Many institutions are developing innovative strategies to address this problem [7, 8], including immediate exercise treadmill testing [9, 10], early noninvasive imaging [11, 12], use of new markers for myocardial injury [13], and creation of emergency departments for chest pain [6, 14, 15]. Although some of these strategies have been evaluated [6], few data describe the expected or actual effects of most strategies on clinical outcomes or resource use. Nevertheless, short observation protocols for patients with chest pain are being implemented throughout the United States [14, 15].

Previous studies [16-18] addressed whether a patient who presents to an emergency department has unstable angina or acute myocardial infarction. An emerging paradigm instead asks whether it is safe to discharge the patient or whether it is safe for the patient to undergo an exercise treadmill test. This paradigm focuses attention on the safety of the management strategy rather than on the diagnosis because diagnosis often cannot be conclusively determined in the time frame available [19, 20].

We developed a comprehensive, evidence-based approach to management of patients with acute chest pain in the emergency department. We used published investigations to define the strategy and resolved differences or gaps in the literature by consensus of local experts. Quality management techniques were incorporated by using critical pathways to reduce variation in patient care [21-23].

Critical pathways define and manage the timing and sequence of the rate-limiting steps in the process of providing health care [21]. Management of this process includes simultaneously monitoring and giving feedback to providers of information about achievement of expected and unexpected patient outcomes. Health care providers can then adjust patient management to prevent or modify unexpected outcomes.

Critical pathways are distinct from clinical guidelines and algorithms [21]. Clinical guidelines define the appropriateness of care [21], and clinical algorithms define the step-by-step procedure for providing care, usually with conditional logic (if-then statements) [22]. Neither guidelines nor algorithms use simultaneous management of the process of care delivery to monitor compliance with recommended practice and identify opportunities for improvement.

The specific goals of our critical pathway were to reduce admission rates and total days of hospitalization among patients who were at low risk for complications of ischemic disease and to limit adverse outcomes. To address these goals, we developed an explicit sequence of care and recommended the daily availability of early exercise treadmill testing in low-risk patients. We retrospectively applied the criteria of the critical pathway to data from a large cohort of patients with acute chest pain to assess the potential effects of implementation of the pathway on resource use and adverse outcomes.

Methods

Top Methods Results Discussion Author & Article Info References

The Critical Pathway

Development

We developed the critical pathway on the basis of scientific evidence and the experience of local opinion leaders and physician-investigators with long-standing interest in the topic. Relevant articles from the medical literature were identified by doing a MEDLINE search for English-language articles from 1990 to 1995 using the keywords chest pain, unstable angina, and myocardial infarction combined with randomized controlled trial (PT) and cohort study. All identified papers were reviewed for relevance. Study quality was not explicitly evaluated because existing quality scales either have not been validated [24] or do not consider the value of Bayesian analyses [25]. Relevant articles were supplemented with a national clinical practice guideline for management of unstable angina [26].

Opinion leaders included staff emergency department physicians, staff cardiologists, and a cardiologist affiliated with a large health maintenance organization. An informal nominal group process was used to reach consensus [27]. The process consisted of a series of meetings and draft documents. Team members contributed recommendations based on the medical literature, and group discussion clarified and evaluated each recommendation. Team members independently reviewed the documents between meetings to comment on the work in progress. Consensus was achieved by unanimity.

Content

The flow of the critical pathway is shown in Figure 1. The pathway consisted of flowsheets with instructions and preprinted orders to facilitate use. Instructions provided suggestions to the end-user (usually an emergency department physician) about what information may facilitate decision making. The preprinted orders were designed to be selected by the physician in response to the instructions.

View larger version (25K): [in this window] [in a new window]   Figure 1. Critical pathway for management of patients with acute chest pain who are at low risk for myocardial ischemia.

Variance tracking was incorporated into the critical pathway to monitor the length of time that patients were present in the emergency department and the disposition and clinical outcomes of patients enrolled in the pathway [28, 29]. Time goals were suggested for the performance of key elements of the care process. Failure to meet these performance standards or occurrence of other explicit events were considered key variances, which may be reviewed later in aggregate to facilitate improvement of quality of care. We recommended recording the time interval between arrival in the emergency department and completion of the following tasks: 1) obtaining an electrocardiogram, 2) deciding to perform exercise treadmill testing, 3) obtaining the results of exercise treadmill testing, and 4) deciding to admit or discharge the patient.

