classroom space, the cost of additional equipment and
supplies, and allocated overhead costs.  The indirect cost of medical
education is the portion of the higher patient care costs at teaching
hospitals thought to be due to such factors as increased diagnostic
testing, increased number of procedures performed, higher staffing
ratios, and increased record keeping.  While indirect medical
education (IME) payments were intended to compensate hospitals for
higher costs attributable to the involvement of interns and residents
in patient care, they are also used to compensate teaching hospitals
for the higher costs associated with their urban location, treating
more severely ill patients, and treating a disproportionate share of
low-income patients.\18 Thus, IME payments were not included in our
analysis because they are sometimes used to compensate for costs
other than teaching costs. 

Both direct and indirect payments are calculated annually for
hospitals based on formulas using fixed base-costs and driven by the
number of full-time equivalent (FTE) residents and the proportion of
Medicare days of care.  (See figs.  3 and 4.) Thus, the amount of
Medicare funds received by each hospital is determined, in part, by
the number of residents that each hospital recruits and the
proportion of training time interns and residents spend in the
institution. 

   Figure 3:  Direct Graduate
   Medical Education Payment
   Formula

   (See figure in printed
   edition.)

Sources:  HCFA and Committee on Ways and Means, U.S.  House of
Representatives. 

   Figure 4:  Indirect Medical
   Education Payment Formula

   (See figure in printed
   edition.)

Sources:  HCFA and Committee on Ways and Means, U.S.  House of
Representatives. 

In fiscal year 1992, Medicare's payments to teaching hospitals for
graduate medical education amounted to about $5.20 billion, of which
$1.64 billion represented payments for direct medical education costs
and other educational activities.\19 In 1991, Medicare payments
equaled about 29 percent of the total direct costs of graduate
medical education.  About $3.56 billion represented payments for the
indirect costs of medical education. 


--------------------
\16 Medicare considers any hospital with residents enrolled in an
approved GME program to be a "teaching hospital."

\17 Following implementation of Medicare's prospective payment
system, the Congress replaced retrospective, reasonable cost
reimbursement for GME (which had applied to direct payments) with
formula payments based on each hospital's per resident costs.  This
change was designed, in part, to restrict the growth in costs per
resident. 

\18 We previously reported that IME payments to teaching hospitals
are too high and that the indirect teaching adjustment should be
reduced from the current statutory level.  Though we are concerned
about the effect of reducing the indirect teaching allowance to
high-charity hospitals, we do not believe that the indirect teaching
adjustment is the appropriate vehicle for addressing hospitals'
charity care burdens.  (See Medicare:  Indirect Medical Education
Payments Are Too High [GAO/HRD-89-33, Jan.  5, 1989].) The
Congressional Budget Office, the Prospective Payment Assessment
Commission, and HHS have also reported that IME payments are too
high. 

\19 Of this $1.64 billion in direct medical education payments, about
$1.46 billion supported physician training programs and about $180
million supported nursing and allied health training programs. 


      CHANGES IN LEGISLATION
      ATTEMPT TO PROMOTE PRIMARY
      CARE TRAINING
---------------------------------------------------------- Letter :3.1

Since enacting Medicare, the Congress has modified the payment method
for GME for several reasons.  Among these reasons was the desire to
enhance the incentives for training in primary care.\20 This was done
because of a perception that Medicare payments were being used to
provide greater support to nonprimary care training.  To this end,
the Congress made three changes to the payment method for direct
costs. 

  The Consolidated Omnibus Budget Reconciliation Act of 1985 limited
     full payment for direct costs associated with training beyond
     initial residency, placing some disincentive on subspecialty
     training.\21

  The Omnibus Budget Reconciliation Act of 1986 authorized Medicare
     to recognize--for payment purposes for direct costs--training in
     nonprovider settings under limited conditions.\22 Prior to this,
     Medicare did not recognize the costs of training in nonprovider
     settings.  Because primary care residents spend more time in
     nonprovider settings, the change was designed to enhance the
     incentives for training in primary care.  In addition, the
     change was important because of the growing trend of treating
     patients in nonprovider settings. 

