Health Care: Employers Urge Hospitals to Battle Costs Using Performance
Data Systems (Letter Report, 10/03/94, GAO/HEHS-95-1).

Many large employers have become increasingly concerned about the wide
variation in hospital costs across their communities. They believe that
they may be paying for care that is not being delivered in the most
efficient way. At the same time, they contend that they lack the
information needed to assess the value of health care they are paying
for. To meet this need, some employers, such as Walt Disney World in
Orlando and Proctor and Gamble in Cincinnati, have organized health care
coalitions to help them make better purchasing decisions. By developing
and sharing comparable information on hospital performance in their
communities, they hope to make health care providers more accountable
for the services they deliver. This report examines (1) the reasons why
employer coalitions and hospitals are using comparative performance
measurements systems and (2) whether these systems report the
information employers and hospitals need to compare outcomes.

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

 REPORTNUM:  HEHS-95-1
     TITLE:  Health Care: Employers Urge Hospitals to Battle Costs Using 
             Performance Data Systems
      DATE:  10/03/94
   SUBJECT:  Quality assurance
             Health care cost control
             Hospital care services
             Cost effectiveness analysis
             Medical information systems
             Management information systems
             Data collection operations
             Data integrity
             Hospital administration
             Comparative analysis
IDENTIFIER:  Cincinnati (OH)
             Cleveland (OH)
             Orlando (FL)
             Cleveland Health Outcome Indicator of Care Evaluation System
             Acute Physiology, Age, Chronic Health Evaluation System
             Patient Viewpoint Survey
             Medical Illness Severity Grouping System
             
******************************************************************
** 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 Requesters

October 1994

HEALTH CARE - EMPLOYERS URGE
HOSPITALS TO BATTLE COSTS USING
PERFORMANCE DATA SYSTEMS

GAO/HEHS-95-1

Hospital Performance Measurement Systems

(101265)


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

  AIM - Acuity Index Method
  APACHE - Acute Physiology, Age, Chronic Health Evaluation
  APR-DRG - All Patient Refined Diagnosis Related Group
  CHOICE - Cleveland Health Outcome Indicator of Care Evaluation
  DRG - diagnosis related group
  HCFA - Health Care Financing Administration
  ICD-9-CM - International Classification of Diseases, Ninth
     Revision, Clinical Modification
  MEDPAR - Medicare Provider Analysis and Review

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


B-252283

October 3, 1994

The Honorable Michael A.  Andrews
House of Representatives

The Honorable Jim Cooper
House of Representatives

The Honorable Charles W.  Stenholm
House of Representatives

Many large employers have become increasingly concerned about the
wide variation in hospital costs across their communities.  They
believe that they may be paying for care that is not being delivered
in the most efficient manner.  At the same time, they contend that
they lack the information to assess the value of health care services
they are paying for.  To meet this need, employers in some
communities--such as Walt Disney World in Orlando, Proctor and Gamble
in Cincinnati, and LTV Steel in Cleveland--have organized health care
coalitions to help them make better purchasing decisions.  By
developing and sharing comparable information on hospital performance
in their communities, they hope to make health care providers more
accountable for the services they deliver. 

Because of growing interest in obtaining comparative information
about health care providers, you asked us to review the experience of
communities in which employer coalitions have encouraged hospitals to
adopt automated performance measurement systems.  Specifically, we
examined (1) the purposes for which employer coalitions and hospitals
are using comparative performance measurement systems and (2) whether
these systems report the information employers and hospitals need to
compare outcomes. 


   BACKGROUND
------------------------------------------------------------ Letter :1

In general, performance measurement systems generate large
computerized databases that compare providers treating similar
patients.  They produce data on various indicators of hospital
performance--typically how long patients stay in the hospital (length
of stay), death rates by category of condition (mortality rates), and
treatment charges.  Earlier approaches to measuring a provider's
performance did not account for the fact that patients of the same
age, sex, and general diagnosis could have different outcomes after
surgery because of the severity of the patients' conditions.  For
example, a 55-year-old woman admitted for colon cancer with an
intestinal obstruction from a large colon tumor is not likely to have
as favorable an outcome as a patient of the same age and sex with
only superficial colon cancer.  Hospitals contended that because the
systems did not adjust for patients' severity of condition, charge
and outcome comparisons were invalid.\1 Hospitals faring poorly in
comparison with other hospitals claimed their patients were sicker
and thus required more costly care and had an increased likelihood of
poor outcomes.  To address this concern, a number of private vendors
have developed computerized data systems designed to account for
differences in patients' severity of illness. 

Cincinnati, Cleveland, and Orlando are cities in which employer
coalitions selected particular vendors and encouraged the local
hospitals to purchase their severity-adjusted performance measurement
systems.  Cincinnati hospitals use Iameter's Acuity Index Method
(AIM).  Cleveland hospitals use three instruments:  (1) the Cleveland
Health Outcome Indicator of Care Evaluation (CHOICE) system for
general medical and surgical admissions, designed by Michael Pine and
Associates; (2) Apache Medical System's Acute Physiology, Age,
Chronic Health Evaluation (APACHE III) system for intensive care
admissions; and (3) the Patient Viewpoint Survey, a patient
satisfaction survey administered by NCG Research.  Orlando hospitals
use MediQual's Medical Illness Severity Grouping System (MedisGroups
II).  Table 1 presents the basic characteristics of the systems used
in each community. 



                                Table 1
                
                     Characteristics of Performance
                   Measurement Systems Used in Three
                              Communities

                  Cincinnati        Cleveland         Orlando
----------------  ----------------  ----------------  ----------------
Systems used      AIM               CHOICE, APACHE,   MedisGroups
                                    and Patient
                                    Viewpoint Survey

Type and source   Administrative    Clinical data     Clinical data
of data           data retrieved    abstracted from   abstracted from
collected         from patient      medical records   medical records
                  discharge forms   and survey of a
                                    sample of
                                    patients

Performance       Charges                             Charges
indicators        Length of stay    Length of stay    Length of stay
reported                                              Morbidity
                  Mortality rate    Mortality rate    rate\a
                                    Patient           Mortality rate
                                    satisfaction

Level of          Major diagnostic  Types of          Specific
clinical detail   categories and    services          diagnoses
reported          specific high-    (intensive care,
                  volume diagnoses  general medical,
                                    and general
                                    surgical) and
                                    major diagnostic
                                    categories

Level of          Hospital and      Hospital only     Hospital and
provider detail   physician                           physician
reported
----------------------------------------------------------------------
\a MedisGroups defines morbidity as occurring if a second severity
rating taken midstay indicates that the patient has not responded to
treatment. 

Systems based on administrative data, as used in Cincinnati, and
clinical data, as used in Orlando and Cleveland, are fundamentally
different.  To make severity adjustments, administrative systems use
information on patient characteristics (age and sex), diagnoses
(primary and secondary), and procedures performed.  Because hospitals
routinely develop administrative data for billing purposes, little
additional cost is associated with collecting these data as inputs to
performance measurement systems.  In contrast, clinically based
systems use more medically precise data, such as laboratory test
values, radiology reports, and physiology readings (patient
temperature, blood pressure, and so on), to make severity
adjustments.  As a result, clinically based systems require trained
staff to review and abstract each medical record and are therefore
more expensive than administrative systems.\2

All of the systems are confined to assessing care for patients
admitted to the hospital.  None is able to assess care provided in
ambulatory settings--such as physicians' offices, hospital outpatient
clinics, or freestanding laboratories; such care accounts for more
than half of most employers' health care costs.  Also, the systems do
not identify whether a hospital admission is appropriate, although
some hospital admissions may be unnecessary or care could be more
effectively provided elsewhere.  Finally, none of the systems
provides a single summary "grade" for hospitals to highlight that
hospitals perform well for some types of care but not for others.\3


--------------------
\1 In 1986, the Health Care Financing Administration (HCFA) started
releasing hospital mortality rates as an indicator of hospitals'
quality of care.  Hospitals complained, however, that HCFA mortality
data were unreliable because they did not accurately adjust for
differences in patients' severity of illness.  HCFA has continued to
refine its methods, but in 1993 delayed release of hospitals'
mortality data because of concerns over the accuracy of the severity
adjustments. 

\2 The burden of collecting clinical data can be simplified if
medical records are maintained electronically.  The Cleveland Health
Quality Choice Program reported that it is beginning to develop an
electronic format for collecting laboratory test values and other
clinical data, but currently most medical records are manually
abstracted. 

\3 For a review of initiatives to provide summary information
comparing health plans' costs and quality of care, see Health Care
Reform:  "Report Cards" Are Useful But Significant Issues Need to Be
Addressed (GAO-HEHS-94-219, Sept.  29, 1994). 


   SCOPE AND METHODOLOGY
------------------------------------------------------------ Letter :2

To address our objectives and in consultation with your staffs, we
selected Cincinnati, Cleveland, and Orlando as case studies. 
Although hospitals nationwide have had experience with performance
measurement systems, the selection and use of severity-adjusted
performance measurement systems in each of these cities resulted from
an employer coalition initiative.  In addition, each community
adopted different systems, reflecting a variety of hospital data
systems currently available.  Because of the small number of sites
analyzed and the unique circumstances of each, the experience of
these three cities may not be representative of performance
measurement projects elsewhere in the nation. 

In each community, we interviewed employer coalition directors,
individual employers' health benefit managers, and hospital and
physician associations' executives.  We also visited several
hospitals in each city that represented a range of sizes and
types--including community and teaching hospitals--and met with
hospital staff, including administrative officers and physicians
serving as clinical department chiefs.\4 We contacted the principal
developers of the severity adjustment systems and met with other
experts in the field of outcomes research. 

