Author + information
- Received February 6, 2017
- Revision received April 13, 2017
- Accepted April 28, 2017
- Published online June 26, 2017.
- Alexander C. Perino, MDa,b,
- Jun Fan, MSb,
- Susan K. Schmitt, PhDb,
- Mariam Askari, BSb,
- Daniel W. Kaiser, MDa,b,
- Abhishek Deshmukh, MBBSd,
- Paul A. Heidenreich, MD, MSa,b,
- Christopher Swan, MDa,b,
- Sanjiv M. Narayan, MD, PhDa,b,
- Paul J. Wang, MDa,b and
- Mintu P. Turakhia, MD, MASa,b,c,∗ ()
- aDepartment of Medicine, Stanford University School of Medicine, Stanford, California
- bVeterans Affairs Palo Alto Health Care System, Palo Alto, California
- cCenter for Digital Health, Stanford University School of Medicine, Stanford, California
- dMayo Clinic, Rochester, Minnesota
- ↵∗Address for correspondence:
Dr. Mintu P. Turakhia, Palo Alto VA Health Care System, Stanford University, 3801 Miranda Avenue, 111C, Palo Alto, California 94304.
Background Atrial fibrillation (AF) occurs in many clinical contexts and is diagnosed and treated by clinicians across many specialties. This approach has resulted in treatment variations.
Objectives The goal of this study was to evaluate the association between treating specialty and AF outcomes among patients newly diagnosed with AF.
Methods Using data from the TREAT-AF (Retrospective Evaluation and Assessment of Therapies in AF) study from the Veterans Health Administration, patients with newly diagnosed, nonvalvular AF between 2004 and 2012 were identified who had at least 1 outpatient encounter with primary care or cardiology within 90 days of the AF diagnosis. Cox proportional hazards regression was used to evaluate the association between treating specialty and AF outcomes.
Results Among 184,161 patients with newly diagnosed AF (age 70 ± 11 years; 1.7% women; CHA2DS2-VASc score 2.6 ± 1.7), 40% received cardiology care and 60% received primary care only. After adjustment for covariates, cardiology care was associated with reductions in stroke (hazard ratio [HR]: 0.91; 95% confidence interval [CI]: 0.86 to 0.96; p < 0.001) and death (HR: 0.89; 95% CI: 0.88 to 0.91; p < 0.0001) and increases in hospitalizations for AF/supraventricular tachycardia (HR: 1.38; 95% CI: 1.35 to 1.42; p < 0.0001) and myocardial infarction (HR: 1.03; 95% CI: 1.00 to 1.05; p < 0.04). The propensity-matched cohort had similar results. In mediation analysis, oral anticoagulation prescription within 90 days of diagnosis may have mediated reductions in stroke but did not mediate reductions in survival.
Conclusions In patients with newly diagnosed AF, cardiology care was associated with improved outcomes, potentially mediated by early prescription of oral anticoagulation therapy. Although hypothesis-generating, these data warrant serious consideration and study of health care system interventions at the time of new AF diagnosis.
Atrial fibrillation (AF), the second most common cardiovascular condition after hypertension (1,2), requires management in diverse clinical contexts and challenges providers of varied backgrounds with complex treatment decisions. Primary care providers and cardiologists, including cardiology subspecialists such as electrophysiologists, are tasked with caring for the majority of patients with AF and managing their increased risk of stroke, myocardial infarction (MI), heart failure, and death (3–7). Oral anticoagulation (OAC) therapy, with warfarin or non–vitamin K oral anticoagulant agents, has been shown to prevent stroke (8,9). However, OAC prescription is not uniform across treating specialties, and cardiology care is associated with higher odds of receipt of OAC therapy (10–13). Similarly, rates of antiarrhythmic prescription are higher in cardiology-treated patients (11,12). Associations of treating specialty with other aspects of AF care (e.g., catheter ablation) have not been adequately studied.
