Symposium on data in a capitated environment. Introduction.

Data about the utilization of health care services have been used in the public sector to serve a number of important purposes. One major purpose has been to gain an understanding of the patterns of health care use. By analyzing utilization of services in relation to sociodemographic variables such as age, sex, race or ethnicity, income level, and insurance coverage, knowledge can be gained about the factors that influence access to health care services. By analyzing morbidity, mortality, and hospital readmissions in relation to the provision of services, information can be developed about the quality of care delivered. When data on utilization and payments are analyzed by certain variables such as type of provider, insight can be gained about the appropriateness of those payments and the effectiveness of different health care delivery systems.

The Health Care Financing Administration (HCFA) has been able to analyze the patterns of health care use by Medicare beneficiaries from the records of bills submitted for payment. These data, augmented by data collected in national surveys and special research and demonstration projects, have influenced the formulation of many policy initiatives for the Medicare program. A major influence on the development of the Medicare hospital prospective payment system, for example, was the knowledge gained about the experience of the Medicare program through data based on claims for payment. Claims data are now being used to evaluate certain aspects of the impact of the new system. In contrast to the extensive data on fee-for-service use, information available to HCFA on the use of services in capitated delivery systems is confined to a few aggregate statistics.

The role of health maintenance organizations (HMO's) in the delivery of services to Medicare beneficiaries has grown with the passage of the Tax Equity and Fiscal Responsibility Act (TEFRA) of 1982 and is expected to increase substantially in the future. To gain an understanding of the use of data in private-sector capitated delivery systems, and the lessons to be learned from their experience to date, the Office of Research and Demonstrations invited six organizations to take part in a written symposium focusing on five broad questions relating to HMO data systems:

The major purpose of their data systems.

The basic features of their files.

How their data systems are being used today.

Personnel and costs needed to operate their systems.

The kinds of data requests made by employer groups.

These organizations were invited to provide information because of their interest and activities in HMO data systems and data issues. The full set of questions is shown preceding the responses of the participants.

The symposium brings together three kinds of viewpoints on these HMO data issues. First, executives from three HMO's respond to questions about their internal data systems. The contributors are Bernard Neeck, Senior Vice President and Chief Financial Officer, and Daniel Kennedy, Vice President and Underwriter, of the Health Insurance Plan of Greater New York, a large nonprofit network model; Joseph LaAsmar, President, and Mary Tierney, Associate Medical Director, of the Chesapeake Health Plan, a for-profit network and independent practice association (IPA) model in Maryland; and Gino Nalli, President, of M.D. IPA, a for-profit IPA model in the Greater Washington-Baltimore area.

A second viewpoint is provided by Gail Warden, Chief Executive Officer, and Michael Wagner, Director, Data Management Center, of the Group Health Cooperative of Puget Sound in Seattle; and by Mark Hornbrook, Senior Investigator, Merwyn Greenlick, Director, and Marjorie Bennett, Research Associate, of the Center for Health Research, Kaiser Permanente in Portland, Oregon. These authors discuss data not so much from the perspective of the systems in place in their own HMO's but from the perspective of the general information requirements of HMO's (Warden and Wagner) or from the point of view of what the Medicare program might need (Hornbrook, Greenlick, and Bennett).

A third point of view is provided by Laird Miller, from Honeywell Inc. in Minneapolis, who is his company's Corporate Director of Health Systems and its representative on the Washington Business Group on Health, the Midwest Business Group on Health, and the Minnesota Coalition on Health Care Costs. He discusses the kinds of information that purchasers of care need to make sound choices among plans as well as the recent change in scope and focus of Honeywell's data collection activities.

Several broad themes and issues emerge from the articles in this symposium. The first theme is the need for data in the increasingly competitive environment in which HMO's find themselves today, an environment in which superior management is becoming a requirement for success. Competition may come from other HMO's, some affiliated with large for-profit or not-for-profit chains, as well as from preferred provider organizations, other modified fee-for-service plans, and traditional fee-for-service plans.

In this highly competitive environment, a primary purpose of current data systems in HMO's is utilization control. HMO's are interested in monitoring how their physicians and enrollees use services, such as referrals to specialists, inpatient care, laboratory tests, and prescriptions. They have instituted reporting systems to track the use of these services.

Nearly all of the respondents note the importance of developing their data reports by the demographic characteristics of their enrollees, such as age and sex. Some have also pointed out that it is necessary to link the diagnosis of the patients to the utilization of physician services by specialty in order to effectively monitor the use of services. These data are used by HMO managers for resource control, for planning, for providing physicians with objective data on their practice styles, and for peer comparison.

A second major theme—mentioned by Miller; Warden and Wagner; and Hornbrook, Greenlick, and Bennett—is the desirability of data on health status and health risk of covered populations and on outcomes of care related to patient risk. Risk factors would include smoking, weight, cholesterol level, and blood pressure. HMO's would use such data to manage resources to produce the best possible outcomes. Purchasers of care would use these data to evaluate the performance of providers and health plans serving their employees. In addition, Miller points out the need for employers themselves to collect data on environmental and occupational risk factors for a number of reasons, including tracking the experience from risk exposure to medical care events and costs.

Another important theme is the need for gathering data periodically on consumer satisfaction. Such data are generally collected by means of surveys. Information important to HMO's includes waiting time for appointments, complaint patterns, and reasons for disenrolling. As noted by Warden and Wagner: “Long-term measures of health status will be meaningless unless the HMO administrator can maintain a reasonably stable enrollment base.”

An interesting fact brought out in this symposium is the low costs of operating the management information systems at the Health Insurance Plan of Greater New York (HIP), Chesapeake Health Plan, and M.D. IPA. The figure cited for HIP is $0.46 per member per month; for Chesapeake Health Plan, $0.48; and for M.D. IPA, $0.22. Recent site visits were made to the latter two plans. Demonstrations of their systems indicated impressive ability for rapid retrieval of longitudinal data on enrollee encounters.

Finally, Miller raises two issues that are important from the standpoint of the employer. One issue is the change in management strategy relating to health. At first, Honeywell worked to collect utilization data primarily to contain health care costs. However, their emphasis has recently shifted from “managing health care costs to managing health.” Their belief now is that, to manage costs, the employer needs to work to “prevent problems from occurring through early identification and evaluation of personal risks, workplace hazards, and environmental and community health problems.” Miller notes that, to achieve this purpose, health plans require sophisticated information systems for collecting data that allow comparisons across plans and providers.

The second major issue raised by Miller relates to the difficulty Honeywell experienced in obtaining data from the fee-for-service insurance carriers that cover 75 percent of their employees and from the HMO's and preferred provider organizations that enroll the remaining 25 percent. With regard to HMO's, Miller notes that some could not “identify by employer group patients who receive care.” In contrast to Honeywell's experience, unique patient identifiers are a basic part of the Chesapeake Health Plan and M.D. IPA data systems, and HIP is planning to incorporate patient identifiers into their encounter system.

We believe that these articles provide valuable insight into the importance and the role of data in the changing health care scene. If future changes in HMO data systems are in keeping with the issues raised in these articles, we anticipate the following developments:

Increasing sophistication in the systems used by HMO's for utilization control and management.

Increasing demands by purchasers of care for data to judge the appropriateness of premiums, quality of care, and health outcomes.

Integration of clinical measures into utilization and enrollment data to develop data bases that allow HMO's to manage resources for production of optimal outcomes and to provide purchasers with evidence of good performance.

Pressure from purchasers of care for common measures to allow performance comparisons across health plans.

Attempts to develop population-based measures of health risk for employee groups and HMO enrollees.

Developments will depend on the nature of the requests from employers for performance data, on technical advances in measuring health risk and outcomes, and also on any decisions that may be made regarding data about Medicare and Medicaid enrollees in HMO's.

What were the major purposes and objectives as you considered a system for maintaining data on enrollment and utilization of services? What kinds of questions and management concerns was the system designed to address?

Please describe the basic features of the enrollment and utilization data systems, including an overview of the types of files maintained. Did your HMO design the enrollment and utilization data system, or did you use the services of a consulting firm? We are interested in both the hardware and systems design. For example, do you have a computerized file of physician visits? What are the major data elements in that file? Can the data be linked for a particular enrollee or physician? Do you have a similar file for hospital admissions and referrals? How might you change the system in the future?

How are the systems used now? How important are they to your HMO? Do you generate periodic reports? What variables are of greatest importance in these reports? For example, do you analyze visits per person and/or visits per 1,000 enrollees? What levels of aggregation are used? Are data aggregated by provider and/or by market area? For what issues might you do special analyses? Do you have any examples of how data analyses led to changes in the way you manage your HMO? Is the system used on line also?

Relating to personnel needed to maintain the systems and systems' costs: Could you tell us about the number of staff needed to update and maintain the information systems? Could you estimate the cost of operating your system on a per enrollee basis?

Employers, both singly and in health care coalitions, are increasingly interested in data on their employees' experience in health plans. What kinds of utilization data have employer groups requested, or do you anticipate such requests in the future?

Introduction

The Health Insurance Plan of Greater New York (HIP) is the country's second largest prepaid group practice plan. HIP began operations in 1947, offering medical care on a prepaid group basis. Members were required to carry hospital insurance, which was purchased through Blue Cross, union, or private plans. Hospital coverage was not offered until December 1, 1978, when HIP was certified as a health maintenance organization (HMO).

Responsibility for HIP policy and operations rests with an unpaid Board of Directors composed of 24 individuals. HIP management is responsible for the operation of the plan.

Of HIP's 900,000 members, 777,000 are enrolled in HIP/HMO. Included in these figures are 75,000 Medicare members, 50,000 of whom are covered through HIP/HMO Medicare supplemental contracts. HIP began enrolling Medicare eligibles in July 1966, the first possible enrollment date after the enactment of Title XVIII of the Social Security Act. At the present time, HIP has a cost contract with the Health Care Financing Administration (HCFA).

HIP has entered into a single contract with nine affiliated medical groups. This contract is an all-inclusive document detailing the contractual relationship between HIP and the medical groups. It encompasses the medical groups' responsibilities in providing appropriate medical care as well as HIP's obligations toward the medical groups, including compensation.

