Charge to the Institute of Medicine
The definition of health has evolved over time from an emphasis on mortality and survival rates to a multidimensional perspective that includes ''an emphasis on the individual's ability to perform daily activities, and ... an emphasis on positive themes of happiness, social and emotional well-being, and quality of life'' (McDowell and Newell, 1996). Similarly, the World Health Organization has updated its definition to emphasize the role of health as a resource for living that applies to individuals, groups, communities, or whole populations. Specifically, the current World Health Organization definition asserts that "for people to reach a state of complete physical, mental, and social well-being, an individual or group must be able to identify and realize aspirations, to satisfy needs, and to change or cope with the environment" (World Health Organization, 1986).
As the paradigm for health has shifted away from acute care interventions and reducing deaths, a model to address serious and complex medical conditions has emerged. Such conditions are frequently persistent in nature and require intermittent interventions to proactively identify and resolve emerging problems and resolve acute exacerbations of illness or disability. There is also a need for coordination of medical and other services to support and sustain individuals' functional ability, as well as their physical, social, and mental health and well-being.
To help provide a framework for the committee's deliberations, a paper was commissioned from David R. Nerenz, Ph.D., director of the Centers for Health Services Research, Henry Ford Health System, Detroit. This paper addressed alternative definitions for serious and complex conditions and is included in this chapter.
Frequently, a complex and comprehensive array of medical and other services is required to manage the multidimensional sequelae of populations with serious or complex medical conditions. Interventions are required not only to respond to the most immediate and acute threats to physical well-being, but also to prevent further complications and to initiate and sustain improvements in overall quality of life. Efforts to accomplish the latter require systems and services to address the long-term physical, social, and mental dimensions included in currently accepted definitions of health.
The nation's health care delivery system has been challenged to accommodate the unique and complicated needs of the growing numbers of Americans with serious or complex medical conditions. Of particular concern are strategies to finance the long-term provision of complicated arrays of health and other services required by these patients. Issues have also arisen about ensuring access to care and quality of care for the growing numbers of individuals with persistent, disabling, or progressively life-threatening medical conditions. Federal health care programs (Medicare and Medicaid) share concerns about these issues with health care plans—including health maintenance organizations, other types of managed care plans, and fee-for-service plans—and a broad constituency of consumers including persons with disabilities.
A leading initiative to address quality of care and access to care issues has been the President's Advisory Commission on Consumer Protection and Quality in the Health Care Industry. One aspect of the commission's work has been development of a Consumers' Bill of Rights and Responsibilities (President's Advisory Commission, 1998a). Of particular relevance to consumers with serious, complex, persistent, disabling, or progressively life-threatening medical conditions is the right to select providers and health care plans that will ensure access to appropriate high-quality health care. Specifically, the Consumers' Bill of Rights and Responsibilities states that "consumers with complex or serious medical conditions who require frequent specialty care should have direct access to a qualified specialist of their choice within a plan's network of providers. Authorization, when required, should be for an adequate number of direct access visits under an approved treatment plan" (President's Advisory Commission, 1998b).
Coincident with the activities of the President's Advisory Commission on Consumer Protection and Quality in the Health Care Industry, the Balanced Budget Act of 1997, Title IV, was passed with implications for Medicare, Medicaid, and Children's Health Provisions (Public Law 105–33, August 5, 1997). A major implication of the Balanced Budget Act of 1997 was the development of a new health care plan for Medicare beneficiaries called Medicare+Choice. According to the interim final regulations for the establishment of the Medicare+Choice program, Section 422.112 addressed issues concerning patient access to services. Specifically, it states that plans must have procedures approved by HCFA for:
- identification of individuals with serious or complex medical conditions;
- assessment of these conditions, including medical procedures to diagnose and monitor them on an ongoing basis; and
- establishment and implementation of a treatment plan appropriate to these conditions, with an adequate number of direct access visits to specialists to accommodate the treatment plan. Treatment plans must be time-specific and updated periodically by the primary care provider.
In response to a request by HCFA, the Institute of Medicine proposed a study to examine definitions of serious and complex medical conditions and related issues. Specific elements of the charge to the Institute of Medicine are:
- Should health plans be required to have a general policy for dealing with serious or complex medical conditions, or should HCFA provide more specific guidance as to the types of beneficiaries to be included in this requirement?
- If specific guidance is to be provided by HCFA, what are the advantages and disadvantages of alternative measurement approaches (e.g., activities of daily living scales, listing of medical conditions)? How feasible would it be for plans to implement such guidance about specific measurement approaches? Should the need for coordination with community and social service agencies be included in the definition of "complex conditions"?
- What are the implications of alternative definitions for Medicaid and commercial programs, when applied to Medicare beneficiaries who are dual eligible?
Consideration of these issues demonstrated unequivocally to the committee the challenges of caring for highly vulnerable populations such as those having serious and complex medical conditions. The challenges ranged from ensuring quality, coordination, and continuity of care for these patient populations from a wide variety of service providers to clarifying issues associated with access to specialty care, patient confidentiality, patients' right to choice and involvement in their health care, provider and plan liability, and appeal and grievance processes.
