Factors associated with rhythm control treatment decisions in patients with atrial fibrillation—Insights from the NCDR PINNACLE registry

Factors associated with rhythm control treatment decisions in patients with atrial fibrillation—Insights from the NCDR PINNACLE registry

Clinical Investigation Factors associated with rhythm control treatment decisions in patients with atrial fibrillation—Insights from the NCDR PINNACL...

706KB Sizes 2 Downloads 34 Views

Clinical Investigation

Factors associated with rhythm control treatment decisions in patients with atrial fibrillation—Insights from the NCDR PINNACLE registry Anil K. Gehi, MD, a Gheorghe Doros, PhD, b,c Thomas J. Glorioso, MS, d,e Gary K. Grunwald, PhD, d,e Jonathan Hsu, MD, f Yang Song, MS, b Mintu P. Turakhia, MD, g Alexander Turchin, MD, MS, b,h Salim S. Virani, MD, PhD, i and Thomas M. Maddox, MD, MS e Chapel Hill, NC; Boston, MA; Aurora, CO; Denver, CO; La Jolla, and Stanford, CA

Background Decisions to use rhythm control in atrial fibrillation (AF) should generally be dictated by patient factors, such as quality of life, heart failure, and other comorbidities. Whether or not other factors affect decisions about the use of rhythm control, and catheter ablation in particular, is unknown. Methods A cohort of all patients diagnosed with nonvalvular AF were identified from the National Cardiovascular Data Registry’s Practice Innovation and Clinical Excellence (PINNACLE) AF registry of US outpatient cardiology practices during the study period from May 1, 2008, to December 31, 2014. Overall and practice-specific rates of rhythm control (cardioversion, antiarrhythmic drug therapy, or catheter ablation) were assessed. We assessed patient and practice factors associated with rhythm control and determined the relative contribution of patient, practice, and unmeasured practice factors with its use. Results

Among 511,958 PINNACLE AF patients, 22.3% were treated with rhythm control and 2.9% underwent catheter ablation. Significant practice variation in rhythm control was present (median rate of rhythm control across practices 22.8%, range 0.2%-62.9%). Significant patient factors associated with rhythm control therapy included white (vs nonwhite) race (odds ratio [OR] 2.43, P b .001), private (vs nonprivate) insurance (OR 1.04, P b .001), and whether a patient was seen by an electrophysiologist (OR 1.77, P b .001). In an analysis of the relative contribution of patient, practice, and unmeasured practice factors with rhythm control, the contribution of unmeasured practice factors (95% range OR 0.29-3.44) exceeded that of either patient (95% range OR 0.46-2.30) or practice (95% range OR 0.15-2.77) factors.

Conclusions One in 5 AF patients in the PINNACLE registry received rhythm control, and 1 in 50 received catheter ablation, suggesting that rhythm control may be underused. A variety of measured and unmeasured practice factors unrelated to patient characteristics play a disproportionate role in the use of rhythm control treatment decisions. Understanding the drivers of these decisions may identify inappropriate treatment variation and better inform optimal use of these therapies. (Am Heart J 2017;187:88-97.)

In atrial fibrillation (AF), rhythm control, using cardioversion, antiarrhythmic drugs, and/or catheter ablation, is a cornerstone of therapy. Current AF management

From the aSection of Cardiac Electrophysiology, Division of Cardiology, University of North Carolina, Chapel Hill, NC, bHarvard Clinical Research Institute, Boston, MA, c Department of Biostatistics, Boston University, Boston, MA, d Department of Biostatistics and Informatics, Colorado School of Public Health, University of Colorado Denver Anschutz Medical Campus, Aurora, CO, eVA Eastern Colorado Health Care System, University of Colorado School of Medicine, and the Colorado Cardiovascular Outcomes Research Consortium, Denver, CO, fSection of Cardiac Electrophysiology, Division of Cardiology, University of California San Diego, La Jolla, CA, gVA Palo Alto Health Care System, Palo Alto, CA, and Department of Medicine Stanford University School of Medicine, Stanford, CA, hDivision of Endocrinology, Diabetes and Hypertension, Brigham and Women's Hospital, Boston, MA, and iMichael E.

guidelines 1 suggest that patient factors—such as diminished quality of life, AF-associated cardiomyopathy, potential efficacy of rhythm-control interventions, patient

DeBakey VA Medical Center and Department of Medicine/Cardiology, Baylor College of Medicine. Funding sources: This research was supported by the American College of Cardiology Foundation's National Cardiovascular Data Registry. Submitted November 21, 2016; accepted February 8, 2017. Reprint requests: Anil K. Gehi, MD, CB 7075, 160 Dental Cir, University of North Carolina, Chapel Hill, NC 27599. E-mail: [email protected] 0002-8703 © 2017 Elsevier Inc. All rights reserved. http://dx.doi.org/10.1016/j.ahj.2017.02.006