Despite these data, there has been limited investigation into the effects of treating specialty assignment on outcomes of AF (14). Therefore, the goal of the present study was to evaluate the association of treating specialty and clinical outcomes in AF by using observational data.
TREAT-AF (Retrospective Evaluation and Assessment of Therapies in AF) is a retrospective cohort study of patients with newly diagnosed AF treated in the Department of Veterans Affairs (VA) national health care system between October 1, 2003, and September 30, 2012 (VA fiscal years 2004 to 2012). Datasets used represent the claims data and electronic health records covering the full denominator of VA users. These include data from the VA National Patient Care Database (15), the VA Decision Support System national pharmacy extract (16), the VA Fee Basis Inpatient and Outpatient datasets, the VA Laboratory Decision Support System extract (17), and the VA Vital Status File that contains validated combined mortality data from VA, Medicare, and Social Security Administration sources (18,19). Methods for cohort creation have been previously described in detail (11,20).
Patients with newly diagnosed AF were defined as having a primary or secondary diagnosis of AF (International Classification of Diseases-9th Revision, codes 427.31 or 427.32) associated with an inpatient or outpatient VA encounter, who had no previous diagnosis of AF within 4 years, and had a second confirmatory diagnosis of AF between 30 and 365 days after the index AF diagnosis. Patients were excluded for the following: 1) if they were not seen in outpatient cardiology or primary care clinics within 90 days of the index AF diagnosis; 2) they were not seen in outpatient cardiology or primary care clinics within the continental United States; 3) they did not receive any outpatient prescriptions within 90 days of the index AF diagnosis; 4) they died within 120 days of the index AF diagnosis; or 5) data on age and sex were not available.
The primary “predictor” was the treating outpatient specialty. The methodology has been previously detailed and applied (11). Patients seen in a cardiology clinic within 90 days of an AF diagnosis, regardless of whether they were seen in primary care clinics, were categorized as having received cardiology care. Patients seen in primary care clinics within 90 days of an AF diagnosis, who were not seen in a cardiology clinic within 90 days of an AF diagnosis, were categorized as having received primary care only. For cardiology clinics, we included encounters from general and specialty cardiology clinics (e.g., heart failure, electrophysiology), which are classified according to a single clinic type code, but not cardiac surgery or hypertension clinics because they are typically not staffed by cardiologists in the VA system. For primary care clinics, we only included encounters from clinics designated as general internal medicine or primary care.
The primary outcome was ischemic stroke. Secondary outcomes included death and cardiovascular hospitalizations (transient ischemic attack [TIA], heart failure, AF or supraventricular tachycardia [SVT], or MI). The incidence of stroke and other cardiovascular hospitalizations was determined from inpatient claims and medical record data by using validated and previously applied algorithms (21). For ascertainment of death, the VA Vital Status file was used; this file has previously been shown to have 97.6% agreement and 98.3% sensitivity for detection of deaths identified by using the National Death Index (19). Because cause of death was not available, deaths preceded by a cardiovascular hospitalization within 30 days were also assessed. International Classification of Diseases, Ninth Revision, codes used to define cardiovascular death and hospitalizations are available in Online Table 1 and have also been previously detailed (22).
Cox regression was performed to estimate hazard ratios (HRs) for outcomes of interest, adjusting for patient demographic characteristics (age, patient distance to clinic/medical center, race, sex, and VA priority status), patient comorbidities (Charlson and Selim comorbidity indices, diabetes, glomerular filtration rate, heart failure, hypertension, MI, and stroke/TIA), cardiovascular non-AF medications (antiplatelet agents, angiotensin-converting enzyme inhibitors/angiotensin receptor blockers, diuretics, niacin/fibrates, statins), and anticoagulation, rate control, and antiarrhythmic medications (amiodarone, anticoagulation [warfarin/non–vitamin K oral anticoagulant agents], beta-blockers, calcium-channel blockers, class I agents, class III agents, and digoxin).
