Institutional Variation in 30-Day Complications Following Catheter Ablation of Atrial Fibrillation.

Background Complications are a measure of procedural quality, yet variation in complication rates following catheter ablation of atrial fibrillation (AF) among hospitals has not been systematically examined. We examined institutional variation in the risk-standardized 30-day complication rates (RSCRs) following AF ablation which may suggest variation in care quality. Methods and Results This cohort study included all patients >18 years old undergoing AF ablations from 2012 to 2017 in Australia and New Zealand. The primary outcome was procedure-related complications occurring during the hospital stay and within 30 days of hospital discharge. We estimated the hospital-specific risk-standardized complication rates using a hierarchical generalized linear model. A total of 25 237 patients (mean age, 62.5±11.4 years; 30.2% women; median length of stay 1 day [interquartile range, 1-2 days]) were included. Overall, a complication occurred in 1400 (5.55%) patients (4.34% in hospital, 1.46% following discharge, and 0.25% experienced both). Bleeding (3.31%), pericardial effusion (0.74%), and infection (0.44%) were the most common complications while stroke/transient ischemic attack (0.24%), cardiorespiratory failure and shock (0.19%), and death (0.08%) occurred less frequently. Among 46 hospitals that performed ≥25 ablations during the study period, the crude complication rate varied from 0.00% to 21.43% (median, 5.74%). After adjustment for differences in patient and procedural characteristics, the median risk-standardized complication rate was 5.50% (range, 2.89%-10.31%), with 10 hospitals being significantly different from the national average. Conclusions Procedure-related complications occur in 5.55% of patients undergoing AF ablations, although the risk of complications varies 3-fold among hospitals, which suggests potential disparities in care quality and the need for efforts to standardize AF ablation practices among hospitals.

Catheter Ablation Versus Antiarrhythmic Drug Therapy for Atrial Fibrillation

risk‐standardized complication rate

What Is New?

Catheter ablation of atrial fibrillation (AF), although superior to medical therapy in restoring sinus rhythm, is associated with a risk of complications. While the incidence and types of complications have been extensively examined, it is uncertain whether complication rates vary among hospitals, which may imply differences in care quality.

We found that 1 in 18 patients undergoing AF ablation experienced a procedural complication within 30 days of hospital discharge, with bleeding and pericardial effusion being the most common complications.

More importantly, the risk of complications varied significantly among ablation centers. Using a hierarchical generalized linear model, a method widely used for profiling hospital performance, we found that the hospital‐specific risk‐standardized complication rate varied nearly 3‐fold (range, 2.89%–10.31%) among 46 hospitals, with 10 having a risk‐standardized complication rate significantly higher (6) or lower (4) than the national average (5.55%).

What Are the Clinical Implications?

There was a clinically meaningful and statistically significant institutional variation in complication rates following AF ablation, suggesting that the risk of complications may be related to care quality and modifiable by improving procedural technique (such as more frequent use of vascular ultrasound or intracardiac echocardiography) and by quality improvement initiatives such as clinical audits or safety checklists.

This institutional variation might be unsurprising, as AF ablation is rapidly disseminating and disparities in the management of AF have been reported before.

Routine implementation of process and outcomes measures, such as those recommended by the Heart Rhythm Society, across all ablation centers may standardize care, reduce variation, and improve quality.

Since its inception in 1998, catheter ablation of atrial fibrillation (AF) has rapidly evolved from an investigational procedure to a guidelines‐recommended therapy for drug‐refractory symptomatic AF. Paralleling this change, worldwide surveys have shown a rapid increase in the number of AF ablations performed. , However, this complex and invasive procedure can cause serious complications such as bleeding, stroke, and cardiac tamponade, which may cause substantial harm to patients and may lead to additional invasive treatments. Reducing the risk of complications is therefore highly desirable to minimize patient harm and improve procedural safety.

Although the incidence of procedural complications following AF ablations has been extensively reported, , , , , little is known about the variation in complication rate among hospitals, which may suggest differences in care quality. Several studies have compared procedural safety in high‐ versus low‐volume ablation centers, , although variation in complication rates among individual hospitals has not been examined in the literature. Significant institutional variation has been reported for well‐established procedures such as cardiac device implantation, , raising the possibility that similar variation may exist for AF ablation. Understanding the risk among individual hospitals is also important in the context of the recent studies reporting rising rates of mortality and complications following AF ablation , which have raised concern about disparities in procedural safety as this procedure disseminates more widely. Indeed, the 2017 consensus guidelines have called for observational data on procedure‐related complication rates in the “real world” to inform patients and clinicians considering AF ablation and to inform hospitals and policymakers seeking to improve procedural quality.