American Heart Journal Volume 187, Number 0

age, and patient preference—should drive the decision to use rhythm control. These factors are especially germane in the decision to use the most recent innovation in rhythm control: catheter ablation. The factors underlying the actual use of rhythm control and catheter ablation in contemporary practice have not been thoroughly studied. Factors affecting treatment decisions may exist not only at the patient level but also at the provider or health system level. Examples of practice and health system variation have been seen in many areas of cardiac care, including heart failure treatment, 2 coronary artery disease treatment, 3 and anticoagulation in AF. 4 Whether similar practice effects occur in decisions to pursue AF rhythm control therapies is unknown. Evidence of significant practice variation and influence on rhythm control decisions would suggest that factors independent of guideline-recommended patient characteristics are driving treatment decisions and thus resulting in unnecessary, and potentially suboptimal, variation. To understand these issues in the US AF patient population, we analyzed data from the National Cardiovascular Data Registry (NCDR)’s Practice Innovation and Clinical Excellence (PINNACLE) Registry. Specifically, we determined the prevalence of rhythm control and catheter ablation among AF patients, their variation by practice, and the contribution of patient and practice factors to their use.

Methods Study population Data for our study were drawn from the NCDR PINNACLE registry of US outpatient cardiology practices. 5 Practices participate in the PINNACLE registry on a voluntary basis but include both academic and private practices from all regions of the United States. All AF patients in these participating practices are included in the PINNACLE registry. Data are prospectively collected through a validated electronic medical record mapping algorithm to capture relevant data elements. Collected data elements include patient demographics, medical conditions, vital signs, medications, cardiac diagnostic testing and interventions, and laboratory values. Practice information, such as practice location, provider number, and provider type, is also collected. Data quality of the PINNACLE registry is maintained through rigorous data definitions, standardized data collection and transmission, and periodic data audits 6 performed at St Luke's Mid America Heart Institute (Kansas City, MO), the PINNACLE registry's primary analytic center. The NCDR operates under a quality improvement model with deidentified information that does not require institutional review board approval. For the present study, the study cohort included all patients diagnosed with nonvalvular AF during the study period from May 1, 2008, to December 31, 2014. Given

Gehi et al 89

that current AF guidelines recommend that rhythm control interventions aside from cardioversion alone be considered only for patients with recurrent AF, patients were excluded from the cohort if they were diagnosed with their first episode of AF to not skew the results toward non–rhythm control. Patients were also excluded if there was a history of left ventricular (LV) assist device or cardiac transplant given the uncertainty of rhythm control efficacy or safety in these patients. Initial analyses included only practices with at least 1 patient receiving rhythm control to avoid skewing the results due to practices with incomplete data ascertainment from electronic medical record mapping.

Outcome variables The 2 outcome variables of interest were (1) evidence of any rhythm control treatments (yes/no) at any time during the study period and (2) catheter ablation treatment (yes/no) versus other rhythm control treatments at any time during the study period. Rhythm control treatments were defined as treatment with any of the following: cardioversion, class I antiarrhythmic drug (flecainide, propafenone), class III antiarrhythmic drug (sotalol, dofetilide), other antiarrhythmic drug (dronedarone, amiodarone), or catheter ablation (excluding AV node ablation alone). Predictor variables The primary predictor variables were patient and practice factors. Patient factors included age, gender, insurance type (private vs nonprivate), race (white vs nonwhite), medical comorbidities (diabetes, coronary artery disease, history of myocardial infarction [MI], history of percutaneous coronary intervention [PCI], diabetes mellitus, hypertension, peripheral arterial disease, heart failure, chronic kidney disease, prior cardiac surgery, history of transient ischemic attack/ cerebrovascular accident [TIA/CVA] or systemic embolism, history of hemorrhage, tobacco use), body mass index (BMI), persistent (vs paroxysmal) AF, CHADS2 score, and CHA2DS2-VaSC score, physician versus nonphysician provider, and whether the patient was seen by an electrophysiologist (EP). Nonprivate insurance included military, Medicare, Medicaid, non-US insurance, state-specified plan, none, and missing. Nonwhite race included black, Asian, American Indian/Alaskan native, native Hawaiian/Pacific islander, mixed, and missing. Practice factors included number of providers (b10 vs ≥10), urban/suburban versus rural practice, and practice region (West, Midwest, Northeast, South). Statistical analysis The rates of rhythm control and catheter ablation use were calculated, and comparisons were made between those with and without rhythm control therapy use. Statistical comparisons were made using χ 2 or Fisher