Eligibility

Four clinical categories of patients were ineligible for enrollment in the critical pathway: patients with ongoing chest pain, patients with congestive heart failure present at initial clinical examination, patients with ischemia or injury present on the initial electrocardiogram (defined as ST-segment elevation, ST-segment depression, or T-wave inversion suggestive of ischemia or infarction and not known to be old), and patients judged by physicians to be at sufficiently low risk for myocardial ischemia to be discharged. At Brigham and Women's Hospital, Boston, Massachusetts, patients with active ischemia or infarction are treated by following other protocols [30]. Patients who clearly do not have ischemic chest pain are considered for direct discharge from the emergency department, as recommended by a national consensus panel [26].

Patients with left bundle-branch block and other electrocardiographic changes who might not be candidates for exercise electrocardiography were not explicitly excluded from eligibility for the pathway; patients could undergo alternative tests, such as exercise echocardiography or nuclear medicine studies, assuming that these technologies were available.

Data

Clinical, demographic, and outcome data were previously collected for a prospective cohort of consenting patients 30 years of age or older from the ongoing Chest Pain Study at Brigham and Women's Hospital [5, 31, 32]. These patients presented to the emergency department with chest pain and had no history of trauma or abnormal radiologic findings between 2 July 1990 and 18 February 1994. Inclusion criteria, definition of outcomes, and verification of outcomes for that cohort study are summarized below [5, 31, 32].

Clinical data were recorded by the emergency department physician at the time of presentation or by a research nurse before he or she had any knowledge of the patient's outcome. These data were recorded on a standardized form that was part of the permanent medical record. A study nurse reviewed the charts of all admitted patients to determine subsequent outcome.

Reviewers who were blinded to baseline clinical data classified patients into three categories on the basis of their final discharge diagnoses [5, 33]: acute myocardial infarction, unstable angina, or other. Acute myocardial infarction was diagnosed in admitted or discharged patients on the basis of characteristic evolution of serum enzyme levels, electrocardiographic changes, or sudden unexpected death within 72 hours of presentation [16]. Acute myocardial infarction was also diagnosed in 1) patients who did not meet these criteria but received acute reperfusion therapy with intravenous thrombolytic agents or primary percutaneous angioplasty and 2) patients who had new ST-segment elevation that evolved over the next day and had total occlusion of the infarction-related artery, an echocardiographic wall-motion abnormality that corresponded to the acute electrocardiographic changes, or an elevated total creatine kinase level and a creatine kinase-MB isoenzyme level greater than 2.5% of the total creatine kinase level, with characteristic evolution. Unstable angina was diagnosed if the original chest pain that led the patient to present to the emergency department was described by the patient as similar to or worse than previous symptoms of angina or myocardial infarction or if unstable angina was diagnosed by the senior clinician associated with the case. Other cardiac diagnoses, such as congestive heart failure or arrhythmias, and noncardiac diagnoses were made by the senior clinician involved with the case.

Life-threatening complications were defined as nonfatal ventricular fibrillation, new type II second-degree atrioventricular block, new complete heart block, new atrioventricular dissociation, pulmonary edema, infarction extension, cardiogenic shock, or cardiac arrest. Other complications included pulmonary emboli, congestive heart failure, and recurrent ischemic pain.

Analyses

Definitions

To analyze the potential effects of implementation of the pathway on resource use, ongoing pain was defined as pain that occurred up to 2 hours after arrival in the emergency department. Early recurrent pain was defined as pain that occurred 2 to 6 hours after arrival in the emergency department. Patients who described pain that was similar to that of a previous myocardial infarction or worse than previous angina were classified as requiring direct admission and were thereby ineligible for the critical pathway.

Twelve-Hour Observation Period

To calculate the potential reduction in admissions, we assumed that all patients who were actually admitted, classified as eligible for the pathway, and had a benign clinical course at 12 hours would be discharged from the emergency department at the end of the observation period. The potential relative reduction in admissions was then calculated by dividing the number of potentially discharged patients by the total number of actual admissions.

To calculate the potential reduction in hospital days, we assumed that the days of hospitalization for all patients who were actually admitted, were eligible for the pathway, and had a benign clinical course at 12 hours would be saved. The potential relative reduction in hospital days was then calculated by dividing the actual days of hospitalization for these patients by the total days of hospitalization for all admitted patients.