Scope and Organization of this Report
In the remainder of Chapter 1, alternative operational definitions of "serious and complex" medical conditions are examined, including discussions of their relative strengths and weaknesses. Chapter 2 outlines the challenges of ensuring timely access to high-quality health care services and resources for patients with serious and complex medical conditions. Chapter 3 presents the committee's overarching conclusions and recommendations in response to specific elements of the charge. Together, these chapters suggest an approach to understanding the implications of defining serious and complex medical conditions for managed care and other types of health plans and make recommendations in eight areas. Finally, there are five appendixes to this report. Appendix A provides a sum-
mary of the workshop proceedings. Appendixes B, C, and D, respectively, include brief biographical sketches of the workshop panel members, the workshop agenda, and a list of workshop participants. Appendix E contains brief biographical sketches for each of the seven members of the committee.
Definitions of Serious and Complex Medical Conditions: Alternatives and Implications
Managed care has the potential to significantly improve the quality of health care provided to health plan members. The original promise embodied in the term '"health maintenance organization" was an emphasis on preventive services, early detection of illness, close coordination of medical and nonmedical health services, and treatment care plans designed to maintain function and minimize disability. For some enrollees and in some plans, this potential has been realized. In the best of circumstances, a fragmented, uncoordinated set of services has been replaced by a more seamless array of services tailored to the needs, values, and priorities of plan members.
There is another side to managed care, though. For members in some plans, managed car e seems to mean a system of hurdles, hoops, and barriers that prevent access to needed services. This side of managed care is the one that has frequently been portrayed in print and electronic media and has captured the attention of politicians and patient advocates who are working on a number of patient protection initiatives.
Access to care can be a significant issue for anyone with an illness, but it is particularly significant for those individuals who have conditions with the potential to cause death, disability, or serious discomfort unless treated with the best available technologies. Access to specialty care or to specific treatment or diagnostic services can become literally an issue of life or death for people with cancer, heart disease, HIV/AIDS, and a number of other acute and persistent conditions.
With these considerations in mind, the President's Commission on Consumer Protection and Quality in the Health Care Industry raised the question of whether patients with serious and complex medical conditions should have a guarantee odirect access to qualified specialists within their plan's network of providers (President's Advisory Commission, 1998b). The Health Care Financing Administration has been charged with developing regulations for Medicare+Choice plans that would implement such a guarantee.
Any such provision written into federal or state law or into the accreditation process for health plans would require a clear, practical definition of the term serious and complex medical condition. The remainder of this chapter outlines conceptual categories of conditions that might be considered serious and complex, examines possible alternative operational definitions to screen for the pres-
ence of serious and complex conditions, and discusses the implications of these definitions for use in the regulatory or legislative process.
Conceptual Overview—Serious and Complex Conditions
A number of criteria can be used to describe medical conditions as "serious and complex." These could include severity of the illness, degree of impairment or disability, and level of need for comprehensive care management. Disability is defined according to the Americans with Disabilities Act (1990) as pertaining to persons with a physical or mental impairment that substantially limits a "major life activity," persons with a record of such an impairment, or persons who believe that others regard them as having such an impairment. The following presents examples descriptive criteria for serious and complex medical conditions. It is important to recognize that these conditions may be serious and complex for some patients at some points during the course of their disease or disability. The conditions will not necessarily be serious and complex for all patients at all times. Further the list is intended to suggest criteria for serious and complex conditions but it is not an exhaustive summary of such criteria.
• Conditions that are life threatening:
(1) cancer, (2) heart disease, (3) stroke, and (4) HIV/AIDS.
• Conditions that cause serious disability without necessarily being life threatening:
(1) stroke, (2) closed head or spinal cord injuries, (3) mental retardation, and (4) congenital malformations.
• Conditions that cause significant pain or discomfort that can cause serious interruptions to life activities:
(1) allergies, (2) migraine, (3) arthritis, and (4) sickle cell disease.
• Conditions that require major commitments of time and effort from care-givers for a substantial period of time:
(1) mobility disorders, (2) blindness, (3) Alzheimer's disease and other dementias, (4) chronic obstructive pulmonary disease, (5) paraplegia or quadriplegia, (6) Down's syndrome, and (7) depression.
• Conditions that may require frequent monitoring:
(1) diabetes, (2) conditions requiring anticoagulation treatment, (3) severe asthma, (4) severe allergies, and (5) schizophrenia and other psychotic illnesses.
• Conditions that predict or are associated with severe consequences:
(1) hypertension (associated with heart disease), (2) depression (associated with suicide), (3) diabetes (associated with blindness, kidney failure), and (4) alcohol and other substance abuse (associated with intentional and unintentional injuries).
• Conditions associated with negative consequences for someone else:
(1) high-risk pregnancy (risk to newborn), (2) HIV/AIDS (risk to sex partner), and (3) tuberculosis (risk to community in general).
• Conditions that affect multiple organ systems:
(1) HIV/AIDS, (2) cancer, and (3) diabetes.
• Conditions that require management to "tight" physiological parameters:
(1) conditions requiring anticoagulation therapy, (2) type I diabetes (possibly type II also), and (3) kidney failure.