American Heart Journal Month 2017

90 Gehi et al

exact test for categorical variables and Student t tests for continuous variables. Standardized differences between groups were also calculated to aid the reader in the evaluation of differences between groups. In general, standardized differences b0.1 are considered to represent differences of small magnitude. 7 Practice-specific rates of rhythm control and catheter ablation use were also calculated, and ranges were reported using median, maximum, and minimum values. To determine patient and practice factors associated with rhythm control, univariate and multivariable hierarchical logistic regression analyses were conducted using the patient and provider factors described above. Hierarchical modeling was used to account for patient clustering by practice. To determine the relative contributions of measured patient and practice factors and unmeasured practice variation to rhythm control use, a reference effect measures (REM) methodology was used to compute and graph the results in an enhanced forest plot. 8,9 To quantify the contribution of all measured patient characteristics to rhythm control use, a risk score using all measured patient characteristics is calculated for each patient. From this, the empirical distribution of odds ratios (ORs) is then calculated by comparing each patient's odds of receiving rhythm control to an average (reference) patient's odds of receiving rhythm control, assuming equal measured and unmeasured practice characteristics. This calculation thus provides the full range of association that measured patient factors contribute to rhythm control. A similar calculation occurs for measured practice characteristics. Then, once distributions of ORs for patient and practice characteristics are known, they can be directly compared. If the patient range of ORs is larger than the practice range of ORs, then patient characteristics have a larger impact on the use of rhythm control than practice characteristics. In addition to estimating the relative impact of measured factors such as patient and practice characteristics, the REM method can also estimate the relative impact of unmeasured characteristics on rhythm control. To do so, the practice random effect distribution is used to calculate ORs comparing a practice to an average (reference) practice assuming all measured patient and practice factors are equal, giving a range of ORs quantifying variation in rhythm control use due to unmeasured practice factors. This range can then be compared with the measured patient and practice OR ranges to understand the relative impact of each group of variables—measured patient characteristics, measured practice characteristics, and unmeasured practice characteristics—on rhythm control use. Unlike prior studies quantifying the amount of practice variation in outcomes due to unmeasured factors using measures such as median ORs, 2-4 this REM approach allows for direct comparison of the relative contributions of patient, practice, and unmeasured practice factors on the same scale and plot as individual patient and practice factors.

To determine the factors associated with catheter ablation, patient and practice predictors of catheter ablation versus other rhythm control therapy were calculated. Using the same methodology as with the rhythm control analyses above, we conducted univariate and multivariable hierarchical logistic regression analyses to evaluate patient, practice, and unmeasured practice predictors of catheter ablation and then used REM methods to estimate and graph results. Several sensitivity analyses were then performed to test the robustness of our primary findings. First, we repeated analyses excluding amiodarone from the rhythm control definition because amiodarone may have been used for the treatment of ventricular arrhythmias rather than AF management per se. Second, because outcomes are assessed throughout the study period, there may be differential follow-up time of individual patients. To account for this, univariate and multivariable frailty Cox regression analyses were used, defining the outcome as the time of initiation of rhythm control therapy at follow-up visit. For this analysis, only patients without rhythm control at index visit were included in the analysis. Third, because a practice without any rhythm control patients may represent actual practice rather than incomplete data ascertainment from the electronic medical record mapping, the primary analyses were repeated including practices with no rhythm control patients. Again, to help quantify and compare fixed and random effects at the patient and practice level, an REM approach was used to compute and graph results in an enhanced forest plot for each of these sensitivity analyses. All analyses were performed with SAS version 9.4 (SAS Institute, Cary, NC). Enhanced forest plots showing patient and practice variation were made in R (R Foundation for Statistical Computing, Vienna, Austria). Funding for this research was provided by the American College of Cardiology Foundation's NCDR. The authors are solely responsible for the design and conduct of this study, all study analyses, the drafting and editing of the paper, and its final contents.

Results Rates of rhythm control and catheter ablation use Of 634,602 patients with nonvalvular AF in the PINNACLE data set, 122,644 (19.3%) were excluded from the cohort (first episode of AF, history of LV assist device or heart transplant, at site with no rhythm control patients), leaving 511,958 patients for the present analysis (Figure 1). Of the 511,958 included in this analysis, 113,972 (22.3%) were treated with rhythm control therapy and 14,847 (2.9% of overall AF population, 13.0% of rhythm control population) underwent catheter ablation.

American Heart Journal Volume 187, Number 0

Gehi et al 91

Figure 1

634,602 Patients with NVAF

122,644 (19.3%) Patients Excluded: 47,669 (38.9%) Patients with First Episode AF 2 (<0.1%) Patients with LVAD 1,437 (1.2%) Patients with Heart Transplant 73,536 (60.0%) Patients in 44 Sites with No Rhythm Control Patients

Analysis Population: 511,958 (80.7%)

No Rhythm Control 397,986 (77.8%)

Rhythm Control 113,972 (22.3%)

Atrial Ablation 14,847 (13.0%)

Non-ablation 99,125 (87.0%)

Flowchart of the analysis set.