Medical care is rendered by more than 1,000 Board-certified or Board-eligible physicians who participate in HIP's nine affiliated medical groups and 53 centers located in the Greater New York area. Members may choose any HIP medical group. Family members, if they wish, may select different medical groups. Within the center, members select their own personal physician. Hospital care for HIP members is provided in two community hospitals affiliated with HIP and in other appropriate hospitals.

HIP has a large computer facility and has many specialized computer systems in operation. Of these, the enrollment system is probably the single most important.

Enrollment system

Background

The enrollment system currently in place at HIP was designed and developed in-house by HIP staff and has been in use since 1968. The current enrollment system is a batch system that is run on an IBM 4381 mainframe computer. HIP is presently developing and implementing a prototype (pilot) system for adding data on line.

The major purposes of the file, as designed, were as follows:

Verification of eligibility of members.

Billing of premiums.

Payment of capitation to medical groups.

Mailing to members.

Maintenance of data on maturing of members.

Production of enrollment statistics.

The management of HIP was concerned about a number of issues, and the system was designed to provide information to address these issues. The major concerns were control of premium income and information for rate setting, the age and sex of membership, membership by contractor, distribution of membership across medical groups, enrollment trends for planning new facilities and new programs, and staffing patterns.

Enrollment file

HIP's computerized enrollment file is the base from which vital information for the operation of the Plan is derived. The file contains the following major data elements for each member:

First name, middle initial, last name.

Month and year of birth.

Home address.

Sex code.

Medical group selected.

Physician number.

Medical chart number.

Medicare number, if applicable.

Coverage effective date.

In addition, for the contract holder only, the file contains contract number (group number), premium period, and rider code.

The file also contains a complete history of changes of data items. These data changes are retained on line for a period of 12 months. Access to further historical data is available on request.

Features of enrollment system

The enrollment file contains a number of features important for the administration of the Plan:

Online checking of member eligibility by policy number, subscriber name, or dependent name.

Interface with HCFA for enrollment and premium payment for the Medicare population.

Interface with the New York City government for enrollment and premium payment for the Medicaid population.

Daily updating of the file.

Weekly issuance of identification cards.

Weekly notification of changes to the medical groups.

Summary enrollment reports, issued monthly, on topics such as major contractor groups, medical group, type of coverage, and age/sex. (A report writer is available so that user departments can obtain their own ad hoc enrollment reports.)

Information on maturing of members (e.g., change of status for dependent children at age 19, eligibility for Medicare).

Reconciliation of enrollment file with contractors' records.

HIP is considering the need to redesign the enrollment master file to handle future new products and further support the need of company management for ad hoc reports. To this end, we are currently evaluating the purchase of an automated data-base management system to handle enrollment and utilization files.

Recently, HIP installed a prototype online system at one of our medical groups. The system permits verification of enrollment, schedules appointments, and produces automated encounter reports. Current enrollment data from the HIP enrollment file are loaded into the system each week.

Support files

Contractor file

This computerized file contains major data items pertaining to the contractor, such as name and address, name of the benefit administrator, name of the HIP marketing representative handling the account, past and present premium rate information, billing data, account anniversary date, and codes describing the benefits purchased.

Physician file

The other major support file is the computerized physician file, which contains data on physicians approved by HIP's Medical Control Board. The Medical Control Board is composed of 21 physician members, one-half of whom come from HIP and one-half of whom are distinguished faculty members of New York teaching hospitals. To be considered as a medical group physician, an applicant submits detailed documentation concerning his or her medical education and background to HIP. This information, together with the curriculum vitae, is submitted to the Medical Control Board for review. Only after a positive vote from the Medical Control Board can the physician work for HIP. Data elements include physician's name, New York State and New Jersey license numbers, HIP identification number, medical specialty codes, board specialty status, medical group affiliations (current and past), hospital affiliations, date of birth, and sex. This file is updated daily and contains information on terminated physicians as well as those currently active.

Utilization system

HIP undertakes a coordinated centralized effort to have an effective utilization reporting system in place for all medical and hospital services. This enables us to monitor costs and recognize emerging utilization patterns. Data on utilization by medical group and service specialty also provide a basis for planning new facilities or programs and determining staffing patterns, as well as providing information required by regulatory agencies and contractors.

HIP has a number of different systems that maintain data on utilization of the following types of services:

Encounters with professional staff affiliated with medical groups (in and out of hospital).

Services provided on a referral basis by consultant specialists providing tertiary care.

Hospital utilization by our HMO membership.

Each system is described separately.

Encounters with medical group staff

Information on face-to-face encounters with professional staff members is manually entered on encounter forms. These forms are processed in the Research and Statistics Department, and aggregate totals of services by individual physicians, nurse clinicians, and physician assistants are entered into a computer system each month. The major data elements in the computerized file are:

Number of visits.

Provider name and specialty.

Month and year of service.

Medical group/center where service was provided.

Location of service, i.e., in or out of hospital.

The system does not contain information on individual members receiving service or details on diagnosis/problem or procedures performed.

The enrollment system and the reports generated enable HIP, on a current basis, to obtain all data relating to services provided to our members. The physician encounters are reported by specialty for each of the medical groups. The system provides HIP management and the medical groups with reports and tables showing the physician utilization rates of services per enrollee, as well as the number of services provided by individual physicians. This information is available for each medical group. Other information details utilization by age and sex for each medical group, as well as the number of services provided in each specialty, including services performed by the Centralized Laboratory Service. Groups that have higher or lower utilization than anticipated are reviewed by our Quality Assurance Program.

The aims of the Quality Assurance Program are not only to identify potential problems but to work with central management and medical group leadership to quickly correct problems. The program's professional staff conducts performance surveys based on protocols covering administrative and clinical activity at the medical group level. A comprehensive profile of each group is provided to both the medical groups' management and senior management at HIP for appropriate joint action.

As part of ongoing efforts to make the system operate more effectively and efficiently, HIP and its affiliated medical groups began planning, in 1984, a major program for automation of medical group support systems. One of the planned programs will result in automated encounter reporting, with data entered at the medical groups and transmitted to Central HIP for additional processing. Some of the data elements that will be captured in the new system are:

Member identification.

Diagnosis.

Reason for visit.

Procedures/services provided.

Provider identification.

The automation program will be operational in pilot locations during 1986 and will be expanded to all medical groups thereafter.

Services by consultant specialists

Rare or highly technical procedures that would be uneconomical to provide in each medical group are paid from the Special Services Fund maintained and financed by HIP. These procedures are performed by consultant specialists under contract with HIP. They include, for example, cardiac surgery, neurosurgery, certain transplant procedures, and reconstructive plastic surgery including skin grafting.

The consultant specialist, on completion of the authorized treatment, submits the medical group's referral form to HIP for processing and payment. These referral forms are also used by HIP to develop utilization reports. The following data are entered into the computerized Special Services Fund file:

Patient's policy number, contract number, medical group, date of birth, sex, and relationship to subscriber.

Date of service.

Codes identifying Medicare or Medicaid status.

Code number and specialty of referring group doctor.

Code number and specialty of consultant specialist.

Current procedural terminology (CPT) code number identifying type of treatment rendered.

Amount of approved payment.

Based on this information, utilization reports on services provided and costs of these services are developed by medical group and procedure.

Hospital utilization

The hospital utilization data for non-Medicare HIP/HMO members are derived from the computerized hospital claims file. This file is updated daily, and reports are periodically run from it. These reports summarize services by medical group, hospital, age and sex of patient, and type of service.

The major data elements in the computerized file are:

Patient's policy number, contract number, medical group, date of birth, sex, and relationship to subscriber.

Hospital code.

Date of admission.

Number of paid hospital days.

Diagnostic information based on the Ninth Revision of the International Classification of Diseases, including type of service.

Amount paid to hospital.

The hospital utilization reports generated from the enrollment system enable HIP to monitor total cost and hospital days per 1,000 enrollees for the Plan. Also, the reports enable HIP to monitor cost and days by medical group as well as diagnosis. Variations from targeted costs and days are investigated.

Medicare claims are processed by Medicare carriers, and utilization data are not provided to individual insurers.

Data requests

HIP receives requests from regulators, contractors, business associations, and other health care coalitions for enrollment, cost, and utilization data. These requests are for aggregate data as well as specific breakdowns, such as for age and sex.

Utilization and enrollment data for Medicaid eligibles enrolled in HIP are provided on a regular basis to the New York City Department of Social Services. A Medicare Cost Report reflecting the reimbursable cost for Medicare beneficiaries enrolled in HIP is sent to HCFA. Enrollment information concerning members insured through the Federal Employees Health Benefit Program is provided to the program.

Over the last few years the requests for data have increased significantly. HIP has responded to these requests whenever possible and will continue to do so in the future.

Management information system operating costs

For the 1986-87 fiscal year, HIP anticipates a total expenditure of $5 million for the Management Information Systems Department. This figure includes the cost of computer rental and professional staff as well as overhead items such as rent and utilities. The present HIP enrollment is 900,000 members, so this results in a cost of $0.46 per member per month. This expenditure estimate includes the enrollment system's expense as well as the costs of providing other data processing services to HIP Central, medical groups, contractors, and regulators.

Background

The Chesapeake Health Plan (CHP) is a nonprofit, State-licensed health maintenance organization (HMO) in the State of Maryland. It began operations on the grounds of Baltimore City Hospital (now Francis Scott Key Medical Center) in July 1976 under a prepaid Title XIX contract with the State of Maryland. In 1979, CHP was licensed to operate as a State-certified HMO. From 1979 to 1981, CHP marketed chiefly to the population eligible for Medicaid (i.e., Maryland Medical Assistance). In 1981, the Plan began the process of marketing to the private sector.

Further, CHP management made the decision to expand from their original five sites, which were located mainly in East Baltimore. Those original sites were operated through a contract for physician services with Chesapeake Physicians, P.A. (CPPA), a large multispecialty group practice. The decision was made to contract with additional existing provider groups and hospitals to provide a comprehensive Baltimore-wide network serving an enrollee patient population from both Medical Assistance and private sector populations. We also decided to enter the private market in Anne Arundel County, starting in Annapolis and expanding through Severna Park and Glen Burnie.

