Correspondence
Outcome of Acute Myocardial Infarction According to the Specialty of the Admitting Physician
N Engl J Med 1997; 336:1607-1609May 29, 1997
- Article
To the Editor:
Jollis and colleagues (Dec. 19 issue)1 report the results of their study on the outcome of acute myocardial infarction in relation to the admitting physician's specialty. Research that compares the quality of care provided by generalists and that provided by specialists is of considerable interest because of the increasing focus on primary care. Unfortunately, the way in which the research questions are posed often stacks the deck against primary care.
Primary care is person-oriented rather than disease-oriented medicine. Its challenges include dealing with vague and difficult-to-resolve problems, most disorders and ailments, and behavior that signifies a high risk. A comparison of primary care with specialty care should include characteristics that are in the purview of primary care as well as those of specialty care.
Competence in dealing with a problem comes not only with training but also with practice; studies have demonstrated that the frequency with which a problem is encountered is an important determinant of the quality of care. Specialists caring for patients with diagnoses that fall within the scope of their specialty will always be found to provide care of better quality for these conditions, because they see more patients with these conditions than do primary care specialists. Would cardiologists or neurologists, however, want to be evaluated according to how well they deal with vague symptoms or diseases or preventive care outside the interests of their specialty? The issue is not only how primary care physicians deal with particular diseases but also how specialists perform in carrying out the functions and tasks of primary care.
One of the challenges of primary care is deciding when to refer a patient to a specialist. The frequency of problems in a practice population, not the interest of specialists in the care of particular diseases, should be the determinant of whether the responsibility lies with a primary care physician or a specialist. Specialists and primary care physicians should decide jointly who should bear the responsibility for addressing certain types of problems and at what stages a referral is indicated. These decisions may vary from area to area as the incidence and prevalence of the problem vary.
An accumulating literature on comparisons of population-based primary care and specialty care indicates the importance of primary care in terms of both cost and outcomes.2-5 Disease-specific comparisons are valuable in focusing attention on the need for collaboration between primary care specialists and other specialists in determining when a problem moves from the purview of primary care to that of specialty care. Such comparisons have no value in assessing the benefits of primary care, since they are inherently biased toward finding it inferior.
5 ReferencesBarbara Starfield, M.D., M.P.H.
Johns Hopkins University, Baltimore, MD 21205-19961
Jollis JG ,DeLong ER ,Peterson ED , et al. Outcome of acute myocardial infarction according to the specialty of the admitting physician. N Engl J Med 1996;335:1880-1887
Full Text | Web of Science | Medline2
Farmer FL ,Stokes CS ,Fiser RH ,Papini DP . Poverty, primary care and age-specific mortality. J Rural Health 1991;7:153-169
CrossRef | Medline3
Parchman ML ,Culler S . Primary care physicians and avoidable hospitalizations. J Fam Pract 1994;39:123-128
Web of Science | Medline4
Shi L . Primary care, specialty care, and life chances. Int J Health Serv 1994;24:431-458
CrossRef | Web of Science | Medline5
Starfield B . Is primary care essential? Lancet 1994;344:1129-1133
CrossRef | Web of Science | Medline
To the Editor:
The study by Jollis et al. was a provocative but not definitive use of data from the Health Care Financing Administration's Cooperative Cardiovascular Project (CCP). The use of these data is not definitive because the validity of the multivariate model is not documented.
The main thrust of the study is that despite many differences in their characteristics, patients admitted by cardiologists appear to have a longer survival than those admitted by generalists, as estimated by a multivariate model modified from that used in the Global Utilization of Streptokinase and Tissue Plasminogen Activator for Occluded Coronary Arteries (GUSTO) study.1 Jollis et al. do not provide any data on the adequacy of this model as applied to the CCP data set, and neither the GUSTO investigators nor Jollis et al. estimated the percentage of the variation in survival that is explained by the models. Normand et al. developed a similar model specifically for the CCP data,2 and they estimated that their model explained only 27 percent of the variation in survival. It is possible that unmeasured characteristics or measured characteristics, such as age, that are inadequately adjusted account for the differences in survival.
2 ReferencesThomas A. Marciniak, M.D.
Health Care Financing Administration, Baltimore, MD 212441
Lee KL ,Woodlief LH ,Topol EJ , et al. Predictors of 30-day mor-tality in the era of reperfusion for acute myocardial infarction: resultsfrom an international trial of 41,021 patients. Circulation 1995;91:1659-1668
Web of Science | Medline2
Normand ST ,Glickman ME ,Sharma RG ,McNeil BJ . Using admission characteristics to predict short-term mortality from myocardial infarction in elderly patients: results from the Cooperative Cardiovascular Project. JAMA 1996;275:1322-1328
CrossRef | Web of Science | Medline
To the Editor:
Jollis et al. provide a thought-provoking and extensively analyzed study of differences in the outcome of acute myocardial infarction in relation to medical specialty. As noted by Dr. Goldman1 in his accompanying editorial, this study is not a randomized clinical trial and therefore does not provide the best evidence on which to base public policy or clinical inferences. Furthermore, it is possible that the authors' conclusions may be compromised by inadequate control of a number of confounding biases (e.g., age, sex, previous coronary artery disease, or diabetes). In addition, a much smaller proportion of the cardiologists' patients were likely to be in rural locations (10.9 percent, vs. 53.3 percent of the patients seen by family-medicine physicians), which arouses concern about differences in education, socioeconomic status, and access to specialty consultation. Also, if specialty training were a primary determinant of the outcome of acute myocardial infarction, one would expect the outcomes to be worse for family physicians than for internists.