Characteristics of patients treated with rhythm control Baseline characteristics of the overall AF study cohort and comparing those treated with or without rhythm control are shown in Table I. Overall, AF patients were elderly (71.7 ± 12.7 years), were mostly white (62.2%), had a number of comorbidities (hypertension 73.3%, coronary artery disease 45.2%, diabetes 22.5%, heart failure 21.9%), were overweight (BMI 29.4 ± 6.3 kg/m 2), and had a moderate-high CHADS2 (1.9 ± 1.3) and CHADS2-VaSC (3.6 ± 1.8) score. Patients treated with rhythm control, compared with those who were not, were younger, were more likely male, were more likely white, were more likely privately insured, had fewer comorbidities, had lower CHADS2 and CHADS2-VaSC scores, and were more likely to be seen by an EP physician. Practice variation of rhythm control and catheter ablation In the cohort, 128 practices had at least 1 patient receiving rhythm control, and their distribution of rhythm control and catheter ablation rates is shown in Figure 2 and Supplemental Figure 1. The proportion of patients receiving rhythm control varied widely (median 22.8%, range 0.2%-62.9%). Of those receiving rhythm

control, the proportion of patients receiving catheter ablation also varied widely across practices (median 0.3%, range 0%-100%). At 59 of the 128 (46.1%) practices where rhythm control was used, no patients received catheter ablation.

Predictors of rhythm control Adjusted predictors of rhythm control therapy at any time in follow-up are shown in Table II and Figure 3. Significant independent patient predictors of rhythm control included younger age, male gender, white race, private insurance, lower CHADS2 score, prior PCI, diabetes, hypertension, heart failure, prior TIA/CVA, nonphysician versus physician provider, and seen by an EP. REM methods showed that there was greater variation due to unmeasured practice factors relative to the measured patient and practice factors (Figure 3). The ORs for rhythm control use at an individual practice compared with an average practice due to unmeasured practice factor variation ranged from 0.29 to 3.44 (95% range). This variation exceeded that due to measured patient (ORs 0.36-2.31) and practice (ORs 0.66-1.12) factors.

American Heart Journal Month 2017

92 Gehi et al

Table I. Characteristics of nonvalvular AF patients by rhythm control use Characteristics Patient factors Age (y) Mean ± SD (n) Male Race 1 White Black/African American Asian American Indian/Alaskan Native Native Hawaiian/Pacific Islander Mixed Missing Ethnicity Hispanic Insurance2 Private Military Medicare Medicaid Other None Missing Comorbidities CAD Prior MI Prior PCI Diabetes Hypertension PAD Heart failure Chronic kidney disease Prior heart surgery Prior TIA/CVA/Embolism Prior hemorrhage Smoker (current or quit within 1 y) BMI Mean ± SD (n) LV ejection fraction Mean ± SD (n) Persistent AF CHADS2-VaSC score 4 Mean ± SD (n) 0 1 2 3 4 5 6 7 8 9 Provider type Physician Nonphysician Patient seen by EP physician

Overall (N = 511,958)

Rhythm control (n = 113,972)

Non–rhythm control (n = 397,986)

71.7 ± 12.7 (511,883) 69.4 ± 11.5 (113,969) 72.4 ± 12.9 (397,914) 56.1% (286,487/510,500) 58.4% (66,412/113,777) 55.5% (220,075/396,723) 62.2% (318,351/511,958) 2.7% (14,065/511,958) 0.6% (3244/511,958) 0.4% (1850/511,958) 0.1% (543/511,958) 0.3% (1684/511,958) 33.6% (172,221/511,958) 5.1% (5888/115,248)

75.7% 2.8% 0.7% 0.5%

(86,302/113,972) (3243/113,972) (796/113,972) (517/113,972)

0.1% (119/113,972) 0.3% (348/113,972) 19.9% (22,647/113,972) 5.0% (1144/22,933)

58.3% (232,049/397,986) 2.7% (10,822/397,986) 0.6% (2448/397,986) 0.3% (1333/397,986) 0.1% (424/397,986) 0.3% (1336/397,986) 37.6% (149,574/397,986) 5.1% (4744/92,315)

Standardized difference

P value

−0.242 0.059

b.001 b.001 b.001

0.377 0.006 0.012 0.032 0.000 0.000 −0.399 −0.005

.354

52.7% (269,671/511,958) 1.9% (9846/511,958) 59.3% (303,845/511,958) 4.7% (23,808/51,1958) 1.3% (6541/511,958) 4.1% (20,904/511,958) 13.3% (67,909/511,958)

56.5% 2.2% 57.4% 3.0% 1.4% 2.8% 13.2%

(64,338/113,972) (2539/113,972) (65,438/113,972) (3388/113,972) (1594/113,972) (3175/113,972) (14,988/113,972)

51.6% (205,333/397,986) 1.8% (7307/397,986) 59.9% (238,407/397,986) 5.1% (20,420/397,986) 1.2% (4947/397,986) 4.5% (17,729/397,986) 13.3% (52,921/397,986)

