Warfarin Use Following Ischemic Stroke Among Medicare Patients With Atrial Fibrillation FREE

OVER THE PAST decade, great advances have been made in stroke prevention.^1- 2 Among the most potent therapies for stroke prevention is the use of anticoagulation therapy for patients with atrial fibrillation. Atrial fibrillation is a common cause of stroke accounting for 10% to 15% of all ischemic strokes and nearly a quarter of strokes for those older than 80 years.³ Primary prevention of stroke in patients with atrial fibrillation is of particular importance since these strokes appear to be more severe and the outcome markedly poorer than in patients with a sinus rhythm.^4- 5 The rate of death or significant neurologic disability is as high as 71%.^5- 6 Among appropriate patients with atrial fibrillation, anticoagulation can reduce the risk of a first stroke by about 68%⁷ and is the recommended therapy in published guidelines.^8- 11

Early studies indicated a dramatic underutilization of anticoagulation for patients with atrial fibrillation in a variety of local settings, including small clusters of community hospitals¹² and academic medical centers.¹³ Larger studies also reported similar findings for office-based physicians,¹⁴ and in long-term care facilities.¹⁵ More recently we reviewed warfarin sodium therapy prior to hospital admission among patients with known atrial fibrillation across a broad range of practices in the community.¹⁶ We found that more than half of patients older than 65 years with known atrial fibrillation were not receiving warfarin. This result was not dramatically changed when we looked at the subgroup who were free of contraindication to anticoagulation and were at increased risk for stroke because of the presence of additional vascular risk factors.

Elderly patients with an ischemic stroke associated with atrial fibrillation are at especially high risk for recurrent stroke with an annual rate of recurrent stroke of more than 10% per year.^5,17- 18 Secondary prevention therapy with warfarin is highly effective in reducing this risk. The recurrence rate can be reduced by two thirds with warfarin therapy.¹⁷ The use of warfarin has also been demonstrated to decrease mortality¹⁹ and to be cost-effective.^19- 20

Given the more intense level of medical treatment usually associated with hospital care, it might be expected that underutilization of well-established vascular therapies was uncommon. Krumholz and colleagues,²¹ however, demonstrated underutilization of both acute and secondary prevention with antiplatelet therapy among patients hospitalized with myocardial infarction. In addition, survey data²² suggested underutilization also occurred for preventive therapies following hospitalization for stroke.

We sought to determine the actual use of warfarin for secondary prevention among a population sample of elderly patients with atrial fibrillation hospitalized for ischemic stroke.²³ A prior report¹⁶ from our group was based on a larger atrial fibrillation project within the Connecticut Peer Review Organization. We measured the use of anticoagulation among patients in the community, before admission, regardless of current or prior cerebrovascular disease. Our article looks at another subset of patients, those who were hospitalized with an acute ischemic stroke associated with atrial fibrillation.²³

The Connecticut Peer Review Organization conducted a chart review of Medicare patients, aged 65 years or older, hospitalized in 1994 with a diagnosis of atrial fibrillation. The methods used are briefly reviewed herein. The details of case ascertainment, data abstraction, and analysis were presented in a previous publication¹⁶ dealing with outpatient care of patients with known atrial fibrillation.

The Connecticut Peer Review Organization conducted a retrospective study to examine use of warfarin among patients with atrial fibrillation. Using reliable techniques previously described for related projects, information was obtained from medical charts by trained nurses and medical record technicians.^21,24

To identify potentially eligible patients, the diagnosis of atrial fibrillation was determined by examining principal and secondary discharge diagnoses using the MEDPAR (Medicare Part A claims) files. Patients with stroke were selected with a principal diagnosis (reason in retrospect for which the patient was admitted) of ischemic stroke using International Classification of Diseases, Ninth Revision (ICD-9) codes.²⁵ The definitions and coding used for the diagnoses have previously been reported.¹⁶

We eliminated those with a potential indication for anticoagulation (other than atrial fibrillation) by excluding those with a primary diagnosis of myocardial infarction or embolic events (other than stroke), and also excluded those with retinal vascular occlusion, peripheral vascular disease, vascular insufficiency of the intestine, and vascular disorders of the kidney.

Variables examined in this study included age, sex, race, risk factors for stroke, classes of medication, possible contraindications to anticoagulation, functional ability, and medical history. The specific items were abstracted both from the coded discharge diagnoses and medical record as previously described.¹⁶ The data abstraction was based on the complete medical record including documented history and physical examinations. In addition to specified contraindications to anticoagulation, any potential items mentioned in the chart were abstracted. Although cases with other potential contraindication were counted, the individual items listed in the medical record were not available for this analysis.