In parallel, we also performed a propensity-matched analysis, applying multivariate logistic regression to develop a propensity score by using all baseline covariates to predict the probability of evaluation in a cardiology clinic. Cox regression was performed in a 1:1 matched cohort using nearest-neighbor matching without replacement. Methods for propensity-matched cohort creation and analysis have been previously described in detail (11,20).
Mediation analysis was performed to determine whether the association of cardiology care on outcomes (direct effect) was mediated by receipt of anticoagulation (OAC) therapy within 90 days of AF diagnosis (indirect effect). Mediation was assessed for in a stepwise fashion by using the Baron and Kenny approach (23).
All analyses were performed by using SAS version 9.2 (SAS Institute, Inc., Cary, North Carolina) and STATA version 11.0 (Stata Corp, College Station, Texas). The study was approved by the local institutional review board.
The full cohort included 184,161 patients newly diagnosed with AF (age 70 ± 11 years; 1.7% women; CHA2DS2-VASc score 2.6 ± 1.7) (Figure 1), of whom 69,901 received cardiology care and 114,260 received primary care only within 90 days of AF diagnosis. Cardiology-treated patients lived closer to medical centers, were slightly younger, and, despite similar CHA2DS2-VASc scores, had a higher baseline prevalence of hypertension, diabetes, coronary disease, MI, and stroke (Table 1) compared with patients treated by primary care alone.
Compared with patients treated only by primary care practitioners, cardiology-treated patients had a substantially higher proportion of 90-day receipt of OAC therapy (70.3% vs. 58.8%; p < 0.0001), AF rate control agents (90.1% vs. 80.5%; p < 0.0001), AF rhythm control agents (20.8% vs. 11.0%; p < 0.0001), and non-AF cardiovascular medications (antiplatelet agents [42.6% vs. 28.6%; p < 0.0001], statins [65.6% vs. 58.1%; p < 0.0001], and angiotensin-converting enzyme inhibitors/angiotensin receptor blockers [66.0% vs. 57.5%; p < 0.0001]) (Table 1). Cardiology-treated patients also had higher VA priority status (benefits) (priority status 1: 19.3% vs. 16.6%; p < 0.0001) and shorter differential distance to care (difference in distance to nearest VA primary care clinic vs. distance to nearest VA medical center facility [cardiology clinic]) (21.4 ± 36.9 miles vs. 28.9 ± 42.4 miles; p < 0.0001).
Patients who received cardiology care had lower unadjusted incidence rates of stroke (7.6 per 1,000 person-years vs. 8.8 per 1,000 person-years; p < 0.0001), overall mortality (69.4 per 1,000 person-years vs. 85.3 per 1,000 person-years; p < 0.0001), and mortality preceded by cardiovascular hospitalization within 30 days (15.5 per 1,000 person-years vs. 18.8 per 1,000 person-years; p < 0.0001) compared with patients treated by primary care practitioners only (Table 2). However, cardiology-treated patients had higher unadjusted incidence rates of hospitalizations for heart failure (53.4 per 1,000 person-years vs. 44.1 per 1,000 person-years; p < 0.0001), AF/SVT (78.2 per 1,000 person-years vs. 43.1 per 1,000 person-years; p < 0.0001), and MI (57.8 per 1,000 person-years vs. 45.7 per 1,000 person-years; p < 0.0001). There were no differences in rates of TIA between cardiology and primary care–treated patients. In the propensity-matched cohort analysis, incidence rates were similar without substantial attenuation. Primary cardiovascular discharge diagnoses for hospitalizations within 30 days of cardiovascular death are presented in Online Table 2.