To assess the robustness of potential reductions in resource use to changes in assumptions about use of the critical pathway, the potential reduction in resource use was calculated as if not all patients would be discharged even if they were stable after initial observation and evaluation [10]. The potential reduction in resource use was also calculated as if some patients who were eligible for the pathway were not enrolled [8]. Finally, because some patients who were actually discharged met criteria for further observation, reductions in resource use were calculated as if these patients had been admitted to the hospital.

We compared the incidence of outcomes for patients who remained on the pathway at 12 hours and those who were ineligible for the pathway. Univariate differences in categorical variables were assessed by using chi-square tests, and univariate differences in continuous variables were assessed by using t-tests.

Six-Hour Observation Period

We calculated the potential reduction in resource use as if all patients who were eligible for the critical pathway and were stable after 6 hours of observation were discharged directly from the emergency department. Clinical outcomes were compared for admitted patients who were eligible to remain on the pathway at 6 hours and those who were not eligible for the pathway.

Results

Top Methods Results Discussion Author & Article Info References

Recommendations for Management during the First 6 Hours after Arrival in Emergency Department

The duration of the initial observation period before performance of noninvasive testing in patients with acute chest pain and low risk for myocardial ischemia has varied widely. Observation periods of 9 to 12 hours have been proposed [15, 32]. However, immediate exercise testing (median < 1 hour) in low-risk patients is safe and effective [9, 10]. Therefore, we proposed an intermediate target observation period of 6 hours for patients who were able to exercise.

The type and timing of markers for myocardial injury to be included in the pathway were determined on the basis of the diagnostic performance of each test. A single measurement of cardiac enzyme levels or other markers is not enough to rule out myocardial infarction in higher-risk patients with chest pain [19]. However, patients who are eligible for the critical pathway have a rate less than 5% for myocardial infarction [16]. Use of a single electrocardiogram, creatine kinase level, and creatine kinase-MB level has been advocated for establishment of a diagnosis in low-risk patients with acute chest pain [34].

Bayesian analysis showed the diagnostic value of early measurement of the creatine kinase-MB level by combining the prior probability of disease (determined on the basis of patient risk) with the sensitivity and specificity of the test to reflect the post-test probability of disease [35]. A single creatine kinase-MB measurement taken at least 3 hours after onset of pain has a sensitivity of 76% for detecting acute myocardial infarction and a specificity of 72% for detecting the absence of acute myocardial infarction [36]. If a patient who is initially thought to be at low risk for acute myocardial infarction (that is, a prior probability of 5%) has a benign clinical course, including a negative result on creatine kinase-MB testing during a 6-hour observation period, the post-test probability of acute myocardial infarction is low enough to allow consideration of noninvasive testing (1.9% [95% CI, 0.9% to 3%]). Conversely, the occurrence of a clinical end point (such as recurrent ischemic pain, congestive heart failure, arrhythmia, or a positive early result on creatine kinase-MB testing) indicates that the post-test probability of acute myocardial infarction is high enough to warrant longer observation and more intensive therapy (13% [CI, 9.7% to 17%]).

The role of newer markers for myocardial injury, such as troponin-T [37], troponin-I [38], creatine kinase isoforms [13], or myoglobin [39], was considered unclear. These markers were considered promising and potentially even better than creatine kinase-MB level. Troponin-T [37] and troponin-I [38] can be considered a reasonable alternative to creatine kinase-MB, but no controlled trials have shown the effects of their use on clinical outcomes. Therefore, a single normal creatine kinase-MB level obtained at least 4 hours after the onset of pain was used in the pathway to complement clinical observations and identify low-risk patients to undergo exercise testing.

We propose criteria for taking a patient off the pathway and formally admitting the patient to a monitored bed (Table 1). Conversely, if no complications occur during the 6-hour observation period, patients should undergo an exercise tolerance test as soon as possible after the observation period is complete.

Two small cohort studies [9, 10] evaluated the safety and sensitivity of exercise testing in patients who present with acute chest pain. Among patients with atypical chest pain, no patient with a normal result on an exercise tolerance test had a cardiac event during the 6 months after evaluation [9]. Lewis and Amsterdam [10] observed no complications after immediate exercise testing in low-risk patients admitted to the hospital with suspected acute myocardial infarction. In their study, 50% of patients had positive results on exercise electrocardiography; 50% of those patients had significant coronary artery disease documented by angiography. One percent of patients had uncomplicated myocardial infarction, and 1% had late acute myocardial infarction. Therefore, exercise tolerance testing seems to be a safe and sensitive diagnostic strategy in patients who present to an emergency department with acute chest pain.