The degree to which this confounding bias is present can be seen in the difference between the crude odds ratios and those in the proportional-hazards model. In the comparison of cardiologists with internists, the crude odds ratio for mortality at one year was 0.80, whereas the adjusted odds ratio was 0.88. The difference represents a 40 percent change in the measure of an effect, suggesting that confounding bias is extremely important in the analysis of these data.
If, in fact, the confounding biases have been fully accounted for, what are the implications of this study for policy? It seems unlikely that large numbers of cardiologists are going to locate their practices in rural areas or that institutions that do not have on-site access to angiography or revascularization procedures will invest in developing it. The authors have the capability of adding to their model the use of treatment with beta-blockers, thrombolytic agents in eligible patients, aspirin, heparin, or nitroglycerin, as well as revascularization procedures, to see whether the effect of increased use of these approaches by cardiologists explains the difference in outcomes according to specialty. If so, then strategies can be developed to ensure that appropriately stratified patients receive these treatments. If they do not explain the differences in outcomes according to specialty, then this is further evidence of a misclassification bias associated with measurements of known and unknown confounding.
1 ReferencesCharles B. Eaton, M.D.
John B. Murphy, M.D.
Vincent R. Hunt, M.D.
Brown University School of Medicine, Providence, RI 029121
Goldman L . The value of cardiology. N Engl J Med 1996;335:1918-1919
Full Text | Web of Science | Medline
To the Editor:
Jollis et al. suggest that the outcome is better for Medicare patients with acute myocardial infarction who are cared for by cardiologists than for those cared for by primary care physicians. The authors report a risk-adjusted 12 percent survival advantage at one year for patients treated by cardiologists.
The authors assumed that the treating physician was the admitting physician, identified as the physician who submitted the Part B Medicare claim for initial hospital care (Current Procedural Terminology [CPT] code 99221-3) or for critical care services (CPT 99291-2) or subsequent hospital care (CPT 99231-3), if a claim for initial care was not submitted. As noted in the article, for 18 percent of the patients, the authors used the specialty of the attending physician noted on the Medicare Part A claim.
I suspect that the risk-adjusted difference is actually greater than 12 percent if primary care physicians who did not seek consultation with the cardiologist (pure primary care cases) are compared with cardiologists. This could easily be determined by eliminating from the primary care group cases in which a bill for consultation (CPT 99251-5) was submitted by a physician whose unique physician-identification number indicated that he or she was a cardiologist. If a cardiology consultant was involved, he or she probably played a major part in the patient's care, which would diminish any difference between the group of patients cared for by primary care physicians and the group cared for by cardiologists alone. The authors did not establish that these cases could be identified by examining Medicare Part A claims submitted by the hospital. This point further strengthens the argument that cardiologists should be involved in the care of such patients.
Daniel C. Belin, M.D.
85 Church St., Middletown, CT 06457Author/Editor Response
The authors reply:
To the Editor: Dr. Starfield makes a number of salient comments about the importance of experience and the need to identify the point along the spectrum of disease severity where patients should be referred to specialists. Our study examined patients at one end of this spectrum, where specialty care is most likely to be associated with improved outcomes. Specialists and generalists should work together in apportioning the responsibility for care. Our study and others like it should help inform such decisions.
As Dr. Marciniak suggests and as we noted in our article, our results could have been confounded by unmeasured risk, a potential problem in all observational research. It is unlikely that additional CCP data beyond the 17 variables included in our models would identify such an unmeasured risk.1 The ideal study would involve randomization, but such a study would be difficult to conduct. Other observational evidence in support of our findings includes the identification of a potential mechanism (differences in treatment) and the recent report of a similar association between specialty and mortality in a study of 36,927 patients in Pennsylvania.2 Contrary to Dr. Marciniak's suggestion, the percentage of variance of a dichotomous end point (death at 30 days) explained by a continuous risk function is not a measure of the adequacy of adjustment; hence, the figures he cites are not directly relevant to the analyses we reported.3
The change in relative risk noted by Dr. Eaton and his colleagues in part reflects the extent to which we attempted to account for differences in survival by factors other than specialty. Our results are not inconsistent with the conjecture of a better performance by internists. Neither hazard ratio cited by Eaton et al. is significantly different from 1, and the confidence intervals clearly overlap, indicating that the difference is too small to be estimated reliably with these data. After adjustment for claims-data characteristics in the national sample (220,535 patients), patients admitted by internists were 5 percent less likely to die within one year than those admitted by family-medicine physicians (P<0.001), suggesting that a larger, more detailed data set may yield the expected relation. Models with adjustment for medications showed a persistence of the survival benefit for patients admitted by cardiologists.
Dr. Belin raises an important issue regarding the interaction between specialties. Noting the potential bias toward longer survival in cases involving consultation with a cardiologist (since patients have to survive long enough for the consultation to take place), we found a significant gradient in the mortality rate, with patients admitted by cardiologists having the lowest rate, patients admitted by primary care physicians without consultation having the highest rate, and patients admitted by primary care physicians with consultation having an intermediate rate.
3 ReferencesJames G. Jollis, M.D.
Elizabeth R. DeLong, Ph.D.
Daniel B. Mark, M.D., M.P.H.
Duke University Medical Center, Durham, NC 277101
Lee KL ,Woodlief LH ,Topol EJ , et al. Predictors of 30-day mortality in the era of reperfusion for acute myocardial infarction: results from an international trial of 41,021 patients. Circulation 1995;91:1659-1668
Web of Science | Medline2
Nash IS ,Nash DB ,Fuster V . Do cardiologists do it better? J Am Coll Cardiol 1997;29:475-478
CrossRef | Web of Science | Medline3
Chassin MR ,Hannan EL ,DeBuono BA . Benefits and hazards of reporting medical outcomes publicly. N Engl J Med 1996;334:394-398
Full Text | Web of Science | Medline