0.098 0.029 −0.051 −0.107 0.018 −0.091 −0.003

b.001 b.001 b.001 b.001 b.001 b.001 .198

45.2% (231,321/511,958) 17.4% (88,942/511,958) 8.9% (45,495/511,958) 22.5% (115,111/511,958) 73.3% (375,368/511,958) 7.9% (40,463/511,958) 21.9% (112,013/511,958) 0.1% (283/511,958) 9.3% (47,445/511,958) 12.9% (65,928/511,958) 0.7% (3611/511,958) 15.8% (63,721/402,742)

42.8% 15.8% 8.6% 19.7% 72.6% 6.3% 19.5% 0.1% 8.0% 9.7% 0.9% 15.5%

(48,808/113,972) (17,966/113,972) (9834/113,972) (22,494/113,972) (82,783/113,972) (7230/113,972) (22,201/113,972) (89/113,972) (9131/113,972) (11,021/113,972) (1000/113,972) (13,919/89,765)

45.9% (182,513/397,986) 17.8% (70,976/397,986) 9.0% (35,661/397,986) 23.3% (92,617/397,986) 73.5% (292,585/397,986) 8.4% (33,233/397,986) 22.6% (89,812/397,986) 0.0% (194/397,986) 9.6% (38,314/397,986) 13.8% (54,907/397,986) 0.7% (2611/397,986) 15.9% (49,802/312,977)

−0.062 −0.054 −0.014 −0.088 −0.020 −0.081 −0.076 0.045 −0.057 −0.128 0.022 −0.011

b.001 b.001 b.001 b.001 b.001 b.001 b.001 b.001 b.001 b.001 b.001 .003

0.131

b.001

54.2 ± 13.3 (143,802) 79.3% (80,648/101,743)

−0.019 0.186

b.001 b.001

3.60 ± 1.75 (511,958) 3.27 ± 1.71 (113,972) 3.70 ± 1.75 (397,986) 3.4% (17,329/511,958) 4.6% (5266/113,972) 3.0% (12,063/397,986) 8.8% (45,145/511,958) 11.6% (13,167/113,972) 8.0% (31,978/397,986) 14.5% (74,200/511,958) 17.4% (19,849/113,972) 13.7% (54,351/397,986) 20.6% (105,716/511,958) 21.8% (24,859/113,972) 20.3% (80,857/397,986) 23.0% (117,702/511,958) 21.5% (24,448/113,972) 23.4% (93,254/397,986) 16.2% (82,806/511,958) 13.6% (15,522/113,972) 16.9% (67,284/397,986) 8.3% (42,298/511,958) 6.2% (7057/113,972) 8.9% (35,241/397,986) 3.7% (19,130/511,958) 2.5% (2813/113,972) 4.1% (16,317/397,986) 1.3% (6510/511,958) 0.8% (861/113,972) 1.4% (5649/397,986) 0.2% (1122/511,958) 0.1% (130/113,972) 0.2% (992/397,986)

−0.248

b.001 b.001

29.4 ± 6.3 (382,207)

30.1 ± 6.3 (85,907)

54.2 ± 13.4 (188,237) 54.0 ± 13.8 (44,435) 80.9% (107,522/132,899) 86.3% (26,874/31,156)

90.8% (464,963/511,957) 9.2% (46,994/511,957) 14.4% (73,513/511,957)

90.3% (102,890/113,971) 9.7% (11,081/113,971) 20.7% (23,546/113,971)

29.2 ± 6.3 (296,300)

91.0% (362,073/397,986) 9.0% (35,913/397,986) 12.6% (49,967/397,986)

0.084 0.121 0.102 0.037 −0.046 −0.092 −0.102 −0.090 −0.058 −0.024 0.024 0.219

b.001 b.001 b.001

American Heart Journal Volume 187, Number 0

Gehi et al 93

Table I (continued)

Characteristics Practice factors Practice ≥10 providers b10 providers Urban Rural Region West Midwest Northeast South

Overall (N = 511,958)

Rhythm control (n = 113,972)

Non–rhythm control (n = 397,986)

Standardized difference

P value

b.001 b.001 .002 .002 b.001

78.2% (400,115/511,958) 21.8% (111,843/511,958) 90.1% (403,227/447,648) 9.9% (44,421/447,648)

77.8% (88,661/113,972) 22.2% (25,311/113,972) 89.8% (89,595/99,753) 10.2% (10,158/99,753)

78.3% (311,454/397,986) 21.7% (86,532/397,986) 90.2% (313,632/347,895) 9.8% (34,263/347,895)

−0.012 0.012 −0.013 0.013

26.4% (135,153/511,958) 27.1% (138,935/511,958) 9.4% (48,339/511,958) 37.0% (189,531/511,958)

25.9% (29,557/113,972) 28.3% (32,232/113,972) 7.4% (8464/113,972) 38.4% (43,719/113,972)

26.5% (105,596/397,986) 26.8% (106,703/397,986) 10.0% (39,875/397,986) 36.6% (145,812/397,986)

−0.014 0.034 −0.092 0.037

CAD, Coronary artery disease; PAD, peripheral artery disease.

