Warfarin Use among Ambulatory Patients with Nonvalvular Atrial Fibrillation: The AnTicoagulation and Risk Factors in Atrial Fibrillation (ATRIA) Study

Atrial fibrillation is the most common, potent risk factor for ischemic stroke. Nonrheumatic atrial fibrillation, the predominant form in the United States, occurs in nearly 6% of persons 65 years of age or older (1). Atrial fibrillation is an independent risk factor for stroke, increasing the annual risk by fivefold and accounting for approximately 15% of all strokes in the United States (2). Over the past decade, multiple randomized trials have demonstrated that warfarin therapy can reduce the average annual risk for ischemic stroke by two thirds, from 4.5% to 1.4%, in patients with nonvalvular atrial fibrillation (2). This benefit was accompanied by a relatively low annual bleeding rate (1.3%) (2-4). However, recent data suggest that these dramatic findings have not been adequately implemented in clinical practice (5-9). Previous studies of warfarin use in atrial fibrillation have primarily been performed among hospitalized patients (10-17) or those in long-term care facilities (18-20). By focusing on hospitalized patients or those receiving long-term care, who may be older and have more comorbid illnesses than ambulatory patients, previous estimates of warfarin use may not be generalizable to the majority of patients with atrial fibrillation, who are treated primarily in outpatient clinics. The few studies that have assessed the outpatient use of warfarin for atrial fibrillation have yielded widely varying results (5, 7, 10, 21, 22). Temporal trend data reveal an increase in warfarin use among outpatients with atrial fibrillation, from 7% in 1980-1981 (before the publication of randomized trials demonstrating the efficacy of anticoagulation [23-27]) to 33% to 50% in 1996 (5, 7). Available studies of ambulatory patients with atrial fibrillation have generally included relatively small numbers of patients (5, 8, 10, 21, 28-30), used varying methods for identifying patients with atrial fibrillation (5, 8, 10, 21, 28-30), or estimated warfarin use on the basis of clinic visits rather than person-level analyses (7, 22). Furthermore, despite the rapid growth of managed care, especially in terms of increased use by elderly persons, little is known about the quality of care for atrial fibrillation provided by these organizations. To address these issues, we assembled a large, contemporary sample of ambulatory patients with nonvalvular atrial fibrillation treated in a health maintenance organization and assessed the prevalence and determinants of warfarin use for stroke prevention. Methods Data Sources and Study Sample Between 1 July 1996 and 31 December 1997, we assembled a cohort of ambulatory patients with atrial fibrillation in the Kaiser Permanente Medical Care Program in Northern California. Kaiser Permanente is a group-model health maintenance organization in which 52% of the 1.9 million adult members were women; 16% of adult men and 15% of adult women were 65 years of age or older. Our goal was to identify patients with atrial fibrillation who were eligible for warfarin therapy. To that end, we identified all patients who had 1) a diagnosis of atrial fibrillation (code 427.31 from the International Classification of Diseases, Ninth Revision, Clinical Modification [ICD-9]] recorded in an automated outpatient database plus an electrocardiogram showing atrial fibrillation found in an electrocardiographic database or 2) more than one outpatient diagnosis of atrial fibrillation during the study period. For each of these groups, the date of the first diagnosis of atrial fibrillation (according to outpatient visit or electrocardiogram) was considered the index date. The electrocardiographic database included physician-confirmed diagnoses obtained since 1994 by using the same model of electrocardiograph (Hewlett-Packard model no. XLi) throughout all facilities in clinics, emergency departments, and inpatient wards. We also reviewed the medical records of a random sample of 50 patients in the subgroup who had serial outpatient diagnoses of atrial fibrillation but no electrocardiogram showing atrial fibrillation in the available database (23% of the total cohort). A high proportion of these patients (80%) had at least one 12-lead electrocardiogram demonstrating atrial fibrillation in their records. Almost all of the electrocardiograms with atrial fibrillation found in medical records were dated before the electrocardiographic database was developed. Therefore, we estimate that nearly all of the cohort identified by the above approaches had atrial fibrillation. In addition, according to review of the medical records of a random sample of 115 patients with nonvalvular atrial fibrillation and serial electrocardiograms, 21% had paroxysmal atrial fibrillation. The target population for warfarin analyses was ambulatory adult patients with nonvalvular atrial fibrillation who were receiving outpatient care after their index date. Therefore, we