Although the responsible physician must manage the patient as he or she believes is appropriate, we made recommendations for triage after the exercise test has been performed (Table 2). Specific arrangements were made for the availability of short-term follow-up of patients at the ambulatory general internal medicine practice of the hospital for patients who were discharged from the emergency department and who lacked a regular source of primary care [43]. Patients with a source of primary care called their own physician for follow-up because their insurers may not have reimbursed another visit to the hospital. However, no patient was turned away from the walk-in program. Follow-up arrangements deserve special emphasis because the long-term prognosis of patients discharged from the emergency department with acute chest pain who have a known history of coronary disease is similar to that for patients admitted with angina or myocardial infarction [31].

The development team discussed which patients with probable coronary disease can be safely discharged from the emergency department. It was generally agreed that not all patients with coronary disease require admission or continued observation, particularly if ischemia does not develop at low levels of exertion (stage I). A national consensus panel made similar recommendations [26]. Such patients may have follow-up as outpatients within 72 hours. However, our consensus was that the final disposition should be at the discretion of the attending physician rather than according to strict guidelines.

Recommendations for Management 6 to 12 Hours after Arrival in Emergency Department

Patients who cannot exercise or whose electrocardiograms are likely to make interpretation of exercise electrocardiographic results difficult undergo a longer period of observation according to our critical pathway. Patients with acute chest pain who are at low risk for ischemic heart disease and who remain stable after a 12-hour observation period were previously shown to have a 0.5% incidence of myocardial infarction and a 0.4% incidence of cardiac arrest [32]. This incidence compares favorably with a 1.4% incidence of myocardial infarction among patients discharged from the emergency department [19]. Therefore, we concluded that patients who are unable to exercise should be observed for at least 12 hours. We propose criteria for taking such patients off the pathway and formally admitting them to a monitored bed (Table 1). At the end of the observation period, clinicians make decisions about further testing (for example, with pharmacologic radionuclide tests) and discharge the patient according to their clinical judgment.

Again, Bayesian analysis shows the diagnostic value of a 12-hour observation period. A strategy that combines an observation period with two sequential creatine kinase-MB immunoassays done at least 2 hours apart has a sensitivity of 94% and a specificity of 91% for acute myocardial infarction [36]. If a patient who is initially thought to be at low risk for myocardial infarction (5%) has a benign clinical course during the observation period, the probability of myocardial infarction decreases to 0.25% (CI, 0.05% to 1.1%). Conversely, if such a patient reaches an end point (such as recurrent ischemic pain, congestive heart failure, arrhythmia, or elevation of creatine kinase-MB levels on one of two assays) during the initial 12-hour observation period, then the probability of myocardial infarction increases to 36% (CI, 28% to 45%).

We thought that evidence supporting the use of noninvasive nuclear imaging [44] or early echocardiography [45] was not sufficient to justify routine inclusion in the pathway because the impact of these technologies on triage decisions is unproven.

Patients who cannot exercise but have a benign clinical course and remain on the pathway at the end of 12 hours can be considered for discharge with short-term outpatient follow-up.

Potential Effects of the Critical Pathway

Characteristics of Patients in Chest Pain Study

Of a total of 4705 patients enrolled in the Chest Pain Study, 91 (1.9%) were excluded from analysis because their descriptions of chest pain were unclear. Twenty-nine patients (0.6%) were excluded from analysis because interpretations of their electrocardiograms were not available. Of the 4585 remaining patients, 2898 (63%) were admitted to the hospital, and 1687 (37%) were discharged from the emergency department. The demographic characteristics of patients in the Chest Pain Study are summarized in Table 1. Few patients had comorbid illnesses. The overall mean length of stay among patients who were admitted to the hospital was 4.5 ± 5.8 days. A total of 2117 (45%) of admitted patients had either acute myocardial infarction or unstable angina.

Among patients who were admitted, 1746 (60% [38% of all patients]) were ineligible for the critical pathway because they described their pain as similar to that in a previous myocardial infarction or worse than usual angina, had ongoing pain, had congestive heart failure on clinical examination, or had changes on electrocardiography that were consistent with ischemia or infarction and not known to be old. These excluded patients had a mean length of stay of 5.5 ± 6.1 days.