After adjustment for covariates, cardiology care was associated with a lower risk of stroke (HR: 0.91; 95% confidence interval [CI]: 0.86 to 0.96; p < 0.001), overall mortality (HR: 0.89; 95% CI: 0.88 to 0.91; p < 0.0001), and mortality preceded by cardiovascular hospitalization within 30 days (HR: 0.88; 95% CI: 0.84 to 0.91; p < 0.0001). Cardiology care was associated with increased risk of hospitalization for AF/SVT (HR: 1.38; 95% CI: 1.35 to 1.42; p < 0.0001) and MI (HR: 1.03; 95% CI: 1.00 to 1.05; p < 0.04). There was no significant association with risk of heart failure hospitalization or TIA. In the propensity-matched cohort, results were similar to the full cohort with the exception of cardiology care being associated with a small, but significant, increased risk of heart failure hospitalization (HR: 1.07; 95% CI: 1.04 to 1.10; p < 0.0001) (Table 3).
In mediation analysis, there was a statistical trend for partial mediation of the association of cardiology care to reduced risk of stroke by 90-day OAC receipt (indirect effect: 0.957; 95% CI: 0.909 to 1.007; p = 0.09) (Online Table 3). Results were similar in the propensity-matched cohort (indirect effect: 0.932; 95% CI: 0.868 to 1.001; p = 0.0538) (Online Table 4). Mediation effects were nonsignificant for death or death preceded by cardiovascular hospitalization within 30 days.
In patients with newly diagnosed AF in the VA health care system, cardiology care (compared with primary care only) within 90 days of an AF diagnosis was associated with a reduced risk of stroke, death, and death preceded by cardiovascular hospitalization within 30 days. However, there was an increased risk of hospitalization for AF/SVT and MI. Early OAC prescription (within 90 days of AF diagnosis) was substantially higher in cardiology-treated patients (70.3% vs. 58.8%; p < 0.0001), which may have partially mediated reductions in stroke but did not mediate reduction in death (Central Illustration).
Previous studies have investigated the effect of treating specialty on OAC prescription, a cornerstone of AF management and plausible mediator for improved outcomes. In a previous analysis of the TREAT-AF cohort using data from 2004 to 2008, cardiology care was associated with a higher rate of warfarin prescription after adjustment for covariates and site-level factors (odds ratio [OR]: 2.05; 95% CI: 1.99 to 2.11; p < 0.0001) (11). Similar results were obtained in patients with AF with previous stroke (CHA2DS2-VASc score ≥2) in the AFNET (German Competence Network on Atrial Fibrillation) registry, which found that noncardiology care was associated with a lower rate of warfarin prescription after adjustment for covariates (OR: 0.40; 95% CI: 0.21 to 0.77; p < 0.001) (13). ORBIT-AF (Outcomes Registry for Better Informed Treatment of Atrial Fibrillation) found a nonsignificant trend toward noncardiology-treated patients having a lower rate of OAC prescription compared with patients treated by general cardiologists (OR: 0.73; 95% CI: 0.49 to 1.09; p = 0.12) (12). Notably, ORBIT-AF stratified electrophysiologists and general cardiologists, reducing power and excluding cardiologists with the highest rate of OAC prescription from comparison with noncardiology-treated patients.
Effect of treating specialty on cardiovascular outcomes has previously been investigated for other disease states, most notably for heart failure. Cardiology-treated patients were found to have reduced rates of readmission and mortality (24–26). In an analysis more reflective of contemporary treatment models, the effect of, and interaction between, treating specialty on outcomes for Medicare patients hospitalized for heart failure found that treatment by primary and cardiology care had lower readmission rates and performed better with respect to heart failure quality of care measures, compared with primary or cardiology care (27). These studies did not evaluate subgroups or outcomes related to AF.
In the present analysis, stroke (the most frequent major adverse outcome associated with AF after controlling for covariates) was reduced 9% in patients who received cardiology care, and cardiology-treated patients were substantially more likely to receive early prescription of OAC therapy. Early OAC prescription by cardiology is a plausible indirect pathway for cardiology care to influence stroke, and mediation analysis is needed to explore variable relationships. In mediation analysis, loss of association between the independent (cardiology care) and dependent (stroke) variables after addition of the mediator variable (early OAC prescription) to regression models is consistent with complete mediation, whereas an attenuated but still significant association between the independent and dependent variables is consistent with partial mediation. Our mediation analyses showed indirect effects of borderline statistical significance, which may still plausibly demonstrate partial mediation of stroke by early OAC treatment.