• Conditions whose management requires coordination of multiple specialties:
(1) breast cancer, (2) depression comorbid with serious medical conditions, (3) multiple sclerosis, and (4) cerebral palsy.
• Conditions whose treatment carries a risk of serious complications:
(1) most cancers, and (2) other conditions requiring complex surgery.
• Conditions requiring adjustment in a "nonmedical environment":
(1) mobility disorders, (2) blindness and other sensory disorders, (3) Alzheimer's disease, and (4) frailty.
These categories are not mutually exclusive, and clinical conditions such as cancer can appear under several headings. Serious and complex medical conditions should reflect the characteristics of the management of the condition rather than some inherent biological complexity. Although there may be some correlation between biological parameters and complexity of management, it is possible to imagine biologically complex conditions that are easy to treat because a very simple treatment (e.g., a once-a-day prescription drug) has been proven effective. Alternatively, some biologically simple conditions (e.g., spinal cord injuries) may be very complex from a management point of view. In this report, the various meanings of serious and complex will have more to do with illness and disability management than with underlying biology.
Potential Operational Definitions and Criteria
None of the above concepts is of sufficient specificity to be used to write or implement regulations about serious and complex medical conditions. They may, however, represent a set of screens to assess whether potential operational definitions are sufficiently inclusive of the range of conditions that might fit in the category of serious and complex medical conditions. A number of possible operational definitions for serious and complex conditions are described below, along with a summary of their strengths and weaknesses related to validity, reliability, and feasibility. Comments on strengths and weaknesses are not meant to be definitive but to highlight points for further consideration.
Disease Code Definitions
ICD-9 Codes Alone.
The five-digit International Classification of Diseases, Ninth and Tenth Editions (ICD-9 or ICD-10) coding system would permit the classification of individual diseases or codes (down to the fifth-digit level if necessary) as either being or not being serious and complex medical conditions (American Medical Association, 1999). Using a set of conceptual criteria like
those presented at the beginning of this chapter, expert panels could make such determinations, perhaps requiring an additional factor to be added to the codes when the codes themselves are ambiguous relative to a description of serious and complex medical conditions. In practice, then, an individual with one or more ICD-9 codes in the "included" list in a defined period would be considered to have a serious and complex medical condition. The codes would typically be found in insurance claims databases. Most managed care plans would be able to use a "presence-of-codes" rule to identify patients with serious and complex conditions.
Rules would be clear and obvious to patients, physicians, health plan representatives, and external regulators. Data would be fairly readily available in existing administrative data sources. Rules could be revised and updated easily since no fundamental changes in data collection or coding technology would be required for changes.
There would be a clear incentive for "gaming" or "upcoding" and "downcoding'' in ambiguous situations in which one code would qualify as a serious and complex medical condition and another potentially valid code would not. Using ICD-9 codes alone fails to capture the severity of disease. Moreover, they do not address the issue of comorbid conditions. Coding may be biased or inaccurate.
ICD-9 Codes Plus Severity Codes.
For some ICD-9 codes, a fifth digit is interpretable as a "severity" code. Future enhancements of coding systems in either inpatient or outpatient systems may involve the use of a fifth digit or some other convention to indicate a severity or complexity level within the basic disease code. Classification of conditions as serious and complex would then depend on both the ICD-9 or-10 "stem" and the severity code attached to it.
The system would carry most of the advantages of ICD-9 disease codes alone and add the precision of a severity level that would serve to make finer distinctions possible for conditions such as diabetes, asthma, or hypertension.
Unless the clinical criteria for various severity codes are very clearly and objectively defined, there will be an incentive for either upcoding or downcoding distortion in the severity field. It may be very difficult to audit or validate such codes without clear, objective definitions. Health plans wishing to avoid the provision of additional services may implement procedures designed to make providers use the lowest possible severity codes. Use of ICD-9 or-10 codes with severity codes will not address the issues of comorbidity, coding errors, and coding biases.
For mental health and substance abuse conditions, psychiatrists may prefer the use of Diagnostic and Statistical Manual for Mental Disorders (Fourth Edition) codes to ICD-9 or-10 codes (American Psychiatric Association, 1994). The DSM-IV classification system may permit a finer-grained distinction among types of condition and severity level. Use of these codes would also require some sort of expert panel process to determine which codes or cate-
gories should be associated with regulations about serious and complex medical conditions.
Some of the same advantages would be found here as with the ICD-9 codes. Once established, the rules would be clear to virtually all parties involved. The coding system is generally well established among mental health and substance abuse providers and would be familiar to those most likely to be assigning codes to individual patients.
The DSM-IV system is not typically used in insurance claims due to concerns about confidentiality; it is therefore much less likely than the ICD-9 system to be feasible to implement from the claims databases of health care plans. Use of the system may require manual medical record abstraction, which would make it extremely expensive and also presents possible compromises to confidentiality. Furthermore, the DSM-IV is applicable only to mental health and substance abuse care.
International Classification of Impairments, Activities, and Participation (ICIDH-2).