At the same time that we were making our expansion plans, the Robert Wood Johnson Foundation announced a grant competition entitled “A Program for Prepaid Managed Health Care.” The grant specifically concerned the Medicaid-eligible sector. Chesapeake Health Plan applied for the grant and was one of 12 awardees from across the country. The purpose of the project is to test whether we can enroll the Medical Assistance population, provide high-quality health care, and develop utilization review programs and techniques for the benefit of patients, providers, and third-party payers alike. Since the award, we have been engaged in the task of expanding our provider base and our Medical Assistance and private sector enrollment. The management information system (MIS) is playing a crucial role in marketing, quality assurance, and utilization review as we expand in metropolitan Baltimore and Anne Arundel County.

Purposes and objectives of MIS

At the time the MIS was first being developed, the purposes and objectives were quite different from our present needs. The objectives considered during the developmental stage included:

Automating and increasing the speed of processing for all types of data.

Enhancing the accuracy of data, data outputs, reports required by various regulatory agencies, and internal reports.

Devising Plan management standards and comparisons through review of utilization data by age, sex, and employer group.

Improving clinical management standards and comparisons (e.g., for physician productivity—visits per week, per month, and so on) through use of data on such variables as length of visits and type of patient and problems seen.

Implementing medical quality assurance by reviewing automated medical records for given diagnoses and comparing the data with standards set by a quality assurance committee.

Developing a utilization review system in which data—including data on specialty referrals, emergency room use, and hospitalization by physician and by health center site—are monitored.

Developing an automated medical record that includes problem lists, pharmaceutical orders, and the interaction of drugs with other drugs and forms of therapy.

Running clinical trials and other basic medical research studies.

Over the years, the objectives of the system have been changed and simplified. Our main concerns are now to provide enrollment and demographic information, utilization statistics, and financial information. Despite the changes in objectives, the system remains an encounter-driven system that allows for clinical and plan management. Unlike some other systems, its main purpose is not solely to address financial concerns such as claims processing and billing.

Description of data

Our management information system is a hybrid that utilizes existing hardware: Nixdorf, IBM, and molecular interfaces. The main system is encounter driven. Demographic information on the enrollee is updated at the time of an encounter. In addition, an encounter form is generated at each visit. The encounter form serves a dual purpose. The first page of the form becomes a permanent part of the medical record. The second page is imprinted through a carbon. Various procedures ordered, such as a complete blood count, are translated on the second page with their corresponding current procedural terminology (CPT) code. The second page of the encounter form is forwarded to be keypunched into the computer. The reports (outputs) generated provide information in the categories of enrollment, utilization, and financial statistics. (Figure 1 outlines the Ambulatory Care Encounter System used by the Chesapeake Health Plan.)

Figure 1

Ambulatory Care Encounter System (ACES) management information system

Enrollment data

The enrollment data provide demographic information such as age, sex, family configuration, location of work and home. For those enrollees who are Medicaid eligible, the system provides information concerning category of eligibility, such as Aid to Families with Dependent Children or Old-Age Assistance. It also indicates at which of our various sites the person is enrolled.

Utilization data

The utilization data provide information for three overall categories: patient profiles, physician profiles, and aggregated utilization statistics.

The patient profiles can be linked for a particular patient/enrollee. Information that can be retrieved on request includes:

Problem lists by individual patient and attending physician.

Detailed automated medical records that include diagnosis for each encounter, pharmaceuticals prescribed by encounter, referrals made by encounter, and attending physician.

The physician files supply reports on individual physicians and include the following:

Monthly activity by provider (i.e., number of patient encounters per month).

Diagnosis by provider.

Specialty referrals made per 100 encounters.

Laboratory procedures ordered per 100 encounters.

X-rays ordered per 100 encounters.

Aggregated utilization files include:

Total number of referrals to various specialists by department—orthopedics, dermatology, etc.

Hospitalization by health center site and provider.

Ancillary services (i.e., prescriptions, laboratory procedures, and X-rays) by site and provider. The average numbers of prescriptions per encounter and laboratory procedures per encounter are calculated for the report.

Statistics by site and by provider. Encounters are calculated by sex, age, and diagnostic class.

Statistics on diagnoses. The top diagnoses file lists the 10 most common diagnoses for adults and pediatric patients.

Financial data

A monthly trial balance is provided, as well as information on claims processed and paid. The system also generates checks for claims to be paid. Unlike many systems now in operation, our MIS was designed in-house, so the Plan did not have to adapt to an existing system sold by outside vendors. Its design was a physician-guided effort. Thus, the major emphasis of the MIS is on clinical and plan management rather than on being a financial claims payment system only.

Uses of data

Specifically, the system is used now for utilization review and quality assurance, billing for coordination of benefits, and claims processing for and reports on specialty physicians and inpatient hospital stays. The MIS allows us to compute capitation payments to provider groups who are providing services on a contractual basis. For example, in obstetrics and gynecology, we total the previous year's actual payment experience by the total membership and derive a capitation payment per member per month. The system's aggregation of the data by provider, by enrollee, and by market area is particularly important for utilization review, financial management, and plan management (such as marketing).

The system is used for marketing in both the private and public sector. Using the demographic information available to us, marketing staff is able to compute rates for the private sector and pinpoint areas of the city to be targeted for intense marketing efforts. We have the capability to review primary and specialty physician utilization and hospitalization rates by employer and other groups (i.e., Medicare, Medicaid). We also have data on age and sex cohorts by these groups. We are considering the possibility of experience rating private employer groups depending on their utilization of services.

Reports are generated on a weekly, monthly, and quarterly basis. The data are aggregated by provider, by site, by specialty of the provider, and by financial class and/or group.

The variables we analyze that are of greatest importance relate to utilization patterns by providers and enrollees. They are:

Number of referrals by provider per 100 encounters.

Number of laboratory procedures by provider per 100 encounters.

Number of X-rays by provider per 100 encounters.

Inpatient utilization by provider, by site, and for the Plan as a whole in terms of days per 1,000 enrollees and admissions per 1,000 enrollees.

The utilization data by provider are supplied to each site medical director, who can use the data to modify provider behavior vis-a-vis specialty referral, laboratory usage, hospitalization, etc. Individual and site utilization patterns are compared with Plan-wide standards, which are set by the Quality Assurance/Utilization Review Committee. For example, we would expect that 10 of 100 pediatric encounters (under 19 years of age) would result in a specialty referral, whereas 30 of 100 adult medicine encounters would be referred to specialists. If a primary care provider's referral rates consistently vary more than 10 percent from the standard, his or her referrals could be reviewed individually by the medical director. In some cases, prior approval for referrals may be necessary.

We also analyze costs per day for all hospitals in the area. These data include the basic facility costs, laboratory charges, and X-ray charges. We have been able to pinpoint certain hospitals that have excessive facility and ancillary services charges. We have several options. For example, overutilization of ancillary services has been of special concern in teaching hospitals. In response, management has met with hospital administrators and chiefs of services with regard to the use of inpatient ancillary services. If the problems cannot be resolved, we have, from time to time, steered admissions to other facilities that provide high-quality care at a more reasonable cost.

We also perform special analyses that include data from patient satisfaction surveys. The analyses may address studies about waiting times for appointments with primary care and specialty providers. We have also run special analyses that allow us to improve patient management. For example, when the new Hemophyllis Influenza B (HIB) vaccine became available, we developed a program to identify all children aged 2-5 years. We then contacted the parents of all these children, advising them to bring the children into the health centers in order to receive the HIB vaccine.

Finally, several of these information subsystems are on line. They include the registration file, which has demographic data. This information is used daily to update information on individual enrollees, such as additional health insurance carried by the enrollee, address changes, and additions to the family. Individual patient health histories, including problem lists, medications, referrals, and hospitalizations, are also on line. Physician profiles of their patient population by diagnosis are available.

Costs

Costs for maintaining these systems and reports are quite reasonable. Costs on an enrollee basis can be determined by aggregating the per-encounter costs generated each time the system supports a face-to-face enrollee visit with a provider. The cost per encounter is approximately $1.45, excluding corporate overhead and indirect charges.

Cost per encounter can be broken down as follows:

Chesapeake Health Plan enrollees average four encounters (including specialty visits) per year. Therefore, the estimated cost per enrollee is $5.80 per year, or $0.48 per enrollee per month.

Requirements for future systems

After 4 years of using our MIS, we have made some changes. Some on-line capabilities of the computerized medical record have been dropped, although the statistics are still available on request. We have also felt a need to revise and simplify our reports. Many of the hard-copy reports are very complex and detailed, making them difficult to utilize by inexperienced persons. We are interested in providing reports which would allow management to review physician practice patterns vis-a-vis referral rates, use of ancillary services, hospitalization rates, and referrals to emergency rooms. In addition, marketing and management need utilization reports by employer group in order to develop rates and to key in on specific marketing areas and groups.

We are now in the process of revising our MIS. An effort is being made to simplify future outputs in order to make them more immediately useful to line management. We anticipate that marketing, utilization, and financial data will remain of utmost importance. At the present time, employer groups have not requested utilization data on their employees. However, we would not anticipate any problems with providing requested information.

Finally, we believe that the development of a management information system is an evolutionary and ongoing process. We intend to reevaluate the system in order to make it useful in the management of our HMO.

M.D. IPA is a health maintenance organization (HMO) serving the Greater Washington-Baltimore area. In operation since January 1981, it now has a membership of approximately 90,000. Services are provided through a network of about 2,000 providers and most of the hospitals in the area. M.D. IPA originally entered into a Medicare cost contract for its approximately 1,500 Medicare members. Since early 1986, however, it has operated under a risk contract. M.D. IPA is a for-profit corporation, its stock owned principally by physicians.

Management information system

On one level, the objective of the M.D. IPA management information system (MIS) is to automate those administrative processes that are necessary for the accurate and timely identification of enrolled members, collection of revenue, and payment of claims. Of equal concern, however, is the need to provide timely data necessary to the management of the prepaid program, particularly with regard to evaluation of our experience with various enrolled populations (i.e., commercial, Federal, individual, and Medicare). This information enables senior management to initiate appropriate program refinements. The information necessary for program management is provided by the data collected through the administrative enrollment and claims payment processes. A final objective is the ability to manipulate and format the data produced through the MIS to respond to a series of “what if” questions. This capability is considered critical to ensuring flexibility in the management process.