  The Omnibus Budget Reconciliation Act of 1993 provided that GME
     payments for interns and residents not in a primary care or
     OB-GYN training program will not receive the GME inflation
     update during fiscal years 1994 and 1995.  This is likely to
     result in a permanent difference in rates between primary care
     and most nonprimary care training programs. 

Under the American Health Security Act of 1993, the administration
has proposed several changes in reimbursement for GME costs in order
to refocus federal support on primary care.  The stated reason is
that ensuring quality health care and access for all Americans
requires shifting the balance in GME from nonprimary care to primary
care. 


--------------------
\20 The other reasons were to restrict the growth in costs per
resident and to limit the participation of less qualified foreign
medical graduates. 

\21 Full Medicare payment was limited to the period required for
initial board certification plus 1 year, not to exceed 5 years
(initial residency), with the exception of geriatrics.  As a
practical matter, this results in full payments for either 4 or 5
years.  After that period of residency, payments are reduced. 

\22 Medicare pays for training in nonprovider settings if a hospital
incurs all or substantially all of the costs of the training program. 


      BARRIERS TO PRIMARY CARE
      TRAINING PERSIST
---------------------------------------------------------- Letter :3.2

Despite legislative changes, barriers to primary care training
persist in Medicare's payment method for the direct costs of graduate
medical education.  First, Medicare continues to rely primarily on
hospitals to determine the specialty distribution of physicians to be
trained.  Hospitals make those decisions based largely on hospital
service needs rather than other methods that might account for the
full range of health and medical needs of the community.  Second,
under current HCFA rules, only hospitals and hospital-based providers
are eligible to receive DGME payments.  An Institute of Medicine
(IOM) study reported that because of changes in the health care
system, hospitals are less suitable than ambulatory settings as
principal training sites, in particular for primary care physicians
who spend most of their career in ambulatory settings.\23 The IOM
study further stated that because payments for health services are
skewed to favor inpatient care and specialty education, it is
difficult for educators to increase the time that residents spend in
outpatient settings.  When residents do train in outpatient or
ambulatory care settings, Medicare does not always recognize the
direct costs of such training; Medicare limits DGME funding for
training in ambulatory care settings to those training programs for
which a hospital incurs all or substantially all of the costs of the
ambulatory care training program.  This places primary care programs
at a financial disadvantage because of those programs' extensive use
of ambulatory care sites for training, including those in nonhospital
settings.  In addition, hospital-based training can create an
environment that may influence residents in internal medicine and
pediatrics to subspecialize, thus diminishing the primary care
pool.\24


--------------------
\23 See Primary Care Physicians:  Financing Their GME in Ambulatory
Settings, Institute of Medicine (1989). 

\24 An HHS study indicated that the training environment exerts an
independent effect in directing residents in internal medicine and
pediatrics into primary care careers.  (See Assessment of the
Development and Support of Primary Care Residency Training:  General
Internal Medicine and Pediatrics, HHS (Sept.  30, 1987).)


   MORE RESIDENTS TRAIN IN
   NONPRIMARY CARE
------------------------------------------------------------ Letter :4

Our analysis of the IRIS dataset for the 1989-91 period revealed that
a greater proportion of interns and residents were receiving training
in nonprimary care specialties than in primary care specialties. 
About 60 percent of interns and residents were receiving training in
nonprimary care specialties while the remaining 40 percent were
receiving training in primary care.\25

It is important to note that these results represent a "snapshot" of
the specialty distribution of interns and residents in training in
the 1989-91 period--not necessarily the specialty distribution of
physicians in practice after training has been completed. 


--------------------
\25 For the purposes of this analysis, we included general internal
medicine, general pediatrics, family practice, general practice
(osteopathic), as well as preventive medicine, and public
health/preventive medicine in the definition of "primary care." All
other specialties and subspecialties were included in the definition
of "nonprimary care."


   MORE MEDICARE DGME PAYMENTS
   SUPPORT NONPRIMARY CARE
   TRAINING
------------------------------------------------------------ Letter :5

Our analysis showed that a greater proportion of Medicare DGME
payments are used by hospitals to support the training of interns and
residents in nonprimary care specialties.  In the 1989-91 period,
about 55 percent of DGME payments were associated with the training
of interns and residents in nonprimary care and about 45 percent of
DGME payments were associated with the training of interns and
residents in primary care.  During this period, the average annual
Medicare DGME payment for the training of interns and residents in
nonprimary care was about $453 million and the average annual payment
for the training of primary care interns and residents was about $380
million.\26


--------------------
\26 DGME payments for physician training programs in 1989, 1990, and
1991 totaled $1.03 billion, $1.073 billion, and $1.10 billion,
respectively.  Because of incomplete data in the IRIS dataset, our
payment estimates are less than the actual amounts. 