Finally, to develop trend data, we sought data on costs, charges, and
length of stay for specific medical conditions from each hospital
participating in the three community projects.  However, difficulties
with access, completeness, and timing limited our use of this
information.  We surveyed each hospital for information on costs,
charges, and payments from 1988 to 1993, but we were unable to use
these data because of partial and inconsistent responses.  We also
obtained hospital data on charges and length of stay collected by the
Florida and Ohio governments.  However, these data were available
only through 1992, and in Ohio the data only included privately
insured patients (Medicare and Medicaid patients were excluded) for
the 100 highest-volume diagnosis related groups (DRG) reported by the
hospital.  Therefore, we were often restricted to anecdotal
information provided during our interviews that could not be
verified.  We conducted our review between May 1993 and July 1994 in
accordance with generally accepted government auditing standards. 


--------------------
\4 To obtain the full cooperation of hospitals involved in these
initiatives, we provided pledges of confidentiality.  Therefore, we
have not identified hospitals by name in this report. 


   RESULTS IN BRIEF
------------------------------------------------------------ Letter :3

In the communities we reviewed, the introduction of severity-adjusted
performance measurement systems has given hospitals an impetus to
initiate efficiency improvements.  Employers plan to use the
information produced by these systems to steer patients to
cost-effective providers, while hospitals can use the information to
help manage treatment costs.  Because employer coalitions have only
recently encouraged communitywide use of the systems, the cost-saving
potential of the initiatives remains an unresolved issue.  Also, in
light of data quality concerns, employers should use caution in
interpreting systems' results. 

Employer coalitions in Cincinnati, Cleveland, and Orlando have made
severity-adjusted performance measurement systems an important
element of their communities' cost containment strategies.  That is,
employers will direct or encourage their employees to hospitals that
demonstrate superior cost performance without compromising quality of
care.  By influencing employer health care purchasing decisions,
performance measurement systems may increase hospital accountability
and competition. 

In the communities we examined, hospitals generally regard the
systems as one of several useful internal tools for identifying
efficiency problems.  Many hospitals have attempted to reduce their
average charge and length of stay by modifying their patterns of
care.  These efforts include (1) shifting certain services into
ambulatory settings (where costs may be lower); (2) using fewer or
less costly supplies and pharmaceuticals; and (3) reducing,
substituting, or rescheduling treatments and services.  Anecdotal
evidence linking the performance measurement systems to cost
reductions for certain services implies a potential to generate
savings for individual hospitals.  However, statistical evidence to
date on the effectiveness of these systems as a sustainable,
communitywide cost containment tool is neither complete nor
conclusive.  This may be explained by (1) the fact that the systems
have not been in use long enough to produce sufficient evidence of
cost savings and (2) the difficulty in isolating the influence of
these systems from other factors contributing to hospital
improvements. 

Hospitals, physicians, and experts in the field of outcomes research
caution employers that the results of these systems should not be the
sole guide for health care purchasing decisions.  Although the
systems allow hospitals to be compared more fairly than previous
approaches because they adjust for severity of illness, the systems'
reports are incomplete sources of information on costs and quality. 
In particular, concerns exist about the selection of performance
indicators reported to employers and the validity of the severity
adjustments applied to the data.  Although the systems offer improved
comparisons of mortality rates and length of hospital stays, many of
the most meaningful patient outcomes for employers, such as
functional status and time needed to return to work, are not
documented by the systems.  Furthermore, certain hospitals may still
be disadvantaged in the comparisons because of their unique
characteristics and limitations in the severity adjustment
methodology applied. 


   EMPLOYERS USE PERFORMANCE
   MEASUREMENT SYSTEMS TO
   STIMULATE HOSPITAL EFFICIENCY
------------------------------------------------------------ Letter :4

Vendors have recently begun to market their performance measurement
systems to employer coalitions, stressing the systems' capability to
highlight opportunities for aggregate hospital cost savings. 
Employers in Cincinnati, Cleveland, and Orlando have succeeded in
prompting most local hospitals to use a performance measurement
system selected by the employers.  Many hospitals have incorporated
the data from these systems in their management of clinical
performance, and some hospitals have begun to make practice
improvements that they attribute to the employer coalitions'
initiatives.  Because many hospitals have only recently adopted these
systems and few employers have yet to use the comparative information
on hospital performance to select providers for their health care
networks, it is too early to fully assess the cost containment
potential of these employer initiatives. 


      COALITIONS PROVIDE STRONG
      INCENTIVES FOR HOSPITALS TO
      IMPROVE EFFICIENCY
---------------------------------------------------------- Letter :4.1

Coalitions have sought to have all local hospitals participate in the
performance measurement programs with the goal of comparing hospital
performance across their communities.  Employers in the coalitions
represent a substantial share of the communities' patients and either
contract directly with hospitals or make selections through insurers'
prearranged network plans.  To increase hospital participation, the
coalition employers announced that they would be reluctant to deal
with hospitals that choose not to use the coalition-sponsored
performance measurement system. 

The threat of losing coalition employers' business has been a
powerful incentive, prompting nearly all of the hospitals in the
communities we visited to purchase the systems.  In Cincinnati, where
individuals insured by coalition employers constitute about 13
percent of the population, all the hospitals that the employers asked
to participate bought the employer-sponsored system.  Similarly, in
Cleveland and Orlando, coalition members insure about 20 to 30
percent of the population, and most local hospitals purchased the
coalition-sponsored systems.\5

To obtain the hospitals' initial cooperation, employers in the
Cincinnati, Cleveland, and Orlando coalitions agreed to refrain for
several years from using comparative data to exclude hospitals from
their provider networks.  Ultimately, individual employers intend to
compare hospitals' performance and select the most cost effective to
participate in their managed care networks. 


--------------------
\5 Hospitals in some communities considered the costs of these
systems to be substantial.  Costs to hospitals varied widely
depending on the characteristics of the system, size of the hospital,
and amount of data requested. 


      HOSPITALS ENCOURAGED TO CURB
      CHARGES AND STAYS
---------------------------------------------------------- Letter :4.2

First-year data received by the employers indicated considerable
variation among hospitals in charges and length of stay.  For
example, average severity-adjusted charges among the 14 Cincinnati
hospitals in 1992 varied by about $900 for pregnancy cases, $3,300
for circulatory cases, $4,000 for respiratory cases, $4,300 for
digestive-tract cases, and $4,600 for musculoskeletal cases. 
Similarly, average length of stay varied by about 0.5 day for
pregnancy cases, 1.5 days for digestive-tract cases, 2 days for
circulatory and respiratory cases, and 2.5 days for musculoskeletal
cases.\6 In many cases, hospitals had inconsistent performance
profiles, scoring well for some patient groups but poorly for others. 

Similarly, the systems' data showed some variation in mortality
rates.  For example, the spring 1994 Cleveland Health Quality Choice
report showed mortality rates for general medical patients indicating
that 2 hospitals had more deaths than expected, 2 hospitals had fewer
deaths than expected, and 25 hospitals were as expected.  For
intensive care patients, 18 of 29 Cleveland hospitals had the
expected mortality rates and the other 11 had fewer deaths than
expected. 

Many hospitals we visited reported efforts to improve their
performance results by modifying treatment patterns.\7

Generally, hospitals use the systems to guide "continuous quality
improvement" efforts.\8 The systems help hospitals identify
particular conditions or clinical departments needing improvement.\9
In addition, hospital officials noted that the systems are useful for
demonstrating to physicians how their performance indicators differ
from those of colleagues caring for comparable patients.  These
comparisons lead hospital administrators and clinical staff to
conduct further reviews of medical records and discuss alternative
care procedures.  Hospitals we visited highlighted various practice
changes they attributed in part to the use of the systems.  The
following are examples: 

  -- Performing more diagnostic tests and procedures in ambulatory
     settings--Some hospitals reported reducing length of stay and
     inpatient charges by increasing their use of ambulatory
     services.  Physicians are more often ordering diagnostic tests
     and procedures before hospitalization so that patients may be
     admitted on the day of surgery, thereby shortening hospital
     stays by as much as 1 or 2 days.  Physicians are also reducing
     length of stay by increasing their use of ambulatory
     rehabilitation services, relying more on home health services
     and discharging patients earlier into other settings, such as
     skilled nursing facilities.  Hospitals also reported renewed
     efforts to avoid hospital admissions by substituting ambulatory
     care.\10

  -- Changing how physicians schedule and use ancillary personnel and
     services--After learning that its actual average length of stay
     for patients admitted for knee joint replacement surgery was
     greater than expected, a Cleveland hospital determined that
     several physicians had been underutilizing physical therapy
     services.  The physicians are now ordering physical therapy for
     these patients earlier and more often.  The hospital increased
     the utilization of this particular service but expected to
     reduce costs through shorter hospital stays. 

  -- Restricting the purchase and use of certain supplies and
     pharmaceuticals--In a Cincinnati hospital, all orthopedic
     surgeons are generally now using a single type of artificial hip
     device instead of choosing from among several available types. 
     This allows the hospital to take advantage of volume discounts. 
     An Orlando hospital noted that it had convinced physicians to
     use less expensive pharmaceuticals by pointing out that other
     physicians used those medications with similarly effective
     results.  Another Orlando hospital estimated that substantial
     savings resulted from switching to a less expensive pulmonary
     treatment--using a simple metered dose inhaler rather than
     extensive therapy--for patients with respiratory problems. 

Some changes hospitals have made in response to the performance data
initiatives may have already achieved their full cost containment
effect.  For instance, once hospitals have eliminated presurgery days
by performing diagnostic and preoperative procedures in an ambulatory
setting, no further reduction in days prior to surgery is possible. 
Similarly, discounts achieved by purchasing pharmaceuticals and
medical equipment from a single supplier may initially reduce prices
but may not continue to affect the rate of cost growth.  To achieve
further savings, hospital officials said they would need to make more
profound changes in their operations, such as reducing bed capacity
and staff. 