Predictors of catheter ablation Multivariable predictors of catheter ablation among those patients receiving rhythm control therapy are shown in Supplemental Table I and Supplemental Figure 2. Patient predictors of catheter ablation included younger age, male gender, white race, private insurance, lower CHADS2 score, presence of diabetes, heart failure, prior TIA/CVA, and seen by an EP. Significant practice predictors of catheter ablation included practices with ≥10 providers. After applying REM methods, the unmeasured practice distribution had greater variation relative to the combined measured patient and practice distributions, again indicating that unmeasured practice factors drove catheter ablation use. The 95% range of ORs for unmeasured practice factors, comparing an individual practice to an average practice, ranged from 0.01 to 165.5, far exceeding the range for the measured patient (95% range OR 0.46-2.30) or practice (95% range OR 0.15-2.77) factors. Sensitivity analyses Sensitivity analyses of both predictors of rhythm control and catheter ablation excluding amiodarone as an antiarrhythmic drug did not significantly change the results of the analysis. Univariate and multivariable frailty Cox regression analyses defining the outcome as the time to initiation of rhythm control at follow-up visit were performed, including only patients with no rhythm control at index visit. The median number of days of follow-up for this analysis was 504 days (interquartile range 115-1059 days). The overall findings of the study were not affected by this analysis (Supplemental Table II and Supplemental Figure 3). At 44 (25.6%) of the 172 practices, no patients were recorded as having received rhythm control. Because this may represent missing data, these practices were omitted from primary analyses to avoid bias. However, because these “zero use” sites may reflect true practice patterns rather than missing data, an analysis was performed including these patients in the cohort. When including these patients, the ORs for rhythm control use at an individual practice

compared with a median practice due to unexplained practice variation increased significantly, ranging from 0.01 to 92.7 (95% range) (Supplemental Figure 4). This unexplained practice variation far exceeded that due to measured patient and practice characteristics. At a given practice, OR for rhythm control use due to measured patient characteristics for an individual patient compared with an average patient ranged from 0.36 to 2.29 (95% range). For a given patient, OR for rhythm control use due to measured practice characteristics for an individual practice compared with an average practice ranged from 0.32 to 2.54 (95% range).

Discussion In our study of N500,000 AF patients, we found that 1 in 5 patients was treated with rhythm control and 1 in 50 patients was treated with catheter ablation. In addition, there was significant practice variation in use of rhythm control, with a large proportion of practices pursuing rhythm control therapy in b5% of patients and some practices pursuing rhythm control in N50% of patients. There was also significant practice variation in catheter ablation use, with a large proportion of practices not pursuing catheter ablation at all and a few practices treating with AF ablation in N30% of patients undergoing rhythm control therapy. Finally, whereas some measured patient and practice factors such as race, insurance, and whether a patient was seen by an EP physician were significantly associated with use of rhythm control, the decision to pursue rhythm control or pursue rhythm control with catheter ablation was most affected by unexplained practice factors, indicated by the wide variation across practices in use of rhythm control therapy, exceeding the impact of measured patient factors. Our study suggests that rhythm control, which can significantly improve AF-associated symptoms, may be underused in many centers in the United States. Moreover, our results suggest that factors besides the guideline-recommended patient factors are primarily influencing treatment decisions. Understanding the causes of these factors and minimizing the role of

American Heart Journal Month 2017

94 Gehi et al

Figure 2

Median

Lower Quartile

Upper Quartile

Range

22.8%

17.7%

28.2%

0.2%-62.9%

Distribution of rhythm control percentage among all 128 practices with at least 1 patient who received rhythm control.

unnecessary treatment variation, if present, in the decision to use rhythm control and catheter ablation are needed. Our study corroborates and extends prior studies of patient factors associated with variation in treatment patterns for AF. Bhave et al 11 demonstrated among N500,000 Medicare beneficiaries with AF that racial and gender differences were evident for both receipt of oral anticoagulation and therapy with catheter ablation. However, our study demonstrates that additional patient factors including insurance type, comorbidities, and physician specialty contribute to treatment decisions as well. Steinberg et al 12 assessed factors determining rhythm control therapy in N10,000 patients enrolled in the ORBIT-AF registry. Similar to our study, the authors found that a minority of patients (32%) receive rhythm control. The authors found that patients managed with

rate control alone were older with more comorbidities and had fewer AF-related symptoms. In addition, patients referred to an electrophysiology were more likely to be treated with rhythm control. However, this study did not look at forms of rhythm control (antiarrhythmic drug therapy vs catheter ablation). In addition, our study furthers these prior studies by comprehensively analyzing patient and practice features to understand their relative independent effects on the rhythm control therapeutic interventions. Our study demonstrates that although patient factors drive treatment decisions, these patient factors are superseded by practice variation in treatment patterns. Our study demonstrates the novel finding that unmeasured practice variation is the biggest factor in rhythm control use. A variety of potential factors may explain these findings, including the availability of dedicated

