Author + information
- Received August 15, 1996
- Revision received October 25, 1996
- Accepted November 5, 1996
- Published online March 1, 1997.
- ↵*Dr. Ira S. Nash, Cardiovascular Institute, The Mount Sinai Medical Center, Box 1030, One Gustave L. Levy Place, New York, New York 10029.
Pressure to lower the cost of health care delivery has fostered widespread efforts to limit patients’ access to specialists such as cardiologists. However, there is concern that diminished specialist involvement may lead to poorer patient outcomes for specific clinical conditions. As part of a state-sponsored effort to improve the quality of health care in Pennsylvania, the Pennsylvania Health Care Cost Containment Council gathered clinical and administrative data on all 40,684 hospital admissions for acute myocardial infarction (AMI) in that state in 1993. They prepared a detailed public report that included risk-adjusted in-hospital mortality and length of hospital stay by physician group, by hospital and by region. These data demonstrate that patients cared for by cardiologists, as a group, had a lower risk-adjusted mortality than patients cared for by either internists (risk ratio 1.26, 95% confidence interval 1.17 to 1.35) or family practitioners (risk ratio 1.29, 95% confidence interval 1.18 to 1.40). The patients of cardiologists also had a shorter length of stay than the other two groups. These data suggest that there is enhanced value in the care provided by cardiologists for patients with AMI and call into question the growing trend toward reliance on generalists instead of specialists.
(J Am Coll Cardiol 1997;29:475–8)
The U.S. health care system is undergoing a dramatic transformation, driven by the intense pressure to lower costs brought to bear on providers by payors. However, there is widespread recognition that low costs alone are no bargain and that the critical issue in evaluating clinical services is not just their cost but their value (). That is, the important question is not simply, How much does it cost?, but, rather, Is the anticipated benefit worth the expenditure? Cost is a sufficient measure only when comparing identical items or services (commodities). Value incorporates the quality of those items or the outcome of those services, or both. Unfortunately, costs are much easier to measure than quality or outcomes, so our ability to draw conclusions regarding the value of particular medical practices remains limited.
One area where value has been difficult to determine is in the provision of care by specialists as opposed to generalists. Acting primarily on the belief that specialists provide costlier care than generalists, many managed care organizations have attempted to reduce their reliance on specialists (). In closed panel health maintenance organizations (HMOs), relatively few specialists have been hired (). In more open managed care arrangements, such as preferred provider organizations (PPOs), access to specialists is restricted by a variety of administrative barriers and gatekeeper physician incentives ([4, 5]). The implications of such restrictions, were they to be universally adopted, are profound () and raise important questions about the appropriate national and regional provider mix of generalists and specialists (). Indeed, the shift away from our current high rates of specialty care by managed care organizations has spawned considerable debate about how to increase the number of primary care providers relative to specialists ([8–11]) but nearly universal agreement that such a shift is desirable ([12, 13]). It has also generated, in response, efforts by specialists to convince the public of the importance of continued access to their services. During the national debate on health finance reform, for example, the American College of Cardiology sponsored advertisements in leading newspapers stating that “direct, unrestricted access to the heart specialist of your choice could save your life” (The New York Times, June 12, 1995).
All these questions—the relative need for generalist and specialist services, the implications that this relative need has on training programs and academic institutions and the value of patient access to specialty services—would benefit from more data on the relative quality and cost of care provided by specialists and generalists for specific conditions (). Some data of this sort, with specific reference to cardiac care, are presently available. In 1993, Borowsky et al. () reported that patients cared for by cardiologists were significantly more likely than patients cared for by noncardiologists to undergo “clinically necessary” coronary angiography, defined by guidelines developed by a RAND Corporation/University of California Los Angeles expert panel that included primary care physicians, invasive and noninvasive cardiologists and cardiothoracic surgeons. Borowsky et al. did not attempt to detect differences in patient outcomes on the basis of observed differences in the use of coronary angiography.
Similarly, Ayanian et al. () reported that cardiologists were more likely than internists or family practitioners to prescribe therapies of proven efficacy in the care of patients with acute myocardial infarction (AMI). For example, 94.1% of surveyed cardiologists reported that they were “very likely” to prescribe a thrombolytic agent, whereas only 82% of internists and 77.3% of family practitioners reported that they would do the same (p < 0.001 for pairwise comparisons with cardiologists). Once again, the study did not attempt to detect differences in patient outcomes traceable to this and other treatment disparities.