The M.D. IPA information system is an in-house, interactive, data-based MIS. Hardware is the Honeywell Level DP-6 with firmware provided by Ultimate Corporation. The PIC operating system is utilized. M.D. IPA purchased access to the software from Comtec Corporation, which developed a basic MIS package for the HMO industry. Comtec provided significant customization of the product to meet M.D. IPA's specific operational configuration. Enhancement of the MIS is a continuing activity as the needs and design of M.D. IPA evolve. These changes are presently accomplished by an in-house MIS staff.

The MIS manager, who reports to the Vice President of Operations, is responsible for day-to-day operations of the system, programming, hardware enhancements, and establishing department objectives and priorities. The MIS staff in 1986 includes two full-time development programmers, one full-time maintenance programmer, a personal computer coordinator, and two part-time operations coordinators. The direct operational costs for MIS in 1986 total approximately $225,000, or $0.22 per member per month.

The basic reference files in our MIS are member, group, provider, procedure and diagnosis, and encounter history. The basic features of each file include:

Member—Demographic information about the subscriber and dependents enrolled in the Plan.

Group—Billing information about each membership group (i.e., rates, benefits, plan, and contract date).

Provider—Descriptive information about contracting physicians and other providers that is necessary to determine payment status.

Procedure and diagnosis—Listing of allowable diagnoses and procedures for coverage.

Encounter history—Data about actual utilization experience and costs for services provided.

Data manipulation within each file can be easily accomplished with an integrated report generator, for example, an analysis of subscriber residences by ZIP Code. Analysis of data among major reference files usually requires specific programming, which is provided by in-house staff.

The MIS enrollment system is derived from the member and group reference files. The utilization system is derived from the member, provider, and encounter history files. Both components of the MIS are critical to the operation of M.D. IPA.

Periodic reports are generated for utilization analysis, rate setting, financial reporting, market analysis, and other specific management inquiries. The online interactive capability of the system permits data aggregation at different levels, depending on the specific need. For example, member utilization by primary care specialty can be reviewed vis-a-vis utilization by all primary care services vis-a-vis all physician services. Standard reports are, in general, directed to monitoring volumes and costs. Aberrations in these monitoring reports often trigger special analysis to determine influencing factors and causes. For example, deviations from budgeted revenues per member per month would result in an analysis of variances in average contract size, average family size, and subscriber contract mix.

Future concerns

As HMO's increasingly penetrate the health care market, employers are demonstrating a growing interest in understanding the pricing, risk sharing, and benefit configuration of HMO's. Self-insured employers are particularly vocal with regard to these issues because premium expenses and any savings as a result of efficient utilization methodologies are outside their control and do not benefit them.

This national employer trend is certainly being felt at M.D. IPA, and we anticipate that the trend will continue to grow. To date, however, employers have been imprecise in articulating specific information requests. Confidentiality issues certainly exist. More fundamentally, however, employers are wrestling with questions concerning what constitutes a reasonable basis on which to deal with HMO's and what constitute the measures by which they can evaluate the relationship. Increasingly, M.D. IPA finds itself working with employers and reaching a consensus as to which data are appropriate to the employer needs as well as available through our organization's MIS.

Physician oversupply, excess hospital capacity, prospective reimbursement practices, and the persistent rate of medical care cost increase are driving health care providers, insurers, and purchasers to develop relationships that blur traditional distinctions. Many aspects of emerging health care arrangements mimic existing health maintenance organizations (HMO's). Information requirements and data systems that meet HMO administrator needs are applicable to these emerging delivery systems. At the same time, HMO information requirements are becoming more complex. What are these requirements and specific data needs and what are the implications for HMO data systems? These questions are answered in the context of a brief overview of differences between HMO's and traditional fee-for-service systems and the effects on HMO's of marketplace and health care policy changes.

HMO's in transition

Health maintenance organizations straddle two businesses; they both provide and insure health care. They are at risk for providing generally comprehensive care within the constraints of fixed prepaid revenues. This dual business perspective and a fixed income per enrollee motivate significantly different behavior and information needs when compared with the traditional fee-for-service health care system.

A few large HMO's have been established for many years: Group Health Cooperative of Puget Sound, a 350,000-member Northwest HMO, celebrates its 40th anniversary in 1987. However, the dominant model within the U.S. health care system has been charge-reimbursed fee-for-service medicine. In the early 1970's, Dr. Paul Ellwood, frequently referred to as the Father of the HMO Act, concluded that charge-based reimbursement practices rewarded providers for intensively treating illness in hospital environments and discouraged the practice of preventive medicine; prepaid HMO's, with a ceiling on revenues, encouraged aggressive preventive care, emphasizing early detection and ambulatory treatment in order to reduce expensive hospitalizations. His observations have been substantially validated by the Rand Experiment in Economical Care (Enthoven, 1985). In this longitudinal study, it was found that Group Health Cooperative provided overall care at 28 percent less cost and had hospital admission rates and days fully 40 percent lower than those for comprehensively insured patients in the fee-for-service medical system. Most importantly, these savings were achieved with equivalent health outcomes in a population not biased toward atypically healthy individuals.

For HMO's, the days of this “natural” cost advantage over the fee-for-service sector are ended. Tax Equity and Fiscal Responsibility Act legislation implemented prospective reimbursement practices for Medicare beneficiaries. Prospective reimbursement has been quickly extended to non-Medicare populations by insurers. In the fee-for-service sector, these changes, coupled with employer initiatives to contain costs, have sharply reduced numbers of admissions and increased the number of outpatient visits. To remain competitive with leaner fee-for-service cost structures, HMO's must increase the sophistication of their information systems.

HMO trends

Health maintenance organizations directly provide care or contract for medical services from primary care and specialty or hospital providers in a variety of arrangements, usually described by model type (staff, group practice, individual practice, or primary care network). Different information needs have traditionally been ascribed to each delivery arrangement. Differences have been overemphasized; data needs are equivalent across models.

The information needed to rate plans is increasingly complex. The 1973 Health Maintenance Organization Act dictated that HMO premiums be based on the average cost to provide care for all enrollees (community rating). Only recently have amendments to the act permitted alternative rating methods. Employers now press for rates based on experience or class risk rather than average cost.

Businesses, seeking to control employee benefit costs, increasingly demand leaner mix-and-match benefit packages. Indemnity insurers are eager to fill this need. HMO administrators must be able to respond flexibly.

In contrast to past conditions, each of these trends increases the need for detailed, rather than aggregate, enrollment, cost, and utilization data across five arenas.

Information requirements

Health maintenance organization administrative data needs may be broadly classified as:

Planning.

Assessing risk and designing and pricing products.

Securing the delivery system.

Managing plan performance.

Assuring quality of care and customer satisfaction.

In addition, HMO administrators provide information for several “publics,” each viewing these areas from its own perspective. The publics include the directors and managers of the HMO; customers, both members and employers; providers; legislative, financing and regulatory bodies; and investors (in for-profit settings). What are the data needs for each area?

Planning

The focus for planning activities within HMO's is shifting away from the traditional emphasis on facility or departmental planning toward program-based planning. Program-based plans emphasize the continuity and outcome of health care across all providers and services. Group Health Cooperative (GHC) has developed a number of program plans addressing such dimensions as eye care, musculoskeletal care, pediatric care, women's health care, and cancer care. For example, GHC's cancer care program plan coordinates risk assessment, screening, and prevention activities with ongoing medical care. Through well-defined protocols, it links primary care physicians and practitioners in specialty, hospital, skilled nursing, home therapy, custodial care, hospice care, and psychosocial service arenas. Care delivery plans are then tied to capital and staff planning and to research activities.

Information systems that facilitate this type of planning require not only a high degree of integration of internal enrollment, clinical, claim, utilization, and cost data, but also the ability to include external data. Examples of external data include trends in marketplace demographics and health risks, characteristics of health care providers operating in the marketplace, market potential, competitor data, expressed employer needs, and public policy directions.

Risk, product design, and pricing

With relaxation of the community rating requirement, accelerating demands for alternative benefit packages, purchaser requests for experience rating, and the production of new competitive products stemming from innovative alliances between commercial indemnity insurance companies and providers, HMO administrators are faced with product definition and pricing problems analogous to those of the commercial insurance industry.

Commercial insurers have broad experience in underwriting procedures that protect against adverse selection. They have developed flexible funding arrangements for large clients. The arrangements utilize, individually or in combination, varied techniques such as experience-rated contracts, retrospective rating plans, “cost-plus” plans, “reserveless” group insurance, minimum premium contracts, and stop-loss risk sharing. HMO administrators must address rating requirements for benefits using commercial insurance underwriting techniques (Witter, Bluhm, and Wang, 1986).

HMO data systems must effectively link exposure (enrollment) data with claim and cost data for predefined underwriting categories. Often this is difficult, particularly for staff and group model HMO's, which do not track the detailed service data.

Exposure data include such characteristics as rating category, number of dependents, and subscriber months exposed and corresponding data on income, group, industrial classification, and geographic area matched with exposure period. Claim data include such characteristics as patient age and sex; subscriber-patient relationship; provider identification; place and date of service; discrete services provided, with associated costs; and diagnosis and adjudication information (plan, group, coverage, copayments, capitation- and claim-paid amounts). Traditional community rating does not require detailed service cost data. This presents a particularly acute problem for staff and group models that formerly had little need for discrete service information.

Appropriate statistical techniques must be applied to exposure and claim data to price health care benefits. This is particularly true for smaller groups for which direct cost experience as measured by last year's claims is not an actuarially valid basis for setting rates. Rates must be based on the risk associated with the exposure. For this reason, it is mandatory that employers provide relevant data. In order to support requests for tailored benefit packages and class- or experience-based rates, the data provided must minimally include age and sex distributions of employees and dependents.

Securing the delivery system

HMO administrators contract with physicians and facilities to provide care in a variety of ways, ranging from directly employing staff and owning and operating facilities to individual practice and preferred provider agreements. In each agreement, HMO administrators and medical directors must focus on cost control while maintaining standards of care.

