Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.
Key Patient Management Topics for Effectiveness Research in Acute Myocardial Infraction PRELIMINARY DISCUSSION AND SELECTION OF MAJOR TOPICS The appendix describes the workshop and its preliminary activities, including a homework exercise to identify provisional patient topics of high priority for effectiveness research. It also briefly describes presentations by HCFA staff on the Medicare data system and an analysis of Medicare data on AMI and cardiovascular disease. At the workshop, the committee reviewed the homework results and the HCFA data analyses and then discussed AMI studies that might be appropri- ate for the Effectiveness Initiative. This session made clear the number and complexity of unanswered questions about the appropriate management of individuals who are at risk for, who are experiencing, and who have had an AMI. Many of these questions still lie in the realm of efficacy rather than effectiveness that is, whether a test, a procedure, or a drug works at all, rather than how well it works in the average practice of medicine. These issues may belong more in the arena of clinical and biomedical research of the sort conducted by the National Institutes of Health (in particular the National Heart, Lung and Blood Institute) than in the area of effectiveness research. Thus, the committee tried to narrow its task to clarifying what more lun- ited set of questions could be satisfactorily addressed in one of three ways: (1) with existing HCFA administrative data of the sort on which the prelimi- nary analyses presented at the workshop were based; (2) with existing HCFA data augmented by clinical data from medical records abstracted by PROs or from specially conducted patient surveys and follow-up studies; and (3) by carefully planned longitudinal studies that combine all necessary 29
30 EFFECTIVENESS INITIATIVE inpatient and ambulatory information (e.g., evidence of prior coronary dis- ease, prehospital care, drug or surgical therapy, rehabilitation impact, and quality-of-life outcomes). To this end, the committee generated numerous topics on which the opening discussions focused. Some of these topics cut across several issues, such as the need to examine patterns of care by geographic and rural or urban area, by hospital and type of practitioner, and by type of intervention (e.g., pharmacologic, invasive, surgical, or some combination). Other topics included risk stratification of patients and the need to study various staging methods and their relation to outcomes. Ten categories of study topics were advanced by committee members and recorded during the early part of the discussion; they are reported, in some- what condensed form, in Table 1. Some relate to clinical issues involving the use of diagnostic and therapeutic modalities such as laboratory testing and catheterization and other invasive interventions. Others relate more to research methods, such as outcome measures, or to crosscutting topics, such as risk stratification, ~eaunent modality, and outcomes. From this discus- sion and listing, members were asked to vote for up to five topics that should receive He highest priority attention in the effectiveness program for AMI. These were tabulated and the highest ranking topics formed the basis for the remaining discussion. The following section elaborates this discus- sion and Be committee's recommendations. TABLE ~ Provisional Patient Management Topics Relating to Acute Myocardial Infarction Risk Stratification after AMI Timing related to MI: early versus late Use of diagnostic or therapeutic procedures to stratify risk Guidelines for stratification based on existing administrative data Prospective study to validate stratification testing and criteria High-risk versus low-risk treatment plans Patients appropriate for no stratification Available data on effectiveness of stratification over age 65 Procedures in Early Diagnosis and Treatment Catheterization, angioplasty, and surgical procedures Echocardiography; positron emission tomography scans Defibrillation Use of coronary care unit (CCU) Aggressive (invasive) versus nonaggressive (medical) treatment Continued use of outmoded diagnostic tests
ACUTE MYOCARDIAL INFARCTION TABLE 1 (continued) 31 Pharmacologic Treatment Th~mbolytics (streptokinase, tPA, and urokinase) Relation of pharmacologic agents to other therapies such as angioplasty Outcomes Mortality: before hospital, during hospitalization, early and late after hospitalization Odder outcome measures: functional status, quality of life, activity, pain As related to: caregiver (primary versus specialist); geographic location; site of care (community versus teaching center); diagnostic and or therapeutic measures; stratification; in-hospital care (CCU); length of stay Variations in Treatment and Outcomes Geographic (location of care; origin of patient); individual physicians and groups; specific hospitals Relationship to outcomes Implications for cost Value as "natural e~cperunents" Post-AMI Issues Prognosis for different risk categories Cardiac rehabilitation Sudden death Prudery Prevention Role of pre-event risk identification Concem for neglect of this area, as primary prevention occurs before Medicare age Prehospital Care Precipitating events Recognition, initial treatment out of hospital Emergency room Physician Education Guidelines for patient management Comparative Analyses of Differem Modalities Aggressive versus nonaggressive management Angiaplasty versus CABG Value of ~rombolytic treatment Additional diagnostic tests