More recent studies also suggest that cardiologists provide higher quality of care than noncardiologists for patients with AMI on the basis of closer adherence to national treatment guidelines.1In addition, Jollis et al. () reported that Medicare patients with AMI admitted by a cardiologist had a lower risk-adjusted 1-year mortality than patients admitted by a primary care physician. Finally, other data cast doubt on whether limiting access to cardiologists saves money, irrespective of process of care quality measures.2The recently released reports from the Pennsylvania Health Care Cost Containment Council (PHC4) ([18–20]) offer a large-scale examination of the differences in the outcomes of patients with AMI (not just the process of care) and treatment costs based on attending physician specialty.
1 Pennsylvania Health Care Cost Containment Council
The PHC4 is a state agency chartered to “collect and publish useful information about the charges and patient outcomes for various medical and surgical procedures” (). Because of the mortality and prevalence of AMI, the PHC4 undertook an analysis of the statewide data collected on all 40,684 admissions for AMI to Pennsylvania hospitals that occurred in calendar year 1993 (). Hospitals in Pennsylvania are responsible for supplying the PHC4 with detailed administratively derived, patient-specific information, including demographics, resource (procedure) utilization, comorbidities, measures of disease severity and the name of a single attending physician. The PHC4 used these data to develop two separate multivariate risk adjustment logistic regression models to predict hospital mortality, after excluding 1,428 patients (3.5%) for prespecified reasons. Exclusionary criteria included patients treated at hospitals that closed after 1993 or provided care for <30 patients with AMI during the study period, patients <30 or >99 years old, patients who left against medical advice, patients involved in two or more hospital transfers and patients who met “clinical complexity” criteria (major trauma, anoxic encephalopathy, metastatic cancer or previous heart transplantation). One risk adjustment model pertained to patients who were directly admitted to a hospital; the other model predicted in-hospital deaths for patients who were transferred from another acute care facility. The models were then used to generate risk-adjusted mortality rates for AMI by region, payor, hospital and physician group (although not by individual physician). Another separate linear regression model was created to predict inpatient length of stay, treating “transfer-in” status as a variable. One of us (D.B.N.) served as the chairman of the Technical Advisory Group, a standing committee responsible for overseeing the PHC4’s methodology and analysis. A detailed multivolume report of the methodology, statistical performance of the risk adjustment models and the principal findings is publicly available ([22, 23]).
2 Principal Findings
Two findings of the study have important implications for the generalist/specialist issue. First, the PHC4 found significant differences in risk-adjusted mortality among the three groups of physicians who, collectively, were identified as the attending physician for nearly 95% of all AMI admissions (Table 1). The adjusted inpatient mortality risk ratios, calculated from the data in the PHC4 report, for treatment by a family practitioner or an internist relative to a cardiologist, are both statistically significantly different from 1. If all 12,960 patients who were treated by internists had instead been treated by cardiologists, application of the different risk-adjusted mortality rates suggests that 285 fewer deaths would have been anticipated. Had cardiologists cared for the 6,971 patients of the family medicine physicians, the data suggest that there would have been 174 fewer inpatient fatalities.
The second finding of the PHC4 that has important implications for the relative value of care provided by cardiologists and other physicians is the mean inpatient length of stay by attending specialty (Table 2). For each physician group, the mean length of stay was predicted using the multivariate regression model developed for that purpose. Fewer patients were included in the length of stay analysis than in the inpatient mortality analysis because deaths, transfers out and patients with a length of stay >40 days were deliberately excluded. Only patients treated by cardiologists, as a group, had an actual mean length of stay statistically shorter than that predicted by the model, whereas patients treated by internists had an actual mean length of stay significantly longer than that predicted by the model, and those treated by family practitioners had a length of stay not distinguishable statistically from that predicted by the model.
The PHC4 acknowledged in its report that “choosing one physician as the ‘attending’ was a difficult task” and that “the methods used by hospitals in assigning [identifying] physician were diverse” (). Physicians were also identified by the hospitals by specialty—cardiology, cardiothoracic surgery, internal medicine, family medicine or “other.” Finally, the PHC4 also contacted attending physicians directly so that they could report their own professional identity. This process of physician specialty identification generated uncertainty concerning the “real” identity of the “responsible” physician, both in terms of actual name and specialty designation. For instance, it is unclear whether a patient admitted by an internist and seen in consultation by a cardiologist would be characterized as having been cared for by the former or the latter, or even if such a designation was made consistently within or among hospitals. Furthermore, the process of physician identification did not involve verification of credentials, so it is possible, for example, that some physicians identified as cardiologists do not have American Board of Internal Medicine certification as such. These uncertainties have the effect of diminishing any true differences in outcome by physician designation that might be apparent were physicians identified by stricter, more consistent criteria. Yet, even with this potential blurring of distinctions, significant outcome differences exist among the physician groupings.