Also, some of the cost savings employers attribute to efficiency
improvements in inpatient hospital care are partially offset by
higher expenditures for ambulatory care.  The hospitals we visited
acknowledged that efforts to reduce charges and length of stay in
inpatient settings have been accompanied by the greater use of
ambulatory care (such as outpatient hospital departments, physicians'
offices, and home care) for testing, visits, and procedures. 
Although ambulatory care costs less than hospital care, it is a large
and growing expense for employers--but one that the systems do not
reveal.  Because the severity adjustment systems do not capture
ambulatory care costs, when a hospital shifts care from an inpatient
to an ambulatory setting, the change registers as a complete savings
instead of a partial savings. 


--------------------
\6 For respiratory care, for example, these ranges represent
variations of 25 percent above and 26 percent below the average
charge and 16 percent above and 12 percent below the average length
of stay in Cincinnati hospitals. 

\7 In addition, hospitals and physicians have also made
administrative changes to improve their performance results.  By
revising their documentation of patient care (medical records and
discharge summaries), they may receive a higher severity-of-illness
classification to justify high charges and length of stay. 

\8 In many hospitals we visited, clinical teams are organized to
create internal consensus among physicians as to what constitutes
optimal care for patients with a given condition.  They often develop
standard treatment protocols (also referred to as care maps,
treatment paths, or practice guidelines) for various conditions. 

\9 Hospitals are also focusing attention on improving their care of
patients with diagnoses of particular importance to local employers,
such as maternity care and back pain.  In addition to such focused
efforts, several hospital officials noted evidence of a "halo" effect
where nontargeted areas also benefit from process improvements
initiated as a result of the systems. 

\10 In some instances, this substitution increases the average length
of stay because patients eligible for ambulatory care would have had
relatively short inpatient stays, whereas patients who continue to
receive hospital care have relatively long stays.  Similarly, many
low-cost surgeries may be performed in alternative settings, reducing
total costs but raising the average cost of operations performed in
the hospital. 


      PERFORMANCE COMPARISONS
      EXPECTED TO CONTRIBUTE TO
      CHANGES IN MARKET STRUCTURE
---------------------------------------------------------- Letter :4.3

Employer coalitions as well as hospitals predict that the periodic
performance ratings will contribute to changes in local hospital
markets.  Over time, use of comparative ratings could result in
individual hospitals choosing to specialize in the services for which
they score relatively well.  For instance, a hospital that receives
high ratings for cardiac services but low ratings for obstetric
services could decide to concentrate on cardiac care and drop
obstetric care.  Such decisions may improve quality of care since
hospitals often have better surgical outcomes for procedures
performed at high volumes.\11 Some would also argue, however, that
patients could lose prompt access to specialized services if fewer
local hospitals were able to serve patients needing immediate care. 

In addition, hospital officials expect to complement their strengths
by forming affiliations.  Under a larger integrated system, hospitals
would try to achieve economies through volume while offering
comprehensive services at convenient locations.  They would also
expect to be in a better position to negotiate with health care
purchasers, which have also been organizing to gain negotiation
advantage.  This would also allow hospitals receiving lower ratings
for some services to maintain their economic viability.  Some
predict, however, that consolidation taken to the extreme could
result in reduced competition and thus less pressure to perform
efficiently. 


--------------------
\11 For example, 11 of 14 studies reviewed by the Office of
Technology Assessment found that lower volume of coronary artery
bypass graft surgeries in hospitals was associated with higher
mortality rates.  See The Quality of Medical Care:  Information for
Consumers, OTA-H-386 (Washington, D.C.:  U.S.  Government Printing
Office, June 1988). 


      TOO SOON TO IDENTIFY COST
      SAVINGS ATTRIBUTABLE TO
      SYSTEMS
---------------------------------------------------------- Letter :4.4

Evidence of the systems' ability to hold down hospital cost growth
has been limited because of the short time these systems have been in
place and the time it takes to identify and implement changes in
practice.  Also, it is difficult to isolate the influence of
severity-adjusted performance data from other factors contributing to
hospital improvements. 

To some extent, efficiency gains may reflect a sentinel effect.  That
is, in anticipation of employers' examination of length of stay and
charge data, hospitals began taking action before the systems' data
were available.  This effect may help explain why some of the changes
hospitals attributed to the systems actually began before hospitals
received or had time to analyze and act on the systems' results. 

For example, one Cincinnati hospital submitted its 1991 annual data
to Iameter in October 1992 and submitted its 1992 data in January
1993.  The Iameter results showed significant improvement for the
hospital between 1991 and 1992.  However, during the 4 months between
data submissions, the hospital would not have been able to analyze
the initial Iameter data, identify and implement improvements based
on the data, and have the changes reflected in the subsequent data. 
Therefore, the identified improvements in charges and length of stay
must have resulted from changes made before the system data were
available. 

Another factor making it difficult to gauge the effect of performance
systems on cost containment is that many hospitals already had
continuous quality improvement programs designed to improve
efficiency.  Such programs seek to streamline administrative
functions and modify clinical practice patterns.  Some hospitals we
visited stated that the performance measurement systems would enhance
their quality improvement efforts by supplying valuable data.  Other
hospitals whose performance indicators improved stated that the
performance information provided by the systems added little to
existing quality improvement efforts. 

Furthermore, hospitals and employers noted that increased enrollment
in managed care plans may have contributed to slower cost growth. 
Under such plans, employers and insurers have become increasingly
aggressive in negotiating reimbursement rates with hospitals.  The
advent of statewide reform efforts to stem rising health care costs
may have also provided an incentive for hospitals to mitigate rate
increases. 


   EMPLOYERS SHOULD USE CAUTION IN
   INTERPRETING PERFORMANCE DATA
------------------------------------------------------------ Letter :5

We found that severity-adjusted performance measurement systems are
in a relatively early stage of development and may not provide
adequate information for accurately comparing hospitals' performance. 
Limitations exist in the indicators that the systems report and the
methods they use to adjust the data for severity of illness.  Because
of these limitations, additional information and methodological
improvements are needed to provide more useful data on which to base
purchasing decisions. 


      SYSTEMS' RESULTS NOT
      SUFFICIENT FOR MEASURING
      HOSPITAL PERFORMANCE
---------------------------------------------------------- Letter :5.1

Used alone, performance indicators measured by these systems are
insufficient for comparing cost effectiveness--a function of cost and
quality of care.  Pressure to reduce charges and length of stay
could, in fact, provide an incentive for hospitals to restrict
services.  As a safeguard against potential adverse effects on
quality, the systems generate information to monitor hospitals for
signs of compromised care.  However, the two performance indicators
generated by the systems to measure quality--inpatient mortality
rates and average length of stay--are considered too narrow to truly
reflect quality.  Similarly, the indicators used for assessing
costs--charges and length of stay--are poor proxies for measuring
resource consumption. 

Inpatient mortality rates are not considered a good indicator of the
quality of care most patients receive because only a small percentage
of hospital stays result in death.  For example, admissions of
particular interest to employers, such as maternity and orthopedic
cases, have extremely low expected mortality rates.  Because death
and major complications for such patients are rare, systems that
measure variation in providers' mortality rates for such services
will not yield very useful information about the quality of inpatient
hospital care.\12

Some system vendors contend that average length of stay serves as an
indicator of both costs and quality.  They assert that longer
hospital stays tend to consume more resources and increase patients'
risk of contracting hospital-induced infections or other
complications.  However, hospital representatives we interviewed
noted that it is difficult to determine a clinically optimal length
of stay.  Some patients may benefit from longer-than-average
hospitalizations.  For example, patients who lack an appropriate
environment for home care may benefit from extra days in the
hospital.  Ever shorter stays may, in fact, result in poorer quality
care if patients are discharged before they are ready. 

Average length-of-stay data are useful but are also limited as an
indicator of resource consumption.  Shorter stays reduce the cost of
care because when patients spend less time in the hospitals, fewer
hours of nursing care and fewer routine tests and treatments are
generally provided.  However, the cost savings are not proportionate
to the reductions in days.  The most expensive days are generally
early in the stay when resource-intensive procedures are performed;
the latter days--when charges represent mostly room and board--are
less costly.  In fact, for most conditions, the final day of a
hospital stay costs less than half the average cost for all days.\13
Therefore, if it is assumed that costs per day remain constant,
estimates of savings based on length-of-stay reductions may overstate
actual cost reductions. 

Similarly, data on hospital charges are not precise indicators of
efficiency.\14 A hospital's overhead costs, which are included in
individual charges, are not distributed across services on the basis
of the actual cost of the service.\15 Therefore, the rise or fall of
charges for cardiac bypass surgery, for example, may have little or
nothing to do with the efficiency with which this procedure is done. 

In addition, charges do not necessarily represent what the health
care purchaser actually pays.  Several employers participating in the
coalitions we visited had managed care plans receiving 20- to
40-percent discounts on hospital charges.  Other plans negotiate flat
rates per day or per case, regardless of the hospital's charge. 
Thus, charges may not be relevant because often the amount the
purchaser actually pays depends on the discount or rate negotiated. 

Many of the experts, employers, and providers we spoke with believe
that other indicators should be developed for a more complete
comparison of quality.  To obtain additional information about
quality of care for individual hospitals, inpatient data on
readmission rates, infection rates, and complication rates could also
be monitored.  Other postdischarge patient outcomes, such as 30- and
60-day mortality rates, functional status, and time needed to return
to work, would also be relevant but are not measured by the systems
we examined.  If additional indicators of quality were assessed,
hospitals would have less of an incentive to unduly focus on
improving a single performance indicator, such as length of stay, to
the potential neglect of other indicators. 