32 EFFECTIVENESS INITIATIVE SUMMARY OF RECOMMENDATIONS The workshop discussion did not (nor was it intended to) propose testable hypotheses or specific research questions. Rather, it was meant to identify a broader set of research priorities that could be pursued by using data avail- able in the near term (e.g., through Medicare files) or data that could be col- lected relatively efficiently by independent effectiveness research projects. A more precise selection of topics depends on a more detailed assessment of data needs, on the one hand, and realistic data available, on the other. The following discussion describes the major groups of questions that the com- mittee felt an effectiveness research agenda might include. The committee recommended attention to four data and methodologic issues and three patient management issues. The former all involve research strategies that might be used to address the three patient management issues. The data and methodologic priorities call for the following: 1. improvement of the clinical and diagnostic information available in the HCFA data bases; 2. improvement of risk stratification methods, with particular attention to the relationship between these methods and the delineation of high- and low-risk subgroups, appropriate interventions for each, and resultant out- comes; 3. more effort to define outcomes, particularly those other than mortality, such as functional status and quality of life; and 4. further examination of variations in the patterns of care and the use of treatment modalities in different settings, for example, by type of facility, geographic region, or type of physician (i.e., specialist or generalist). The three patient management or clinical topics for early priority in effectiveness research are 1. selection of diagnostic and therapeutic procedures, particularly angio- plasty and invasive techniques or surgery; 2. selection of pharmacologic therapies, including thrombolysis; and 3. use of diagnostic tests and guidelines to stratify risk and guide treat- ment. In this context, "guide treatment" alludes to decisions about aggres- sive versus nonaggressive treatment, including no intervention. (By "aggressive," the committee referred to invasive diagnostic, therapeutic, and surgical interventions, and by "nonaggressive," to conservative medical and pharmacologic management.) Other patient management issues receiving attention in the discussion and in the second round of committee voting on key topics included primary prevention, prehospital care, rehabilitation, and physician education about clinical management.
ACUTE MYOCARDIAL INFARCTION METHODS ISSUES Improvement of Clinical Information in HCFA Data Bases 33 The committee developed two specific recommendations regarding clini- cal information in the HCFA data base. The first involves coding accuracy; the second, the validation of PRO data collection methods. The committee recommended the development of methods that will more accurately identify acute myocardial infarction patients for any effectiveness studies. These would involve both the nature and the extent of damage to the heart. The literature demonstrating a high incidence of false-positive diagnosis of AMI through the coding system, when compared to clinical documenta- tion, occasioned great concern among committee members. This problem is compounded by the recognized incidence of AMI that occurs without acute ECG changes, although evidence corroborating the infarction may well emerge after the acute hospitalization and therefore not be captured in He data coded for the acute event. The committee recognized the significant limitations of the ICD-9-CM4 classification (code number 410) for this purpose. Although improvements have been made in this area, considerable reassurance about accurate diag- nosis will be required before any effectiveness studies using these hospital data bases will have satisfactory credibility, at least to the practitioner com- munity. The committee also noted that the effect (or the incentives) of reim- bursement must be considered. For instance, reported increases of AMI in administrative databases may arise from efforts by hospitals to improve reimbursement levels by "up-coding." The current coding system is limited in its ability to adjust to new thera- pies that are critical in determining effectiveness. Thrombolytic therapy is considered a drug, and this is not currently captured in the inpatient data set. Also, when angioplasty was introduced as a frequent intervention, it was almost two years before codes were agreed upon. Therefore, a process must be established, perhaps assisted by a group of acknowledged experts in He field, that would identify state-of-~e-art changes and rapidly develop cod- ing to capture these types of critical information. The committee recommended clarification and validation of the PRO methodology for obtaining routine clinical data that may be essential to the DHHS effectiveness research program. The administrative data sets available to Medicare rely chiefly on clanns data. Abstracted clinical records reviewed by PROs may be another source 41ntemational Classification of Diseases, ninth revision, clinical modification.