excluded patients who met any of the following criteria: no health plan membership after diagnosis of atrial fibrillation, age younger than 18 years, transient atrial fibrillation secondary to a recent cardiac surgery, mitral stenosis or mitral or aortic valve repair or replacement, concomitant hyperthyroidism, or no outpatient internal medicine or cardiology care during the 12 months after index date. Inpatient diagnoses used for exclusion purposes were identified by using comprehensive databases of discharge diagnoses from Kaiser Permanente hospitalizations and billing claims for members treated on an emergent basis at non-Kaiser Permanente facilities. Outpatient diagnoses were identified by using a database of assigned diagnoses for all ambulatory encounters (outpatient clinics, urgent care, and emergency department). Transient, perioperative atrial fibrillation was defined as having one of the following procedures up to 30 days before a single diagnosis of atrial fibrillation: coronary artery bypass surgery (ICD-9 codes 36.10 to 36.19), pericardial surgery (ICD-9 codes 37.10 to 37.12, 37.31 to 37.33, or 37.40), or structural cardiac repair (ICD-9 codes 35.00 to 35.04, 35.31 to 35.39, 35.41 to 35.42, 35.50 to 35.56, 35.60 to 35.63, or 35.70 to 35.73). Valvular heart disease was defined as an inpatient or outpatient diagnosis of mitral stenosis or prosthetic heart valve (ICD-9 codes 394.0, 394.2, 396.0, 396.1, 396.8, V43.3, or V42.2) or mitral or aortic valve repair or replacement (ICD-9 codes 35.10 to 35.14 or 35.20 to 35.28). Concomitant hyperthyroidism was defined as having any of the following factors within 12 months before the index date: an inpatient or outpatient diagnosis of hyperthyroidism or thyrotoxicosis (ICD-9 codes 242.0 to 242.9); a low serum level of thyroid-stimulating hormone (<0.03 g/mL) found in laboratory databases; or a prescription for an antithyroid agent (methimazole or propylthiouracil) found in pharmacy databases. Predictor Variables Patient age and sex at the index date were identified from administrative files. On the basis of the corresponding ICD-9 codes, we searched the ambulatory visit database between 1 June 1994 and 31 December 1997 and hospital discharge and billing claims databases in the 5 years before the index date to identify risk factors for stroke (besides older age) (31) and potential contraindications to warfarin therapy. For risk factors for stroke, inpatient and outpatient sources were used to identify previous stroke and congestive heart failure and outpatient sources were used to detect hypertension. Information on diabetes mellitus and ischemic heart disease (possible risk factors for stroke) was obtained from outpatient sources only. For potential contraindications to warfarin, inpatient and outpatient databases were used to detect a history of cirrhosis, hepatitis, or seizure disorder; inpatient sources were used to detect previous intracranial hemorrhage, gastrointestinal hemorrhage, hospitalization for other bleeding, and hospitalizations involving a mechanical fall; and the outpatient database was used to detect dementia. Renal insufficiency was determined on the basis of diagnoses in the outpatient database or a serum creatinine concentration of 221 mol/L or more ( 2.5 mg/dL) in the laboratory database. To assess the utility of the outpatient database, we used the statistic to compare selected diagnoses (previous stroke, congestive heart failure, hypertension, diabetes, ischemic heart disease, and dementia) found in this database with data from outpatient medical records in 295 randomly selected patients with atrial fibrillation (32). Warfarin Use Warfarin use was defined as having either a filled prescription for warfarin or dicumarol in the pharmacy database, more than one outpatient international normalized ratio (based on prothrombin time measurement), or a diagnosis of Coumadin therapy (ICD-9 code V58.61) in the ambulatory visit database in the 3 months before or after the first identified diagnosis of atrial fibrillation during the study period, or any combination of these factors. More than 99% of prescriptions were for warfarin; thus, use of warfarin or dicumarol was referred to as warfarin use. More than 95% of the patients had a pharmacy benefit, and all patients were fully covered by insurance for laboratory tests. The 1% of patients without a pharmacy benefit or an outpatient international normalized ratio were considered to have unknown warfarin status and were excluded from all analyses. We used the statistic to validate our approach to assigning warfarin status by chart review of 98 randomly sampled patients (45 users and 53 nonusers) (32). Statistical Analysis The Student t-test or Wilcoxon rank-sum test for continuous variables and the chi-square test for proportions were used to compare demographic characteristics and the prevalence of risk factors for stroke and potential contraindications to therapy in warfarin users and nonusers. Descriptive analyses of warfarin use were confined to eligible patients who h