Trained abstractors reviewed each record, and determined whether atrial fibrillation was noted in the official interpretation of any electrocardiogram. When atrial fibrillation was not noted in any official electrocardiographic interpretation, a board-certified cardiologist reviewed the record. Because both are associated with an increased risk of stroke, persistent and intermittent atrial fibrillation were both accepted as a diagnosis.⁵ A history of atrial fibrillation was determined by the cardiologist's review of physician admission notes, consultation notes, progress notes, or discharge summary. As with previous studies done by Connecticut Peer Review Organization, reabstraction was performed on 4% of cases.^24,26 The interabstractor agreement was more than 96% among all abstracted items.

First, we evaluated the bivariate association between the use of warfarin and a family of variables that included demographic, clinical factors, medications, and possible contraindications to anticoagulation. For each variable, a χ² test was used to determine whether the proportion of patients prescribed warfarin at discharge differed significantly from those not prescribed warfarin. Clinically relevant variables that were associated with the use of warfarin with a value of P<.10 in the bivariate analysis were entered into a multivariate logistic regression analysis model. For each association of interest, an adjusted odds ratio and 95% confidence interval were calculated.

Second, we separately investigated those factors associated with initiating anticoagulation by repeating these analyses for the subgroup of patients who were not receiving warfarin at the time of admission.

The study cohort was assembled as part of a statewide project at the Connecticut Peer Review Organization.¹⁶ There were 635 patients (402 women; 585 white; 218 ≥85 years old; 147 with a new diagnosis of atrial fibrillation; 554 with at least 1 risk factor for stroke) included in our original study cohort. Among these patients, there were 334 who had stroke listed as a principal diagnosis. Of these, 282 (84%) of 334 were alive at discharge. We excluded 4 patients who were transferred to another acute care facility at discharge from their index admission. This resulted in 278 patients with atrial fibrillation, discharged alive after an ischemic stroke, who were the primary cohort for this article. The demographics, clinical characteristics, and potential contraindications present at the time of admission are presented in Table 1.

In addition, factors present on admission, and their association with the use of warfarin at discharge, are shown in Table 1. Overall, only 148 (53%) of 278 patients were prescribed warfarin at hospital discharge. There were differences by demographic and clinical characteristics. Patients who were older, had prior bleeding, dementia, or were limited in their activities of daily living were all less likely to be prescribed warfarin at discharge. Patients who had been receiving warfarin, β-blockers, or angiotensin-converting enzyme inhibitors, or had been living at home before admission were all more likely to be prescribed warfarin at discharge from the hospital. Among those with none of the listed possible contraindications to anticoagulation (168 patients), there was a slight, but statistically significant higher rate of warfarin use among 100 (60%) of 168 patients (P<.01); however, the presence of risk factors for stroke (other than the index ischemic stroke) was not associated with a higher rate of warfarin use at the time of discharge. Furthermore, among the 130 (47%) of 278 patients not prescribed warfarin at discharge, 81 (62%) of 130 patients were also not prescribed aspirin. Even among those without a contraindication to anticoagulation (168) who were not prescribed warfarin (68 [40%] of 168), 42 (62%) of 68 were not even prescribed aspirin at discharge.

Among the patients not prescribed warfarin who did not have 1 of the possible contraindications listed in Table 1, 15 (22%) of 68 had a reason documented in the medical record (regardless of appropriateness) for not using warfarin. Similarly, among the patients not prescribed warfarin who did not have 1 of the possible contraindications and had 1 or more of the risk factors for stroke listed in Table 1, 10 (19%) of 54 had a reason documented in the medical record (again, regardless of appropriateness) for not using warfarin.

Table 2 lists characteristics of the hospital admission and their association with the rate of warfarin use at discharge. Only a low serum albumin level (associated with a lower rate of warfarin use) and being discharged home (with or without the need for assistance) were associated with a statistically significant higher rate of warfarin use (P<.05).

The items included in Table 1 and Table 2 were selected for their clinical relevance, and those with a P≤.10 in the bivariate analysis were entered into the multiple logistic regression analysis. The results are listed in Table 3. Patients older than 85 years, with congestive heart failure, or who had an albumin level of 30 g/L or lower were all less likely to be prescribed warfarin. Those who were receiving warfarin at admission, β-blockers at discharge, or had been admitted from home were all more likely to be prescribed warfarin at discharge. These factors from the multivariate model, however, accounted for only 30% of the variation seen for the use of warfarin therapy at discharge.

