Consultation Between Cardiologists and Generalists in the Management of Acute Myocardial Infarction

Background: The rapid expansion of managed care in the United States has increased debate regarding the appropriate mix of generalist and specialist involvement in medical care. Objective: To compare the quality of medical care when generalists and cardiologists work separately or together in the management of patients with acute myocardial infarction (AMI). Methods: We reviewed the charts of 1716 patients with AMI treated at 22 Minnesota hospitals between 1992 and 1993. Patients eligible for thrombolytic aspirin, b -blockers, and lidocaine therapy were identified using criteria from the 1991 American College of Cardiology guidelines for the management of AMI. We compared the use of these drugs among eligible patients whose attending physician was a generalist with no car-diologist input, a generalist with a cardiologist consul-tation, and a cardiologist alone. Results: Patients cared for by a cardiologist alone were younger, presented earlier to the hospital, were more likely to be male, had less severe comorbidity, and were more likely to have an ST elevation of 1 mm or more than gen-eralists’ patients. Controlling for these differences, there was no variation in the use of effective agents between patients cared for by a cardiologist attending physician and a generalist with a consultation by a cardiologist. However, there was a consistent trend toward increased use of aspirin, thrombolytics, and b -blockers in these patients compared with those with a generalist attending physician only ( P , .05 for b -blockers only). Differences between groups in the use of lidocaine were not statistically significant. The adjusted probabilities of use of thrombolytics for consultative care and cardiologist attending

was no variation in the use of effective agents between patients cared for by a cardiologist attending physician and a generalist with a consultation by a cardiologist.However, there was a consistent trend toward increased use of aspirin, thrombolytics, and ␤-blockers in these patients compared with those with a generalist attending physician only (PϽ.05 for ␤-blockers only).Differences between groups in the use of lidocaine were not statistically significant.The adjusted probabilities of use of thrombolytics for consultative care and cardiologist attending physicians were 0.73 for both.Corresponding probabilities were 0.86 and 0.85 for aspirin and 0.59 and 0.57 for ␤-blockers, respectively.
Conclusions: For patients with AMI, consultation between generalists and specialists may improve the quality of care.Recent policy debates that have focused solely on access to specialists have ignored the important issue of coordination of care between generalist and specialist physicians.In hospitals where cardiology services are available, generalists may be caring for patients with AMI who are older and more frail.Future research and policy analyses should examine whether this pattern of selective referral is true for other medical conditions.Med.1998;158:1778-1783   T HE RAPID expansion of managed care in the United States has increased public and scientific debate regarding the appropriate mix of generalist and specialist involvement in medical care.Previous research suggests that cardiologists have better knowledge than generalists concerning efficacious therapies in the treatment of acute myocardial infarction (AMI). 1 However, previous studies 2 also suggest that there is little relationship between knowledge or self-reported practice and actual behavior.Despite intense controversy, few studies 3 exist comparing the care provided to comparable patients by generalists and specialists.

Arch Intern
The focus on comparative performance of generalists and specialists ignores opportunities for sharing of knowledge and experience through formal and informal consultation.In 1 study, 4 the quality of psychoactive drug prescribing in nursing homes was highest among generalists who reported frequent consultations with psychiatrists.A recent study 5 of a US health maintenance organization showed that an intervention fostering collaborative care between generalists and psychiatrists improved adherence to antidepressant regimens, patient satisfaction, and other outcomes in patients with major depression.
In contrast to previous studies 1,6,7 that used survey data to measure differences be-

ORIGINAL INVESTIGATION
The affiliations of the authors are listed in the Acknowledgments section at the end of the article.
tween generalist and specialist care, we investigated differences in quality of care actually provided to patients with AMI when generalists and cardiologists work separately and in consultation.Quality of care was defined as care consistent with nationally recognized evidence-based practice guidelines.Specifically, we examined the use of drugs known to reduce morbidity and mortality in eligible patients (aspirin, thrombolytics, and ␤-blockers), [8][9][10][11] and nonindicated use of lidocaine, which may cause increased mortality. 12We compared use of these drugs among patients whose attending physician was (1) a generalist with no cardiologist input, (2) a generalist with cardiologist consultation, and (3) a cardiologist.