   A GREATER PROPORTION OF
   MEDICARE DGME PAYMENTS SUPPORTS
   NONPRIMARY CARE TRAINING WHEN
   BRANCHING IS CONSIDERED
------------------------------------------------------------ Letter :6

The proportion of interns and residents in nonprimary care training,
and associated DGME payments, increased when branching was
considered.  In this case, the proportion of interns and residents
categorized as nonprimary care was about 75 percent versus 25 percent
categorized as primary care; the proportion of DGME payments
associated with nonprimary care training was about 72 percent versus
about 28 percent with primary care. 

Our objective for this analysis was to determine the distribution of
interns and residents, and associated DGME payments, according to the
type of training they would ultimately complete (i.e., primary care
or nonprimary care).  In our original analysis, we categorized
interns and residents in general internal medicine and general
pediatrics as training in primary care.  However, a proportion of
these interns and residents who train in primary care specialties
will ultimately complete their training in nonprimary care
subspecialties, a phenomenon referred to as "branching."\27
Therefore, we estimated the number of primary care interns and
residents (in general internal medicine and general pediatrics) who
will branch, and we reallocated them to nonprimary care.  (See app. 
I for a detailed description of our methodology.)

When branching is considered, the proportion of interns and residents
categorized as training in nonprimary care increased from about 60
percent to 75 percent.  The proportion of interns and residents
categorized as training in primary care decreased from about 40
percent to about 25 percent. 

The proportion of DGME payments associated with the training of
nonprimary care interns and residents increased from about 55 percent
to about 72 percent.  The proportion of DGME payments associated with
the training of primary care interns and residents decreased from
about 45 percent to about 28 percent.  This change represented about
$148 million in annual DGME payments:  an increase in DGME payments
for nonprimary care from about $453 million to about $601 million,
and a decrease in DGME payments for primary care from about $380
million to about $232 million. 


--------------------
\27 We used American Board of Medical Specialties data on the number
of general and special certificates awarded to estimate the
proportion of residents in general internal medicine and general
pediatrics who pursue additional subspecialty training. 


   CONCLUDING OBSERVATIONS
------------------------------------------------------------ Letter :7

Medicare's stated purpose for supporting graduate medical education
is to meet community needs for trained health personnel.  During the
1989-91 period, Medicare funds were used to support a training
distribution of 75 percent nonprimary care interns and residents
versus 25 percent primary care.  This distribution is based primarily
upon hospital service needs.  To the extent that "community needs"
are reflected by hospitals' service needs, Medicare payments support
community needs.  There is reason to question, however, whether
hospitals should be the primary decisionmakers in determining
physician training distributions and, in effect, in defining
community need. 

Health care reform is expected to place a greater emphasis on managed
care; and, as a result, the types of physicians needed and the
settings in which they are trained are expected to change.  The
definition of community need as it relates to Medicare graduate
medical education payment policy may need to be reassessed as the
need for primary care physicians increases. 


   AGENCY COMMENTS
------------------------------------------------------------ Letter :8

HHS officials reviewed a draft of this report and generally agreed
with our findings.  (See app.  II.) HHS officials concur that
Medicare payments are being driven by hospitals' decisions regarding
their residency programs.  They noted that the Council on Graduate
Medical Education\28 is concerned that the payment methodology
provides an incentive to add residency positions based on hospital
service needs rather than societal and educational needs.  They
further noted that the administration's proposed Health Security Act
supports increasing the amount of residency training that is
performed in nonhospital settings.  They also provided technical
comments, which we incorporated as appropriate.  We provide some
additional clarification on our methodology in our response to HHS'
letter in appendix II. 


--------------------
\28 The Council on Graduate Medical Education is administered by the
Public Health Service and reports to the Secretary of HHS and the
Congress on matters related to graduate medical education. 