This classification system is similar to the ICD-9 system in that it seeks to provide a standard terminology for use in describing specific patient characteristics and is supported by the World Health Organization (World Health Organization, 1997). The difference is the subject matter. Whereas ICD-9 and-10 focus on diseases or injuries, ICIDH-2 focuses on impairments; the acronym reflects the original system's use of the terms disabilities and handicaps. In the present context, individuals having specific impairments would be considered to have serious and complex medical conditions.
The ICIDH-2 system provides a standard nomenclature and typology for the entire range of impairments and disabilities. To the extent that any of these specific problems are valid markers of serious and complex medical conditions, the ICIDH-2 system provides a clear way of assigning codes to the impairments to facilitate accurate classification. The ICIDH-2 system may represent some disabilities better than do current ICD-9 codes.
The ICIDH-2 system is not nearly as well known or widely used in the United States as the ICD-9 system. It has not been linked to reimbursement by public or private payors, and it has not served as the basis for other well-known patient classification systems (e.g., ICD-9 codes have been used to build diagnosis-related groups). There is really no infrastructure in place to support the use of this system, even if there are clear matches between specific ICIDH-2 codes and the concept of serious and complex medical conditions.
Functional Status Definitions
Karnofsky Performance Status.
The Karnofsky scale is one of the oldest functional status measures in use today (Karnofsky and Burchenal, 1949). It is a single 0–100 scale (generally used in increments of 10 for individuals) used to describe patients' functional ability. Individuals at the low ranges of the scale are
presumably there because of some serious medical condition, although the scale does not strictly require that there be a distinct medical reason for a given level of function. In the context of describing serious and complex medical conditions, a threshold (or possibly multiple thresholds for different ages or classes of patients) would need to be established.
The Karnofsky scale is generally well known and accepted in a variety of clinical areas, particularly oncology. Although there may be some potential for gaming, the definitions for various levels of the scale provide some opportunity for validation or auditing.
This scale is not typically found in administrative databases and is not often used to describe patients outside of research contexts or specific clinical areas (e.g., oncology) in which it has come into common use. Although the scale in general has proven validity, the assignment of an individual to one of two or three adjacent levels can be somewhat arbitrary. If being described as having a serious and complex medical condition depended on being below a certain threshold, there could be an incentive to put a patient either above or below the threshold that would bias "gray area" cases.
Another key problem is that the scale does not directly refer to any specific medical condition, so a serious and complex medical condition is evident only. indirectly in the scale score. This approach may be useful only if combined with some more explicit coding of disease states (e.g., some cancer patients with Kamofsky scores of <50 may be judged to have a serious and complex medical condition, whereas not all cancer patients or all patients with Kamofsky scores of <50 would be so labeled). In addition, individual scores may not be stable over the period required to gather data, assemble a database, and determine eligibility for individual patients.
SF-36, SF-12, or Similar "Profile" Scales.
Scales such as the Short Form-36 (SF-36), Short Form-12 (SF-12), Dartmouth COOP Chart, and so forth, are similar to the Karnofsky scale in that they assign numbers to individuals to reflect varying levels of functional status or health-related quality of life (Ware and Sherbourne, 1992). The scales differ in that they contain two or more specific dimensions on which individuals receive scores. It is possible, then, to select specific dimensions rather than an overall score, or perhaps to identify a specific profile of scores that would be consistent with the concept of a serious and complex medical condition. Again, the scores do not reflect the presence of any specific medical condition. Most of the scales are designed to apply to patients with any medical condition or with no condition at all.
The SF-36 and SF-12 in particular are widely known and used frequently for performance measurement, quality improvement, and research purposes. The scales do not take long to administer. They are becoming more widely used in managed care plans on a full-membership rather than a disease-specific basis. The Health Care Financing Administration already requires this sort of survey in Medicare+Choice plans for "report card" and consumer choice
purposes, so plans serving Medicare beneficiaries are familiar with these instruments.
Most of the concerns about the use of the Kamofsky scale apply to these other scales as well. In addition, the scales are almost always completed on a self-report basis, so patients must be engaged in the process of data collection. The Karnofsky scale, by contrast, is based on clinician evaluation and is easier to incorporate into a claims-based data collection process. Scale scores in and of themselves do not identify serious and complex medical conditions, but the scale scores used in combination with other information (e.g., ICD-9 codes in claims databases) might add precision to the classification process.
Quality of Well-Being, Health Utilities Index, or Similar Utility-Based Scales.
A number of scales are designed to evaluate individual patients' health status on a scale from 0 to 1, where 0 is death or the worst imaginable health status and 1 represents full function and absence of any pain or disability (Kaplan et al., 1989; Torrance, 1987). The purpose of this particular scaling format is to permit the calculation of quality-adjusted life years (QALYs) or disability-adjusted life years (DALYs) that adjust the value of a year of life according to the quality of life reflected in the 0–1 scale. A year at a level of 1 is worth 1; a year at a level of .4 is worth .4, or is equivalent to .4 of a year at level 1. These scale scores would presumably be used in the same way as the Karnofsky, SF-36, or other scores. Individuals below a certain defined threshold would be considered to have (or at least be likely to have) a serious and complex medical condition.