Do cardiologists “do it better”—that is, provide higher quality and less costly care for patients with AMI than internists and family practitioners? Although the data suggest that they do, limitations of the PHC4 report should prompt some caution. The lower mortality rates achieved by cardiologists may be primarily a reflection of the relative number of patients with AMI treated by each practitioner. Inpatient mortality is dependent on events early in the hospital course, such as treatment initiated by the emergency department staff, which may occur before the involvement of the attending of record. If such early treatment were linked in some way to patient assignment, this could confound the analysis. In addition, although inpatient mortality is certainly an important dimension of the quality of care, it is not the only appropriate measure. The PHC4 report is silent on other potential quality measures, such as patient satisfaction, functional status at discharge, physician-directed modification of risk factors to prevent another coronary event and late mortality. Likewise, length of stay is, at best, a crude measure of the overall cost of care provided. No physician-stratified data are available from the PHC4 report on other important “drivers” of cost, such as resource utilization (e.g., whether cardiologists order more echocardiograms than other physicians for patients with AMI) or referral for catheterization and revascularization procedures. The differences in length of stay are small in absolute terms and could potentially be overshadowed by some of these other costs. Finally, the conclusions of the PHC4 would be strengthened by repeating the analysis using attending physician designation as an independent variable and assessing directly its impact on patient outcomes. Only further analysis of these data, along with the examination of other, contemporary, large data bases can address these issues.
These limitations, although important, do not negate the principal findings that a comprehensive examination of all hospital admissions of patients with AMI in the state of Pennsylvania in 1993 found that cardiologists achieve a lower mortality with a lower ratio of actual to predicted inpatient length of stay than either internists or family practitioners. The reasons why this is so are not clear. However, what is clear is that more comparative data such as these, even down to the physician level, will be demanded of physicians and hospitals in the years ahead (). The challenge we face is to collect () and use () these data in a responsible way, for they may be critically important in helping consumers, providers and payors make well informed decisions about the appropriate level of access to specialized cardiac care and, by extension, the number of cardiologists needed to provide it.
We thank Ernest Sessa, Executive Director of the Pennsylvania Health Care Cost Containment Council for his encouragement and Jayne Jones, also of the Council, for technical support. The views expressed here are solely those of the authors.
↵1 Wilson D, Soumerai SB, McLaughlin T, et al. Cardiologist and generalist involvement in the care of acute MI—implications for quality [abstract]. Presented at the 13th Annual Meeting of the Association of Health Services Research, Atlanta, Georgia, June 1996.
↵2 Warner CD, Weintraub WS, Rask K, Saunders C. Does the gatekeeper model of managed care affect access to specialists for patients with chest pain? [abstract]. Presented at the 13th Annual Meeting of the Association of Health Services Research, Atlanta, Georgia, June 1996.
- acute myocardial infarction
- Pennsylvania Health Care Cost Containment Council
- Received August 15, 1996.
- Revision received October 25, 1996.
- Accepted November 5, 1996.
- The American College of Cardiology
- Nelson EC,
- Greenfield S,
- Hays RD,
- Larson C,
- Leopold B,
- Batalden PB
- Goldstein S,
- Pearson TA,
- Colwill JM,
- Faxon DP,
- Fletcher RH,
- Moodie DS
- Lee TH
- Goldstein S,
- Pearson TA,
- Colwill JM,
- Faxon DP,
- Fletcher RH,
- Moodie DS
- Borowsky SJ,
- Kravitz RL,
- Laouri M,
- et al.
- ↵(1996) Focus on heart attack in southeastern Pennsylvania: a 1993 summary report for health benefits purchasers, health care providers, policy-makers, and consumers (Pennsylvania Health Care Cost Containment Council, Harrisburg (PA)).
- (1996) Focus on heart attack in central and northeastern Pennsylvania: A 1993 summary report for health benefits purchasers, health care providers, policy-makers, and consumers (Pennsylvania Health Care Cost Containment Council, Harrisburg (PA)).
- (1996) Focus on heart attack in western Pennsylvania: A 1993 summary report for health benefits purchasers, health care providers, policy-makers, and consumers (Pennsylvania Health Care Cost Containment Council, Harrisburg (PA)).
- ↵(1996) Focus on heart attack in Pennsylvania; research methods and results (Pennsylvania Health Care Cost Containment Council, Harrisburg (PA)).
- ↵(1996) Focus on heart attack in Pennsylvania; the technical report—1993: Parts A and B (Pennsylvania Health Care Cost Containment Council, Harrisburg (PA)).