In this study, we used population‐wide data from hospitals in Australia and New Zealand to determine the incidence of procedure‐related complications following AF ablation occurring up to 30 days after discharge. We further estimated the hospital‐specific risk‐standardized complication rate to identify if there were meaningful differences in complication rates among hospitals that may suggest disparities in care quality.

Methods

Because of the sensitive nature of the data collected for this study, requests to access the data sets from qualified researchers trained in human subject confidentiality protocols may be sent to the Human Research Ethics Committee of each state and territory in Australia and the New Zealand Ministry of Health.

Data Source

We used hospitalization data from all public and most (80%) of private‐sector hospitals and day procedure centers using each Australian state and territory’s Admitted Patient Collection and the New Zealand National Minimum Dataset (Hospital Events) from 2012 to 2017. These data sets record all in‐patient and day‐only admissions, including all outpatient procedures, irrespective of age and payer. A standard set of variables is collected for each patient encounter, including patient demographic characteristics, primary and secondary diagnoses, all procedures performed, and the patient status at discharge. Both countries use the International Classification of Diseases, Tenth Revision, Australian Modification (ICD‐10‐AM) and the Australian Classification of Health Interventions for coding of diagnoses and procedures, respectively. Validation against medical records has shown >85% coding accuracy, with cardiovascular diagnoses and procedures being particularly well coded. When such data were used for surveillance of adverse events in other fields, >90% agreement with clinicians was reported. Within each state or territory in Australia, hospitalizations were linked to subsequent hospitalizations and each region’s Registry of Deaths to track hospital readmission and postdischarge deaths. Greater than 99% accuracy is reported for the linkage of health records using probabilistic matching techniques based on multiple patient identifiers. In New Zealand, hospital encounters are linked nationally using a unique National Health Index number, and all deaths are recorded in the National Health Index sociodemographic profile.

Study Cohort

We included patients >18 years old hospitalized with a primary diagnosis of AF (ICD‐10‐AM codes I48, I48.0‐2, and I48.9) and underwent catheter ablation as defined by Australian Classification of Health Interventions procedure codes 38287‐01, 38287‐02 and 38290‐01. The use of the AF diagnosis code together with catheter ablation code has shown high specificity (100%) and sensitivity (87.3%) in identifying AF ablation procedures.

We excluded patients who had other arrhythmias as a secondary diagnosis to ensure the catheter ablation was for AF; had an implanted cardiovascular implantable electronic device (pacemaker, implantable cardioverter defibrillator, or cardiac resynchronization therapy pacemaker or defibrillator) during the index or previous admissions to avoid including patients undergoing atrioventricular nodal ablation for AF rate control; underwent open (surgical) ablation; were discharged against medical advice; had prior catheter ablation within 30 days since a complication may relate to the previous rather than the index ablation; or lacked at least 30 days follow‐up after the procedure to assess complications. We also excluded acute (unplanned) hospitalizations to ensure that the complications were procedure related rather than attributable to the underlying acute illness. Table S1 provides a full description of diagnoses and procedure codes used to define inclusion and exclusion criteria.

Outcome

The primary outcome was the occurrence of ≥1 procedure‐related complications identified from the prior literature, , , , expert clinical opinion, and empirical examination of patient records. Specific complications included (1) death, (2) cardiorespiratory failure and shock, (3) stroke or transient ischemic attack, (4) pericardial effusion, (5) hemothorax or pneumothorax, (6) bleeding (hemorrhage or hematoma formation, internal organ bleeding [bleeding from the gastrointestinal, pulmonary, or urinary system], or requirement for blood transfusion), (7) vascular injury or intervention, (8) infection (pneumonia, sepsis, or endocarditis), (9) pericarditis, (10) acute myocardial infarction, (11) venous thromboembolism, (12) acute kidney injury, (13) complete atrioventricular block, and (14) complications requiring cardiac surgery. Consistent with prior studies, we considered complications occurring in‐hospital and within 30 days after discharge as procedure related. In‐hospital complications were identified on the basis of the secondary diagnoses and procedures performed during the hospital stay. Postdischarge complications were defined as postdischarge death or any readmission with a complication coded as the primary discharge diagnosis. Table S2 lists all relevant codes used to define complications.