Scaling is similar to the Karnofsky in that it is a single dimension with intuitively interpretable values. The translation of 0–1 scores into QALYs, DALYs, and similar measures may allow the incorporation of life expectancy or life expectancy without disability into the calculation of serious and complex medical conditions.
Most of the same objections to functional status scores listed above apply here as well. The scores may not be stable over long periods of time for specific individuals with serious and complex conditions. The scores do not directly reflect the presence of any specific medical condition, and they may be subject to gaming if the underlying data collection methods permit it. The Quality of Well-Being scale requires patients to answer approximately 40 questions; other scales such as the Euro-Qol scale involve as few as 5 items. In any case, however, the direct involvement of patients is required for data collection.
ADL or IADL Scales.
Activities of daily living (ADLs) or instrumental activities of daily living (IADLs) involve the assessment of patients' ability to perform defined tasks (e.g., getting dressed, bathing, using a telephone) (Katz et al., 1963; Lawton and Brody, 1969). The items are somewhat more specific than those involved in some of the functional status scales described above, but the intent of scaling is similar—to classify or score individuals according to their level of function or independence. In determining eligibility for some disability benefit programs, ADL or IADL scores are expressed in terms such as "limited
in two or more areas,'' so that the issue is not the overall score but the number of areas of impairment.
There is some precedent for using ADLs and IADLs as criteria for determining eligibility for specific benefits or services (e.g., long-term-care services). The criteria are reasonably well defined, and there is a long history of use of the scales for research and clinical purposes.
As with other functional status measures, ADLs and IADLs do not directly indicate the presence of a serious and complex medical condition although individuals with severe ADL or IADL limitations may have such conditions. Alternatively, patients with a low number of limitations on these measures may also have a serious and complex medical condition. The ADLs or IADLs are used in the context of care management programs, rehabilitation programs, or long-term-care programs, but they are not typically used more broadly among the general population of health plan members. Some method of data collection would have to be developed for large numbers of patients who are not currently assessed.
Patient Classification System Definitions
Diagnosis-related groups (DRGs) are familiar to most plans and providers as the basis for payment to hospitals by Medicare and many other public and private payors. The DRGs were designed to group patients together who are similar in terms of length of stay and other measures of resource consumption in the inpatient setting. The "D" in DRGs indicates that the system is based on ICD-9 diagnostic codes, but information on age, surgical procedures, and complications or comorbidities is also used to assign patients to DRGs for payment purposes. The DRGs could conceivably be used as a means of determining whether patients have serious and complex medical conditions, and the information beyond diagnostic code may provide a more precise classification than diagnosis alone. Presumably, a diagnosis with complications or comorbidities is inherently more serious and complex than one without them.
The DRG system is essentially universal in acute care medical and surgical hospitals and may be applicable to some psychiatric, substance abuse, rehabilitation, or long-term-care facilities. Because the system is used for payment purposes, there is a means available for auditing DRG coding to prevent most intentional gaming. The DRG information is contained in virtually all administrative databases on inpatient care.
The DRG system is a population-based measure of very general categories and is not applicable to any specific individual. The system is not useful for patients who have not been admitted to a hospital. It may be possible to map a set of ICD-9 codes used in outpatient settings to the algorithms used to assign DRGs from inpatient codes, but it is not clear that the diagnoses associated with long lengths of stay or high resource use in the inpatient setting are necessarily serious and complex in the outpatient setting. Since most health plan
members at any specific time are not admitted to a hospital, no DRG information would be readily available for most plan members at any particular time. Likewise, persons with serious and complex medical conditions that are managed successfully on an outpatient basis would not have any DRG information available.
Diagnostic cost groups (DCGs) are an alternative set of patient groupings (Ash et al., 1989; Ellis et al., 1996; Lamers, 1999). In this case, the goal is to classify patients together who are similar in terms of annual expenditures across all settings rather than length of stay or expenditure in just the inpatient setting. The original DCG modeling was based on inpatient diagnoses, but subsequent versions have incorporated both inpatient and outpatient diagnoses in model development. The DCGs would be useful only to the extent that annual expense is associated with the concept of serious and complex medical conditions.
The DCGs have been well researched since the mid-1980s and are beginning to be used in prominent applications. For example, HCFA has decided to use DCGs—albeit for inpatient services only—to implement provisions of the 1997 Balanced Budget Act regarding risk-adjusted capitation payments to Medicare+Choice plans. Some large private employers (e.g., Ford Motor Company) are beginning to use DCGs in the analysis of health care costs and quality for their employees. The grouping algorithms use generally available administrative data that plans have in their billing or encounter data systems. The DCG system has "face validity" for the current application to the extent that users accept an association between the categories of serious and complex medical conditions and annual expenditures.
The DCG system was designed for a specific purpose—to model annual expenses based on diagnoses—and this purpose may not exactly match the description of persons with serious and complex medical conditions. If there are conditions that are not expensive but are very serious, or conditions that are expensive but not serious in any other sense, then the DCG system would yield misleading results. It has also been noted that the DCG system can reward inefficient patient management because of its assumption that higher costs are associated with greater complexity of care. Such higher costs could, in fact, be the result of inefficient management. Although HCFA has made a commitment to use the system for Medicare+Choice plans, it is not widely used in other settings for risk-adjusting capitation payments to plans.