34 EFFECTIVENESS INITIATIVE of information (see He discussion of He UCDS in the appendix). The pool of PRO-abstracted records, however, is based partly on a random sample and partly on samples of varying sizes of a wider array of cases that might have quality, utilization, or other problems. The validity and usefulness of the resulting data for research are subject to considerable question. For example, how random is the sample of AMI patients chosen for review? Because charts are reviewed retrospectively, can issues of dining and risk stratification be evaluated? The committee was concerned that the current design and implementation of data acquisition by the PROs would not be adequate to support certain elements of effectiveness research, such as the analysis of risk stratification. Independent validation of the methods is required to provide convincing recommendations about effectiveness to the scientific and business communities. In a related finding, the committee strongly urged HCFA (and DHHS, more broadly) to involve experienced clinicians in the development and analysis of data for the effectiveness program. Physicians show consider- able enthusiasm for the opportunity to develop clinical data that will support a determination of effectiveness, and every effort should be made to capital- ize on this interest. The committee was concerned with the perception that experienced clinicians in private and academic practice were not being ade- quately involved in certain elements of the effectiveness initiative within HCFA. For instance, evaluation of existing data bases and development of critical new data bases such as He UCDS appeared to be proceeding within HCFA with only limited participation and review from practicing clinicians. In addition, the committee concluded Hat He methods adopted for effective- ness evaluations, and the use of administrative mechanisms such as the PROs to collect data, should be subjected to peer review by qualified clini- cians and heals service researchers outside HCFA. Having stated the caveats noted above, however, He committee strongly supported HCFA initiatives to provide access to its data bases for qualified heals sciences investigators at minimal or no cost. The committee was convinced that such access to data files would allow creative investigators to identify new projects and important issues and problems Hat might not overwise be examined. Improvement of Risk Stratification Methods The committee recommended the funding of extramural research to validate stratification and triage decision methods that would permit determination of effectiveness in clinically important subsets of patients.
ACUTE MYOCARDIAL INFARCTION 35 Physicians use many methods to stratify risk and to determine acute and subsequent care for their AMI patients. They may apply these stratification methods systematically, or idiosyncratically and variably, depending on availability of tests or delay in access to testing. The methods themselves may be based on published information, data found in the individual patient record, or personal experience. Relevant clinical data include results of echocardiography, radionuclide studies, exercise tests, other physiologic information, and the presence of hypertension and other coexisting condi- tions. At the present time, none of these methods has been viewed as ideal for stratifying risk in the Medicare age group, and none by itself adequately explains the marked variations in mortality of elderly AMI patients. In the absence of a reliable and valid risk stratification method, review agencies will find it difficult to identify homogeneous subgroups of patients within which to measure the effectiveness of interventions. Patient stratification methods can vary across geographic locations, between rural and urban areas, between teaching and community hospitals, and among physician specialties. Thus, it is particularly important to assess how the "local" environment in which care is provided, which is essentially equivalent to local stratification schemes, may correlate with He appropri- ateness of the intervention. For example, clinical trials indicate that angioplasty provides no immedi- ate benefit in the early stages following thrombolytic therapy. Hence, the use of angioplasty in different settings could be followed to ascertain whether this procedure is being performed in elderly patients soon after ~rombolytics, despite the findings in these clinical trials; whether this uti- lization varies by setting; and what implicit or explicit local stratification method explains this variation. In sum, the variety of methods used by physicians to stratify risk and aid their therapeutic decisionmaking should be tracked because these methods may influence the use of interventions such as CCUs, thrombolytic therapy, angioplasty, or surgery, as well as the outcomes of these interventions. Particular attention should be paid to decisions on resource utilization that become inevitable once a particular risk has been determined. As noted with the angioplasty example, this research could begin with an analysis of existing data sets, identify "natural experiments" in the differing risk stratifi- cation methods that exist in practice, and then generate relevant clinical hypotheses that could be tested by ongoing surveillance or more targeted research, such as one or more "pilot" studies. The committee was con- vinced, however, that a prospective study will eventually be needed to pro- vide a sufficiently reliable, valid, and practical method for stratifying AMI patients that will allow appropriate and unambiguous interpretation of effec- tiveness data.