Outcomes researchers have increasingly emphasized the importance of
gaining patients' perspectives when monitoring quality.  Patient
satisfaction surveys may provide additional information on quality
for individual hospitals and are relatively easy to conduct.  For
these reasons, the Cleveland Health Quality Choice Program has
complemented its severity-adjusted systems with a patient
satisfaction survey.  However, critics of patient satisfaction
surveys note that quality of care as perceived by patients is
subjective and often associated with characteristics of the provider
that make care more "personal" rather than improve outcomes. 


--------------------
\12 A maternal death during childbirth would be considered a sentinel
event that warrants further hospital review regardless of the
availability of the performance measurement systems. 

\13 See Grace M.  Carter and Glenn A.  Melnick, How Services and
Costs Vary by Day of Stay for Medicare Hospital Stays (Santa Monica,
Calif.:  RAND Corporation, Mar.  1990). 

\14 In our contacts with hospitals, we found that many hospitals do
not have cost accounting systems to estimate their actual operating
costs for discrete types of services and that little consistency
exists in their cost estimates. 

\15 Charges are also a poor measure of expenses because they are
believed to overvalue procedures, such as surgeries, and undervalue
evaluation and management services. 


      SEVERITY ADJUSTMENTS NOT
      ADEQUATE FOR COMPARING ALL
      HOSPITALS
---------------------------------------------------------- Letter :5.2

Many experts and hospitals noted that the appropriateness of the
severity adjustments remain suspect.  If the severity adjustment
methods are incomplete or flawed, then hospital comparisons may be
skewed.  For example, some researchers have found that the systems
tend to overestimate the expected mortality rate among hospitals with
relatively healthy patient populations and underestimate it among
hospitals with more severely ill patient populations.\16 Another
expert noted that severity adjustment "is a necessary but inherently
imperfect tool."\17

Hospitals we visited were wary that the severity adjustment
methodology is often a "black box" because little independent
validation occurs.  This apprehension about proprietary severity
adjustment systems led Cleveland hospitals to participate in
designing a new severity-adjusted performance measurement system. 
Despite their involvement in the design, some Cleveland hospitals
continue to criticize the severity adjustment methods for not
accounting for important factors that may influence their severity
ratings.\18 In particular, the Cleveland-designed system has been
criticized for not properly adjusting for seriously ill patients
transferred from other hospitals. 

Because most of the measurement systems were initially developed for
internal hospital use, severity adjustment limitations are compounded
when the systems are used to make comparisons across hospitals. 
Several large inner-city and teaching hospitals we visited contended
that the systems do not accurately compare the severity of illness of
their patients with that of patients treated in smaller community
hospitals.  Inner-city hospitals often treat patients who have
socioeconomic characteristics that severity adjustment systems do not
capture.  For example, a downtown Cincinnati hospital believes that
it is put at a disadvantage in the comparisons because it cares for
many drug-abusing patients who are not so identified in the data used
for severity adjustment.  Similarly, teaching hospitals often treat
patients with rare conditions that may not be adequately accounted
for in severity adjustments, and their costs may also be higher than
community hospitals because of their teaching mission. 

Hospitals that treat a large number of uninsured individuals may also
be disadvantaged in the comparisons.  Much of the health care
received by the over 37 million Americans who are currently uninsured
is not paid for, prompting providers to recover those costs from the
insured.  Hospitals that provide more services to uninsured patients
may charge higher rates to patients with private insurance and
thereby show poor performance results.  Thus, because performance
measurement programs give hospitals an incentive to reduce their
charges to compare more favorably on the systems' indicators, the
programs also could create an unintended incentive to reduce services
to patients without private insurance. 


--------------------
\16 See, for example, Harry P.  Selker, "Systems for Comparing Actual
and Predicted Mortality Rates:  Characteristics to Promote
Cooperation in Improving Hospital Care," Annals of Internal Medicine,
Vol.  118, No.  10 (May 15, 1993), pp.  820-22. 

\17 Lisa I.  Iezzoni, "Risk Adjustment for Medical Outcome Studies,"
Medical Effectiveness Research:  Data Methods, Agency for Health Care
Policy and Research, Public Health Service, U.S.  Department of
Health and Human Services, 1992. 

\18 Some outcomes researchers noted that systems achieve greater
precision by specifying unique severity adjustment factors for
specific conditions.  For example, adjusting for severity of illness
for coronary-artery bypass graft patients may require a very
different methodology than for hernia repair patients. 


   CONCLUSION
------------------------------------------------------------ Letter :6

Collecting and sharing data among health care purchasers and
providers is the goal of the employer coalitions in our review.  The
severity-adjusted measurement systems used to meet this goal are
designed to produce more precise and relevant information.  Because
of the systems' limitations, however, it is important that coalitions
not overestimate the capabilities they offer. 


---------------------------------------------------------- Letter :6.1

For a more detailed description of the severity adjustment systems we
reviewed and how they are being used in the communities we visited,
see appendixes I to III.  We asked a representative of each
performance measurement initiative to review the appendixes.  These
representatives generally agreed with the technical content, and we
incorporated their comments as appropriate. 

As arranged 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 interested parties and make copies available to others on
request. 

Please call me on (202) 512-7119 if you or your staff have any
questions about this report.  Other major contributors are listed in
appendix IV.

Mark V.  Nadel
Associate Director, National and
 Public Health Issues


CENTRAL FLORIDA HEALTH CARE
COALITION
=========================================================== Appendix I

Recognizing that the Orlando area's health care costs were rising
faster than elsewhere in the nation, a group of local employers
developed the Central Florida Health Care Coalition in 1984 with a
charter to "limit the increases in health-care expense without
reducing the quality of health care." As part of its cost containment
efforts, the Coalition began encouraging local hospitals to purchase
MediQual's MedisGroups severity-adjusted performance measurement
system in 1990.\19 MedisGroups uses information from medical records
to adjust for patients' severity of illness and compares providers'
charges, length of stay, mortality rates, and morbidity rates to
national norms.  Several local hospitals reported that they are using
MedisGroups data in their continuous quality improvement programs. 
The Coalition believes that the initiative has helped contain the
increase in Orlando's hospital costs. 

During our visit to Orlando, we met with the president and the
executive director of the Central Florida Health Care Coalition;
officers representing the area's two hospital chains and an
independent hospital; the vice president of the Florida Hospital
Association; and health benefits managers for several large Orlando
employers.  We also contacted executives with MediQual, the president
of the Orange County Medical Society, and representatives of other
Orlando-area hospitals. 


--------------------
\19 "MedisGroups" stands for Medical Illness Severity Grouping
System. 


   HISTORY OF THE COALITION'S
   INITIATIVE
--------------------------------------------------------- Appendix I:1

The Central Florida Health Care Coalition was formed in 1984 by 10 of
the Orlando area's largest employers, including Walt Disney World,
the Orange County Public Schools, Martin Marietta, and General Mills
Restaurants.  The Coalition has expanded to include about 100 local
employers, most having fewer than 300 workers.  Currently, the
Coalition represents about 180,000 employees.  Including dependents,
the Coalition members insure about 380,000 individuals in the Orlando
area, nearly 30 percent of the area's population.\20

The Coalition encouraged local hospitals to purchase the MedisGroups
severity adjustment system to provide comparable performance
information.  At the time of our visit, the Orlando hospital market
was dominated by two hospital chains--Orlando Regional Healthcare
System and Florida Hospitals--representing 9 of the 12 local
hospitals.  Hospitals affiliated with the Orlando Regional Healthcare
System contracted with MediQual in 1990, and Florida Hospitals
contracted with MediQual in late 1991.  Because both hospital chains
phased in the use of MedisGroups, complete data for these hospitals
were not available until early 1993. 

Since our visit, most of the remaining independent Orlando-area
hospitals have been purchased by Columbia Hospital System to form a
third hospital chain.  Although at least one of the hospitals in this
new chain had been using MedisGroups, other Columbia-affiliated
hospitals had decided not to participate in the Coalition's
MedisGroups initiative, in part because of the costs of using the
MedisGroups system.  As of July 1, 1994, Columbia Hospital System had
not yet decided whether the chain, as a whole, will purchase
MedisGroups. 

Some Orlando-area hospitals are also using other severity-adjusted
performance measurement systems, including Iameter's Acuity Index
Method (AIM).  Several hospitals had begun using these severity
adjustment systems before the Coalition's request that they use
MedisGroups uniformly.  Also, some area hospitals participated in a
Florida Hospital Association project using 3M's All Patient Refined
Diagnosis Related Groups (APR-DRG) system.  Several hospitals that
use multiple systems reported that the systems are complementary
because they have different relative strengths.  For example, one
hospital noted that MedisGroups is preferred for assessing mortality
and morbidity results, whereas AIM more precisely measures length of
stay and charges. 


--------------------
\20 The Central Florida Health Care Coalition has recently merged
with the Space Coast Labor Management group, including employers such
as Lockheed.  However, because these employers are in a neighboring
county, they are not actively involved in the Orlando-area
MedisGroups initiative. 


   DESCRIPTION OF MEDISGROUPS
   SYSTEM
--------------------------------------------------------- Appendix I:2

After reviewing various severity adjustment systems, the Coalition
selected MedisGroups, in part because it preferred a clinically based
system.  The database tracks length of stay, hospital charges,
morbidity rates, and mortality rates, and compares each hospital's
performance with national standards.\21 MedisGroups generates data on
these indicators by hospital and physician for individual conditions
and compares them with the expected rates given the patient's
severity of illness.  MedisGroups assesses severity of illness once
within the first 2 days of admission and then again during days 3 to
7 of the stay.  To make the severity adjustment, trained nurses and
administrative staff abstract "key clinical findings," including
patients' vital signs, laboratory test values, and radiology reports,
from the medical record.  These data are used to categorize the
patients by their likelihood of suffering a major organ
failure--"severity 0" indicates patients are not expected to have
major organ failure, whereas "severity 4" indicates the presence of
organ failure.\22

MedisGroups omits some pertinent information that may affect the cost
of treatment during the hospital stay.  Although the system tracks
severity of illness, it does not take into account the patient's
medical history, such as whether it is the patient's first or second
heart attack.  The system is also unable to distinguish between an
emergency coronary bypass surgery, a repeat coronary bypass surgery,
and a coronary bypass surgery following angioplasty. 