Ambulatory care groups (ACGs) are similar to DCGs in that they are designed to group patients together who are similar on a measure of annual cost or expenditure (Weiner et al., 1991). As the title implies, the focus of ACGs is on outpatient rather than inpatient care, so the system is perhaps better suited to physician profiling or partial capitation payment applications in the outpatient arena than to more global payment applications. The ACGs might conceivably be a better conceptual fit than DCGs to the concept of serious and complex medical conditions. It would be an empirical question whether ACGs or DCGs
more accurately predict the need for a broad strategy of care management, as judged by some independent criterion.
A number of health plans (the exact number is uncertain and can change frequently) use ACGs for purposes of physician profiling or adjusting capitation payments to individual physicians or physician groups. Most plans would at least be familiar with ACGs and would be able to assign patients to ACGs using data routinely available in administrative or encounter databases.
Annual expense in the outpatient setting is almost certainly correlated with the underlying descriptions of serious and complex medical conditions, but the relationship is not perfect and there may be some misleading results. Given the relatively small number of ACGs, one could expect some heterogeneity within specific ACGs on the extent to which patients really do have serious and complex conditions.
Kaiser Clinical-Behavioral Classification System.
The Kaiser system was developed in the 1960s as part of a program of research on health care utilization in Kaiser's Center for Health Research in Portland (Hurtado and Greenlick, 1971). The system is similar to both DCGs and ACGs in that it tries to categorize patients into groups that are homogeneous on overall resource utilization. Also like ACGs and DCGs, the system is based on diagnostic categories but adds other information that helps create homogeneous and clinically meaningful groups.
The Kaiser system is based primarily on diagnostic codes that would typically be found in plans' administrative databases. It was designed for a single managed care plan to explain patterns of utilization within that plan.
The Kaiser system is not as well known now as the DCG, ACG, or DRG systems and has not been widely used outside Kaiser. Although many of the categories seem to have a conceptual link to categories of serious and complex medical conditions, the system was not designed specifically with this issue in mind. Consequently, it shares many of the same problems as the DCG or ACG systems regarding possible mismatch of concepts.
Total Annual Expenditure or Cost.
The classification systems just described (DCGs, ACGs, Kaiser) use annual expense in one way or another to define homogeneous groups. The measure of cost or expense itself could be used to identify a group or groups of persons who are likely to have serious and complex medical conditions. For example, one could apply a screen to identify individuals in the top 5 or 10 percent of annual expenditures in a health plan.
Most plans can use existing administrative data sets to identify patients with high absolute or relative expenses in any defined period. The data requirements are less than for the DCG or ACG approaches since diagnosis or procedure code information is not necessary to calculate costs.
Costs in a previous period do not necessarily predict either future costs or costs associated with current needs for specialty care. Patients with acute injuries who are fully recovered might have needed specialized services at the time of the injury (before they would have been identified in an administrative database due to lag time in claims submission and processing), but would not necessarily need such access in the future if the injuries are fully healed. Expense information might have to be combined with diagnostic or other information to accurately identify patients with serious and complex medical conditions.
Above-Threshold Number of Hospital Admissions per Year.
Patients who have multiple hospital admissions in a period of time such as a year are a statistically unusual group and are probably dealing with serious and complex medical conditions. Multiple admissions reflect either persistent conditions under poor control, serious illnesses such as cancer that require frequent admissions for administration of complex treatments, serious anatomical problems that require multiple surgeries, or general frailty.
Data needed to assess multiple hospital admissions are usually available in plans' administrative databases. There is generally little ambiguity about the occurrence of an admission, so opportunities for gaming are minimized. The measure might be highly specific in that very few people with multiple admissions would not pass some other test of the presence of a serious and complex medical condition.
The criterion would not be very sensitive because only a few patients with serious and complex medical conditions would be admitted to hospitals multiple times in a given year. As a stand-alone definition or criterion, multiple admissions would probably not be useful, but its predictive value might increase when combined with diagnostic or other criteria (e.g., patients with type I diabetes and two or more admissions for a diabetes-related diagnosis).
Prior Use of Specialty Care.
Provider codes in billing or encounter databases could be used to identify patients who have used specialized services in the recent past. To the extent that prior use can be interpreted as continuing need, previous use of specialists and specialized services could be another indicator of individuals with serious and complex medical conditions.
Data are probably readily available, although some provider codes may not accurately reflect the provision of specialized services.
Past use is not necessarily a marker of current or future need, particularly in situations in which the appropriateness of past use might have been questionable.
Use of Specific Treatments, Procedures, or Services.