Potential Effects of the 12-Hour Observation Period

Of the 1152 admitted patients who were eligible for the critical pathway, 84 (2.9% [1.8% of all patients]) had recurrent pain, elevated levels of creatine kinase-MB isoenzymes, or a complication during the first 12 hours after arrival in the emergency department (Figure 2). The remaining 1068 admitted patients (37% [93% of eligible patients and 23% of all patients]) were stable after at least 12 hours of observation and had a mean length of stay of 2.8 ± 4.8 days (P < 0.001 compared with mean length of stay of admitted patients who were ineligible for the pathway).

View larger version (18K): [in this window] [in a new window]   Figure 2. Classification of patients by application of criteria for 12-hour observation period.

If only 47% of patients eligible for early exercise testing were discharged from the hospital after completion of the test [10], the critical pathway would potentially reduce admissions by 505 (17%) and days of hospitalization by 1407 (11%). If only 69% of eligible patients were enrolled in the critical pathway, the pathway would potentially reduce admission rates by 346 (12%) and days of hospitalization by 971 (7%) [8].

Among patients who were discharged, some met criteria for further observation on the basis of initial presentation (n = 96 [5.7%]) or occurrence of an end point (n = 5 [0.3%]) while they were still in the emergency department. Of these 101 patients, 1 (1.0%) had a final diagnosis of myocardial infarction and none had a final diagnosis of unstable angina. Even if these patients had been admitted and had had a length of stay of 2.8 days, the projected net decrease in admissions would be 457 (16%) and the decrease in days of hospitalization would be 1125 (9%).

Among the 1068 patients who were stable after at least 12 hours of observation, 4 (0.4%) had acute myocardial infarction (P < 0.001 for comparison with 306 ineligible patients [18%]) (Table 3), and 163 (15%) had unstable angina (P < 0.001 for comparison with 793 ineligible patients [45%]).

Life-threatening complications occurred in only 5 patients (0.5%) who remained on the pathway after 12 hours of observation (P < 0.001 for comparison with ineligible patients) (Table 4). Other complications occurred in 74 patients (6.9%) (P < 0.001). One patient who was admitted but eligible for early exercise treadmill testing and possible early discharge underwent coronary bypass grafting after admission and then had cardiac arrest.

Potential Effects of the 6-Hour Observation Period

Figure 3 shows the classification of patients using the criteria of the critical pathway with a 6-hour observation period. Of the 1152 patients admitted to the hospital who were eligible for the critical pathway, 67 (2.3% [1.5% of all patients]) had recurrent pain, elevated levels of creatine kinase or creatine kinase-MB enzymes, or a complication during the first 6 hours after arrival in the emergency department. The remaining 1085 admitted patients (37%) remained on the pathway after 6 hours of observation and were eligible for early exercise treadmill testing. These patients had a mean length of stay of 2.8 ± 4.8 days (P < 0.001 for comparison with length of stay among admitted patients who were ineligible for the pathway). Thus, in the unlikely event that all of these patients were discharged, the admission rate would decrease by 37% and days of hospitalization would decrease by 3067 (24%).

View larger version (20K): [in this window] [in a new window]   Figure 3. Classification of patients by application of criteria for 6-hour observation period.

Among the patients who were stable after at least 6 hours of observation, 13 (1.2%) had acute myocardial infarction (P < 0.001 for comparison with 306 ineligible patients [17.5%]), and 163 (15%) had unstable angina (P < 0.001 for comparison with 793 ineligible patients [45.4%]). Life-threatening complications occurred in only 5 patients (0.5%) (P < 0.001 for comparison with 102 ineligible patients [5.8%]). Other complications (recurrent ischemia or congestive heart failure) occurred in 76 patients (7.0%) (P < 0.001 for comparison with 542 ineligible patients [31.0%]).

Of the patients excluded from the analysis because of missing data, 75 were admitted and had a mean length of stay of 4.2 ± 4.2 days. Classification of these admitted patients as either eligible or ineligible did not substantially change the potential effect of the pathway on resource use.

Potential Economic Effects of Pathway Implementation

In the United States, about 2 million persons are admitted to the hospital each year to rule out a diagnosis of myocardial infarction [10]. The average cost of hospitalization of low-risk patients is $5000 [5]. If hospitalizations were reduced by 17%, the potential savings from implementation of the critical pathway would be more than $500 million annually.

