Session E: Ethical Issues in Health Care Delivery
Moderator: Lynn Kazemekas, Ed.D., AHCPR
Panel members: Philip J. Boyle, Ph.D., Park Ridge Center, Chicago; James E. Sabin, M.D., Harvard Center for Ethics in Managed Care
Physicians' Use of Outcomes Data: Moral Conflicts and Potential Resolutions—Philip J. Boyle, Ph.D.
The Hastings Center conducted a 2-year study on how physicians receive and use outcomes data
and practice guidelines. Evidence has shown that outcomes data and practice guidelines improve
quality of care, yet physicians sometimes ignore them. If managed care requires physicians to rely
on outcomes data and guidelines, problems may arise when they decline to follow them.
At four medical centers in different geographic areas, each with a diverse patient population, the
study addressed the following questions: (1) Are physicians aware of outcomes data and practice
guidelines? (2) Have their institutions begun to use them? (3) Are physicians paying attention to
them? and (4) What are physicians' hesitations in using them?
Three categories of objections were raised against the use of outcomes data and practice
- Physicians disputed the reliability of data or the objectivity of the practice guideline
- Physicians found the data reliable, but overrode the data because of a patient's request, the
inapplicability of the data, or the fear of a lawsuit.
- Physicians had undefined motivations (both professional and technology specific) that require
Another example of physician objection to the use of guidelines was found in the GUSTO (Global
Utilization of Streptokinase and t-PA for Occluded Coronary Arteries) trial. Here, physicians
were concerned with biased and defective methodology.
Physicians may object to using a guideline for moral reasons if recommendations that are implicit
or explicit in the document are based on the values of those who developed them. An option for
avoiding "value-driven" inferences is to publish data without any recommendations.
Doctors will not resist for moral reasons if only facts are presented, but their use of the data may
be limited. Alternatively, the author of a guideline document could acknowledge that it may
contain embedded values and attempt to persuade the physician to follow the guideline
Physicians may have other reasons for ignoring outcomes data and guidelines, such as when
outcomes data do not show an effect. In such situations, three options are available:
- Defer to the patient. Giving the patient the recommendations eliminates physician
resistance to guidelines.
- Create constraints or incentives by lowering reimbursements when physicians ignore
outcomes data or practice guidelines.
- Remove the physician from the decisionmaking process.
The following conclusions resulted from the study:
- The integrity of the physician is important: many will resist using outcomes data and
practice guidelines if they feel overly pressured or believe their own qualifications surpass the
- Although outcomes data and practice guidelines promote the best practices, the risk of
overcompliance has not yet been studied.
- When the staff of HMOs and managed care organizations regularly refer to practice
guidelines, they tend to understand them better and are thus more likely to follow them.
- Physicians do not always read outcomes data and guideline documents because of the
overwhelming amount of information presented in them. Doctors often make judgments without
- Physicians need rules on when deviation from outcomes data is acceptable.
Insurance Coverage for Promising But Unproven Last Chance Therapies: Ethical Issues and Practical Policies—James E. Sabin, M.D.
Treatments that might save, prolong, or improve life, but have not been proven to work can pose
a dilemma for the policymakers of insurers and managed care organizations. If a treatment proves
effective, withholding it can lead to premature death or harm, which violates the physician's
fidelity to the patient. Conversely, if the treatment is ineffective, using it squanders resources.
In 1991, Aetna developed a policy to manage decisions about unproven cancer treatments. Under
this policy, a request for an unproven treatment is channeled through headquarters to a suitable
specialist. If appropriate, the request is referred to an organization called the Medical Care
Ombudsman Program (MCOP). The program forms a panel of experts experienced in the
particular clinical area who have no attachment to the insurer or the health care delivery site. If
any of the experts consider the treatment appropriate, Aetna will cover it. MCOP determines only
if a treatment is appropriate, but does not make treatment recommendations.
In 1994, Northern California Kaiser Permanente instituted a program for unproven but promising
last-chance technologies. Kaiser does not authorize its physicians to prescribe questionable
treatments, so patients are guided through a series of consultations within the organization.
Should the patient and family be dissatisfied with the outcome of the internal consultations, Kaiser
offers a consultation or appeal outside the organization, a policy that triggered much internal
controversy. In 3 years, with a patient base of 2.5 million, only 6 consultations went outside the
organization. The Kaiser staff observed that when the concerns of all parties are adequately
addressed, a collaborative outcome can usually be reached.
