Methods
Patients
We identified patients with AF newly diagnosed during FY 2002 through 2011 (i.e., between October 1, 2001, and September 30, 2011), using the VA Outpatient Care files and Inpatient Treatment files and International Classification of Diseases, 9th Revision, Clinical Modification [ICD-9-CM] diagnosis code 427.31. Patients were deemed to have AF if they had at least 2 encounters with a primary or secondary AF diagnosis within 120 days of each other. Episodes were deemed new if the patient had no other AF-related encounters during the 12 months prior. A total of 406,168 episodes met these criteria. Episodes in which patients were not receiving care in the VA for at least 12 continuous months before the AF diagnosis date were excluded (n = 94,561) because this lack of enrollment precluded us from knowing that the AF episode was truly new. In addition, approximately 1% of patients had >1 eligible AF episode during the study period (ie, patients had ≥2 AF episodes separated by at least 12 months). We selected the first episode for patients with multiple episodes, leaving 297,611 episodes during the 10-year period. Finally, 20,635 episodes that were preceded by oral antiarrhythmic medication or pharmacologic or electric cardioversion within 90 days before the AF diagnosis were excluded, leaving 275,559 eligible episodes.
Variable Definition
We searched VA Decision Support System National Pharmacy Extracts for prescriptions filled by VA pharmacies. Drugs were identified by the VA identification number, which corresponds to national drug codes associated with generic names of drugs. Three types of rate control medications were identified with input and guidance from the clinicians (A.M. and P.C.) and pharmacist (E.C.) involved in this project. Medications were classified according to mechanism of action: β-blockers (eg, atenolol, carvedilol, metoprolol, propranolol), calcium channel blockers (eg, diltiazem, verapamil), and digoxin. Oral rhythm-controlling medications included amiodarone, dofetilide, dronedarone, flecainide, profafenone, sotalol, disopyramide, and quinidine. We confirmed that all oral medications were in doses approved for maintenance therapy because some antiarrhythmics (eg, flecainide) have been approved for cardioversion in high doses. We defined the use of each medication type as a binary variable indicating any use within 90 days of the AF diagnosis date.
Patients' demographic characteristics were identified using information in the VA administrative data, including age, sex, race (categorized as white, black, Hispanic, other nonwhite, and missing), region of the country where the patient resided (categorized as Northeast, South, Midwest, and West), and VA enrollment priority (categorized as patients with service-connected disabilities, low income, and other). Patient comorbid conditions were identified on inpatient and outpatient claims during the 12 months before the AF diagnosis date. Using algorithms originally developed by Elixhauser et al and updated by Quan et al, we identified the presence of 31 possible comorbid conditions (eg, diabetes, hypertension, renal disease). Additional conditions relevant to AF were identified, including other dysrhythmias (ICD-9-CM codes 427.xx, excluding 427.31), cardiomyopathy (ICD-9 codes 425.x), cardiac conduction disorder (eg, bundle-branch block, ICD-9 codes 426.x), previous implantable pacemaker (ICD-9 codes V45.01 and V53.31), and previous implantable cardiac defibrillator (ICD-9 codes V45.02 and V53.32). Previous cerebrovascular events were identified using a previously published algorithm and included cerebral infarction (ICD-9-CM codes 433.01, 433.11, 433.21, 433.31, 433.81, 433.91, 434.01, 434.11, 434.91) and transient ischemic attacks (ICD-9-CM code 435.x). We also calculated the CHADS score for each patient, as has been done in prior studies based on administrative data. This score represents the risk of stroke based on a point system reflecting the presence of congestive heart failure (1 point), hypertension (1 point), age ≥75 years (1 point), diabetes (1 point), and previous stroke (2 points).
Statistical Analysis
Characteristics of patients with AF episodes during FY2002-2003 and FY2010–2011 were compared using χ statistics and analysis of variance. The proportion of patients receiving at least 1 medication in each medication category and the proportions receiving individual medications were calculated in 2-year increments over the study period. Trends in the use of each medication type as well as medication combinations were evaluated using the Cochran-Armitage test for trend. Additional analysis used logistic regression to estimate risk-adjusted rates of medication use while controlling for patient demographic and comorbid conditions, using a statistical criterion of P < .01 to identify variables eligible for inclusion in multivariable models. The final models for receipt of oral rate and rhythm control medication are shown in the online Appendix. The method of recycled predictions was used to calculate the risk-adjusted rates of medication use. This method allows comparison of rates over time by calculating predicted medication use rates assuming a constant pool of patients over time.
Finally, all analyses were repeated separately for patients who did and did not undergo cardiac ablation or cardioversion (electrical or pharmacologic) within 90 days after AF diagnosis. These analyses were conducted to determine whether these interventions have impacted the use of oral rate or rhythm control. Cardiac ablation was defined using Current Procedural Terminology code 93651 or ICD-9-CM procedure code 37.34. Cardioversion was defined by Current Procedural Terminology codes 92960 to 92961; ICD-9-CM procedure code 99.61 to 99.62; or by administration of amiodarone, ibutilide, or procainamide intravenously. For each patient, we identified cardioversion and ablation occurring within 90 days after the AF diagnosis. Because we wanted to control for the use of these procedures as the first treatment strategy (ie, excluding procedures that occur after initiation of oral medication), cardioversion or ablation that occurred after an oral maintenance medication was dispensed were not included.
We used SAS version 9.3 (SAS Institute, Cary, NC) for all analyses. This study was approved by the institutional review board of the Iowa City VA Medical Center.
This work was supported by a grant from the Agency for Healthcare Research and Quality (AHRQ K18HS021992). The authors are solely responsible for the design and conduct of this study, all study analyses, the drafting and editing of the manuscript, and its final contents.