---------------------------------------------------------- Letter :8.1

As agreed with your office, unless you publicly announce its contents
earlier, we plan no further distribution of this report until 30 days
after its issue date.  At that time, we will send copies to others on
request.  If you have any questions about this report, please call me
at (202) 512-7119.  Other major contributors are listed in appendix
III. 

Sarah F.  Jaggar
Director, Health Financing
 and Policy Issues


OBJECTIVES, SCOPE, AND METHODOLOGY
=========================================================== Appendix I

One purpose of this review was to determine the extent of Medicare's
support for the direct costs of the graduate medical education of
primary care and of nonprimary care physicians.  We divided this goal
into three subobjectives, as follows: 

  determine the distribution of interns and residents training in
     primary care and in nonprimary care,

  determine the amount of DGME payments made in support of training
     in primary care and in nonprimary care, and

  estimate the distributions of interns and residents and of DGME
     payments that account for additional subspecialty training by
     residents in internal medicine and pediatrics. 


      CREATING A COMBINED DATABASE
------------------------------------------------------- Appendix I:0.1

To accomplish these objectives, we combined information from several
HCFA databases for fiscal years 1989, 1990, and 1991.  Specifically,
we used HCFA's Intern and Resident Information System, Hospital Cost
Report Information System (HCRIS) Minimum Data Set, and Second
National Graduate Medical Education Data Collection (NGMEDC). 

IRIS was developed by HCFA to monitor intern and resident activity
affecting Medicare direct and indirect payments for graduate medical
education.  IRIS data records contain information on training
rotations of interns and residents, including chief residents and
fellows.  Among other things, each record includes information on the
type of residency, year of residency, location of training, and
percentage of time working at that location. 

The HCRIS Minimum Data Set contains cost, financial, and other
information from the Medicare Hospital Cost Report.  The NGMEDC
contains information on graduate medical education costs and each
hospital's Medicare GME per resident reimbursement amount, as well as
information on the weighted number of full-time equivalent interns
and residents. 

To create a combined database for our analyses, we added variables
from the NGMEDC and the HCRIS Minimum Data Set to the variables from
the IRIS dataset.  We created a datafile from the IRIS dataset that
included the following information for each intern and resident: 
residency designation (specific specialty or subspecialty), year of
residency training, location (provider number) for training, and the
duration and the percentage of time at the training location.  We
linked variables from the other datasets to the IRIS datafile using
provider numbers.  Specifically, we added hospital GME per resident
amounts from the NGMEDC and the ratio of Medicare inpatient days to
total inpatient days from the HCRIS Minimum Data Set. 

For the purposes of these analyses, we categorized interns and
residents as either primary care or nonprimary care as follows: 
primary care included those who were receiving training in general
internal medicine, general pediatrics, family practice, general
practice (osteopathic), preventive medicine, and public
health/preventive medicine; nonprimary care included the interns and
residents in all other specialty and subspecialty training programs. 


      ANALYSIS OF THE DISTRIBUTION
      OF INTERNS AND RESIDENTS
------------------------------------------------------- Appendix I:0.2

To determine the distribution of interns and residents, we ran
frequencies on the combined database to determine the number and the
proportion of interns and residents in primary care residencies and
in nonprimary care residencies, for each of 3 years of data.  We then
computed the proportion of interns and residents in primary care and
the proportion in nonprimary care for the 1989-91 period. 


      ANALYSIS OF THE DISTRIBUTION
      OF DGME PAYMENTS
------------------------------------------------------- Appendix I:0.3

To determine the distribution of DGME payments, we calculated DGME
payment amounts for training in primary care and in nonprimary care
for each of 3 years of data.  We based these calculations on HCFA's
DGME payment formula: 

weighted number of FTEs x the hospital's per resident amount updated
by the Consumer Price Index x ratio of Medicare inpatient days to all
inpatient days. 

For each year in our analysis, we determined the value for the
factors in the payment formula in three steps.  First, we determined
the FTE status of each intern and resident at each hospital, based on
HCFA's rule for calculating FTE, using information on training
rotations.\1 We then determined a weight to be assigned to each FTE
intern and resident, based on HCFA's rule, using data on the year of
residency training.\2 Second, we computed each hospital's per
resident amount by updating the base period per resident amount with
the Consumer Price Index-Urban Consumers.  Third, we calculated each
hospital's ratio of Medicare inpatient days to all inpatient days. 