The costs of MedisGroups are paid by the hospitals, with the amount
varying by size of the hospital.  The hospitals we visited paid
initial purchase fees of $230,000 to $330,000 for licensing,
software, and equipment, and ongoing annual fees of $28,000 to
$70,000.  Another major expense of using MedisGroups is the staff
time required to abstract clinical findings from medical records and
enter the data in the MedisGroups system.\23 A relatively small
hospital with 300 acute-care beds dedicates 3 full-time staff to
MedisGroups data collection at an annual cost of $72,000.  A large
hospital chain has 15 full-time staff abstracting MedisGroups records
for 1,500 hospital beds at an annual cost of $500,000.  The Coalition
estimates that MedisGroups costs average about $8 to $11 per hospital
admission. 


--------------------
\21 MedisGroups defines morbidity as occurring if a second severity
rating taken midstay indicates that the patient has not responded to
treatment. 

\22 For a more thorough technical discussion of MedisGroups, see Lisa
I.  Iezzoni and Mark A.  Moskowitz, "A Clinical Assessment of
MedisGroups," Journal of the American Medical Association, Vol.  260,
No.  21, (Dec.  2, 1988), pp.  3159-63. 

\23 The training to become a MedisGroups' abstractor is extensive,
with abstractors required to meet a 95-percent consistency standard
before becoming qualified by MediQual.  One hospital reported that
abstractor training required 3 to 6 months. 


   HOSPITALS' USE OF MEDISGROUPS
   RESULTS
--------------------------------------------------------- Appendix I:3

Hospital administrators can analyze the MedisGroups-adjusted data to
identify the performance of individual physicians for particular
conditions.  In addition, they can compare their hospital's
performance with that of others in the MedisGroups database.  Table
I.1 provides an example of the MedisGroups data provided to a
hospital.  It indicates the hospital's actual charges compared with
their expected charges, estimated by MedisGroups on the basis of the
patients' severity of illness.  The ratio of the actual value divided
by the expected value is defined as the standard ratio.  The system
compares whether the standard ratio is significantly different from
1:  a standard ratio significantly greater than 1 is worse than
average, whereas a standard ratio less than 1 is better than average. 



                         Table I.1
          
          GAO Illustration of Typical MedisGroups
                     Report Information


                            Number          Expect
                                of  Actual      ed  Standa
                            patien  averag  averag      rd
DRG   Description               ts       e       e   ratio
----  --------------------  ------  ------  ------  ------
391   Normal newborn           276    $796    $861   .92\a
371   Cesarean section          64  $6,635  $5,188  1.28\a
       without
       complications
209   Major joint/limb          57  $18,20  $19,64   .93\b
       operation of lower                4       5
       extremities
98    Bronchitis, asthma,       48  $3,335  $3,334  1.00\c
       ages 0 to 17
127   Heart failure and         48  $7,239  $8,919   .81\b
       shock
----------------------------------------------------------
Note:  "Standard ratio" is an index that is created by dividing the
actual value by the expected value. 

\a Standard ratio is statistically significant at the .01 level. 

\b Standard ratio is statistically significant at the .05 level. 

\c Standard ratio is not significantly different from 1 at the .05
level. 

Several local hospitals have incorporated the MedisGroups data in
their quality improvement processes.  By using MedisGroups data to
identify specific procedures for which performance is significantly
worse than expected, the hospital can target its improvement efforts,
as the following examples illustrate: 

  -- One hospital established physician task forces to review
     cesarean sections, pneumonia, and back surgery.  The hospital
     believes that the task force recommendations, such as switching
     to a less expensive pharmaceutical following back surgeries,
     have saved the hospital money. 

  -- Another hospital has established committees to focus on acute
     myocardial infarctions, congestive heart failure, cesarean
     sections, and respiratory care.  One committee recommended using
     a metered dose inhaler rather than aerosol treatment for
     patients receiving respiratory therapy.  The hospital estimates
     that changing to the inhaler will save about $300,000 per year,
     largely because the inhaler requires only about 3 minutes of a
     respiratory therapist's time, whereas the aerosol treatment
     requires about 20 minutes. 

Several hospitals noted that their quality improvement programs had
been active before they received the MedisGroups or other severity
adjustment systems' data.  One hospital cited an aggressive quality
improvement program as a factor in decreasing inpatient hospital
charges.  Even without the MedisGroups data, it might have had a
similar quality improvement program.  But the hospital believes that
the employers' interest in the data has provided additional
motivation to the hospital's quality improvement process. 


   EMPLOYERS' USE OF MEDISGROUPS
   RESULTS
--------------------------------------------------------- Appendix I:4

Hospitals summarize the MedisGroups results and their continuous
quality improvement efforts in regular reports to the Central Florida
Health Care Coalition.\24 For the most part, employers have
encouraged hospitals to continue using the MedisGroups data for
reviewing their medical practices rather than directly using the data
to select the better performing providers.  However, several
employers have made hospital participation in the MedisGroups
initiative a condition for inclusion in their health networks: 

  -- One large employer intends to drop a hospital from its managed
     care network because it has not purchased MedisGroups. 

  -- A purchasing alliance established by members of the Central
     Florida Health Care Coalition and other Florida employer
     coalitions has also made participation in the MedisGroups
     project a criterion for hospitals to be included in their health
     plan. 

  -- Another large company has used MedisGroups and other
     risk-adjusted data in establishing a subset of "most preferred"
     providers within its existing managed care network.  Employees
     receive financial incentives to use the subset of providers. 


--------------------
\24 At least one hospital also reanalyzes the MedisGroups data to
develop customized reports for specific large employers. 


   CLAIMED SAVINGS
--------------------------------------------------------- Appendix I:5

Several hospitals and employers have identified hospital cost savings
that have occurred since purchasing MedisGroups: 

  -- One Orlando hospital reduced its average length of stay by 13.8
     percent in 1992, leading to a 1.7-percent reduction in average
     costs per patient.  This compares with a statewide average
     length-of-stay decline of about 2.7 percent during the same
     period. 

  -- Another hospital has identified savings for specific conditions
     and procedures.  For the individual conditions that hospital
     task forces have assessed, costs per case declined by between 3
     and 43 percent, charges by between 8 and 28 percent, and length
     of stay by between 3 and 32 percent. 

  -- One employer, large enough to have independent clout, also cited
     benefits from its participation in the coalition.  From 1989 to
     1991, its premium increases averaged 11 percent per year; since
     1991, the premium growth has been less than half that rate.  The
     employer estimated a savings of $1 million in anticipated
     inpatient hospital costs due to the MedisGroups project. 

Our analysis of data collected by the Florida Agency for Health Care
Administration indicates that Orlando's hospital charges have grown
more slowly since the implementation of the MedisGroups program (in
1991 and 1992) and relative to other hospitals in Florida.  (See fig. 
I.1.) Similarly, Orlando's average hospital length of stay has
declined more rapidly compared with prior years and with other
Florida hospitals.  (See fig.  I.2.)

   Figure I.1:  Growth in Hospital
   Charges in Orlando and the
   Remainder of Florida, 1989 to
   1992

   (See figure in printed
   edition.)

Notes:  The Orlando data exclude three hospitals that were not using
the MedisGroups system during this period.  These hospitals' data are
included in the "remainder of Florida" category. 

The use of MedisGroups was phased in by hospitals.  By 1992,
hospitals representing about 85 percent of Orlando hospital beds were
implementing MedisGroups. 

Source:  GAO calculation of data from Florida Agency for Health Care
Administration. 

   Figure I.2:  Decline in
   Hospital Patient Length of Stay
   in Orlando and the Remainder of
   Florida, 1989 to 1992

   (See figure in printed
   edition.)

Notes:  The Orlando data exclude three hospitals that were not using
the MedisGroups system during this period.  These hospitals' data are
included in the "remainder of Florida" category. 

The use of MedisGroups was phased in by hospitals.  By 1992,
hospitals representing about 85 percent of Orlando hospital beds were
implementing MedisGroups. 

Source:  GAO calculation of data from Florida Agency for Health Care
Administration. 

Some hospitals and employers recognize that it is difficult to
isolate the impact of MedisGroups on Orlando's hospital cost trends. 
For example, one hospital noted that its ongoing quality improvement
program and lower payments from Medicare and managed care plans also
contributed to its shorter stays and slower cost growth.  In
addition, some members of the Central Florida Health Care Coalition
are participating in a statewide employers' insurance purchasing
alliance.  The Coalition attributes cost savings in the area to this
purchasing alliance's leverage in achieving hospital discounts as
well as to the MedisGroups initiative. 

Employers also indicated that recent reductions in their inpatient
hospital expenditures have been matched by increases in ambulatory
care costs, with their total health care expenditures continuing to
increase at about the same rate as in prior years.  For example,
between 1991 and 1992 one large Orlando employer's inpatient hospital
payments per employee declined by about 14 percent, whereas its
payments for hospital outpatient care, physician care, and
pharmaceuticals rose by nearly 17 percent.  Overall, the employer's
health care payments per employee still grew by about 4 percent. 
Similarly, another large Orlando employer's average charge per case
for inpatient hospital care declined by about 9 percent from 1990 to
1992, but because of increasing expenditures for ambulatory care
overall health plan costs increased 5 percent.\25


--------------------
\25 This compares to an average 8 percent increase in health plan
costs reported for over 300 large Florida employers between 1991 and
1992.  See William M.  Mercer, Inc., 1992 Florida Health Care Costs
and Benefits:  Survey Results (Tampa, Fla.:  1992). 