36 EFFECTIVENESS INITIATWE Definition of Outcomes Because definitions of outcomes are so central to effectiveness research the committee recommended further efforts to develop more sensitive and comprehensive measuring of the range of experiences patients may have. The committee proposed three related priorities: (~) more compre- hensive definitions of the elements of outcomes in line with recent devel- opments in the field of health status and quality-of-life measurement; (2) better definitions of alternative "best outcomes" (from the patient's point of new) that take health status and quality of life into account; and (3) expansion of techniques for acquiring data to measure outcomes for acute myocardial infarction. The committee recommended further that the Department of Health and Human Services give considerable priority to understanding and developing these methods for use throughout its effectiveness research effort (not just for acute myocar- dial infarction). Several committee members noted that survival alone may not be regard- ed by some patients (or their families or physicians) as the best outcome fol- lowing AMI. To the extent this is true, death or length of survival after AMI may be an incomplete indicator of outcome, because it gives no indi- cation of morbidity (including chest pain), functional status (including capacity for physical activity and activities of daily living), psychological and emotional well-being, social functioning, support networks, and general outlook on health status. For many patients, these measures may be more important than survival per se; that is, some patients may see the highest possible quality of life after AMI as the best outcome. For this reason, it becomes important to differentiate between outcomes and patient prefer- ences for outcomes. Information on patient preferences, given different potential outcomes, must be obtained. In general, the committee judged that these dimensions of health status should be viewed as independent of the need or use of health care services, although measures based on utilization could form part of more comprehensive sets of outcome variables. The committee recommended that the Department of Health and Human Services solicit outside expert opinion to define an adequate, appropriate set of outcome measures and to propose instruments for measuring these outcomes in effectiveness research both generally and for acute myocardial infarction. Consensus is emerging ~at, for health status and quality-of-life measure- ment, use of reliable and valid "generic" measures of health status coupled win selected "disease-speciD'c" measures is an appropriate, desirable, and practical research strategy. The committee concluded that existing HCFA
ACUTE MYOCARDIAL INFARCTION 37 data bases lack the necessary range of measures of functional status, although it was given to understand that HCFA is attempting to add a sim- ple, physician-based functional status measure to its claims forms. In the short run at least (or in the event that the measure selected for the claims form is inadequate), this information would have to be obtained directly from patients (or proxies). All in all, the committee believed that the appli- cation of patient-based measures of acceptable reliability and validity must be an integral part of effectiveness research over the longer term. With regard to AMI, the committee called attention to several specific concerns. First is the need to identify specific outcome measures such as exertional pain, exercise tolerance, recurrence of infarction, and congestive heart failure Hat must be captured in the longitudinal data base. Little change has occurred in outcomes such as mortality after introduction of "life-saving" interventions, and more specific outcomes of these treatments must be identified. The possibility that a variety of treatments may produce the same outcome must also be considered. The committee also noted that several common outcome indicators that might be valid in younger age groups, such as return to work or ability to drive a car, may not be valid or desirable in Medicare patients following AMI. Also, evaluation of exercise capacity or exertional pain may be thwarted by the inability of elderly patients to exercise or in some cases to feel pain in the same ways or to the same extent as younger patients. Thus, the committee strongly urged DHHS support to extramural work that will produce reliable and valid instruments for functional assessment and quality of life in the Medicare population not only win regard to acute myocardial infarction, but also as part of a broader charge to define outcomes satisfactorily for all projects carried out under the aegis of effectiveness. Examination of Variations in the Patterns of Care and Use of Treatment Modalities in Different Settings The committee recommended that work on understanding patterns of care be supported and be linked to issues relating to the three high priority patient management topics identified for the Effectiveness Initiative. A major dimension of effectiveness research is analysis of the way in which patterns of care change over time as a function of numerous epidemi- ologic, clinical, and health care system factors. These factors include geog- raphy, type of community, type of hospital or other facility, type of practice and physician specialty, cost, and reimbursement. This aspect of effectiveness research reflects the extraordinary concern about three general problems: wide, unexplained variations in population-based rates of use of
38 EFFECTIVENESS INITIATIVE services; high levels of use of inappropriate services and procedures; and differences in practice styles that appear to reflect areas of professional dis- agreement. The committee fully supported the use of HCFA databases, aug- mented as appropriate by information obtained from patient records, in this effort. With respect to AW, several factors discussed below should be part of a comprehensive set of analyses of patterns of care and variations. Many of these variations are cultural; others are matters of habit. For example, it was noted that physicians have a tendency to use many previously employed diagnostic tests, even when newer or better tests become available and com- mon practice. Type of hospital including size, teaching involvement, and role as prima- ry or referral facility may have a profound effect on the interventions that are available to the patient and on outcomes. Length of stay, care in special- ized units such as the CCU, use of remote monitoring, and access to bypass surgery or state-of-the-art technologies can be compared through data from facilities in similar geographic areas serving similar patients. Particular attention to those facilities that have made major investments in highly spe- cialized units employing invasive technology may be valuable. Differences in practices of different types of physicians may also prove instructive, particularly the care provided by generalists such as family physicians and internists compared with that rendered by cardiologists. Group practices, particularly those that try to influence clinical practices or that use formal clinical guidelines and specialty cardiology practices that use advanced technology, may demonstrate marked differences in patterns of care, patient outcomes, and costs relative to other types of practices. Geographic location of both patient and treating institution should be examined, particularly with regard to distance between the patient's home (perhaps by ZIP code) and hospital, the geographic region of the country (e.g., East versus West Coast), or location in a rural or urban area. Allowances must be made for differences in outcome that are influenced by the sophistication of evaluation and treatment before He patient reaches a hospital or by the circumstances surrounding a transfer from one hospital to another after initial evaluation and stabilization. Small area analysis may provide a valuable starting point for identification of geographic variations. These studies may not determine the effectiveness of the treatments employed, but they will be a useful starting point for detailed prospective studies. Of particular interest will be the identification of patients who did not have an invasive procedure, determination of why the procedure was not undertaken, and clarification of the relationship among length of stay, mor- tality, morbidity, and clinicians' use of tests to determine risk stratification in these patients.