Through these calculations, we determined DGME payment amounts
associated with each intern and resident's training for a given year. 
We totaled DGME payments for those in primary care residencies and
for those in nonprimary care residencies, in each year 1989, 1990,
and 1991.  We then computed the proportion of total DGME payments in
1989-91 that supported interns and residents in primary care and the
proportion that supported those in nonprimary care.\3


--------------------
\1 No individual may be counted as more than one FTE.  If a resident
spends time in more than one hospital or in a nonprovider setting,
the resident counts as a partial FTE based on the proportion of time
worked at the hospital to the total time worked.  A part-time
resident counts as a partial FTE based on the proportion of time
worked compared with the average time spent by other residents
working in the same specialty program. 

\2 For interns and residents in the initial residency period (i.e.,
the number of years necessary to satisfy the requirements for
certification in a specialty or subspecialty, plus 1 year, not to
exceed 5 years, with the exception of geriatrics whose initial
residency may last up to 2 additional years), the weighting factor is
1.0.  For residents not in an initial residency, the weighting factor
is 0.5. 

\3 We calculated DGME payments for only the interns and residents in
the IRIS dataset.  Because the IRIS dataset was incomplete, our DGME
payment estimates for the 3 years analyzed were less than the DGME
expenditures reported by HCFA.  However, our estimates represented
about 78 percent of reported DGME payments. 


      ESTIMATES OF THE
      DISTRIBUTION OF INTERNS AND
      RESIDENTS AND DGME PAYMENTS,
      GIVEN BRANCHING
------------------------------------------------------- Appendix I:0.4

IRIS contains information on interns and residents and their training
assignments during the 1989-91 period.  In a given year, interns and
residents in the dataset are at different stages in training (ranging
from training year 1 to 9).  Our objective for this analysis,
however, was to determine the distribution of interns and residents
according to the type of training they would ultimately complete
(i.e., primary care or nonprimary care).  We estimated this
distribution using the following methodology: 

1.All interns and residents who were not receiving training in
general internal medicine, general pediatrics, family medicine,
preventive medicine, public health/preventive medicine, or
osteopathic general practice were counted as nonprimary care. 

2.Interns and residents in family medicine, preventive medicine,
public health/preventive medicine, and the osteopathic specialties of
general internal medicine, general pediatrics, and general practice
were counted as primary care. 

3.We allocated the remaining number of residents (those in allopathic
internal medicine and pediatrics) between primary care and nonprimary
care using estimated "branching" rates.\4 These branching rates
reflect the estimated proportion of allopathic residents being
trained in internal medicine and pediatrics (heretofore classified as
primary care) who will ultimately complete their training in
nonprimary care.  Specifically, these estimates were based on the
ratio of special certificates to general certificates awarded between
1982 and 1991 as reported by the American Board of Medical
Specialties.\5 We calculated and applied separate estimates of
branching rates for internal medicine and pediatrics (66 percent and
18 percent, respectively).\6 (See fig.  I.1.)

   Figure I.1:  Estimating the
   Distribution of Interns and
   Residents While Accounting for
   "Branching"

   (See figure in printed
   edition.)

Note:  PC = primary care.
IM-PEDS = internal medicine and pediatrics. 

We also estimated DGME payments associated with training of interns
and residents for primary care and for nonprimary care.  For each
year, we computed the average DGME payment per primary care intern
and resident before adjusting for branching and multiplied this
amount by the estimated number of interns and residents who branched,
as determined in the previous analysis.  We reallocated this amount
from the total DGME payments for primary care to the total DGME
payments for nonprimary care.  This resulted in the revised
distribution of DGME payments for primary care and nonprimary care
for the 1989-91 period.  (See fig.  I.2.)

   Figure I.2:  Estimating the
   Distribution of DGME Payments
   While Accounting for
   "Branching"

   (See figure in printed
   edition.)

Note:  PC = primary care.
IM-PEDS = internal medicine and pediatrics. 