These findings warrant serious consideration of care pathways for patients with AF soon after diagnosis, identification of additional mediators of improved outcomes and exploration into the scalability of these interventions across health care settings, and innovative health care delivery models. Streamlined care pathways for patients with AF will require integration into currently operating health care systems and adaptation to ongoing system reform. Investigation using both qualitative and quantitative methods would be informative in determining barriers to appropriate cardiology care. Even so, these data highlight notable differences, and health care system interventions may be warranted. AF specialty clinics, focusing on patient education and using decision support software based on AF guidelines, have been evaluated in small observational and randomized studies, with promising reductions in cardiovascular mortality, hospitalization, and stroke (14,28). In addition, these care models have been shown to be cost-effective (29). With increasing emphasis on evidence-based systems reform to provide high value care, point of care or pragmatic trials evaluating AF specialty clinics and novel care models for newly diagnosed AF should be considered as a next step.
We also found reduced mortality in cardiology-treated patients (11% after controlling for covariates). There are several plausible mechanisms that may explain this observation. Central to the care of patients with AF is appropriate OAC prescription, which in addition to preventing stroke has been associated with reductions in all-cause mortality and stroke-related mortality (30–32). We did not find that OAC prescription mediated reduced mortality among cardiology-treated patients, which may be attributable to limitations in our mediator variable (which does not capture duration and adequacy of anticoagulation). Lack of mediation of mortality preceded by cardiovascular hospitalization within 30 days may also be due to anticoagulation only being protective against specific cardiovascular mortality subtypes.
In addition to reducing stroke and mortality through AF care, cardiology care may improve outcomes through improved care of non-AF cardiovascular conditions. Cardiology-treated patients in our cohort had higher rates of statin prescription and antiplatelet agents, which can reduce stroke and improve survival (33,34). These findings could indicate that the AF diagnosis acts as an entry point to cardiology care and subsequent management of cardiovascular risk factors. Also, survival advantages might be conferred from collaborative cardiology and primary care (as opposed to cardiology care only), with primary care able to focus limited resources on optimal management of noncardiovascular conditions when AF (and non-AF cardiovascular conditions) is managed by cardiology.
Despite cardiology-treated patients having reduced overall mortality and mortality preceded by cardiovascular hospitalization within 30 days, a paradoxical increase in hospitalization for AF/SVT and MI in both cohorts and heart failure in the propensity-matched cohort was found. Explanations for this increased risk of hospitalization include differential practice patterns and confounding by severity, with highly symptomatic patients with AF more likely to be referred to cardiologists (independent of other baseline characteristics), who may use hospitalization for antiarrhythmic drug initiation. Regarding practice patterns, cardiologists may be more aggressive in using hospitalization to intervene on disease progression (e.g., heart failure optimization), potentially improving long-term outcomes. Notably, increased rates of MI and heart failure in cardiology-treated patients, although statistically significant, may not be clinically significant (MI full cohort HR: 1.03; MI propensity-matched cohort HR: 1.08; heart failure propensity-matched cohort HR: 1.07).
These data have significant limitations owing to the observational design. Motivations for referral to cardiology care may not be captured in structured data. For example, likelihood of referral could be related to AF symptom severity. Also, differences in distance to care suggest assignment to cardiology care is likely not random. Even after accounting for observed baseline differences through 2 sets of techniques, there is substantial risk of unidentified confounders, which could include patient motivation (or medication and behavioral adherence), self-efficacy, or healthy-user effect of specialty care. Cardiology care may be a marker, rather than a cause, of improved patient health status, including lifestyle factors such as diet and exercise. Also, we could not ascertain non-VA specialty care, which could lead to misclassification of primary care–only patients if they were receiving cardiology care using private insurance, for example. Finally, the cohort is from an integrated health care system with a mostly male population. As such, these data remain hypothesis-generating.