There are specific tests, treatments, or procedures that may identify persons with serious and complex medical conditions. Hospital and health plan credentialing committees address this issue on an ongoing basis, as individuals or groups seek approval to perform
various tests or procedures. To the extent that the rules used to grant privileges and credentials can be generalized across plans and parts of the country, a patient's receipt of specific services (or, more specifically, the projected future receipt of such services) could be a criterion for identifying that patient as having a serious and complex condition. Organ transplants, for example, might indicate not only the presence of serious disease in the past but also the continuing need for a broad care management approach and the presence of serious and complex conditions in the future.
A number of therapeutic procedures (e.g., organ transplants, bone marrow transplants, "cocktail" treatments for HIV/AIDS) are done only in the presence of serious and complex conditions that can be presumed to continue as long as the patient is alive. This sort of definition or criterion, then, would be highly specific in such cases. Although the intensity of need for care management (and the underlying seriousness and complexity of disease) may diminish over time, it may never go away totally. Procedure codes are readily available in administrative databases since they are used for reimbursement.
At best, this is a partial solution to the problem of defining serious and complex conditions. Not all conditions are associated with distinct treatments. Any tests or procedures done to detect or rule out disease would not be useful since they may find the absence of serious illness rather than its presence.
Physiological Measure Definitions
Laboratory Test Values.
Some inpatient severity of illness systems (e.g., APACHE III) rely on laboratory tests to assess the concept of physiological "derangement" and severity of illness. The models are built on a statistical association between extreme laboratory test values and either risk of mortality or inpatient cost. In the inpatient context, patients with extreme lab values may be considered to have serious and complex medical conditions, although not all values would have this connotation. Some laboratory test results may be interpretable as markers of serious and complex conditions over longer periods of time in the outpatient setting. A high prostate-specific antigen value, for example, may indicate the presence of prostate cancer and the need for a broad care management plan for further diagnostic and treatment purposes far into the future.
Test values are quantitative, objective, usually precise and accurate, and generally not subject to gaming. A number of tests are used for the purpose of detecting and monitoring serious and complex conditions, so the presence of an abnormal value has a clear, precise meaning.
Only a fraction of all serious and complex medical conditions can be identified by specific lab tests. Most of the mental health domain would be excluded from this criterion, and there would be major debate about the interpretation of lab tests for other common conditions. Glycosylated hemoglobin, for example, is the "gold standard" for glycemic control in diabetics, but it is not clear that a specific value or range of values would indicate "serious and com-
plex" diabetes rather than some alternative. Plans would have to develop methods of obtaining actual lab test values rather than administrative data on whether a test was done or not.
Other Physical Exam Values (e.g., blood pressure, weight).
This set of criteria is conceptually similar to the lab values just discussed. There are some widely used clinical measures that may be markers for the presence of serious and complex medical conditions. Hypertension is identified through blood pressure readings, and unusually high readings could be used as markers of serious and complex hypertension. Weight is a measure of obesity (one might prefer to use body mass index instead), and again, unusually high values might reflect more serious and complex conditions.
Measures such as weight, blood pressure, visual acuity, and so forth are common in clinical practice, generally accurate, inexpensive to obtain, and reasonably well interpreted as markers of serious and complex conditions in defined contexts. These measures may be particularly useful in identifying persons with comorbid conditions.
Health plans do not have this information routinely available in administrative databases, and provider groups may not have the information available in electronic form. Although this information is inexpensive to obtain in a clinical context, it may be very expensive to obtain for large numbers of people to ensure proper identification of those with serious and complex medical conditions. There may be some potential for gaming for individuals at or near a threshold value. There is some potential for misinterpretation of almost all measures especially for those individuals with significant comorbidities.
The Medicare Home Health Care Quality Assurance and Improvement Demonstration Outcome and Assessment Information Set (OASIS-B) data set has been designed for HCFA to use in monitoring physiological parameters of patients enrolled in home health care programs. While its primary purpose is to assist in stratifying populations by morbidity to set prospective payment rates, this instrument is also expected to be used for quality improvement and other functions. It contains fields to collect a primary and up to five secondary diagnoses, with each entity rated on a 0 through 4 severity scale. It catalogues areas such as prognosis, living arrangements, supportive assistance, sensory status, activities of daily living, and use of medications.
This data instrument is used fairly extensively and directly addresses issues concerned with the identification of patients with serious and complex medical conditions. The use of a severity rating scale for each diagnosis coupled with an assessment of functional status and the cataloging of physiological dysfunctions could be very useful in identifying and tracking the progress of individuals with serious and complex conditions.
This instrument is time intensive to administer and is now mandated for use only in the home health arena. The accuracy of the information when collected in community settings is unknown. There is little experience at
this point with systematic use of these data to manage populations of patients and potential impact upon their physiological and functional status.
Care Management Guidelines or Other Entitlement Program Definitions
Eligibility for Medical Disability Payments.
Individuals who are eligible for disability payments or related services through private employers, Medicaid programs, Veterans Administration programs, or other public programs may already have had to demonstrate the presence of serious and complex medical conditions. Children eligible for Medicaid benefits through the Children with Special Health Care Needs program, for example, may not need to pass any other tests to be labeled as having a serious and complex medical condition.
As long as the criteria and definitions fit, the work of data collection and the use of data to classify individuals have already been done. No separate process has to be invented or used.