To achieve these objectives, a variety of delivery and contractual arrangements have been developed. Primary care physicians typically perform a “gatekeeping” function. The intent of gatekeeping is to deliver as much care as possible in the primary care setting, reserving more costly specialty and hospital resources for only those patients requiring such care. Gatekeeping with salaried staff has traditionally been accomplished through internal protocols and peer review. Group, network, and individual practice arrangements usually rely on capitation agreements by which some portion of the capitation is set aside in a risk pool. This risk pool is distributed only if specialty referral and hospital costs are within expectations. Where volumes justify, capitation arrangements may also be negotiated with specialists, hospitals, and long-term care facilities.

HMO data systems must therefore produce information to support distribution of incoming premiums to capitated provider categories and provide for corresponding risk pools. At GHC, this is accomplished by linking each enrollee in the membership system to the enrollee's primary care physician. Simultaneously, the HMO data system must be able to track the gatekeeping physician's management of primary, specialty, referral, and hospital utilization. GHC employs internal and external specialty referral protocols and has developed corresponding computer systems for referral management. Case management systems are used to coordinate care under program plans.

Utilization data simply tabulated by physician are inadequate for administering risk pools when the demographic characteristics of the physician's “panel” (assigned patients) varies. GHC defines age and sex cohorts with associated utilization factors to standardize primary care panels. Primary care panels are inadequate predictors of specialty care utilization, however. Case-mix characteristics are required if specialty care is capitated. Therefore, data on diagnosis, diagnoses-related group (DRG), and severity of illness should be linked to specialty utilization patterns.

Preferred- or designated-provider agreements are commonly used to secure hospitalization and other facility-based care. Claims and utilization data systems must provide information to support multiple reimbursement schemes, including such combinations as reasonable and customary, fixed discount, volume-based discount, and geographic-area-specific schedules.

Managing plan performance

HMO administrators and medical directors must manage plan performance in several domains. These domains include marketing and membership objectives, financial performance, health service delivery, product line performance, practitioner and provider performance, and quality of care.

Operational performance data

The most complete distillation of critical performance measures and required data for HMO administrators is described in a monograph for HMO managers (Birch & Davis Associates, 1983). In this study, conducted for the Office of Health Maintenance Organizations, management reports solicited from 45 HMO's were summarized. Marketing–membership, financial management, and health services delivery domains are addressed. Representative examples of summary data include:

Summaries and analyses of membership, contracts, and accounts.

Analyses of revenues, expenses, and profits per member per month.

Analyses of hospital, physician, and other outpatient medical costs.

Key financial and operational indicators.

Analyses of hospital, physician, and other outpatient service utilization.

Most existing HMO data systems are insufficiently integrated to easily produce these indicators.

Product line performance data

As multiple benefit structures evolve, product lines emerge. Few HMO data systems have the capability to monitor the performance of specific product lines. Data to support such analyses require development of benefit categories and levels typical of commercial insurance products. Examples include describing benefits (products) in terms such as basic health plan, dental coverage, vision coverage, prescription drug coverage, and long-term care insurance. Within product categories, levels of coverage must be specified, such as a split between high and low options of the basic health plan or between hospital and medical/surgical benefits. Adjudication systems for processing claims must attach costs and assign services to specific utilization categories corresponding to benefit structures. Cost data by category must then be matched to corresponding allocations of incoming premiums. Among the more difficult problems in this area is the typical lack of mature cost-accounting systems in staff or group model HMO's. In the community-rating era, sophisticated cost accounting was an unnecessary “overhead” expense. GHC is installing enrollment and claims systems derived from the insurance industry that are capable of tracking our various product lines and benefit plans. Although predominantly a staff-model HMO, GHC tracks both claims and internal encounters and is employing increasingly sophisticated cost-accounting techniques.

Practitioner and provider performance data

Practitioner and provider performance management hinges on the ability of the HMO administrator and the medical director to influence specific practitioner behavior while maintaining standards for quality of care. With the strong tradition of physicians autonomously determining practice styles and being reimbursed on the basis of production of services, this represents a unique challenge, particularly in nonstaff models. It is the critical variable in achieving cost control.

Physicians must be given constructive, direct, and useful feedback on their practice styles. Among the more effective approaches is production of peer-comparison reports for selected utilization measures. GHC uses a number of peer-comparison reports to provide physicians with objective data by which they can evaluate their own practice styles.

In the primary care setting, key variables include referral rates to specialists, hospitalization rates and lengths of stay by DRG, use of substitute facilities for inpatient care such as surgery centers, and use of major ancillary department services (particularly prescribing patterns for pharmaceuticals). To be meaningful, utilization and cost data must be standardized for the demographic characteristics of patient panels.

In specialty and hospital settings, key variables include unusually costly ambulatory procedures, secondary referral rates to other specialists, hospitalization rates and lengths of stay by DRG, use of substitute facilities for inpatient care such as surgery centers, and use of major ancillary department services. To be meaningful, utilization and cost data must be standardized for the case-mix characteristics of specialty workloads.

To be useful both primary and specialty physicians, data must be succinctly presented. Therapeutically equivalent but less costly alternatives must be identified. This is particularly true for drugs in nonstaff models where a formulary is not in place and where wide variations in the costs of equivalent drugs prevail.

Assuring quality and customer satisfaction

The bottom line to health care is its efficacy in maintaining or improving health status and satisfying the customer at acceptable cost. Efficacy of care is measured in terms of outcomes that span long periods of time. HMO administrators must take a longer view than the next quarter's profit performance in this area. This is part of the impetus behind program planning, as described previously. It represents a fundamental shift from the past focus of managing care on the basis of discrete events such as the cost of a visit or a hospitalization. Little research has been done in this area to define specific data needs. However, general statements can be made.

Measures of health risk and health status are required. Both measures are critically dependent on careful integration of clinical data systems with administrative systems. Risk factors include lifestyle habits such as patterns of smoking and alcohol use as well as clinical parameters such as cholesterol levels. Health status data include measures such as bed-disability days as well as clinical parameters such as blood pressure and fitness levels. HMO data systems do not deal with these types of data today. Computerized medical records are mandatory before active use of health risk and health status measures is realistic. GHC presently employs several systems to assess risk and health status and to support preventive care protocols. For example, a comprehensive breast cancer screening system is used to assess risk and initiate periodic appointments to primary care and mamography services based on GHC's cancer care plan protocol.

Consumer satisfaction is critical; the member makes the decision to continue care with an HMO. Long-term measures of health status will be meaningless unless the HMO administrator can maintain a reasonably stable enrollment base. Key data required to assess satisfaction include measures of access such as appointment waiting times, complaint patterns, attrition reasons, and malpractice suits. Most often, data in this area are obtained via survey techniques and may be limited. GHC routinely monitors access parameters such as appointment and telephone wait times. In addition, we systematically conduct controlled consumer opinion surveys to assess satisfaction.

Purchasers of health care need to know how effectively HMO benefit plans offered to employees are performing. Financial performance as well as quality of care and employee satisfaction are key indicators. With other factors equal, the employer relies on employee feedback to decide which plans to offer. Employer requests for actual utilization and cost data require careful statistical interpretation. Utilization seldom correlates well with risk for a given period unless based on very large populations. Consequently, purchasers and HMO managers are struggling to identify useful comparators. The Group Health Association of America is facilitating this effort between employers and HMO managers.

Comparative information

The needs described previously relate primarily to the internal management of the HMO and the situation-specific needs of customers and providers. Other publics (legislative, financing, and regulatory bodies and investors) need access to overall comparative data. This need is analogous to the need for hospital industry data that is satisfied largely through data bases and summary reports of the American Hospital Association and Commission on Professional Hospital Activities.

Comparative information needs for the HMO industry are being addressed through the development of Group Health Association of America's National Comparative Database (Brudevold, English, and Reeves, 1986). This data base will contain three major data categories encompassing enrollment, utilization, and financial measures.

Enrollment data will include such measures as source of membership, contract type, and age and sex distributions. These data will permit analysis of the effects of enrollment on operations.

Utilization data will include such measures as inpatient services, discharges, and patient days; nonadmission services that substitute for inpatient care such as surgery centers or inhome care; and ambulatory encounter information. These data will permit comparisons with industry and area norms.

Financial data will include such measures as capital structure and profitability data to provide insight into access to funding, rates of growth, and financial viability. Detailed distribution of expense data will provide insight into corporate strategies as they relate to the U.S. health care market.

Implications for HMO systems

HMO administrator data needs dictate a sophisticated approach to the design of information systems. As the U.S. health care system evolves and traditional distinctions blur, the integration of administrative and clinical data becomes critical.

Current HMO information systems reflect simpler times. Relatively little integration exists among various applications, making it difficult or impossible to respond to both internal and external data needs. Integration, as it exists, is largely based on the need for two operational systems to communicate for routine business transactions. A simple example is the need for the claims processing application to reference eligibility and benefit data maintained by means of the membership application. Clinical systems are poorly integrated. Nonstaff HMO model types normally do not control or link to the clinical systems used by the HMO's various providers. Clinical data are kept, as in the traditional fee-for-service sector, by individual physicians and are not accessible to the HMO except by request or chart audit. The HMO relies on the “bill” submitted by the physician for needed utilization data. This is satisfactory for administrative purposes, but it is only marginally useful for clinical purposes.

HMO administrators need to require vendors and their own management information system departments to address the critical need for integrated systems. The ability to compete effectively will hinge on the degree to which information technology supports operational and strategic planning functions. This requires a data-driven systems design that tightly links marketing, membership, clinical information, claims processing, and financial functions. At GHC, a comprehensive strategic information plan is the basis of a major redesign of information systems.

Implications for health policy

Managed health care systems have the potential for improving the cost effectiveness of health care. A recent survey for the Henry J. Kaiser Family Foundation by Louis Harris and Associates found that 85 percent of the business executives polled agreed that HMO's are effective in containing the cost of health care, up from 59 percent in a 1980 survey (Kosterlitz, 1985).