The Oregon Blue Cross/Blue Shield (BC/BS) approach to policymaking in this area was to create
the role of transplant coordinator, a skilled individual who functions as a pre-problem
ombudsperson. If a patient is in a last-chance situation, the coordinator interfaces between the
patient, doctors, providers, and clinical researchers to find a solution. Oregon BC/BS is
committed to covering unproven treatments if they are part of a medical trial and will go to great
lengths to find such a trial. The organization does not pay for an unproven treatment that is not
part of a trial.
In conclusion, three key values are at stake in managing last-chance health care: fidelity to the
patient, stewardship, and science (which is attached to evidence). Additionally, motivation,
integrity, and reliability of both the data and those who use it are important issues for guidelines
and policy decisions. Last-chance technologies can serve as a good litmus test for collaborative
decisions and informed consent. The above examples show that when insurers address patient
concerns, ethical managed care is possible.
Return to Contents
Session F: Evidence-based Medicine: Implementation in Managed Care—Lessons Learned
Moderator: David Atkins, M.D., AHCPR
Panel members: Ronald J. Gebhart, M.D., Division of Primary Care, Department of Veterans
Affairs; Marianne Laouri, Ph.D., PacifiCare, Washington; and George Isham, M.D.,
Translating Evidence into Practice at the Department of Veterans Affairs—Ronald J. Gebhart, M.D.
The Department of Veterans Affairs (VA) seeks to provide quality care by functioning as an
integrated health system, that includes primary, specialty, and extended care as well as social
support. The VA provides care to more than 3 million military-connected individuals
At the recommendation of the VA Undersecretary for Health, the VA is developing and adopting
guidelines to improve consistency and effectiveness of treatment in this very large system. In
1996, 22 VA clinics served as pilot sites for guideline development. Steps were taken to maximize
the acceptability and implementation of the guidelines, once developed or adopted. Each was
tested in one site for a 6-month period before nationwide implementation. Local modification to
promote "buy-in" was encouraged, as long as it did not violate the scope of the
Guidelines developed June 1996 through April 1997 cover many conditions. Covered conditions
include: stroke, amputation, ischemic heart disease, major depressive disorder, major depressive
disorder with post-traumatic stress disorder, major depressive disorder with substance abuse,
dementia, diabetes mellitus, and psychosis. Since then, the VA sites have been developing
guidelines on Helicobacterpylori, prostate disease, asthma, anxiety, degenerative joint
disease, and gout.
In addition, pharmacologic treatment guidelines were completed in December 1996 for chronic
obstructive pulmonary disease, HIV/AIDS, hyperlipidemia, hypertension, and non-insulin
dependent diabetes. Currently in the works are pharmacologic guidelines for congestive heart
failure, depression, peptic ulcer disease, glaucoma, benign prostatic hypertrophy, and degenerative
Guidelines already developed by other agencies or organizations were adopted for many
conditions. For example, treatment guidelines for HIV were adopted from AHCPR, and
guidelines for cancer staging were adopted from the American Cancer Society.
Currently, data are being collected on potential improvement in patient outcomes; results will be
Translating Evidence into Practice at PacifiCare, Washington—Marianne Laouri, Ph.D.
PacifiCare, a network-model managed care organization (MCO), serves 3.5 million members.
With competing needs for its limited resources, PacifiCare must measure and improve care
effectiveness. Company-wide quality improvement programs attempt to do so under the direction
of Quality Measurement and Research unit. In December 1996, the Quality Initiatives Selection
Committee, which was made up of employers, providers, medical directors, and community
members, selected the following areas for measurement and improvement: maternal and child
health, diabetes management, cardiovascular health, and depression. Baseline data were collected
for each area, including clinical, administrative, and survey data. National guidelines were
implemented for each area, and followup data will be collected 1 year later to determine the
impact of the guidelines on quality of care.
The nationally implemented guidelines sometimes conflicted with locally developed guidelines. At
such times, feedback from physicians helped make the national guidelines more user-friendly. The
eventual success of national guideline implementation yielded several conclusions: (1) local
standards of care are not needed, (2) evidence-based guidelines can be used to increase
consistency of care, and (3) existing measures and guidelines can be used for quality
improvement, even in a large MCO or HMO.
Translating Evidence into Practice at HealthPartners, Minneapolis—George Isham, M.D.