We also abstracted any documented reason for not using anticoagulation, regardless of whether appropriate. Only 26 (20%) of 130 patients discharged without warfarin listed a reason other than those shown in Table 1.

For subsequent analysis, we restricted our analysis to those who had not been prescribed warfarin at the time of admission (ie, initiation of warfarin therapy following their stroke). Table 4, Table 5, and Table 6 are based on the 203 patients with atrial fibrillation and admitted with stroke who had not been prescribed warfarin at the time of admission for their stroke. They parallel the items and analyses of Table 1, Table 2, and Table 3.

Table 4 lists factors present on admission and their association with the initiation of warfarin therapy at discharge. There were differences by demographic and clinical characteristics. Patients who were older, had dementia, hypertension, or prior stroke, or were limited in their activities of daily living were all less likely to have warfarin therapy initiated at discharge. Patients who had been receiving β-blockers, or living at home before admission, were more likely to be initiated on a regimen of warfarin therapy. Among those with none of the listed possible contraindications, there was a trend toward a higher rate of warfarin use (P=.10); however, similar to Table 1, the presence of risk factors for stroke did not increase the rate of warfarin use at the time of discharge.

Among the patients not prescribed warfarin who did not have 1 of the possible contraindications listed in Table 4, 14 (23%) of 62 had a reason documented in the medical record (regardless of appropriateness) for not using warfarin. Similarly, among the patients not prescribed warfarin who did not have 1 of the possible contraindications and had 1 or more of the risk factors for stroke listed in Table 4, 9 (19%) of 48 had a reason documented in the medical record (again, regardless of appropriateness) for not using warfarin.

Among those not receiving warfarin at admission who were not prescribed warfarin at discharge, 69 (58%) of 118 (34% of the total in this subcohort) were also not prescribed aspirin at hospital discharge. Even among the patients without 1 of the listed contraindications to anticoagulation, only 55 (47%) of 117 were placed on a regimen of warfarin therapy. Of those not receiving warfarin, 37 (60%) of 62 (or 32% of this category) were also not prescribed aspirin at discharge.

Table 5 lists characteristics of the hospital admission and their impact on the rate of warfarin use at discharge among those not receiving warfarin therapy at the time of admission. Only a low serum albumin level (associated with a lower rate of warfarin use), a new diagnosis of atrial fibrillation, or discharge home (with or without the need for assistance) achieved statistical significance (P<.05).

The items from Table 4 and Table 5 with a P ≤.10 were entered into the multiple logistic regression analysis, and the results are listed in Table 6. Patients older than 85 years or who had an albumin level of 30 g/L or lower were less likely to be prescribed warfarin. Those who had a new diagnosis of atrial fibrillation or had been admitted from home were more likely to have a new prescription for warfarin initiated at hospital discharge; however, these factors from the multivariate model account for only 29% of the variation seen for the initiation of warfarin therapy.

Previous reports have documented the underutilization of anticoagulation in the community. The focus of earlier work has been largely primary prevention. Because all our patients had a recent ischemic stroke, they are at especially high risk for recurrent stroke and are the most likely to benefit from anticoagulation. Among these patients, there is little controversy among consensus statements and guidelines. Compared with primary prevention, those with a history of stroke had slightly higher rates of warfarin use.^12- 13

Our work demonstrated, even for this very high-risk population, anticoagulation therapy appears underused. Although patients with ischemic stroke associated with atrial fibrillation are at greatly increased risk for stroke, contraindications to anticoagulation exist. However, even when we restricted our analysis to those without a contraindication to anticoagulation, it appears as if only half of eligible patients are receiving anticoagulation therapy.

In addition, although not the focus of this analysis, aspirin was also dramatically underused, with more than 62% of all patients not discharged on a regimen of warfarin therapy also not given aspirin. Aspirin reduces the rate of recurrent stroke, but only by about 15% compared with placebo.¹⁷ In addition to affording some protection against stroke associated with atrial fibrillation, aspirin is a standard part of preventive therapy following ischemic stroke. Although contraindications to antiplatelet therapy exist, the low rate of use, even among those without contraindications to anticoagulation, suggests that there is also a true underuse of this well-established therapy.

We have identified an important therapy that is not being used. The reasons why this occurs and how physician practices can be improved are critical next steps. In our analyses few factors predicted the use of warfarin, and these only accounted for a small portion of the variation.

Those in older age groups are less likely to receive warfarin; however, even in the youngest age groups the rate of anticoagulation appears disappointingly low. Furthermore, it is not clear that the strong age gradient seen is appropriate. The concern is that older individuals are at higher risk for bleeding complications, especially intracranial hemorrhage²⁷; however, these same patients are also at increased risk for recurrent stroke.