ACUTE MYOCARDIAL INFARCTION PATIENT MANAGEMENT TOPICS 39 Selection of Diagnostic Tests and Therapeutic Procedures The committee recommended that research continue to be supported to identify the effectiveness of specific diagnostic and treatment modali- ties in the Medicare age group, whether alone or in groupings based on risk stratification methods. Further, it recommended that such investi- gations proceed from observation of variations of practice patterns and from surveillance of specific risk stratification strategies linked to appropriate outcomes. Workshop discussion touched repeatedly on the wide array of diagnostic tests and treatment procedures commonly used in AMI. These tests and procedures are often employed in "packages" or sequential groupings, depending on the relative risk assigned to the patient, as described earlier in the section on the clinical knowledge base. This complex armamentarium led to several general questions about the role of individual or combinations of diagnostic and therapeutic interventions in AMI. As a case in point, given the fakings of recent clinical trials, the choice of thrombolysis or angioplasty (or a combination) as the most appropriate treatment in many patients is a major decision point. These trials did not involve many patients in the Medicare age range (particularly individuals over age 75) and, therefore, may not be applicable to patients who are 75 or 85 years of age, for example. Nevertheless, the research findings are likely to be followed as a rationale for the treatment of Medicare patients. Because these important results may not apply as well to Medicare patients, however, the committee judged that the outcomes of thrombolysis and angioplasty must be observed and analyzed. For example, Medicare data could be ana- lyzed to document the rate of cerebrovascular accidents or other bleeding complications after the use of thrombolytics (or angioplasty, or both) in the elderly. Another concern was raised about the use of certain diagnostic tests such as catheterization, exercise testing, echocardiography, radionuclide imaging, scintigraphy, and more recently PET scanning and MRI for risk stratif~ca- tion or prognostic purposes. Although these procedures enjoy widespread use based on published research and local customs, there is no clear indica- tion of either efficacy or effectiveness in Medicare patients, and the cost and especially the risk to the elderly patient are not well understood. Furthermore, as newer modalities are added, older methods are not dropped, even though the purpose and value of these procedures has been supplanted. This adds a further complexity to the patient management topics warranting investigation through electiveness research.
40 EFFECTIVENESS INITIATIVE Selection of Pharmacologic Therapies The committee recommended that the use of pharmacologic agents, particularly thrombolytics, be monitored in both the inpatient and the outpatient settings. As for other interventions, individual drugs or packages of drugs, alone or in combination with invasive procedures, should be analyzed with regard to variations in outcomes. A variety of pharmacologic agents is used to treat AMI, either alone or in combination with other drugs or treatments. The drugs employed in the acute period are often superimposed on long-term medications used to treat ischemic heart disease both before and after infarction. Although the effica- cy and appropriate use of these medications is well described in the biomed- ical and clinical literature, how effective they are for AMI patients of all ages, especially the elderly, is less well documented. Beta blockers, nitrates, calcium blockers, nitroprusside, and anticoagulants are used in different combinations and sequences; choices of medications, sequences, and combi- nations vary by types of institutions, practitioners, and presenting clinical state and comorbidities of patients. Also, these agents may be used in con- junction with other interventions such as catheterization, angioplasty, and coronary artery bypass surgery, a phenomenon that adds another important layer of complexity to these questions of effectiveness. Of particular interest is the role of thrombolytic therapystreptokinase, tPA, or related compoundsin the Medicare population. Although He effi- cacy of these drugs in preventing death and reducing morbidity from AMI has been demonstrated recently in clinical trials, these studies did not include many elderly patients. Anecdotal experience of committee mem- bers suggested a similar benefit in the elderly under age 75, but the risks may be greater, and fewer elderly patients may present with a clinical pic- ture suitable for the use of thrombolytics. One useful tactic might be to identify cohorts of elderly AMI patients who received no interventions, those who receive only thrombolysis, those who received only invasive or other procedures, and those who received bow thrombolytic drugs and procedures. Another might be to obtain specif- ic data on the extent to which thrombolytic therapy is used in the Medicare population. Information about thrombolytic therapy as a function of patient age, location and type of provider, use of other pharmaceuticals and hospital setting would be important background information for future studies. The HCFA a~Tninistrative databases do not currently capture information on drug use In He ambulatory setiing.5 Because thrombolytics are catego- SAt the tune of dais workshop, the committee anticipated that better data on medication use would become available when the Medicare Catastrophic Coverage Act took effect. This legislation was reversed, but the need for these data did not diminish, and for effectiveness research purposes, the problem of data collection simply worsened.