RESULTS
Seventeen (85%) of the 20 hospitals had similar distributions of involvement of generalists and specialists in the management of patients.Two hospitals made little use of cardiologist consultations and had no patients treated by a cardiologist attending physician.At 1 hospital, almost all patients with AMI were under the care of a cardiologist attending physician.
A description of patient characteristics, stratified by level of generalist and specialist involvement in the care of that patient, is provided in Table 2. Patients cared for by a cardiologist were younger and had less severe comorbidity.They were more likely to be male, to be taking lipid-lowering medications prior to admission, to have presented to the hospital within 6 hours of onset of symptoms, and to have electrocardiographic findings clearly indicating AMI (ie, ST elevation Ն1 mm) (Table 2).
Controlling for differences in patient characteristics, cardiologist involvement in care, whether as a consultant or as the attending physician, was associated with a statistically significant increase in the use of ␤blockers only, with a consistent trend toward increased

PATIENT SAMPLE
The patient sample was drawn from a larger study of 37 hospitals participating in a randomized controlled trial to improve quality of care in the treatment of AMI in community hospitals. 13Data reported herein were collected prior to the intervention.In this study we included 20 hospitals in Minneapolis, St Paul, Duluth, and St Cloud, Minn, which had both cardiologists and generalists on staff.We excluded the 17 rural hospitals that did not have a cardiologist on staff because of confounding of the characteristics of rural practice and cardiologist availability.The 20 study hospitals represented 91% of the number of community hospital beds (adults) in the 4 urban areas.
The medical records of 1839 patients admitted to the 20 hospitals for AMI between September 1992 and August 1993 were abstracted by trained nurse-abstractors.We identified patients with an admission diagnosis of AMI or suspected AMI who met 2 or more of the following 3 criteria 14 : (1) clinical signs and symptoms consistent with AMI (arm or shoulder pain, chest pain, diaphoresis, dyspnea, nausea or vomiting, or neck/jaw pain); (2) electrocardiographic evidence of AMI; or (3) laboratory evidence of AMI (elevated levels of serum creatine kinase and its isoenzyme MB subfraction).We excluded patients who had sustained a myocardial infarction in the 2-week period prior to the present hospital admission, were dead on hospital arrival, or were transferred from a nonstudy hospital.
For each drug, a subsample of eligible subjects was identified, based on criteria developed from the 1990 American College of Cardiology/American Heart Association (ACC/ AHA) guidelines for the management of AMI, which were in effect in the study period. 15A detailed description of the eligibility criteria for each target therapy is listed in Table 1.

PHYSICIAN SAMPLE
We recorded the unique identification code and specialty of the attending physician, along with evidence of any cardiologist consultation.We defined 3 categories of generalist and cardiologist involvement in patient care: (1) generalist only, if the attending physician was an internist, family physician, or general practitioner, and there was no cardiologist consultation during the hospitalization; (2) consultative care, generalist attending with cardiologist consultation; and (3) cardiologist attending physician alone.In 6.7% of cases (n = 123), the attending physician was some other specialist and was excluded from the analysis.

DATA QUALITY
A complete description of the overall quality control methods have been reported elsewhere. 13Interrater agreement concerning the identity of the attending physician was 86%.

DATA ANALYSIS
To determine whether patient characteristics were similar, we compared several demographic and clinical characteristics of patients across the 3 categories of generalist/ cardiologist involvement (Table 2).
Next, in separate models, we regressed the use of each study drug on the category of generalist/specialist involvement using binary logistic regression analysis, 16 modeling the 3 groups with 2 indicator variables, and controlling for potentially confounding covariates.Covariates included patient age, sex, presence of severe comorbidity (Greenfield and coworkers' Index of Coexistent Disease 17 ), time from onset of symptoms to presentation, use of lipid-lowering drugs prior to admission, method of presentation (emergency medical services transport, emergency department, or other), location of AMI (anterior, inferior, or other), and presence of ST elevation higher than 1 mm.Patient age and sex were included in all statistical models.Other control variables were retained if for the Wald statistic PՅ.10 in all models.Age of the attending physician was initially included in the model, but dropped because it was not a significant predictor of use of the study drugs.Finally, we adjusted for the hospital-level nesting of the binary dependent variable using generalized estimating equations. 18se of other effective medications (Table 3 and Figure 1).Compared with generalist attending physicians with no cardiologist input, odds ratios for use of aspirin for consultative care and cardiologist attending physicians alone were 1.5 (95% confidence interval [CI], 1.0-2.2) and 1.4 (95% CI, 0.8-2.6),respectively.Odds ratios were 1.9 (95% CI, 1.0-3.9)and 2.0 (95% CI, 0.8-4.8),respectively, for thrombolytic agents; 2.3 (95% CI, 1.3-3.9)and 2.1 (95% CI, 1.1-3.9),respectively, for ␤blockers.To provide a better representation of the actual differences in prescribing of study medications, we converted odds ratios into adjusted probabilities, controlling for patient level covariates (Figure 1).The adjusted probability of use of thrombolytics for consultative care and cardiologist attending physicians was 0.73 for both.Corresponding probabilities were 0.86 and 0.85 for aspirin; 0.59 and 0.57 for ␤-blockers, respectively.
There was a trend toward increased use of lidocaine with increasing cardiologist involvement, but this was not statistically significant (Table 3; 95% CI, 0.6-2.6 and 0.6-4.7 for cardiologist consultation and cardiologist attending physician, respectively).Because this trend was unexpected, we conducted further analyses of specialty-associated variations in the use of study drugs at the hospital level.
When we plotted levels of use of aspirin, thromboagents, and ␤-blockers by hospital, the pattern of increased use of these effective agents among patients in the consultative care and cardiologist-treated groups was consistent across all hospitals.On the other hand, most of the increase in use of lidocaine was concentrated among 14 cardiologists in 2 hospitals (Figure 2).In these 2 hospitals, lidocaine was given along with thrombolytic therapy in 92% of cases and without thrombolytic agents in 29% of cases.Corresponding figures for other hospitals were 60% and 20%.Thus, the questionable use of lidocaine in these hospitals may have been due to prophylactic use of lidocaine for arrhythmias associated with reperfusion.