   FUTURE DEVELOPMENTS IN
   ORLANDO'S HOSPITAL MARKET
--------------------------------------------------------- Appendix I:6

As previously discussed, the Orlando hospital market has recently
become more consolidated with the development of a third hospital
chain.  In the future, specialization is expected to occur within
these hospital groups, with hospitals concentrating on providing the
services at which they excel.  For example, a hospital administrator
noted that two affiliated hospitals may decide that one hospital will
provide pediatric care while the other hospital will provide cardiac
care.  In this way, each hospital can reduce hospital beds for some
services while the hospital chain can continue to offer a
comprehensive package of hospital services to employers. 

As part of a broader health reform initiative to make outcomes data
from Florida hospitals available to the public, Florida's law
requires the use of a uniform severity-adjusted performance
measurement system.  In 1993, the state selected an administratively
based system, APR-DRG, which had been piloted by the Florida Hospital
Association.  The hospital association advocated this system because
of its relatively low cost--it estimated that APR-DRG cost hospitals
less than $10,000 per year.  Thus, many of the Orlando-area hospitals
participated in the Florida Hospital Association pilot and are using
APR-DRG in addition to MedisGroups.  Because the Central Florida
Health Care Coalition prefers the advantages of a clinically based
system, it is continuing to encourage Orlando-area hospitals to use
MedisGroups despite the state's mandated use of APR-DRG. 


CINCINNATI PAYER INITIATIVE
========================================================== Appendix II

Frustrated by annual double-digit health plan cost increases, four
large Cincinnati employers collaborated to launch the Cincinnati
Payer Initiative.  The employers view this 3-year initiative as a
mechanism to combat hospital cost increases by using
severity-adjusted data to identify less efficient hospitals.  The
potential market clout of the 4 large employers encouraged 14 local
hospitals to purchase Iameter's AIM severity adjustment system.\26
AIM provides a method to make severity adjustments for providers'
patient populations and compare inpatient charges, length of stay,
and mortality rates on a communitywide basis.  Hospital officials
reported that the data assist in identifying clinical areas on which
to focus their cost containment efforts. 

We contacted each of the 14 Cincinnati hospitals that have purchased
AIM and conducted more detailed visits with 4 major hospitals.  We
also contacted several of the employers that began the Cincinnati
Payer Initiative; the Greater Cincinnati Employer Health Care
Alliance; the Greater Cincinnati Hospital Council; the Academy of
Medicine of Cincinnati; and representatives of Iameter. 


--------------------
\26 All of the hospitals that the employers asked to participate
agreed to purchase AIM.  The initiative did not include several
Cincinnati-area hospitals serving specific types of patients,
including a children's hospital, a veterans' hospital, a burn center,
and a psychiatric hospital. 


   HISTORY OF THE INITIATIVE
-------------------------------------------------------- Appendix II:1

In 1984, the Greater Cincinnati Employer Health Care Alliance was
established as a forum to exchange information and collaborate on
special projects with the goal of promoting cost containment and
quality care.  Currently, the Alliance represents nearly 80 large and
small employers.  In 1990, the Alliance began examining various types
of severity-adjusted performance measurement systems.  Although
several experts in the field of performance measurement provided
guidance on selecting severity adjustment systems, the project never
progressed to the point of actually comparing systems on the market. 

Meanwhile, health care costs continued to escalate.  From 1989 to
1991, Cincinnati employers experienced annual health care inflation
rates averaging 10 to 14 percent.  Disappointed by these cost
increases and the slow pace of the Alliance's project, four large
employers initiated an independent examination of severity-adjusted
performance measurement systems in 1991.  The four
employers--Cincinnati Bell, General Electric Aircraft Engines,
Kroger, and Proctor and Gamble--represent about 168,000 employees and
dependents, constituting over 13 percent of the local health care
market. 

After examining several severity adjustment systems, the employers
selected Iameter's AIM.  Iameter performed an initial comparison of
hospital performance based on publicly available Medicare Provider
Analysis and Review (MEDPAR) data.\27 The employers shared these
preliminary comparisons with representatives from 14 local hospitals
in March 1992.  After this meeting, each hospital agreed to contract
with Iameter and submitted discharge data for all of their patients
for 1991. 


--------------------
\27 MEDPAR data account only for Medicare hospital patients and do
not include patients with Medicaid or private coverage.  Medicare
patients represented nearly half of hospital discharges nationally in
1991. 


   IAMETER'S AIM
-------------------------------------------------------- Appendix II:2

AIM reports a comparison of charges, length of stay, and mortality
rates for two levels of specificity:  (1) for five major disease
categories, such as circulatory care, and (2) several dozen
high-volume diagnosis related groups (DRG), such as coronary-artery
bypass grafts.  AIM is one of several administratively based systems
that uses routinely collected discharge data to adjust for severity. 
At discharge, the attending physician writes each patient's diagnosis
and procedures on an attestation statement.  For billing purposes,
the hospital's medical records department translates these narratives
into numeric codes (DRGs and the codes specified in the International
Classification of Diseases, Ninth Revision, Clinical Modification
[ICD-9-CM]).  To make its severity adjustments, AIM collects patient
data on age, sex, length of stay, total charges, discharge status,
principal diagnosis, secondary diagnosis, and procedures performed. 
Using this information, AIM classifies patients with the same
diagnosis into one of five severity levels. 

The employers selected AIM largely because of its relatively low cost
and ease of implementation.  Several hospitals we visited reported
that data for AIM's analysis are readily retrievable from their
internal management information systems.  The hospitals required
little or no additional staff, staff training, or equipment to
develop the data for AIM's analysis, so minimal additional costs were
incurred.  Hospital officials reported that Iameter offers several
packages of services ranging in cost from $20,000 to over $80,000. 
The low-cost package consists only of data comparing hospitals at
broad levels of services, while the high-cost package includes
condition-specific and physician-specific comparisons and Iameter's
consulting services.\28

Cincinnati hospitals we visited reported that since the Cincinnati
Payer Initiative began they have paid more attention to the codes
documenting patient services.  Because AIM relies on these codes to
make severity adjustments, differences in coding practices by the
hospitals may significantly influence the patients' severity
classifications and hence the hospitals' relative standing.  Because
caring for more severely ill patients requires more resources and
entails greater risk, hospitals that obtain high severity ratings
appear justified in having higher charges, length of stay, and
mortality rates.  As a result, several Cincinnati hospitals noted
that they have conducted seminars and other efforts to help
physicians and other staff code more "effectively." Other hospitals
have changed administrative coding processes by introducing
concurrent coding in which services and procedures are coded on
billing forms as they are delivered instead of after the patient has
been discharged. 

After the information is collected and analyzed, Iameter provides
participating employers and hospitals with the comparative results. 
Figure II.1 is our adaption of an Iameter chart comparing hospitals'
performance for circulatory care.  The hospitals in the upper right
quadrant had shorter patient stays and lower charges, whereas the
hospitals in the lower left quadrant had longer patient stays and
higher charges for equivalent cases.  Positive numbers on the chart
reflect performance better than the community norm. 

   Figure II.1:  GAO Illustration
   of a Typical Chart in an
   Iameter Report (Length of Stay
   and Charges for Circulatory
   Care)

   (See figure in printed
   edition.)

Note:  Average charges and average length of stay are stated as the
difference from the community norm, which is shown as the "0" point
on each axis. 

Standardized for severity, Iameter's 1992 comparisons demonstrated
wide variations in hospital performance in Cincinnati.  For example,
average charges across Cincinnati's 14 hospitals varied by about $900
for pregnancy cases, $3,300 for circulatory cases, $4,000 for
respiratory cases, $4,300 for digestive-tract cases, and $4,600 for
musculoskeletal cases.  Similarly, average length of stay varied by
about 0.5 day for pregnancy cases, 1.5 days for digestive-tract
cases, 2 days for circulatory and respiratory cases, and 2.5 days for
musculoskeletal cases.\29 In many cases hospitals had variable
performance profiles, scoring well for some patient groups but poorly
for others. 


--------------------
\28 In the comparisons, data from the other hospitals are coded to
protect confidentiality. 

\29 For respiratory care, for example, these ranges represent
variation of 25 percent above and 26 percent below the average charge
and 16 percent above and 12 percent below the average length of stay
in Cincinnati hospitals. 


   HOSPITALS' AND EMPLOYERS' USE
   OF AIM DATA
-------------------------------------------------------- Appendix II:3

Many of the hospitals we visited are using AIM's data in their
ongoing continuous quality improvement programs.  Because AIM's data
report variations among physicians, hospital officials said the
information helps them identify where more detailed examination of
processes, procedures, and physician practices is needed.  Many
hospitals believe that AIM does not provide enough data on how to
improve performance.  However, information on variation is viewed as
a useful tool to help physicians identify problems and to target
hospital quality improvement efforts.  Several local hospitals
contract with Iameter for additional consulting services to identify
opportunities for improvement. 

Hospitals and physicians have begun to make changes to improve
efficiency as these examples show: 

  -- AIM reported that physicians at one hospital had average stays
     for pneumonia patients ranging from 7 days longer to 6 days
     shorter than the community norm and average charges differed by
     nearly $12,000.  In an effort to reduce this variation by
     standardizing care, a team of physicians determined that
     treating patients with a less expensive antibiotic and involving
     respiratory therapists to collect sputum samples earlier in
     treatment could decrease charges and length of stay.  The team
     developed treatment protocols based on the practices of the
     best-performing physicians that are used to guide treatment for
     all pneumonia patients. 

  -- Orthopedists at one hospital agreed to use just one or two
     brands of artificial hips in hip replacement operations rather
     than choosing from among the many versions on the market.  This
     allows the hospital to save money with volume purchases from a
     single vendor and helps physicians standardize treatment
     protocols, which should reduce variation in charges and length
     of stay. 