ACUTE MYOCARDIAL INFARCTION 41 rized as a drug, their use as a major intervention in the early stage of hospi- tal or prehospital care does not appear in administrative discharge data. Consequently, the current Medicare databases will be of limited value in evaluating the effectiveness of pharmacologic or, specifically, thrombolytic therapy. To address these questions, therefore, effectiveness research will need to reach beyond insurance claims files to find several other types of informa- tion. For instance, the percentage of patients discharged with a transmural (Q wave) AMI who are treated with thrombolytic agents should (in princi- ple) be reasonably uniform across institutions. If this is so, then in the pres- ence of "equivalent" diagnostic ECG changes, the time between arrival and administration of thrombolytics should be indicative of the quality of care in a particular hospital. For example, if all patients in a given hospital received a ~rombolytic agent no earlier than three and one-half to four hours after first appearing in the emergency room, this would suggest inor- dinate delays in patient assessment and administration of a thrombolytic agent. Such information might also be useful in examining the effectiveness of different thrombolytics in relation to ECG changes. The routine use of other drugs or the uniform use of agents that are appropriate only under specific conditions would suggest excessive use of these pharmacologic agents. In some hospitals, patients with non-transmu- ral (non Q-wave) myocardial infarction would routinely undergo cardiac catheterization even with favorable response to medical treatment. In other settings the response to exercise on medical management would be used as an indicator for cardiac catheterization. Variations in current practices between regions and hospitals should lead practitioners, investigators, and policymakers to focus on strategic issues such as risk stratification when choices about employing therapeutic and diagnostic measures are to be made. Use of Diagnostic Tests and Guidelines to Stratify Risk and Guide Treatment The committee recommended that explicit attention be directed at ways to assign or stratify risk to individual patients with AM] as an ele- ment in the choice of diagnostic and therapeutic interventions. The role of risk stratification in choosing among diagnostic and therapeu- tic options for AMI patients is difficult to overstate. Approaches to risk stratification may be used (or not used), and well (or poorly) used, in patient care decision making every day across the nation. Effectiveness studies must, therefore, identify stratification methods presently in use (which, as
42 EFFECTIVENESS INITIATWE noted earlier, vary enormously) and correlate them with appropriate out- comes. Within risk groups, then, the outcomes of competing interventions could be measured to establish which methods best define patients for whom one treatment is more effective than another. Given some understanding of adequate risk stratification, variations in practice might serve as early "natural experiments." Trends in the outcomes of, for example, angioplasty or thrombolysis in the elderly could be identi- fied and associated with decision making based on different approaches to risk stratification. Such information might, over time, be fed back to physi- cians or otherwise disseminated, with the hoped-for result being progressive alteration in the practice of physicians, better use of good risk stratification methods, and the associated better interventions in appropriate patients. These comparative analyses could focus on the selection of appropriate aggressive treatment (versus conservative management) for patients identi- fied as being, for instance, at high (versus low) risk of death or complica- tions. More precise definitions of risk categories would provide a frame- work for surveillance of both existing and new therapies. The committee noted that such surveillance data should be obtained as a function of age and risk-adjusted for severity of illness and comorbidities. In short, proven risk stratification methods are essential for appropriate decision making about patient care. Because such methods are not well established, the committee concluded that developing such approaches was, in fact, a critical patient management issue both amenable to and neces- sary for effectiveness research. A model using stratification parameters based on these comparative analyses could be evaluated as a predictor of outcomes in another group of Medicare patients. If successful, these strati- fication methods could then be applied prospectively.