COMMENT
Overall, our results reinforce previous evidence suggesting that consultation between generalists and specialists results in improved quality of care.In our study, the quality of consultative care for patients with AMI, measured as the proportion of eligible patients who received highly effective drugs, was almost identical to the quality of care under a cardiologist attending physician, and somewhat higher than the quality of care provided by generalists without cardiologist consultation.
Our data are generally consistent with the survey data of Ayanian and colleagues, 1 who found that cardiologists were more willing to prescribe thrombolytic agents, aspirin, and ␤-blockers for patients with AMI when compared with internists or family physicians.However, our data on actual practice in the sample of physicians in urban and suburban hospitals in Minnesota suggest smaller differences between generalists and cardiologists in the use of these drugs.Also, our estimates of drug use are somewhat lower than the selfreported behavior, 1 after adjusting for important clinical variables.Differences between these studies suggest a shortfall in the translation of knowledge into action.To some extent, the differences may be due to social desirability bias on the part of respondents to surveys.However, we may also have failed to capture some aspects of the patients' presentation that caused physicians to refrain from using these drugs.
Our findings regarding use of lidocaine differ qualitatively from those of Ayanian and colleagues, 1 who reported that cardiologists were less likely to prescribe prophylactic lidocaine.We found a trend toward increased use of lidocaine for nonindicated purposes among cardiologists when compared with generalists.The trend seemed to be explained by a higher rate of inappropriate use among cardiologists in only 2 hospitals.The marked difference in use of lidocaine at these 2 sites presumably reflected the common practice, among cardiologists at these institutions, of routinely administering prophylactic lidocaine for the possible development of arrhythmias associated with thrombolytic infusion.
We observed important patient differences across the 3 categories of generalist and cardiologist involvement in patient care, consistent with previous studies. 19,20Overall, generalists had older, more frail patients.These results question the assumption that cardiologists care for sicker and more complex patients, at least in the setting of AMI.One hypothesis is that pa-tients with the best prognosis for survival are more likely to be referred to specialists. 19In another report, 21 we found that otherwise eligible patients with more severe comorbidity were less likely to receive aspirin and thrombolytic therapy.Alternatively, this observed difference may be because younger, healthier patients are less likely to have an established relationship with a generalist physician and are, therefore, assigned to a cardiologist.Future studies should examine whether this pattern of selective referral exists for other medical conditions.

LIMITATIONS AND ALTERNATE EXPLANATIONS
As in any observational study, our results may be biased by variables that are either difficult to measure or are unmeasured.For example, in our study generalists cared for patients with more severe comorbidity, which is independently associated with reduced use of study medications and increased mortality risk. 21If we adjusted incompletely for comorbidity, our results would be biased toward greater specialty-associated differences in use of the study drugs.This confounding may be even more serious in studies comparing postmyocardial infarction mortality rates that did not collect information on severe comorbidity. 20Only a randomized controlled trial could adequately address these limitations.However, random allocation of patients to different levels of involvement of generalists and cardiologists would not be feasible.
The presence of a cardiologist may influence quality of care in several ways other than through direct consultation.For example, cardiologists may have been instrumental in developing evidence-based protocols for the management of AMI at individual hospitals.Also, a generalist who had previously consulted with a cardiologist for a similar case would not necessarily consult formally with the cardiologist over the same matter in the next case.We have not included these considerations in our model.
An additional limitation is that our study findings are derived from a retrospective chart audit of the medical record.While richer in detail than hospital discharge and claims data, 20 it is possible that important factors relating to the use of these drugs may not have been recorded in the charts.If this measurement error occurred at random, then we have underestimated the actual difference in the use of study therapies.We do not expect that there would be any differences in recording of important clinical contraindications by specialty, particularly because data on medical history and clinical contraindications were gathered from all sources in the chart and not solely from physicians' notes.Finally, this study was conducted in 1 state and may differ from other states in ways that affect the generalizability of our findings.