American Heart Journal Volume 187, Number 0

Gehi et al 95

Table II. Multivariable hierarchical logistic regression: rhythm control versus no rhythm control Predictors Patient factors Age (per 10 y) Sex: male Race: white versus nonwhite⁎ Insurance type: private versus nonprivate† CHADS2 score Physician versus nonphysician Patient seen by EP physician CAD Prior MI Prior PCI Diabetes Hypertension PAD Heart failure Prior heart surgery Prior TIA/CVA/embolism Practice factors No. of providers b10 Region: West (ref) Region: Midwest Region: Northeast Region: South

OR (95% CI)

P value

0.903 1.016 2.429 1.043

(0.896-0.910) (1.002-1.031) (2.385-2.473) (1.025-1.060)

b.001 .026 b.001 b.001

0.698 0.819 1.765 1.046 0.972 1.073 1.230 1.590 0.858 1.351 0.981 1.690

(0.685-0.711) (0.796-0.842) (1.731-1.800) (1.028-1.064) (0.950-0.994) (1.044-1.103) (1.199-1.262) (1.551-1.629) (0.834-0.883) (1.316-1.386) (0.955-1.008) (1.615-1.768)

b.001 b.001 b.001 b.001 .012 b.001 b.001 b.001 b.001 b.001 .159 b.001

1.091 (0.874-1.361)

.442

0.890 (0.651-1.217) 0.603 (0.393-0.924) 0.944 (0.710-1.255)

.466 .022 .693

⁎ Nonwhite includes black, Asian, American Indian/Alaskan Native, Native Hawaiian/Pacific Islander, mixed, and missing. † Nonprivate includes military, Medicare, Medicaid/IHS, non-US insurance, state-specified plan, none, and missing.

treatment facilities, physician preferences and opinions regarding the benefit of rhythm control therapy, and cost considerations regarding technologically demanding therapies. In addition, because rhythm control therapies are primarily offered to improve quality of life, 10 the difficulty in characterizing AF-specific quality of life and clear guidelines on when to pursue rhythm control may contribute to this variation. Furthermore, our finding that race, insurance, and whether a patient was seen by an EP physician are significant patient and practice factors associated with rhythm control further suggests that differential access, potentially by socioeconomic status, may be a driver of differential treatments. Our finding of large variation in approaches to rhythm control across practices warrants further investigation to ensure that appropriate access and patient considerations are driving the decisions to use rhythm control and catheter ablation treatments. The proportion of patients who should be treated with a rhythm control strategy is unclear, but one would expect that, given current guidelines for AF management, the proportion would be relatively consistent across practices and more influenced by patient factors. The large variation in therapeutic approach suggests that there may be a substantial proportion of patients lacking treatment that can significantly improve quality of life. Awareness of factors affecting treatment decisions in rhythm control is

critical to identify those factors that are appropriate (eg, patient preference) and those that are not (eg, provider preference, avoidance of particular therapies, or practice variation). Quality improvement programs such as the Get with the Guidelines Heart Failure program have shown that adoption of quality improvement tools can help to reduce unnecessary or inappropriate practice variation across participating sites. 13 There are several limitations of our study. First, one of the primary criteria that may affect the decision to adopt a rhythm control strategy for AF is the severity of AF symptoms. 1 AF symptom severity was not measured in the PINNACLE registry. Symptom severity may account for some of the patient variation seen in rhythm control treatment decisions. However, it is unlikely that systematic differences in patient symptom severity could explain the large practice variations of rhythm control use seen in our study. Second, the length of time since diagnosis of AF is not captured in the PINNACLE registry. There may be some patients, for example, who previously failed rhythm control therapy and are now classified as non–rhythm control patients. Therefore, length of time bias may account for some differences in treatment across patients. However, the sensitivity analysis including only patients with a new diagnosis of AF and defining the outcome as time to rhythm control demonstrated no difference in findings. Third, it is unclear whether the large number of sites with no rhythm control patients was due to true practice variation or improper mapping of the data set from the electronic health record, as the mapping process for PINNACLE can differ at each site. However, the primary analysis included only centers with at least 1 rhythm control patient. In addition, a sensitivity analysis including all centers demonstrated no difference, suggesting that the difference in use of rhythm control likely represents true practice variation. Fourth, although unexplained practice variation accounted for a large proportion of the overall difference in care, we cannot separate whether unexplained practice variation originated at the patient or practice level. That is, there may be an imbalance of unmeasured patient factors associated with the outcome across practices, such as differing socioeconomic status. Such imbalances may result in unexplained variation across practices originating from patient characteristics and not differences across practices. In addition, the large contribution of unexplained practice variation may be partially explained by incorrect modeling assumptions brought about by sites with zero interventions. However, when excluding these sites from the analysis, we still found a large contribution of unexplained practice variation in the decision for rhythm control. Finally, cardiovascular practices that participate in the PINNACLE registry are a self-selected subset of all US cardiovascular practices and thus may not represent those practices not included in the registry. Specifically, the PINNACLE population has a low sample of patients from minority and low socioeconomic status backgrounds. Future studies may be needed to understand trends in minority populations, as data