As indicated above in the discussion of ADLs and IADLs, disability per se does not necessarily indicate the presence of a serious and complex medical condition.
The National Hospice Organization (NHO) has disseminated a set of guidelines for appropriateness of hospice care (National Hospice Organization, 1996). These guidelines have been used to make decisions about the care of individual patients to judge their appropriateness for the hospice benefit in Medicare. The guidelines are based on combinations of diagnostic, functional, lab test, and other criteria the underlying concept of which is the likelihood of death in a period of 6 months or less. Virtually all patients meeting the NHO guidelines would be considered to have a serious and complex condition.
This is a criterion with very high specificity. People meeting the NHO guidelines clearly have serious and complex conditions.
The NHO guidelines affect only those people with terminal illnesses, who make up only a small subset of those in the health plan population with serious and complex conditions. Even where the guidelines are used, they are not generally found in administrative databases because they do not produce a single score that can be entered and stored readily.
There is no reason why any of the above definitions would have to be used in isolation. In fact, the best operational definition (as judged by assessments of ''false positives" or "false negatives") would almost certainly be one that combined different types of information to come as close as possible to identifying categories of patients with serious and complex conditions. As long as the necessary data elements were available, relatively sophisticated combination rules
could be developed from the above list to create a more precise definition that would minimize both false positives and false negatives. For example, although not all diabetics (ICD-9 codes 250.xx) are necessarily considered to have serious and complex medical conditions, those type I diabetics with one or more early signs of complications (e.g., diabetic retinopathy) might be more appropriately included in this category.
No single criterion or definition in the above lists will completely meet the needs of HCFA to define patient populations with serious and complex medical conditions. It is almost certain that some sort of combined criteria will have to be developed to reflect underlying medical condition, current disability, severity of a medical condition or disability, and perhaps the presence of a treatment that will continue into the future and that inherently requires access to broad strategies of care management.
It is the opinion of the committee that the state of current clinical and research literature does not yet support the selection of specific categories of patients or service needs to define serious or complex medical conditions. Disease classification systems such as DRGs, DCGs, and ICD-9 lack the sensitivity and specificity required to identify patients with serious and complex medical conditions. The validity of patient classification systems and functional status definitions has yet to be established. Efforts to develop case-mix-adjusted reimbursement systems that ensure cost-effectiveness and high-quality interventions for patients with serious and complex conditions are still in the early stages or are limited to a single disease category. Similarly, efforts to develop standardized protocols for issues of care coordination, continuity of care, care management, and service reimbursement have not met with great success. Furthermore, significant variation on these issues is apparent for different types of health care plans, provider organizations, and patient populations.
Furthermore, several limitations are shared by the suggested list of potential operational definitions for serious and complex medical conditions. First, the measurement of specific dimensions of health and health status will change and improve in the coming years as researchers advance their knowledge about the technologies associated with methods of data collection, management, analysis, and interpretation. Health care plans and HCFA should ensure that state-of-the-art measurement techniques are adapted as quickly as possible to improve identification of patients with serious and complex medical conditions.
It is also important to acknowledge that a system of screening patient populations for the presence of serious or complex conditions will be limited by the frequency with which such assessments are made. Too frequent assessments may not allow sufficient time to elapse for measures to maintain a level of sensitivity to detect change. In contrast, infrequent assessments may fail to detect changes that occur. In either case, patient populations will be at risk of misclas-
sification for having or not having a serious or complex medical condition. Careful consideration must be given to determining the appropriate intervals between identification and screening of patient populations for the presence of a new condition that might be serious and complex, the continuation of an existing serious and complex condition, and the resolution of a condition previously determined to be serious and complex in nature.
Health care plans and HCFA must also be cognizant of the limitations of data that is collected, analyzed, or interpreted retrospectively. Examples of such data might be submission of information about claims and claims resolution data. Reliance on retrospective data may result in significant delays in identifying beneficiaries with serious and complex medical conditions. Furthermore, use of retrospective data to predict future events has the potential for misclassification of patients as having or not having a serious and complex medical condition.
Health plans and HCFA should also be aware of the challenges inherent in the process of data collection, management, analysis, and interpretation. The potential for errors and biases is present at each step and could result in misclassification of patients with serious and complex medical conditions. The collection of timely information that is of sufficient sensitivity and specificity to reduce the number of false-positive and false-negative classifications as serious and complex imposes significant burdens and will require allocation of adequate resources and professionally trained staff.
Clinical research relevant to defining patient populations as having serious and complex conditions and developing the most effective care protocols to respond to patient needs is evolving rapidly, but only preliminary results are available. For example, research in correlating patient volume with patient outcomes, differentiating access to services as a function of health plan benefits, and addressing variations in quality of care has just begun to yield meaningful results about particular groups of patients. The committee anticipates that this research will continue to evolve and will provide more definitive conclusions to guide the efforts of health plans to identify patients with serious and complex medical conditions and ensure appropriate levels of treatment and care management. As the science base is strengthened, so too will be the ability of HCFA and health plans to correctly identify patients with serious and complex medical conditions who would benefit from broad strategies of care management.