Recent public policy has concentrated on cost-containment strategies such as those represented in Tax Equity and Fiscal Responsibility Act legislation. For example, the diagnosis-related group prospective reimbursement mechanism focuses on discrete events of hospital care. The aggregate effect of these strategies has slowed hospital cost growth rates. In a limited sense, implementation of DRG's may represent only a shifting of costs to the less visible ambulatory arena. Proposals to apply prospective reimbursement schemes to ambulatory care using ambulatory visit groups (AVG's) are an extension of the current cost-containment strategy. AVG's may result in subsequent cost shifts to other arenas.

Rapid membership turnover is costly, so Group Health Cooperative encourages long-term enrollment. This results in our bias toward practices that contain present costs without shifting them to later years. GHC's emerging experience suggests that lifetime health monitoring has much potential to highlight the benefit of preventive health care to the member and the employer; to discourage overemphasis by providers on event-based productivity measures; and to encourage program planning and operational decisions that effect coordinated care over time and across providers.

If Group Health Cooperative's emerging experience is indicative of the experience of other health care providers, investigating alternative data, financing, and reimbursement mechanisms that focus attention on long-term health status measures, rather than discrete events of care, may be in order.

Information is necessary for efficient and equitable reimbursement of health maintenance organizations (HMO's) participating in Medicare. The answers to three fundamental questions are needed for each HMO: Who is receiving care? Is the payment adequate? What is being paid for? However, like any economic good, information is costly and resources used for its collection and analysis have competing uses. The aim of this article is to develop some guides to appropriate data collection for capitation payment of HMO's for Medicare beneficiaries.

From the perspective of the Health Care Financing Administration (HCFA), one significant consequence of enactment of the Tax Equity and Fiscal Responsibility Act of 1982 (TEFRA) is that the detailed claims expense and utilization records maintained by HCFA for fee-for-service (FFS) beneficiaries and providers are no longer available for Medicare capitation enrollees. Very little information, beyond eligibility data, is being collected by HCFA on capitation beneficiaries. Hence, HCFA program management staff have little knowledge about who or what is being treated by participating HMO's.

On the Other hand, from the HMO's perspective, it is significant that TEFRA frees HMO's from the need to collect costly and detailed utilization and expense data that are used solely to meet Medicare data requirements. Imposition of data specifications derived solely from FFS experience may counteract many of the gains of the TEFRA reforms. An optimal HMO data collection system would meet HCFA's need to know who and what is being paid for and what the appropriate payment level is by utilizing, insofar as possible, data that are internally useful to HMO's. To require HMO's to collect data that are not pertinent to their operations, when HCFA could equally well use data that are pertinent, is suboptimal. Inefficiencies are introduced into the administration of HMO's, thereby increasing the total cost to society of providing health care to Medicare beneficiaries. As Donabedian has suggested, HMO's carry a unique ability, and, perhaps, a social imperative, to develop nontraditional approaches to providing high quality health care at reasonable cost. He states: “Not the least among the glittering prospects that the words 'health maintenance organization' evoke is the opportunity to reshape our thinking about what constitutes quality in everyday medical practice and about how that quality might be protected, nurtured, and made to grow” (Donabedian, 1983). Medicare reimbursement should support and encourage this imperative.

Thus, we argue that it is necessary to understand the internal data requirements of HMO's, especially the ways in which they differ from the requirements of the FFS sector, before devising data reporting requirements to meet HCFA's needs. Any data requirements promulgated by HCFA ought to help HMO's perform their basic mission of curing, maintaining, and caring for Medicare beneficiaries in an effective, efficient, and equitable manner. In this presentation, we address each issue in turn.

HMO data requirements

The essential distinction of an HMO is integration of ambulatory care, inpatient care, and risk pooling (i.e., insurance) within a single organization. For a fixed capitation payment, an HMO accepts the risk for all Part A and Part B services required by a Medicare beneficiary. Thus, capitation establishes a defined service population and specifies a total budget constraint. These two features present unique demands and opportunities for HMO administrators and clinicians.

The essential analytic impact of this form of medical care organization is that it expands the provider's scope of control to include treatment of the entire illness episode as well as responsibility for preventive services for avoidable illness episodes. This feature distinguishes an HMO from FFS multispecialty group practice, although many HMO's may in fact operate in nearly identical fashion to their FFS counterparts.

In theory, then, HMO's are able to take advantage of all possibilities for saving health care resources by substituting preventive care, ambulatory care, home care, and other lower cost modes of care for more expensive forms of care. This should result in lower overall health care costs than a series of suboptimal solutions determined by multiple providers operating independently, but it imposes information requirements on HMO managers that differ from those on FFS managers; that is, HMO managers need to control the complete spectrum of medical technology and require information systems that support this control. We contend that the procedure-specific focus of FFS providers on price-cost margins means that their information requirements are inappropriate from an HMO perspective.

HMO's have three unique data requirements stemming from the distinctive nature of their output. First, measures of illness episodes are necessary for efficient allocation of acute and chronic medical care resources; second, health risk registries are necessary for efficient provision of preventive services; and, finally, measures of membership satisfaction and health status provide the requisite data for an HMO quality assurance system.

First, supplementing the traditional department-based cost and output accounting systems, HMO's need to have systems that enable development of resource use and outcome profiles for illness and health problem episodes (Hornbrook, Hurtado, and Johnson, 1985). The episode measure represents the essential intermediate output of an HMO. It is required for development of appropriate linkage of resources to health problems in a population with multiple comorbidities and multiple modes of care.

The fundamental concept in any resource allocation scheme is “shadow price,” or opportunity cost. Every use of a resource imposes the costs of the forgone value from use of that resource in another application. Even though the HMO does not charge a price for each service or treatment, except for small fixed copayments, every expenditure of resources has an opportunity cost, the benefit that the HMO could have obtained from the next best use of those resources. For example, the shadow price of a coronary bypass procedure is the benefit that could have been obtained by using those resources to treat some other illness episode or to operate a smoking cessation program.

An HMO management information system should assist managers and clinicians to identify these constraints and options. In order to do this appropriately, an episode approach is required. This approach enables tracking of physicians' propensity to hospitalize and to prescribe revisits while controlling for the underlying nature and severity of illnesses, thus enabling definition of opportunity costs in terms appropriate to an HMO. Measurement of hospital case mix via diagnosis-related groups (DRG's) or ambulatory case mix via ambulatory visit groups (AVG's) is quite appropriate for FFS providers. HMO's, in contrast, require an episode case-mix measure. DRG's and AVG's do not provide any information on productivity across illness episodes because each readmission and each followup visit is counted as a separate output. The HMO advantage is not in running hospitals or clinics more efficiently than FFS providers do but in the composition of treatments for illness episodes (Hornbrook and Berki, 1985).

To obtain the highest quality of data possible on episodes, a sampling approach is desired. This could take the form of a comprehensive episode-tracking system for a 5-percent sample of HMO members. Thus, HCFA data requirements should not stipulate that data be furnished on every illness episode because this is not necessary for HMO operations.

Second, a system of health risk registries is required to govern allocation of preventive services in HMO's. Distinct from FFS providers, for which the criterion for delivery of preventive services is whether consumers are willing to pay for them, HMO's are responsible for developing optimal health screening and detection strategies for their memberships. Only a small subset of preventive services ought to be prescribed to an entire population. Most preventive services are more cost effective when administered to persons who meet specified risk profiles (Russell, 1986).

A health risk registry system, including disease registries, provides a systematic approach for detecting those persons among an HMO's members who are eligible for primary and secondary prevention programs, rather than relying on accidental discovery and prescription during physician office visits. Consumer ignorance may result in underconsumption or overconsumption of preventive services, and the HMO offers a means of attaining higher efficiency in this area. Some risk factors that could be included are smoking habits, athletic activities, seat-belt use, oral health habits, breast disease risk, blood pressure, and blood lipid levels.

Finally, a quality assurance system is necessary for an HMO to monitor its overall success or failure. Donabedian (1983) argues that the core of a quality assurance system is client satisfaction. Members' preferences are the ultimate measure of the contribution of an HMO to welfare and well-being. Satisfaction to an HMO member means two things. First and foremost, it means satisfaction with the outcome of care, which is reflective of the technical effectiveness of the medical care system. Did the patient recover completely from the limitations in social and physiological functioning imposed by an illness? Was the recovery timely? Satisfaction also means comfort, convenience, access, and personal attention—the psychosocial aspects of care. Was care available when and where requested? Was it provided in pleasant surroundings by professionals who demonstrated concern for individuals' needs and desires? Because HMO's control the entire spectrum of medical technology, as well as the organization and delivery of services, they have a correspondently greater responsibility for member health status and satisfaction than do FFS providers. Description of an adequate quality assurance system is beyond the scope of this article and has been covered elsewhere. (For example, see Freeborn and Greenlick, 1973.) Some critical elements of an HMO quality assurance system are highlighted in Figure 1.

Figure 1

Elements of a health maintenance organization quality assurance system

Given these data needs unique to HMO's, the question of how they relate to HCFA data requirements is addressed next.

Medicare data requirements

The minimum data requirements imposed on HMO's for participating in the Medicare program should relate to administering the reimbursement program and meeting the goals set by social policy. Three critical data needs for HCFA can be identified:

Administration of the payment system—ongoing determination of the risk levels for HMO memberships.

Formulation of reimbursement policy—setting the basic value of the core Medicare benefit package for the “average” beneficiary.

Program oversight—evaluating the accessibility, cost, and quality of care provided.

Each of these needs is discussed.

We begin with the premise that it is appropriate to require that specific data be supplied to HCFA to meet specific well-defined program needs. This is an alternative to requiring the establishment of a general-purpose data base that would be used to address a series of questions to be defined at some future date. It can be argued that there is a potential social cost of wrong or superfluous data availability in terms of inappropriate analyses that may lead to misguided policies (Greenlick, 1975).

Administration of payment system

Under the current TEFRA provisions, Medicare beneficiaries have a choice of FFS Medicare or HMO options where the latter are available. TEFRA also requires that the HMO capitation rate be based on Medicare payments on behalf of beneficiaries in the same geographic area in the FFS sector. These two requirements create the possibility of overpayments or underpayments to HMO's based on favorable or adverse selection in HMO enrollment. To reduce these problems, capitation payments are adjusted to account for differences among HMO's and between HMO and FFS sectors in the health risk of beneficiary subgroups.