96 Gehi et al

American Heart Journal Month 2017

Figure 3

Odds ratios for individual patient risks and practice characteristics, combined patient risks and combined practice characteristics, and unmeasured variation for use of rhythm control strategy in nonvalvular AF patients. *Only practices with at least 1 rhythm control patient are included. The individual patient and practice factors compare patients with and without those factors. The combined patient and practice factor distributions show the distributions over the patients in this study and so also take into account prevalence of risk factors.

from the PINNACLE registry may not generalize to the US population. In conclusion, our study demonstrates the possible underutilization and wide variability in treatment strategies with respect to rhythm control in a broad US AF patient

population. Although current guidelines suggest that patient factors, such as AF symptom severity, comorbidities, and potential therapeutic efficacy, should guide treatment decisions, there is evidence of large unexplained practice variation and unnecessary patient factor variation in rhythm

American Heart Journal Volume 187, Number 0

control and catheter ablation decisions. By identifying the factors that lead to differences in treatment strategies, we can determine their appropriateness and better align rhythm control use to optimize patient outcomes in AF.

Gehi et al 97

4.

5.

Disclosures Dr Gehi reports receiving honoraria from Zoll Corp, Biotronik, and St Jude Medical. No other relevant disclosures. Dr Hsu reports receiving honoraria from Medtronic, St Jude Medical, and Biotronik and has served on Advisory Boards for Janssen Pharmaceuticals and Brisol-Myers Squibb. Dr Turakhia reports research funding from Janssen Inc, Medtronic Inc, Gilead Science, and SentreHeart and honoraria from Medtronic Inc and St Jude Medical.

Appendix. Supplementary data

6.

7. 8.

9.

Supplementary data to this article can be found online at http://dx.doi.org/10.1016/j.ahj.2017.02.006.

10.

References

11.

1. January CT, Wann LS, Alpert JS, et al. 2014 AHA/ACC/HRS guideline for the management of patients with atrial fibrillation: a report of the American College of Cardiology/American Heart Association Task Force on Practice Guidelines and the Heart Rhythm Society. J Am Coll Cardiol 2014;64:e1-76. 2. Peterson PN, Chan PS, Spertus JA, et al. Practice-level variation in use of recommended medications among outpatients with heart failure: insights from the NCDR PINNACLE program. Circ Heart Fail 2013;6:1132-8. 3. Maddox TM, Chan PS, Spertus JA, et al. Variations in coronary artery disease secondary prevention prescriptions among outpatient cardi-

12.

13.

ology practices: insights from the NCDR (National Cardiovascular Data Registry). J Am Coll Cardiol 2014;63:539-46. Chan PS, Maddox TM, Tang F, et al. Practice-level variation in warfarin use among outpatients with atrial fibrillation (from the NCDR PINNACLE program). Am J Cardiol 2011;108:1136-40. Chan PS, Oetgen WJ, Buchanan D, et al. Cardiac performance measure compliance in outpatients: the American College of Cardiology and National Cardiovascular Data Registry's PINNACLE (Practice Innovation and Clinical Excellence) program. J Am Coll Cardiol 2010;56:8-14. Messenger JC, Ho KK, Young CH, et al. The National Cardiovascular Data Registry (NCDR) data quality brief: the NCDR data quality program in 2012. J Am Coll Cardiol 2012;60:1484-8. Sullivan GM, Feinn R. Using effect size—or why the P value is not enough. J Grad Med Educ 2012;4:279-82. Glorioso TJ, Grunwald GK, O'Donnell C, et al. Using reference effect measures to identify sources of variation in 30-day readmissions for percutaneous coronary interventions. Circ Cardiovasc Qual Outcomes 2015;8:A316. Lingsma HF, Roozenbeek B, Perel P, et al. Between-centre differences and treatment effects in randomized controlled trials: a case study in traumatic brain injury. Trials 2011;12:201. Wyse DG, Waldo AL, DiMarco JP, et al. A comparison of rate control and rhythm control in patients with atrial fibrillation. N Engl J Med 2002;347:1825-33. Bhave PD, Lu X, Girotra S, et al. Race- and sex-related differences in care for patients newly diagnosed with atrial fibrillation. Heart Rhythm 2015;12:1406-12. Steinberg BA, Holmes DN, Ezekowitz MD, et al. Rate versus rhythm control for management of atrial fibrillation in clinical practice: results from the Outcomes Registry for Better Informed Treatment of Atrial Fibrillation (ORBIT-AF) registry. Am Heart J 2013;165:622-9. Al-Khatib SM, Hellkamp AS, Hernandez AF, et al. Trends in use of implantable cardioverter-defibrillator therapy among patients hospitalized for heart failure: have the previously observed sex and racial disparities changed over time? Circulation 2012;125: 1094-101.