POLICY IMPLICATIONS
The changes that are currently occurring in the organization and financing of health care in the United States are influencing how specialists' services are used in the care of large patient populations.Instead of the open, patient-and physician-driven system of traditional fee-forservice practice, many of the conditions under which generalists now collaborate with specialist colleagues are largely determined at the organizational level, where financial incentives, prior authorization requirements, and other managed care policies are developed.These changes have the potential to affect the quality of patient care.However, evidence to evaluate the added value of specialist care within these various arrangements is not yet avail-  able to help inform organizational policy decisions that impact on the access of patients to specialists.
Our findings suggest that, when a generalist physician is managing the care of a patient with AMI, the involvement of a cardiologist consultant offers an opportunity to enhance the quality of care.Although most of the recent health policy debate has focused solely on questions of patient access to specialty care, 22 a narrow focus on an either/or dichotomy between specialist and generalist care ignores the important issue of coordination of care between generalist and specialist physicians.We believe that our data should encourage further research to delineate the optimum strategy for ensuring effective coordination between generalists and cardiologists for the care of patients with AMI.January 15, 1998.From

Table 2 . Characteristics of Study Patients by Category of Involvement of Generalists and Cardiologists*
Ellipses indicate not applicable; HMO, health maintenance organization; AMI, acute myocardial infarction; NS, not significant; ED, emergency department; EMS, emergency medical services; ICU, intensive care unit; CCU, coronary care unit; LOS, length of stay; PTCA, percutaneous transluminal coronary angioplasty; and CABG, coronary artery bypass graft.†Based on 2 test.‡Other indicates indemnity/commercial insurance.§Based on Greenfield and colleagues' Index of Coexistent Disease. 17Does not include patients without indications (ST elevation Ն1 mm or symptom onset Ͻ12 hours).
*The study patients were eligible for aspirin, thrombolytic, or ␤-blocker therapy, or had no indication for lidocaine.All values are number (percentage).
Nonindicated use of lidocaine by specialty of the attending physician.Comparison of 2 outlier hospitals with all others.

Table 3 . Use of Study Drugs by Level of Generalist/Cardiologist Involvement*
Controlling for age, sex, ST elevation 1 mm or more, severe comorbidity, AMI location, presence of heart failure, use of lipid-lowering agents before hospitalization, presentation to hospital within 6 hours of symptom onset, and location of first contact with hospital (N = 1136).
the Department of Ambulatory Care and Prevention,HarvardMedicalSchoolandHarvardPilgrimHealthCare, Boston, Mass (Drs Willison, Soumerai, McLaughlin, and Pearson and Ms Gao); Centre for the Evaluation of Medicines, St Joseph's Hospital and McMaster University, Hamilton, Ontario (Dr Willison); the Meyers Primary Care Institute, University of Massachusetts Medical Center and the Fallon Healthcare System, Worcester (Dr Gurwitz); the Department of Health Care Policy, Harvard Medical School (Dr Guadagnoli); Cardiovascular Division, Brigham and Women's Hospital, Boston (Dr Hauptman); and Healthcare Education and Research Foundation, St Paul, Minn (Ms McLaughlin).This study was supported by grant HSO 7357 from the Agency for Health Care Policy and Research, Rockville, Md, and a grant from the Harvard Pilgrim Health Care Foundation, Boston, Mass.Dr Gurwitz is supported by a Clinical Investigator Award K08 AG00510 from the National Institute on Aging, Bethesda, Md.At the time of writing the manuscript Dr Willison was supported by a Post-Doctoral Research Fellowship from the Medical Research Council of Canada.Presented in abstract form at the Annual Meeting of the Association for Health Services Research, Atlanta, Ga, June 10, 1996.Reprints: Stephen B. Soumerai, ScD, Department of Ambulatory Care and Prevention, Harvard Medical School and Harvard Pilgrim Health Care, 126 Brookline Ave, Suite 200, Boston, MA 02215 (e-mail: oumerai@warren .med.harvard.edu).