Misperception of age-related efficacy and hemorrhagic risk is thought to influence the use of therapies among patients with acute myocardial infarction.²⁸ A similar effect may exist for stroke therapy with anticoagulation. Data from a combined analysis of 4 atrial fibrillation trials suggest that the overall rate of intracranial hemorrhage, with a target range of an international normalized ratio of 2.0 to 3.0, is low (0.3% per year), but is higher among older age groups.^29- 30 The lower rate of warfarin use in older adults may reflect an uncertainty about the relative efficacy and safety of anticoagulation in these patients; however, available data suggest that even with higher rate hemorrhage than seen in older patients, that net benefit of anticoagulation remains.³¹

Many of the factors we found associated with the use of anticoagulation may be a marker for the general health of an individual. These included type of residence before or after discharge and serum albumin level.

The use of medications such as β-blockers and angiotensin-converting enzyme inhibitors was associated with a higher rate of warfarin use in the univariate analyses; however, this may be a marker for a more aggressive management style or the recognition of cardiac disease (and the risk of cardioembolism). These factors did not remain in the multivariate model. Not surprisingly the use of warfarin therapy before admission was associated with a higher rate of warfarin therapy at discharge, and a new diagnosis of atrial fibrillation during the index admission was associated with the initiation of warfarin therapy following stroke.

We did not see a significant variation by hospital type in our study, nor was variation by academic affiliation seen in a study of long-term care facilities reported by Gurwitz and colleagues.¹⁵ Others¹² have, however, reported lower rate of anticoagulation from community hospitals.

Furthermore, we did not observe significant differences across medical specialties, nor did a national survey of physician attitudes to hypothetical clinical scenarios.³² Stafford and Singer¹⁴ documented a higher rate of warfarin prescription among cardiologists and general internists in office-based practices included in the National Ambulatory Care Survey. Regardless of these differences, as with therapies for myocardial infarction,³³ there appears to be room for improvement across all acute care settings and among all specialty groups.

With the intense care associated with hospitalization, it may seem surprising that anticoagulation is underused. For other forms of vascular disease, such as cardiovascular disease, underutilization has been documented for acute therapies such as those for myocardial infarction^24,34- 35 and also for simple therapies such as aspirin for secondary prevention following myocardial infarction in many common,²¹ but not necessarily all, clinical settings.³⁶

Even in the acute treatment of myocardial infarction more than half of patients do not receive aspirin in the emergency department, and for those receiving therapy there are often substantial delays to initiation.³⁷

Less work has been done related to the medical care associated with stroke; however, it appears there are ample opportunities for improvements in care. In a hospital-based review of patients admitted with stroke in Leicestershire, England, more than half of patients were not receiving antiplatelet therapy or antihypertensive medications (among those with a history of hypertension).³⁸ Significant variances from stroke management guidelines exist for acute hypertension treatment, deep vein prophylaxis, diagnostic evaluation, and antithrombotic therapy at discharge.³⁹ In addition, the setting of care for patients with stroke appears to influence outcome.⁴⁰ Further work, analogous to what is being done for cardiovascular disease, is clearly needed.^26,28,41

When new clinical results are reported, there is often a several-year lag, significant geographic variation in practices, and an underuse in the steady state.^14,35 More than 5 years had passed since the first clinical trial results and the time of our study. In a national survey of attitudes using clinical scenarios, 96% of physicians reported that they were "very" or "somewhat" likely to use anticoagulation therapy for a 65-year-old individual with an embolic stroke. The rate decreased only slightly to 93% for a 75-year-old individual.³² We conclude that physician awareness and acceptance are not likely to explain a large portion of the underutilization we observed.

HOW CAN the underutilization of anticoagulation (and antiplatelet) therapy be addressed? Clinical management is becoming increasing complex. Clinicians must integrate data from many different sources with the results of new clinical trials and management guidelines. The amount of data, especially in the acute care setting, can be overwhelming. A hospital-based study by Bedell and colleagues⁴² reported that about 1 in 10 cardiac arrests could be prevented if clinicians had paid closer attention to available data. It is likely that a similar effect contributes to the underutilization of anticoagulation since, despite low rates in practice, most physicians report knowledge and acceptance of the importance of anticoagulation.

Simple, individualized reminders can change physician behavior. Reuben and colleagues⁴³ documented this in the outpatient setting. A simple low-cost intervention using personalized feedback via telephone or written follow-up could achieve high implementation and adherence rates with clinical guideline treatment recommendations.