Discussion

Top Methods Results Discussion Author & Article Info References

We developed a critical pathway for management of low-risk patients with chest pain in the emergency department to facilitate rapid determination of whether it is safe to discharge the patient or safe for the patient to complete an exercise treadmill test. Retrospective analysis suggests that implementation of the critical pathway may be associated with clinically important reductions in rates of admission and days of hospitalization.

Several methods are available for achieving consensus in the development of clinical guidelines, and there is no agreement about which method is correct [27]. We used an informal method to combine evidence with local opinion. More formal methods offer more structure but may be considerably more onerous [27]. National practice guidelines for management of acute chest pain were also developed with comprehensive literature review and evaluation of the quality of scientific evidence [26]. However, the effectiveness of guidelines developed by external bodies may be limited by lack of local availability of the recommended technology [46]. Conversely, including local opinion leaders increases compliance with a practice guideline [47] and may ensure that recommendations reflect the local availability of technology.

Other guidelines for management of patients with acute chest pain have had variable results [4, 48, 49]. Because the effects of clinical guidelines are not consistent [50], we included multiple interventions in the critical pathway to change physician practice: the input of local opinion leaders, standard orders, audit and feedback of care providers, and increased availability of a diagnostic test. In effect, this pathway substitutes a relatively inexpensive test (exercise testing) for a more expensive test (additional time in the hospital).

The methods that we used to develop the critical pathway have some limitations. First, other tests or observation periods may be equivalent to those that we recommend. Second, we did not formally evaluate the quality of each study. Finally, the individual patients may have soft findings that increase risk but are not considered in the critical pathway. However, the individual physician can respond to such cues because he or she makes the final decision about patient management.

Although few data describe the effects of critical pathways in the emergency department, a case–control study of the introduction of management protocols in the inpatient setting was associated with a 15% reduction in resource use [51]. This reduction is consistent with our estimate of the potential effect of implementation of the pathway in the emergency department for patients with acute chest pain who are at low risk for ischemia. Uncontrolled studies [52-55] of the effect of critical pathways for hospitalized patients suggest that pathways improve efficiency and quality. However, Falconer and colleagues [56] did not find improved length of stay or cost with controlled introduction of a critical pathway to inpatient care. No controlled studies show that the availability of critical pathways reduces length of stay or use of resources [21]. Collectively, studies show that it is essential to conduct controlled evaluations of critical pathways.

Our evaluation of the potential effects of the critical pathway also has some limitations. First, a relatively high proportion of patients had a complication after 6 hours of observation. Most complications were recurrent ischemia. Because our pathway uses an exercise test at the 6-hour time point, it is likely that these patients would be identified and admitted. Because the safety and sensitivity of early exercise tolerance testing has not yet been evaluated in a large prospective study, further evaluation is necessary to examine the role of early testing. In addition, potential cost reductions may be partly offset by the increased cost of observing patients in the emergency department or making exercise tolerance testing available daily. Because the incremental cost of performing exercise tolerance testing in the emergency department is less than $500 [57], it is unlikely that the costs of increased exercise tolerance testing would exceed the savings that result from decreased hospitalization rates.

In summary, we developed a critical pathway for patients with acute chest pain that is based on published research and expert opinion to improve quality and efficiency of care for patients at low risk for complications of ischemic heart disease. Retrospective analysis suggests that this critical pathway may substantially reduce resource use. Prospective study is needed to ensure that efficiency increases without an increase in adverse outcomes.

Dr. Walls: Department of Emergency Medicine, Brigham and Women's Hospital, 75 Francis Street, Boston, MA 02115.

Dr. Goldman: Department of Medicine, University of California, San Francisco, 505 Parnassus Avenue, San Francisco, CA 94143-0120.

Dr. Pearson: Division of Ambulatory Care and Prevention, Harvard Pilgrim Health Care, 126 Brookline Avenue, Boston, MA 02115.

Drs. Hartley, Antman, Teich, Cannon, Johnson, and Kuntz: Department of Medicine, Brigham and Women's Hospital, 75 Francis Street, Boston, MA 02115.

Dr. Stockman: Harvard Community Health Plan, 1 Fenway Plaza, Boston, MA 02215.

Dr. Lee: Partners Community HealthCare, Inc., Prudential Tower Suite 1150, 800 Boylston Street, Boston, MA 02199-8001.