HealthPartners is a not-for-profit group of 550 physicians who participate in an integrated system
of research, managed care, clinical care, training of medical personnel, and development of
practice guidelines and technology assessments. The philosophy of the system is to maximize
communication and cooperation among the various groups.
Guidelines are developed through the Institute for Clinical Systems Integration. Physicians are
recruited to choose topics, examine evidence, draft guidelines, and pilot test them. Then begins an
improvement cycle in which baseline and followup measures of treatment effectiveness are
gathered and compared. Buyers of health plans or employers sit on the Institute board and send
representatives to join guideline committees to provide the employers' perspectives.
Lessons learned to date in this process include the following:
- Physician involvement is crucial if physicians are to be comfortable with adopting the
- The marketplace must be right for successful implementation.
- Participation by purchasers and patients is practical and valuable.
- Guideline implementation is difficult and should begin in appropriate service areas.
- Measuring for improvement is unfamiliar to many.
- Annual review and revision by physicians is recommended.
Return to Contents
Session G: Using Information Systems to Support Evidence-based Practice
Moderator: Tina Murray, R.N., M.A.S., AHCPR
Panel members: J. Michael Fitzmaurice, Ph.D., Center for Health Information Technology, AHCPR; Cary Sennett, M.D., Ph.D., National Committee on Quality Assurance; and Peter McMenamin, Ph.D., Lewin Group
Using Information Systems to Support Evidence-based Practice—J. Michael Fitzmaurice, Ph.D.
The quality and cost of patient care can be improved by using information technology to link
patient information and medical knowledge. AHCPR supports research that identifies productive
uses of information technology; measures its benefits and costs; analyzes the impact on quality,
cost, and access; and identifies the barriers to information technology use. It also investigates how
information technology supports the changes occurring in our health care delivery system.
Providers resist computerization. Much of the clinical information needed for evidence exists only
on paper. Systems are not interoperable; for example, a hospital pharmacy system is not linked
with the radiology system or with the nursing station. Instead, computer information is printed out
and added to the patient's paper record.
Privacy of electronic medical information is of major concern. Debate is ongoing now about use
of unique patient identifiers in place of individual names. The Social Security number, however,
can link health information with employment records or Internal Revenue Service records.
Concerns to protect proprietary data are another part of the reality. Data warehouses (e.g., inside
a hospital or managed care organization) provide market power, because they can be used to
assess the impact of the organization's services.
Patients' rights are not yet defined in regard to access to and use of their medical records. This is
now being debated in Congress in discussion of the Federal Privacy Law. Currently, the owner or
holder of a record controls its use within the scope of that organization's policy, within the scope
of the patient's authorization, and within the State's law. This limits use by accreditors of
organizations, institutions, and providers; uses by researchers; and uses by public health
policymakers—all of whom collect evidence-based information to supply it to
Research on information technology in health care has shown that physicians who receive
computerized prompts on ordering tests and drugs change their ordering behavior, thereby
reducing costs without any apparent reduction in quality of care. Adverse drug events can be
detected and reduced. On the other hand, do-not-resuscitate orders in Living Will prompts often
are not carried out. Preventive service reminders for hospital inpatients are not very effective.
Barriers to using medical information systems include different code vocabularies (code systems
for recording patients' presenting conditions, physician diagnoses, physicians' treatment, and
patient outcomes are lacking); privacy, confidentiality, and security (no Federal laws, only a few
State laws); lack of obvious productivity gains in individual physician practices; and education and
training needs much greater than expected.
A 3-year AHCPR program, Computerized Decision Support Systems (CDSSs) for Health
Providers, is investigating how to link guideline/evidence-based information with patient data and
bring all to the attention of the physician and the patient at the time decisions are made. Three
focal points are (1) using guidelines in CDSSs while maintaining confidentiality of patient care
data; (2) looking at the impact of CDSSs on the patient care process, outcomes of care, and costs
of care; and (3) identifying and testing factors that influence practitioner use of CDSSs.
Specificity of guidelines should be improved, perhaps through evidence reports.
Implementing Guidelines in Systems of Care—Cary Sennett, M.D., Ph.D.