In the current methodology, known as the adjusted average per capita cost (AAPCC), age, sex, welfare status, and institutional status are used as risk-adjustment factors. The risk-adjustment process requires symmetrical data from FFS and HMO sectors, because the basic issue is whether the underlying nature of the health problems treated in any HMO differs from the nature of problems seen by the local FFS providers. Capitation payment is a prospective mechanism, so risk-adjustment factors must predict utilization rather than being derived from observed current use patterns.

Although the specific risk factors used may be altered over time, any useful risk factor will relate to individual attributes that are predictive of future disease and propensity to use health care services. Such risk factors are inherent to HMO internal operations by their inclusion in health risk registries and budgeting criteria.

The general requirement for administering a risk-adjustment mechanism is that HMO's furnish two types of data: data for individual beneficiaries on the prediction factors that are included in the risk-adjustment model and data on the total annual cost of health care for individual beneficiaries. The former are part of HCFA's current data requirements, and the latter are not. Moreover, annual cost data must be adjusted for selection bias in case the risk-adjustment model does not account for all of the difference in average annual per capita Medicare expense between the HMO and the local FFS sector populations (Hornbrook, Greenlick, and Bennett, 1986).

Expense data do not need to be furnished on the total population of Medicare beneficiaries. The average capitation payment to the HMO must be correct, not the amount paid for any specific individual with a specific set of risk factors. Data on annual expenses can be achieved via scientifically designed sampling methodology applied to both HMO and FFS sectors. Indeed, a sampling approach is preferred because it enables use of higher quality data on the risk factors. Cost data need only be valid measures of real resource use, not measures of accounting cost. Hence, proxy measures of comprehensive utilization patterns may be employed if they are sufficiently correlated with the true measure.

In sum, given the policy objectives of TEFRA and the methods specified in the TEFRA regulations, an HMO ought to be required to furnish data on risk factors and annual per capita cost for a random sample of its Medicare beneficiaries in order to participate in Medicare. Development of a valid health risk classification system for a population will help both HCFA and HMO's do their respective jobs better.

Formulation of reimbursement policy

Under the current TEFRA provisions, the basic level of capitation payment for Part A and Part B benefits is set by the adjusted average per capita cost of claims in the FFS sector, less 5 percent. This cost differential is not based on any scientific determination of the relative efficiency of HMO's in treating Medicare beneficiaries. Available evidence suggests that a figure of 12 to 40 percent savings for the nonelderly population is possible (Hornbrook and Berki, 1985), and the possible savings for the elderly population are still unknown. The basic problem of establishing the absolute level of Medicare capitation payment has not been satisfactorily solved. The true efficiency gains of HMO's should be reflected in the program costs of Medicare.

The adjusted community rate (ACR) methodology enables use of market forces to determine the basic capitation payment. The ACR is based on the HMO's community rate, which is the cost per member per month to provide health care services to enrolled members and is the basis of the rates that are charged to various groups in the community. The theory behind use of the ACR methodology is that competition among health plans for subscribers will obtain efficiency in providing a standardized package of benefits. This ensures that market prices (i.e., the community rates and the ACR's) are appropriate indicators of the economic cost of providing care to defined populations. The ACR is derived by adjusting the community rate for the age-sex composition of an HMO's Medicare enrollment.

HMO's are currently required to furnish their ACR's to HCFA because, under TEFRA, if the ACR is lower than 95 percent of the AAPCC, the HMO must pass the savings to the beneficiary in the form of additional benefits or return them to HCFA. Use of the ACR instead of the AAPCC to set capitation payments for Medicare enrollees would help to address the issue of selection bias because payment would be based on the members actually enrolled in the HMO. Clearly, the ACR is vital to internal HMO management, so no inefficiency is imposed by its use by HCFA.

Under current regulations, HMO's are required to enroll a majority of non-Medicare members. One problem with this requirement is that it prevents HMO's from specializing in the Medicare population. Such specialization could have significant economic and social benefits. Moreover, the requirement makes it necessary for new HMO's to establish a given level of market penetration before beginning to enroll Medicare beneficiaries. Finally, it increases the minimum market size required for an HMO to operate in any given locale, thus reducing access to HMO's for beneficiaries residing in sparsely populated areas. If this restriction were lifted, enabling an HMO to enroll up to 100 percent Medicare beneficiaries, the ACR approach would no longer be appropriate because the “community” might be composed entirely of Medicare beneficiaries. We argue that there are social benefits to be gained from abolishing the enrollment constraint, albeit this would raise some additional administrative issues.

One possibility is to arbitrarily establish a fixed amount for the Medicare voucher and let the beneficiary pick up the remainder of the premium cost. However, this would involve regional inequities because of differences in the cost of medical care inputs. Beneficiaries living in high-cost areas would have to pay more out of pocket for health care or receive fewer benefits than would those living in low-cost areas. To deal with this issue, the voucher could be adjusted for differences in the cost of the market basket of HMO inputs. Such adjustment would require HMO's to furnish information on the prices, quantities, and types of labor and capital inputs. However, data ought to be collected on the basis of a stratified random sample so that not every HMO would face this burden, and HCFA ought to pay for the data collection costs so as not to create inequities across HMO's. It is important to emphasize that the purpose here is not to regulate the purchase of inputs by HMO's but to adjust for cost differences beyond their control so that reimbursement inequities among regions of the country are prevented.

Finally, the problem remains of establishing the value of the basic package of benefits for sparsely populated areas that cannot support the competitive HMO market structure needed to create efficient ACR's. In such areas, it is difficult to separate higher costs attributable to smaller scale of operations from higher costs attributable to inefficiency. This creates a need to provide guidance to both HMO's and HCFA regarding the minimum level of the ACR required to provide acceptable quality of health care in rural settings. At the least, an urban-rural adjustment derived from appropriately specified HMO cost functions could be applied to the value of a Medicare voucher. Much more research is required on this problem. Such research would require comprehensive cost and output data on a nationally representative sample of HMO's. The cost of this effort should be borne by the HCFA research budget.

Program oversight

The TEFRA reforms have significant implications for information needs for Medicare program oversight. Under retrospective cost reimbursement, incentives are placed on providers to increase the allowable cost base by shifting to a more resource-intensive style of practice. Hence, a primary policy priority under this system was utilization review to control the expansionist forces inherent in the system. Under capitation payment, providers face incentives to minimize costs, perhaps to the point of undermining the quality of care, by controlling the overall volume of treatments and reducing the resource intensity of treatments. Hence, program monitoring must focus on quality assurance to guard against undermining the health status of Medicare beneficiaries enrolled in HMO's.

A simple and functional approach to this problem would be to require HMO's to furnish evidence that an acceptable quality assurance system is operational within the HMO. This does not mean that every HMO will have the same system, nor does it mean that every HMO will be providing identical quality of care. Rather, the essential requirement would be to show that each HMO is using a system of quality control that is consistent with the principles and methods of the current state of the art in quality assurance. We suggest the components of an HMO quality assurance system in Figure 1.

To make this approach work effectively, it would be necessary to require that a sample of HMO's participate in a systematic quality assurance monitoring system financed by HCFA. Such a system would assess health status, access, and satisfaction for the eligible population; would evaluate the technical quality of prevention strategies and care of illness episodes; and would employ sampling and statistical methods to the greatest extent feasible. The ultimate evaluation criterion is the overall effect on health status of HMO membership for Medicare beneficiaries. This system should complement the quality assurance systems of individual HMO's, again providing an illustration of how HCFA data needs can stimulate improved HMO performance.

To summarize, the data requirements outlined earlier are tailored to the specific needs of HMO's operating under capitation payment for publicly financed beneficiaries. They are summarized in Table 1. As can be seen, these requirements are quite divergent from those required under the FFS system.

Table 1

Proposed data requirements for health maintenance organizations (HMO's) participating in Medicare

Risk adjustment
Data collection strategy:	Sample
Unit of observation:	Individual beneficiary
Variables:	Risk-adjustment factors
	Total annual health care cost
Valuation of the benefit package
Option 1—adjusted community rate
Data collection strategy:	Universe of HMO's
Unit of observation:	HMO
Variables:	Adjusted community rate for standardized benefit package
Option 2—market basket
Data collection strategy:	Sample
Unit of observation:	HMO
Variables:	Prices and quantities of purchased inputs
Program evaluation
Data collection strategy:	Universe of HMO's
Unit of observation:	HMO
Variables:	Evidence of an acceptable quality assurance system
Data collection strategy:	Sample of beneficiaries in a sample of HMO's
Unit of observation:	Individual beneficiary
Variables:	Health status, satisfaction with care, access, and other pertinent system attributes

Conclusions

New advances in information technology are being continually incorporated into the delivery of health care and the practice of medicine. Many observers anticipate the complete automation of patient medical records and use of large-scale disease-specific data bases to guide clinical practice. These innovations are being adopted by HMO's, and every HMO will probably have some form of automation in its clinical information systems within a decade.

From an HMO perspective, investment in such advances must be supported by increases in the quality and cost-effectiveness of health care from a population-based perspective. This allocation criterion will help in controlling total health care costs. Nevertheless, even if a fully automated HMO could provide utilization and cost data to HCFA at very low marginal cost, the full opportunity cost of maintaining and using such a general purpose data base should be counted. This opportunity cost includes threats to individual confidentiality and organizational competitive positions as well as erroneous and misguided analyses.

However, Medicare data requirements focusing on population health risk factors, annual per capita cost of care, and quality assurance systems within appropriately specified sampling strategies can operate to support efficiency in both the health care delivery system and the payment system. By looking to the future and asking what will help HMO's perform better, HCFA data policies can actually stimulate adoption of appropriate information system technologies. A focus limited solely to HCFA needs is shortsighted and likely to be extremely costly to society in the long run.

As employers invest more and more money purchasing health care services for their employees, they are becoming increasingly interested in evaluating the effectiveness of various vendor programs (e.g., health maintenance organizations, preferred provider organizations, and fee-for-service programs) in meeting corporate health management goals. Many corporations share similar priorities: improving or maintaining the health status of employees and their dependents, reducing risks in home and work environments, and managing health care costs. Likewise, they are requiring vendors to offer evidence that vendor programs are maximizing improvements in employee health and using cost-effective treatment technologies.