Whereas a hospital sees only how it compares with other unnamed
hospitals, the employers' reports identify each of the hospitals in
the communitywide comparison.  The employers expect that trends in
hospital performance will begin to appear after 3 years of data are
compiled, and they will then start using the AIM information in
making their purchasing decisions.  One employer reported that it has
shared the AIM data with managed care plans that bid for a contract,
asking that the plans consider the AIM data in selecting hospitals
and physicians for their networks.  However, the employer reported
that other factors are more important in selecting a managed care
plan, including the plan's general management structure.  For this
reason, the employer has maintained contracts with its existing
managed care organization although some hospitals that are highly
rated by AIM are not in the plan's network. 


   CLAIMED SAVINGS
-------------------------------------------------------- Appendix II:4

On May 24, 1994, the hospitals received their most recent
AIM-adjusted reports reflecting 1993 performance.  Iameter reported
that between 1992 and 1993, hospital charges decreased by about 1
percent and length of stay decreased by about 10 percent.\30 Iameter
estimated that these changes have led to savings among the hospitals
of $200 million between 1991 and 1993. 

Although Iameter and employer coalition leaders have claimed cost
savings, we found little pattern to the growth rate in charges over
the period.  We asked hospitals and employers about their cost trends
since using the system and were given the following examples: 

  -- A Cincinnati hospital estimated that a 0.6-day average annual
     reduction in its average length of stay for all patients between
     1991 and May 1993 could have saved as much as $5 million in
     nursing care, food, drugs, and supply costs in 1993 (5 percent
     of the hospital's operating costs).  By comparison, the
     statewide average length of stay in Ohio declined by only 0.2
     day between 1991 and 1992. 

  -- Another Cincinnati hospital estimated that reducing its average
     length of stay by about 0.5 day since the Cincinnati Payer
     Initiative began led to cost savings of between $1.3 million and
     $4.8 million during a 15-month period (about 1.3 to 4.8 percent
     of the hospital's inpatient revenues).\31

Comparing charges and length of stay before and after the employer
coalition's initiative began indicates mixed results.  Data reported
by the Ohio Department of Health indicate that Cincinnati's hospital
charges grew at a faster rate the year after hospitals initially
purchased AIM than the year before.  However, both years had
significantly lower rates of growth than 2 years before.  The 1992
growth in charges in Cincinnati was somewhat lower than the rate
experienced by other hospitals in Ohio, but continued at double-digit
rates.  (See fig.  II.2.) Length of stay declined by 4.2 percent
between 1991 and 1992, a greater decline than reported for the 2
years before AIM was used.  A similar rate of decline (3.7 percent)
was experienced by other Ohio hospitals.  (See fig.  II.3.)

   Figure II.2:  Growth in
   Hospital Charges in Cincinnati
   and the Remainder of Ohio, 1989
   to 1992

   (See figure in printed
   edition.)

Notes:  Ohio Department of Health data exclude Medicare and Medicaid
patients. 

Hospitals report only the 100 highest-volume diagnosis related
groups. 

Hospitals in the "remainder of Ohio" include those in Cleveland,
which were developing their severity-adjusted performance measurement
project during this time. 

Source:  GAO calculations based on data obtained from Ohio Department
of Health. 

   Figure II.3:  Change in
   Hospital Patient Length of Stay
   in Cincinnati and the Remainder
   of Ohio, 1989 to 1992

   (See figure in printed
   edition.)

Notes:  Ohio Department of Health data exclude Medicare and Medicaid
patients. 

Hospitals report only the 100 highest-volume diagnosis related
groups. 

Hospitals in the "remainder of Ohio" include those in Cleveland,
which were developing their severity-adjusted performance measurement
project during this time. 

Source:  GAO calculations based on data obtained from Ohio Department
of Health. 

Hospitals and employers noted that other factors could have
contributed to recent changes in Cincinnati hospitals' cost trends. 
Employers acknowledged that their general, increased attention to
health care costs, including an increased use of managed care plans
and other cost containment strategies, also influenced the Cincinnati
health care market.  According to a representative of the Greater
Cincinnati Employer Health Care Alliance, a local health maintenance
organization that is not involved in the Cincinnati Payer Initiative
also claimed credit for the declining length of stay and slowed
growth in hospital charges. 

In addition, the timing of the initial AIM reports suggests that
hospitals were making changes before receiving their AIM results. 
Iameter's analysis of the data hospitals provided for 1991 was made
available in fall 1992.  The hospitals then had only a few months to
make performance improvements before submitting their 1992 data in
January 1993.  Thus, although the AIM data indicated improvement by
many hospitals between 1991 and 1992, changes in hospital performance
may have resulted from a sentinel effect--that is, from hospitals'
anticipation of the employers scrutinizing their performance.  The
hospitals' ongoing quality improvement efforts, initiated before the
receipt of AIM data, may also have contributed to the differences in
the 1991 and 1992 comparisons. 


--------------------
\30 Iameter did not share the most recent detailed performance
results with us for 1993. 

\31 The hospital estimated the range in savings based on varying
assumptions of how much a shorter average length of stay reduces the
average cost of a day of hospital care, ranging from saving the daily
average cost ($946) to saving only a fraction of the average cost
($250). 


CLEVELAND HEALTH QUALITY CHOICE
PROGRAM
========================================================= Appendix III

The Cleveland Health Quality Choice Program has several unique
characteristics.  In contrast to the Cincinnati and Orlando
initiatives, the program has pursued a slower but a more
collaborative approach among the local physician and hospital
associations and several employer coalitions.  Rather than choosing a
single off-the-shelf system, the program custom-designed a three-part
system.  It compares local hospitals' performance in mortality,
length of stay, and patient satisfaction, but omits hospital charges. 
The severity-adjusted data do not directly make cross-hospital
comparisons, but rather report only whether a hospital performs as
expected for the type of patient it treats. 

During our site visit, we interviewed the executive director of the
Cleveland Health Quality Choice Program as well as representatives of
several local hospitals, employer associations, and insurers.  We
also discussed the program with representatives of the Greater
Cleveland Hospital Association; the Academy of Medicine of Cleveland;
and Michael Pine and Associates, which developed the Cleveland Health
Outcome Indicator of Care Evaluation (CHOICE) system.  Ohio
Department of Health data for 1993 will not be available until fall
1994.  Because the first hospital comparisons from the Cleveland
system were issued in April 1993, we were unable to compare the
changes in Cleveland hospital charges and length of stay before and
after the implementation of the system as we did with Orlando and
Cincinnati. 


   HISTORY OF THE PROGRAM
------------------------------------------------------- Appendix III:1

During the late 1980s, Cleveland employer coalitions began to
negotiate more aggressively with local hospitals, in part as a result
of analyses and anecdotes indicating Cleveland's health care costs
were relatively high.  A Foster Higgins study reported that the cost
of care in Cleveland was the fourth highest among metropolitan areas
in the United States.  This study noted that hospital services
costing $100 in Chicago would cost about $140 in Cleveland. 
Cleveland employers also began telling hospitals that for some
high-cost services, it would be less expensive to fly an employee to
the Mayo Clinic in Rochester, Minnesota, for treatment than to have
the employee treated at a local hospital. 

The employers told the hospitals that they would begin using Ohio
Department of Health charge data and Medicare mortality reports to
select low-cost hospitals that had satisfactory mortality outcomes. 
The hospitals responded that these comparisons would be unfair.  They
echoed the argument made by hospitals elsewhere that by not
adequately adjusting for differences in the average severity of
illness among the hospitals' patients, hospitals with "sicker"
patients would be disadvantaged in the comparisons.  Consequently,
the hospitals, physicians, and employers collaborated to develop the
Cleveland Health Quality Choice Program in 1989. 

The directors of the program include representatives from the Academy
of Medicine of Cleveland, the Greater Cleveland Hospital Association,
and three employer groups.  These employer groups are Cleveland
Tomorrow, a group of chief executive officers from the 50 largest
Cleveland corporations; the Health Action Council of Northeast Ohio,
a coalition of about 100 employers; and the Council of Smaller
Enterprises, a buying cooperative of 12,000 small businesses.  These
employer groups provide health coverage to about 350,000 individuals
in the greater Cleveland area, about 20 percent of the population. 
All 29 of the major hospitals in the Cleveland area (except for a
veterans' hospital) are participating in the program. 

As part of its collaborative approach, the Cleveland Health Quality
Choice Program includes the local hospital and physician associations
as members.  Both groups participated in the development of the
program's three-part system, including the design of the CHOICE
system and the selection of the Acute Physiology, Age, Chronic Health
Evaluation (APACHE III) system and the Patient Viewpoint Survey.  As
a voting member, the Academy of Medicine of Cleveland successfully
opposed including physician-specific data in the program.  In
addition, the health care provider groups have ensured that published
reports meet high standards of statistical validity. 


   PERFORMANCE MEASUREMENT SYSTEMS
   USED BY CLEVELAND PROGRAM
------------------------------------------------------- Appendix III:2

The members of the Cleveland Health Quality Choice Program decided in
1990 that for comparing inpatient hospital services, the
severity-adjusted performance measurement systems on the market were
not suitable.  The program members were concerned that the
methodology of many of the systems was a "black box" because little
rigorous independent evaluation of its validity had been conducted. 
Thus, the program opted for a three-part approach to evaluate
hospital performance, including the new, custom-designed
severity-adjusted performance measurement system developed with the
participation of Cleveland hospitals and physicians.  For intensive
care outcomes, the program's representatives selected the APACHE III
system, which has been widely used and independently evaluated.  The
program also contracted for the Patient Viewpoint Survey to assess
patient satisfaction with hospital services. 

Most of the costs of the Cleveland Health Quality Choice Program have
been borne by the participating hospitals.  The development and
implementation of the new program cost the hospitals about $8 million
over 3 years, with the participating employer groups contributing an
additional $1 million.\32 The Cleveland Health Quality Choice Program
estimates that the hospitals' overall costs average about $8 per
discharge. 