The National Committee on Quality Assurance (NCQA) is a not-for-profit nongovernmental evaluator of managed care plans serving as a source for public information on their quality and performance. The performance measurement program, Health Plan Employer Data and Information Set (HEDIS), is a set of standardized performance measures for assessing important results achieved by health plans. It contains information on immunization of children, adolescents, and the elderly; mammography, Pap tests, and diabetic screening rates; use of beta-blockers after acute myocardial infarction; and use of appropriate antibiotics in children with otitis media. HEDIS is now being revised.
The NCQA Information Systems Work Group was formed to communicate with the managed
care industry on ways to upgrade its information systems to achieve optimum performance
measurement. To assess the gap between current information systems environments and NCQA's
envisioned goal, NCQA conducted more than 60 interviews with members of the managed care
industry. A report on this effort is titled the "Information Systems Road Map" (and
now included in the latest HEDIS version). The report contains precise recommendations to
encourage the industry over a period of 3 to 5 years to achieve more capable information systems.
The recommendations detail seven elements that should compose the framework of an
- Data, including demographic characteristics of enrollees, important diseases, previous history, and
physical findings; results of diagnostic tests and procedures; therapeutic procedures; medications;
baseline and serially assessed functional/health status; patient and provider experiences of care;
treatment-related morbidity; and mortality.
- Linkages, consisting of communications links among data repositories; contracts that permit
access; rapid merging capability; and reduction of duplicate data entry and storage.
- Standardization, which involves unique identifiers, standard definitions of all terms; universal
medical record structure; code mapping, and messaging and data transfer standards.
- Data quality, achieved through verification and audit methods and data quality improvement
- Confidentiality and security, a matter of great public concern, controlled by means of strict
limitation of access.
- Automation, through which to create the electronic medical record. This goal of the electronic
medical record is seen as a long-range goal, given the financial barriers.
- Data-sharing capability, which must be increased among health plans, providers, and
databases of public health agencies (e.g., linking a health plan dataset to a SEER dataset using a
common identifier). To this end, Internet technology could be put to use.
The benefits resulting from improved information systems are that patient needs can be identified
more effectively; best practices can be identified and transferred; decision support tools for
providers can be implemented; recordkeeping efficiency and effectiveness can be improved;
resource management can be improved; and inappropriate variations in service can be
How Does Practice Translate Into Evidence?—Peter McMenamin, Ph.D.
Electronic claims data are being used to examine patterns of treatment practice, and allow
conclusions derived from this information to be translated into evidence. Existing datasets can be
used to examine and document current patterns of practice; the results are then fed back into
improvements in performance. Patterns of practice can be translated into cost-of-illness studies
(costs to treat persons with certain diagnoses or using certain procedures). Work can be done
now on interim outcomes, that is, short-term impacts that can be inferred by a continuing pattern
of receipt of services (or lack of additional services). Analysis of many records reveals patterns of
practice that can be used as benchmarks—how people actually are being treated.
The largest source of data is the physician dataset (which also might include ambulance drivers,
laboratories, and medical equipment suppliers). The most conventional record in use by providers
is the HCFA 1500. Hospital claims data are the second largest source, and another large dataset
comes from pharmacy claims.
A linked set of records has more value than a single record to illustrate common diagnoses and
subdiagnoses and services that seem to follow one another or be associated with one another.
Occasionally a record does not fit into discovered patterns and can be dismissed. Linking records
requires common elements such as: patient elements such as a nominal ID, the time of birth or
age, gender, and sometimes race (patient health status, clinical content, and health history are
usually missing); physician elements including a nominal identifier such as a TIN, a medical record
number, or a license number; hospital elements from hospital datasets or individual State files to
identify the hospital's facilities; information on pharmacies as providers; and insurance claims
Identifying from claims data the services being provided to patients is a challenge. In the United
States, an extensive nomenclature is used to code physician services—CPT. Procedure codes
in IC and ICM are much less specific than CPT and correspond only roughly. National Drug Code
(NDC) codes are a very detailed set of nominal identifiers; each code has two parts—one for
the manufacturer identity and the other for the drug (a sequence number). These codes are
difficult to use because therapeutic class cannot be determined from the code. Diagnosis-related
groups (DRGs) are used in hospitals but do not reveal much information on services.
Diagnosis data are coded in many ways, depending on how data are entered, what the insurer
requires, how the patient presents, and whether the physician is trying to rule out some alternative
condition. Patterns of diagnoses yield useful evidence. It is important to remember that any one
patient may have several diagnoses at the same time, particularly in Medicare, the source of the
Return to Contents
Proceed to Next Section