--------------------
\4 While our analysis is based on both allopathic and osteopathic
interns and residents, osteopathic physicians are not included in the
American Board of Medical Specialties certification data on which our
estimated branching rates were based.  Thus we did not estimate the
effect of branching on osteopathic primary care interns and
residents.  However, osteopathic primary care interns and residents
equaled only about 7 percent of total primary care interns and
residents allocated. 

\5 General certification is the initial or basic certification
conferred on individuals who meet the requirements for certification
in a specified field of medical practice.  Special certificates
designate special training in a subspecialty field.  For example, an
individual may be granted a general certificate in the primary care
specialty of internal medicine and then granted a special certificate
in one of the approved internal medicine subspecialties. 

\6 Our branching estimate for internal medicine is comparable to
other estimates stated in the literature; however, our branching
estimate for pediatrics is somewhat lower than others stated in the
literature. 


      DATABASE LIMITATIONS
      PERTINENT TO OUR ANALYSIS
------------------------------------------------------- Appendix I:0.5

At the time of our analysis, HCFA indicated that the IRIS dataset was
incomplete because some hospitals, although required, did not provide
HCFA with data for fiscal years 1989 through 1991.  We estimated,
however, that DGME payments for interns and residents in our dataset
represented about 78 percent of reported total DGME payments. 

Another shortcoming in IRIS indicated by HCFA was the possibility of
problems with the coding for "residency type." HCFA is not certain
whether the residency type reported for some residents is the
resident's initial residency or a prerequisite that is required for
the resident's initial residency.  For example, ophthalmology
residents (nonprimary care) are required to train first for 1 year in
internal medicine (primary care).  During this first year, some of
these residents may have been inaccurately reported as internal
medicine residents.  This type of error would result in overstating
the number of interns and residents in primary care and understating
the number in nonprimary care. 




(See figure in printed edition.)Appendix II
COMMENTS FROM THE DEPARTMENT OF
HEALTH AND HUMAN SERVICES
=========================================================== Appendix I



(See figure in printed edition.)

See comment 1. 

See comment 1. 

See comment 1.  See also p.  13. 

Now on p.  12. 

See comment 2. 



(See figure in printed edition.)

See p.  8. 

Now on p.  8. 

Now on p.  12. 

See comment 3. 


The following are GAO's comments on the Department of Health and
Human Services' letter dated January 4, 1994. 


   GAO COMMENTS
--------------------------------------------------------- Appendix I:1

1.We agree and acknowledge in the report that the distribution of
Medicare DGME payments is being driven by hospital decisions
regarding their residency programs.  We conclude that DGME payments
have been used by hospitals to fund a training distribution in which
a greater proportion of interns and residents are in nonprimary care
training than are in primary care training. 

2.We began our analyses in March 1993; the Omnibus Budget
Reconciliation Act of 1993 was approved in August 1993.  We recognize
that the act, for Medicare payment purposes, includes geriatric
medicine in its definition of primary care.  Because geriatric
medicine was not included in the definition of primary care in our
analyses, we reviewed the effect that including geriatric medicine
would have on our results.  We concluded that our results would not
be materially affected since interns and residents in geriatric
medicine equaled less than 1 percent of the interns and residents in
primary care. 

3.We are aware that the weight assigned to a resident for payment
calculations is 0.5 after the initial residency; however, this fact
does not have a material effect on our estimations of the DGME
payment distribution that accounts for "branching" (i.e., "leakage"). 
The objective of this analysis was to estimate the distribution of
current DGME payments for primary care versus nonprimary care based
on the type of training that residents would ultimately complete. 
Accordingly, a proportion of residents in their initial residency
period, and the associated DGME payments (which are derived using 1.0
as the assigned weight per resident), were allocated to nonprimary
care based on the estimation that they would ultimately complete
their training in nonprimary care (i.e., "branching" would occur). 
The fact that the weight assigned to such residents, for payment
calculations, is 0.5 after "branching" (a future event) is not an
issue given our methodology. 


MAJOR CONTRIBUTORS TO THIS REPORT
========================================================= Appendix III

Rose Marie Martinez, Assistant Director
Andrew K.  Bhak, Evaluator-in-Charge, (202) 512-7134
Carolyn Cocotas
Robert DeRoy