--------------------
\32 The program has recently contracted with Apache Medical Systems,
Inc., to market the CHOICE system to other hospitals; Apache will pay
further system development costs. 


      CHOICE
----------------------------------------------------- Appendix III:2.1

The CHOICE system was designed by Michael Pine and Associates in
collaboration with Cleveland-area health care providers.  The system
underwent a 3-year start-up phase beginning in 1990; the first report
was issued in spring 1993.  A clinically based system, CHOICE
estimates patient severity of illness by abstracting data from
medical records.  In addition to a patient's age, sex, race, and
diagnoses, the system retrieves information such as vital signs,
radiology results, and laboratory test values. 

Using the CHOICE model, the program reports both the expected and the
actual mortality rates and average length of stay for each hospital. 
The CHOICE results are reported for medical and surgical patients
admitted for 14 specific diagnoses or surgical procedures.\33 In
spring 1994, the program included summary performance for two general
categories (medical patients and surgery patients) and five narrower
clinical areas (cardiovascular, gastrointestinal, neurologic,
respiratory, and coronary-artery bypass graft patients).  In future
reports, the program hopes to also show comparisons of additional
specific services, including obstetrics. 

The CHOICE system has been criticized for making some hospitals
appear worse than their actual performance warrants.  For example,
one hospital has criticized the CHOICE model for neglecting to
distinguish between patients who are transferred from another
hospital and patients who initially seek care at the hospital.  This
may result in worse scores and unfair comparisons.  The hospital
contends that the severity of illness of a patient transferred from
another hospital may be understated because the patient's clinical
status is stabilized before the transfer.  However, the patient's
underlying conditions remain and may deteriorate soon after
admission.  An analysis conducted by Lewin-VHI indicates that the
hospital's mortality rates would improve if transferred patients were
separately accounted for in the data.\34 The hospital also notes that
its share of patients transferred from local hospitals increased
sharply in 1992. 


--------------------
\33 The medical diagnoses include patients admitted with acute
myocardial infarction, congestive heart failure, stroke,
gastrointestinal hemorrhage, pneumonia, and chronic obstructive
pulmonary disease.  Surgical procedures include coronary artery
bypass, peripheral vascular repair or bypass, lung resection, lower
bowel resection, laminectomy, reduction of hip fracture,
prostatectomy, and hysterectomy. 

\34 The Lewin-VHI study acknowledges that accounting for transfer
patients may not change the predicted mortality rates for all
hospitals, but could affect predicted mortality rates for some
hospitals that treat many transferred patients.  See Robert J.  Rubin
and William A.  Gold, "An Assessment of the Cleveland Health Quality
Choice Mortality Models" (Fairfax, Va.:  Lewin-VHI, Apr.  27, 1993). 


      APACHE III
----------------------------------------------------- Appendix III:2.2

APACHE III compares Cleveland hospitals with national norms for
intensive care mortality rates and length of stay.  APACHE III
develops a severity score for intensive care patients, ranging from 0
to 200, from which it computes expected mortality rates and length of
stay.  To estimate the patient's severity of illness, APACHE uses
medical records data, including 16 clinical values (such as
temperature, heart rate, and blood and urine test results) and
information regarding the presence of chronic health problems (such
as acquired immunodeficiency syndrome and cancer). 

The Cleveland Health Quality Choice Program reports both the
predicted and actual mortality rates and average length of stay for
each intensive care unit.  In future reports, after a sufficient
number of patients are included in the program's database to meet
statistical validity standards, the program intends to report
mortality and length of stay results for specific categories of
intensive care patients, such as cardiovascular, gastrointestinal,
respiratory, and neurologic patients.\35


--------------------
\35 The Cleveland Health Quality Choice Program recently recalled
APACHE data for these specific services that had been printed in the
June 1994 report.  The program attributed the mistaken inclusion of
these comparisons to a computer programming error. 


      PATIENT VIEWPOINT SURVEY
----------------------------------------------------- Appendix III:2.3

In addition to the severity-adjusted hospital performance comparisons
described in the previous two sections, the Cleveland Health Quality
Choice Program reports patient satisfaction comparisons among the
hospitals.  The Patient Viewpoint Survey is mailed to 600 randomly
selected patients from each hospital for each of the program's
semiannual reports.  In addition to such global satisfaction
questions as "Would you recommend the hospital to a friend?" the
survey also asks about 11 particular components of care, such as the
availability of doctors, the responsiveness of the nursing staff, the
patient's experience with the admitting and billing departments, and
the patient's satisfaction with food and housekeeping services. 

Patient satisfaction information has been criticized as being
subjective and often associated with characteristics of the hospital
that makes care more "personal" rather than improving outcomes of
care.  However, experts in quality assessment are increasingly
finding patients' evaluations of their care to be meaningful. 


      REPORTING RESULTS
----------------------------------------------------- Appendix III:2.4

Figure III.1 illustrates a Cleveland Health Quality Choice Program
comparison for surgical patients' average length of stay.  (In the
figure, the hospital names are coded for anonymity, but the program's
reports include hospital names.) As shown, the program assesses
hospital performance as above, at, or below the expected levels for
each performance measure (for example, surgical length of stay).  The
program cautions that hospital results should not be compared with
those at other hospitals, but only with its expected performance. 

   Figure III.1:  GAO Illustration
   of a Typical Chart in a
   Cleveland Health Quality Choice
   Report (Length-of-Stay Data
   Over Time for Surgical
   Patients)

   (See figure in printed
   edition.)

The program also summarizes hospitals' performance across indicators,
as illustrated in figure III.2. 

   Figure III.2:  GAO Illustration
   of a Typical Chart in a
   Cleveland Health Quality Choice
   Report (Summary Data on
   Satisfaction and Outcomes)

   (See figure in printed
   edition.)

In contrast to the Cincinnati and Orlando systems, the Cleveland
program does not report comparisons of hospital charges.  Although
such data are of interest to Cleveland employers, hospital and
physician representatives have resisted the inclusion of information
on charges in the program.  A task force is considering whether
hospital charges should be compared in the future. 

Of the employer-sponsored systems we reviewed, the Cleveland
coalition is the only one to have formal safeguards to prevent
hospitals from manipulating their data.  The Cleveland program
routinely audits a sample of records from each participating
hospital.  If a hospital is found to have overestimated its patients'
severity of illness, its medical records are reabstracted. 


   HOSPITALS' USE OF CLEVELAND
   HEALTH QUALITY CHOICE DATA
------------------------------------------------------- Appendix III:3

Several hospitals provided us with examples of changes they have made
in areas identified for improvement with the Cleveland Health Quality
Choice Program information: 

  -- One hospital noted that it has focused on developing treatment
     protocols for ordering laboratory tests and prescribing
     medications.  For example, the hospital has begun more
     aggressively treating pneumonia patients with antibiotics. 
     Although this change may not directly reduce costs, the hospital
     believes that it will improve the quality of care for these
     patients. 

  -- Another hospital recognized that it had not been using physical
     therapists efficiently for knee joint replacement patients.  The
     hospital is attempting to shorten stays for knee joint
     replacement patients by extending the hours physical therapists
     are available, changing the time of the surgery, and involving
     discharge planners earlier in the admission. 

  -- A Cleveland hospital is establishing clinical teams to develop
     treatment protocols for four conditions--pneumonia, chronic
     obstructive pulmonary disease, strokes, and congestive heart
     failure--for which the hospital had longer than expected stays. 

As in the other communities we visited, several Cleveland
participants said they expect that hospitals will become more
specialized and many will affiliate.  Several hospitals and employers
noted that the Cleveland hospital market has too many hospital beds,
but did not expect widespread hospital closings.  Instead, hospitals
and employers expect centers of excellence to be established for
specific clinical services.  Also, hospitals are expected to
affiliate so that a patient can stay within a single hospital system
for a range of health services--from primary care through
tertiary-level care. 


   EMPLOYERS' USE OF CLEVELAND
   HEALTH QUALITY CHOICE DATA
------------------------------------------------------- Appendix III:4

In most cases, employers are waiting until several Cleveland Health
Quality Choice reports are available before identifying trends in
performance and changing their purchasing decisions.  In a few cases,
payers have already used the Cleveland Health Quality Choice
information: 

  -- Aetna Health Plans of Ohio is using data from the Cleveland
     Health Quality Choice Program as a factor in establishing a more
     restrictive managed care network than it currently offers.  The
     insurer's existing contracts with hospitals, claims data, and
     geographic location of hospitals were also important factors in
     selecting hospitals for this managed care network. 

  -- One large employer indicated that the Cleveland Health Quality
     Choice information confirmed a decision it had previously made
     to remove a hospital from its managed care network. 

It is difficult to attribute any cost savings to the Cleveland Health
Quality Choice Program at this point because the first report was
issued in April 1993, probably too recently for hospitals to have
changed their practices in response to the program.  However, the
program's executive director estimated that participating hospitals
have saved nearly $20 million per year because patient stays have
been reduced by a total of 21,500 days.  But he acknowledged that
many factors contributed to this reduction.  The Greater Cleveland
Hospital Association noted that--even before changes associated with
the program occurred--hospital costs rose more slowly in Cleveland
than elsewhere in Ohio.  The hospital association credits the
increasingly aggressive health plan negotiations by business groups
for this slower growth.  Furthermore, a local insurer contends that
it has begun negotiating more effectively with hospitals in its
managed care networks. 


MAJOR CONTRIBUTORS TO THIS REPORT
========================================================== Appendix IV

Rosamond Katz, Assistant Director, (202) 512-7148
John Dicken, Evaluator-in-Charge, (202) 512-7135
Paula Bonin
Hannah Fein
Trisha Kurtz


*** End of document. ***