In the hospital setting, computer-based alerts seem an ideal way to improve compliance with treatment guidelines and allow clinicians to better integrate scientific knowledge with information about individual patients.⁴⁴ In a study of computerized alerts at an academic medical center, Rind and colleagues⁴⁵ reported a 55% drop in the risk of serious renal impairment among patients with elevated creatinine levels who were receiving nephrotoxic medications. In addition, this type of feedback may help counteract the perception of physicians who may be biased to the recall of adverse events such as hemorrhage, as opposed to the much larger group of patients who are seen without adverse events.^35,46

An additional step to further future performance improvements would be to have physicians clearly document their reasons for not using therapy. Our results indicate that a cause is usually not given. By improving this documentation, feedback could be provided not only to clinicians but also to investigators and educators to assist in refining educational/quality improvement efforts (ie, documenting most common characteristics associated with the underuse of warfarin). Focused efforts targeting the most common misperceptions may help increase compliance.

Even when there is a decision to use warfarin in atrial fibrillation, there is often inadequate levels of anticoagulation.¹⁶ Decision support systems can shorten the time to initiate effective therapy and increase the portion of time in the therapeutic range.⁴⁷

There are several limitations to our study. It was restricted to a single state; however, we included all non–Veterans Affairs hospitals in Connecticut. Because of this, it was less likely to be subject to the selection biases inherent in studies restricted to particular practices or related medical centers. The limitations of the International Classification of Diseases (ICD-9-CM) in identifying the diagnosis of stroke have been documented,⁴⁸ but this approach identifies the majority of stroke events.⁴⁹ In addition, as with any retrospective study, there may be reasons for the practice patterns we observed that were not documented in the medical records. We did not differentiate between intermittent and persistent atrial fibrillation. This should not significantly influence our results as the risk of stroke is significantly elevated in both circumstances,⁵ and published guidelines^8- 11 recommend anticoagulation for both.

Increased potential benefit was not associated with a greater use of warfarin. The lack of effect of additional vascular risk factors on increasing the rate of warfarin use suggests that the underuse of therapy is not a reflection of perceived risk of stroke. Similarly, an increasing risk of anticoagulation was not associated with use of warfarin. The slight effect of risk factors for hemorrhage on lowering the rate of warfarin use suggests that the underuse reflects in small part clinical judgments. The judgments and the overall low rate of anticoagulation and antiplatelet therapy are not consistent with published trial and national guidelines.

Anticoagulation of elderly stroke patients with atrial fibrillation, even among ideal candidates, is underused. Although this rate is higher than for primary prevention, the low use even among ideal candidates indicates that there is room for the wider application of these preventive therapies following stroke. Acute care hospitalization is a ready-made opportunity to initiate anticoagulation, and the increased use of warfarin among these patients represents an excellent opportunity for reducing the risk of reinfarction in this high-risk population. Our work provides a benchmark in defining current practices and suggests opportunities for future performance improvement efforts, education, and treatment.

Innovative educational and system improvement projects must be devised and tested for increasing the use of clinical and cost-effective preventive therapies. Those that are shown to be the most effective should be implemented as part of governmental and public health care programs to improve outcome, reduce mortality, and increase the cost-effectiveness of care for patients at risk for vascular disease.

Accepted for publication June 2, 1998.

This study was funded as part of the ongoing quality improvement programs of the Connecticut Peer Review Organization, Middletown.

The analyses on which this publication is based were performed under contract No. 500-96-P549, titled "Utilization and Quality Control Peer Review Organization for the State of Connecticut," sponsored by the Health Care Financing Administration, Department of Health and Human Services.

The content of this publication does not necessarily reflect the views or policies of the Department of Health and Human Services, nor does mention of trade names, commercial products, or organizations imply endorsement by the US government. The author assumes full responsibility for the accuracy and completeness of the ideas presented. This article is a direct result of the Health Care Quality Improvement Program initiated by the Health Care Financing Administration, which has encouraged identification of quality improvement projects derived from analysis of patterns of care, and therefore required no special funding on the part of this contractor. Ideas and contributions to the author concerning experience in engaging with issues presented are welcomed.

Dr Krumholz is a Paul Beeson faculty scholar.

Presented in part at the 22nd International Joint Conference on Stroke and Cerebral Circulation, Anaheim, Calif, February 7, 1997.

Reprints: Lawrence M. Brass, MD, Yale Stroke Program, Department of Neurology, LCI-700, Yale University School of Medicine, 15 York St, PO Box 208018, New Haven, CT 06520-8018 (e-mail: lawrence.brass@yale.edu).