Honeywell Corporation, like other private purchasers of health care services, is developing programs and strategies to evaluate vendor services. Honeywell, which comprises diverse business activities in many locations throughout the world, has about 65,000 U.S. employees. The corporate benefits office pays claims through six insurance carriers and also pays premiums to 125 health maintenance organizations (HMO's) and preferred provider organizations (PPO's). The six carriers, which handle fee-for-service claims, administratively cover approximately 75 percent of Honeywell's U.S.-based employees as well as their dependents. HMO's and PPO's enroll the remaining 16,000 employees. Enrollment in HMO's and PPO's is concentrated in the Twin City (Minneapolis-St. Paul) area, where Honeywell's corporate headquarters is located; 13,000 of the 16,000 Honeywell employees enrolled in HMO's and PPO's live in the Twin Cities.

Honeywell's effort to evaluate vendor services began in 1983, when senior management, alarmed by the corporation's rapidly rising health bill, chartered a task force to determine why medical expenditures were increasing so rapidly and to recommend corrective action. Honeywell, which has a self-insured but not self-administered health benefits plan, was experiencing double-digit increases in vendor payments. From 1978 to 1983, payments for health services rose 19 pecent each year, and forecasts indicated that this trend would continue.

The task force determined that the reasons for rising costs were complex and recommended to Honeywell's senior management that four positions be added to design and implement cost-containment strategies. The four positions, three of which eventually became centralized in the Corporate Health and Environmental Resources Department, were in the areas of benefits design, health promotion and wellness, provider relations, and health data management.

The new four-person team set out to establish processes for reducing costs, at the same time improving or maintaining employee and dependent health. First, however, they needed accurate and comprehensive information about how Honeywell employees were using health services and benefits programs and the associated costs of those services.

Honeywell contracted with Health Data Institute (HDI), which analyzed Honeywell's insurance claims data for 1980–82. Information that the Honeywell team sought from the analysis included employee and dependent demographic data, average annual claims expenditures by vendor and location, hospital claims data per vendor, and frequency of surgical procedures by vendor and location. In addition, the team wanted to compare their utilization data with those of other employers. HDI produced 17 reports for the correspondent Honeywell “locations” plus a corporate overview.

Although the HDI reports provided information for the divisions to use in planning benefit designs, promoting health activities, and developing communication with providers and vendor administrators, the HDI analysis showed that little uniformity existed in how the various vendors stored, collected, and reported information. In addition, vendors had interpreted requests for seemingly standard data items in different ways, and some vendors were not able to provide information. Reasons that the vendors gave for not providing data ranged from concerns for confidentiality to inability to obtain the information. Some HMO's, for example, neither “cost out” individual patient services nor identify by employer group patients who receive care. Consequently, these HMO's cannot provide employers with data on claims, billing, and individual patient costs. The insurance carriers had similar problems in providing the requested data items. In summary, then, the HDI analysis affirmed the need for developing an improved system for collecting, analyzing, and reporting Honeywell's health care data.

Health care data needs

As a result of the HDI analysis, the four-person team developed a proposal for an improved Honeywell health care information system. This new system would supply responses to local division management as well as corporate management, fulfilling both their unique and common information needs. The system would modify, but not overhaul, current employee data systems and existing relationships with health care vendors and insurance carriers.

The following is a list of proposed system requirements:

Both division management and corporate management should be able to use the system.

Any system measurements should be comparable over time and across population groups. Detailed backup information should be available to support any measurements.

The system should track corporate-wide trends in cost and vendor utilization.

System setup and maintenance costs should be reasonable.

The system must protect the confidentiality of individual health data.

The system must be compatible with existing and planned data systems, but it also must be flexible enough to allow tracking of other indexes if required.

The information system must include a method of identifying the company's insured population by location and division as well as by medical plan, age, and sex.

The information system needs key indicators for summarizing current performance, accompanied by statistics that will permit comparisons. In addition, there must be a method of retrieving detailed information to facilitate investigation and resolution of specific cost or utilization problems.

The information system must be sufficiently flexible and comprehensive to include fee-for-service indemnity plans, HMO's, and other alternative health care delivery systems.

The information system should be able to provide summary measurements that can be analyzed by medical plan groups within and across divisions.

These summary measurements include the following:

Inpatient days per 1,000 covered population per year (age and sex adjusted).

Inpatient admissions per 1,000 covered population per year (age and sex adjusted).

Average length of stay per inpatient admission by major service category: medicine, surgery, maternity, psychiatry, chemical dependency, rehabilitation.

Number of surgical procedures performed per 1,000 covered population per year (age and sex adjusted).

Number of outpatient encounters per person per year.

Mortality rates per 1,000 covered population (age and sex adjusted).

Mortality rates per 1,000 admissions (case-mix adjusted).

Readmission rates per year.

Proportion of Cesarean sections among live births.

Proportion of hysterectomies in women under age 35.

Frequency, by provider, of patients staying more than two times the average length of stay for a diagnostic group.

Total premiums paid.

Average premium paid per employee.

Total charges.

Average charge per claim.

Total claims paid.

Average claim paid.

Total hospital inpatient room and board charges.

Total hospital ancillary charges.

Total surgical claims paid.

Total outpatient claims paid.

Diagnosis- and procedure-specific information by the 25 most frequently occurring groups.

Provider-specific (facility and physician) information for any provider with 10 or more patients.

Demographic-specific information, i.e., age, sex, employee, dependent, spouse, retiree group.

Information on salary/hourly groups.

Detailed report components

In July 1984, the team issued an update on the current information system. None of the proposed changes had yet been implemented. The purpose of this report was to pinpoint problems and to suggest resolutions to these problems.

At the time of the report, patient eligibility for health care benefits was generally determined from the payroll system. Patient eligibility data were updated only when the employee changed medical plans. The eligibility payroll data base described only whether the employee had family or single coverage. Eligible dependents were not listed, nor were any demographic factors identified. Moreover, additional problems were found. Division identification and employee geographic location were not always accurate, and historical plan information was not recorded. For example, no records were kept when an employee switched plans or added dependent coverage.

The quality of utilization data varied by carrier. For example, five of the six insurance carriers did not record employee use of emergency room or intensive care unit services. Some carriers indicated only their own diagnostic or procedure codes; others did not indicate any diagnostic or procedure code. Most insurance carriers did not include detailed outpatient information, name of the hospital, physician charges, or any identifier for the attending physician. Two insurance carriers did not maintain records of when the employee received treatment. HMO utilization data were similarly nonexistent, and in the majority of cases it was impossible to extract meaningful comparisons either across HMO's or between HMO's and fee-for-service carriers.

Given the quality of available data, collecting uniform comparable utilization data from HMO's, PPO's, and insurance carriers was proving to be a challenge. Other employers in the Twin Cities were experiencing similar difficulties. As a result, Honeywell and other employer representatives active in the Minnesota Coalition on Health Care Costs formed a subcommittee that included HMO representatives. The purpose of the subcommittee was to develop an HMO data set for use throughout the HMO community. A summary of the data set follows this section. Honeywell and other participating employers will require that HMO's report the requested information on HMO participants as a condition of doing business. Similar initiatives focusing on obtaining the same data from insurance carriers are also planned.

On a national level, Group Health Association of America is working with member plans and with industry representatives like Honeywell to develop uniform reporting requirements. One model that Group Health Association of America is evaluating is the Minnesota Coalition's data set.

Summary of data set

The data set includes both company-specific and aggregate information on employees and their dependents in the following areas.

Membership:

Membership by age and sex.

Member months.

Number of single and family contracts.

Average family size.

Hospitalization:

Admissions, days, and average length of stay for medical, surgical, obstetrical, chemical dependency, and extended care diagnoses and for all diagnoses. Obstetrical admissions by type of procedure.

Hospital outpatient services:

Emergency room treatment—adult or pediatric.

Type of service.

Other outpatient services:

Adult or pediatric.

Type of service.

Clinical/medical center services:

Adult or pediatric.

Type of service.

Outpatient mental health and chemical dependency services:

Individual or group/family session.

Private-duty nursing visits.

Prescriptions.

Change in focus

As Honeywell's corporate team worked to collect utilization data and to contain health care costs, they realized that their ultimate goals could be realized in the long term only by managing health. Their emphasis evolved, therefore, from managing health care costs to managing health.

Corporate goals of improving health, reducing risks, and cost containment are now being worked from a different angle. Honeywell is focusing on identifying risks, personal and environmental, that will affect health negatively in the present and in the future. Honeywell feels that the ultimate way to manage costs is to prevent problems from occurring through early identification and evaluation of personal risks, workplace hazards, environmental and community health problems.

This effort, of course, requires an information system that goes beyond just collecting health care utilization data. The Honeywell information system will identify population risk factors and establish normative data. It will help divisions and subsidiaries to eliminate risks. It will also help location medical departments to put treatment plans together and to better manage cases. In this model, Honeywell shares with medical providers and employees the responsibility for managing problems associated with ill health and for returning the employee to healthy or appropriate status as soon as possible.

The following is a list of some examples of the data that the system will collect and how the data will be used:

Occupational health and safety

Data to be collected—Health outcomes from claims; location and job code data for individual employees; exposure data; biomechanical risk data sorted by job code; program participation records. Data will be used to—Track experience from risk exposures through to medical care events and costs; track performance of identified groups before and after occupational health and safety programs; and compare local results with those of similar groups not involved in control efforts.

Environmental affairs

Data to be collected—Facilities profile, including chemicals inventory; cost per spill data; facility replacement cost data.

Data will be used to—Monitor regularly scheduled facility inspections by chemical type; develop location environmental risk reports; and project future costs under a variety of environmental control scenarios.

As employers' programs evolve from a focus on health care cost containment to a focus on health management, it will be necessary for their data capabilities to evolve into integrated health management information systems. However, prior to developing such capabilities, we must evaluate our informational needs, answering such questions as:

What are our health management goals?

Will our programs help us meet the goals?

What data do we need to evaluate the programs?

What will we do with the data?

Why are the data important?

This can be done effectively only after we have defined our roles and responsibilities in corporate health management.