In this era of rigorous healthcare cost containment, neurologists need creative new ways to identify the value of the services we provide. Therefore, it is valuable to consider the outcomes and costs of hospitalized stroke patients under the care of neurologists compared with those under management by other physicians. In this issue of Stroke, Mitchell et al examine 90 days of claim data for 20% (38 612) of all Medicare patients hospitalized in 1991 with one of two stroke diagnoses (International Classification of Diseases, 9th revision, codes 434 and 436). Because the data come from a national database, we can look at a large, diverse Medicare population, identify services regardless of site of care or provider, and track individuals during and subsequent to hospitalization for a stroke.
Many hypotheses can be generated to explain the results of the analysis, which shows that when neurologists serve as the attending physicians, stroke patients have better outcomes but at a higher cost than when the principal care is provided by other physicians. Many neurologists may conclude that the additional expenses incurred by neurologists are probably justified because newer, more costly technology helps us provide more accurate diagnosis and effective treatment. But can we be sure that the outcomes and costs in this study truly reflect neurologists' value-added services, or are there other explanations for the findings?
Most of us are accustomed to weighing therapeutic decisions by analyzing the results of prospective controlled trials. Because the collection of that kind of primary outcome data is very time consuming and expensive, health-service researchers and managed care organizations (MCOs) also rely on less expensive secondary data sources, such as the Medicare administrative data that serve as a basis for this report. Unfortunately, there are disadvantages to using administrative data sets to analyze the nation's healthcare investment. To avoid drawing unsubstantiated conclusions from the data, it is important that we consider alternative explanations for the observations.
In this study, the only outcomes recorded were mortality and discharge destination. Because medical records were not abstracted and patient functional status and quality of life were not available, these outcome measures are very crude proxies for patient health status. The authors try to adjust their data for patient characteristics that might account for differences in cost and outcome, but the only information available to them is secondary coded diagnoses. It can be very difficult to determine the complexity of coexisting medical illnesses in stroke patients by merely identifying the frequency of coded comorbidities. Furthermore, even if these coded diagnoses are a true reflection of severity of illness, the accuracy of secondary diagnoses is always suspect when the information does not influence provider payment.
What if, as has been shown in other studies, there was a selection bias and the sickest stroke patients were not primarily managed by neurologists? Although inconclusive, there are data in this study to suggest that there was triaging of patients that was based on clinical characteristics. The outcomes of stroke patients managed by neurologists in geographical areas with fewer neurologists were better than in areas with a greater number of neurologists. One explanation for this observation could be that the sicker stroke patients in these “underserved” regions were more frequently managed by other physicians. Given these inconclusive findings, we must at least entertain the authors' alternative hypothesis that neurologists manage different Medicare-eligible stroke patients than other physicians and that these patient differences, rather than the physicians' neurological expertise, account for better outcomes at a higher cost.
In this study, a creative example of the use of administrative data is the analysis of prothrombin tests as a proxy for warfarin use. Neurologists obtain more postdischarge prothrombin tests, suggesting that they are more likely than other physicians to administer anticoagulation to their patients. Their concomitant greater use of brain imaging studies may help them to identify the subpopulation of stroke patients who do not have hemorrhages and benefit most from anticoagulation. With MCOs exerting enormous pressure to reduce the cost of medical care, neurologists need information that enhances their ability to identify all such cost-effective strategies for patients who have a stroke. Although prospective controlled studies will always be the gold standard for comparing alternative therapies, retrospective analysis of management decisions in large, unselected populations is useful in helping to define current medical practice.
A neurologist was listed as the attending physician for only 11.3% of the stroke patients in this study, but in another 24.7% of patients a neurologist and primary-care physician both billed for routine hospital visits. Not unexpectedly, fewer stroke patients were cared for by neurologists in community hospitals than in larger medical centers, but even within the centers many patients were not followed up by neurologists. We have no information as to how many of the remaining patients were seen by neurologists in consultation, but it is clear that the majority of stroke patients did not receive ongoing neurological care. Why did this happen if the care neurologists provide to patients is valuable? Since the patients all were treated in US hospitals, researchers did not influence practice or introduce any selection bias. This snapshot of American stroke treatment should be most disturbing to those stroke victims who did not have the benefit of ongoing neurological care if, in fact, neurologists provide value-added services.
What practical lessons can we learn from this study when the data are imperfect and there is pressure from MCOs to provide immediate answers? First, we must become increasingly familiar with the utility of such health-service research methods. Although this study of stroke outcomes and costs raises as many questions as it answers, it provides some indication of the neurological care throughout the country for hospitalized stroke patients. With increasing restrictions on referral of patients to specialists, the number of stroke patients managed by neurologists today may be even less than it was 5 years ago when this study was done.
Second, we cannot take for granted that our ability to understand and interpret information and to use it to the best advantage of the patient will dictate the treatment options chosen by MCOs. It no longer is sufficient for medical educators to report medical discoveries, randomized trials, review articles, and unusual case reports. We also need evidence-based approaches and cost-effective analyses that help us to become better problem solvers and that are convincing adjuncts for managing day-to-day care of patients. Developing these skills takes time and requires increasing resources and commitment from our academic leaders.
Third, our primary focus should be on defining the optimum care for patients with stroke, regardless of whether a neurologist or other physician is in charge. With this emphasis, critical appraisal of neurological diagnostic and treatment choices will assist payers in accurately identifying cost-effective management of stroke patients. Only then can we expect neurological consultation and care of the stroke patient to be valued accurately and our services to be actively sought.
The opinions expressed in this editorial are not necessarily those of the editors or of the American Heart Association.
- Copyright © 1996 by American Heart Association