In patients with newly diagnosed AF, cardiology care was associated with a reduced risk of stroke, death, and death preceded by cardiovascular hospitalization within 30 days and increased risk of hospitalization for AF/SVT and MI. Early OAC prescription was substantially higher in cardiology-treated patients, which may have mediated reductions in risk of stroke but not mortality. These data warrant serious consideration and study of health care system interventions at the time of new AF diagnosis.
COMPETENCY IN SYSTEMS-BASED PRACTICE: Clinical outcomes differ for patients with AF based on the specialty of the treating physician, with lower risks of stroke and mortality when care is provided by cardiologists within 90 days of diagnosis.
TRANSLATIONAL OUTLOOK: Pragmatic trials evaluating novel care pathways are needed to identify and incorporate high-value approaches typically used by cardiologists for the expanding population of patients with AF.
For supplemental tables, please see the online version of this article.
Dr. Turakhia is supported by a Veterans Health Services Research & Development Career Development Award (CDA09027-1) and an American Heart Association National Scientist Development Grant (09SDG2250647). Dr. Narayan has ownership interest in Abbott. Dr. Wang has received research grants from Medtronic, Siemens, CardioFocus, and ARCA; has received research support from Medtronic, St. Jude Medical, Boston Scientific Corporation, and Biosense Webster; has received honoraria from Janssen Pharmaceuticals, St. Jude Medical, Medtronic, and Amgen; has ownership interest from VytronUS; and has served as a consultant/advisory board member for Janssen Pharmaceuticals, St. Jude Medical, Medtronic, and Amgen. Dr. Turakhia has received research grants from Medtronic; a research grant from Janssen Pharmaceuticals; and has served as a consultant/advisory board member for Medtronic, St. Jude Medical, and Abbott. All other authors have reported that they have no relationships relevant to the contents of this paper to disclose. P.K. Shah, MD, served as Guest Editor-in-Chief for this paper. Ivan Ho, MD, served as Guest Editor for this paper.
- Abbreviations and Acronyms
- atrial fibrillation
- confidence interval
- hazard ratio
- myocardial infarction
- oral anticoagulation
- odds ratio
- supraventricular tachycardia
- transient ischemic attack
- Veterans Affairs
- Received February 6, 2017.
- Revision received April 13, 2017.
- Accepted April 28, 2017.
- Miyasaka Y.,
- Barnes M.E.,
- Gersh B.J.,
- et al.
- Turakhia M.P.,
- Hoang D.D.,
- Xu X.,
- et al.
- Fosbol E.L.,
- Holmes D.N.,
- Piccini J.P.,
- et al.
- Hendriks J.M.,
- de Wit R.,
- Crijns H.J.,
- et al.
- VA Information Resource Center
- Arnold N.,
- Sohn M.-W.,
- Maynard C.,
- et al.
- Turakhia M.P.,
- Santangeli P.,
- Winkelmayer W.C.,
- et al.
- Turakhia M.P.,
- Ziegler P.D.,
- Schmitt S.K.,
- et al.
- Jong P.,
- Gong Y.,
- Liu P.P.,
- Austin P.C.,
- Lee D.S.,
- Tu J.V.
- Reis S.E.,
- Holubkov R.,
- Edmundowicz D.,
- et al.
- Selim A.M.,
- Mazurek J.A.,
- Iqbal M.,
- Wang D.,
- Negassa A.,
- Zolty R.
- Hendriks J.,
- Tomini F.,
- van Asselt T.,
- Crijns H.,
- Vrijhoef H.
- Johnsen S.P.,
- Svendsen M.L.,
- Hansen M.L.,
- Brandes A.,
- Mehnert F.,
- Husted S.E.