Statistical Analysis

Categorical variables are presented as frequencies and percentages. Continuous variables are presented as mean±SD or as median and interquartile range. The student t test or Mann‐Whitney U test was used to test differences between groups for continuous variables, and the χ2 or Fisher’s exact test was used for categorical variables. When estimating rate of overall complications, patients who experienced multiple events were counted only once.

To evaluate institutional variation in complication rates, we calculated the risk‐standardized complication rate (RSCR) for each hospital using a hierarchical (2‐level) generalized linear model, adjusting for differences in hospital case mix and clustering of patients. This method has been widely used to quantify institutional variation in outcomes and public reporting. , , , First, we identified patient characteristics independently associated with the risk of complications using a logistic regression model. Candidate variables included age, sex, hospitalization for AF in the preceding year, prior AF ablation, ablation of both atria, and comorbidities with a statistically significant (P<0.25) association with complications. Comorbidities were identified using the Condition Category classification that grouped ICD‐10‐AM codes into 180 clinically meaningful conditions using secondary diagnosis codes from the index admission and primary and secondary diagnosis codes from admissions within the preceding 12 months (see Table S3 for list of comorbidities used for model development). To select the final variables, we included all candidate variables and then applied purposeful backward elimination as described by Hosmer and Lemeshow until the model contained only variables significant at P<0.05. Model performance was evaluated by estimating model discrimination (C‐statistic) and calibration. In keeping with best‐practice recommendations, model discrimination was validated by calculating the optimism‐corrected C‐statistic using bootstrapping resampling with 100 replications. The optimism is estimated as the difference between model’s performances in bootstrap and original samples, and the corrected C‐statistic equates the difference between the original C‐statistic (derived from modeling using the original data set) and the average optimism.

We then used the hierarchical generalized linear model to estimate a random‐intercept term that reflects each hospital’s contribution to the risk of the outcome, based on its actual complication rate, the performance of other hospitals with similar case mix and its sample size. The RSCR is the ratio of predicted complication rate over the expected complication rate multiplied by the cohort average complication rate. The predicted complication rate was calculated on the basis of the hospital’s case mix and the estimated random intercept, while the expected complications rate was calculated using the hospital’s case mix and the cohort average rate. We used bootstrapping with 1000 replications to empirically construct the 95% CI for each hospital’s RSCR using the percentile method. A hospital was classified as significantly different from the national average if the entire 95% CI was above or below the average rate. To ensure robust estimates of the RSCR, all hospital analyses were limited to those that performed at least 25 ablations during the study period. A detailed description of the RSCR calculation and bootstrapping algorithm is provided in Data S1.

A two‐sided P value <0.05 was considered statistically significant. All analyses were performed using SAS 9.4 (SAS Institute Inc., Cary, NC). The Human Research Ethics Committees of all Australian states and territories provided ethical approval to undertake the study with a waiver of informed consent to use deidentified patient data. Deidentified data from New Zealand were obtained under a data user agreement with the Ministry of Health.

Results

We identified 32 739 eligible patients with a primary diagnosis of AF undergoing catheter ablation. Of these, we excluded 7502 patients (see Figure 1 patient selection flow diagram), and the main reasons for exclusion were having a current or previous cardiovascular implantable electronic device (4104 patients) or unplanned hospitalizations (1972 patients). The final study cohort consisted of 25 237 patients who underwent AF ablation at 67 unique hospitals, of which 46 performed at least 25 procedures in the study period.

Figure 1

Patient selection flow diagram.

AF indicates atrial fibrillation; and CA, catheter ablation.

Cohort Characteristics

The study cohort had a mean age of 62.5±11.4 years and 30.2% were women (Table 1). The median length of stay was 1 day (interquartile range, 1.0–2.0 days). Of these patients, 62.8% had a prior hospitalization for AF or atrial flutter, and 12.2% had a prior catheter ablation. Hypertension (11.3%) and diabetes (11.3%) were the most common cardiac and noncardiac comorbidities, respectively.

Table 1

Characteristics of the Study Cohort

Variables	Overall (N=25 237) n (%)	Any complication (N=1400) n (%)	No complication (N=23 837) n (%)	P value
Patients’ demographics
Age (mean±SD)	62.5±11.4	64.1±11.0	62.5±11.4	<0.001
Age group, y
18–34	497 (2.0)	20 (1.4)	477 (2.0)	<0.001
35–49	2657 (10.5)	121 (8.6)	2536 (10.6)
50–64	10 419 (41.3)	540 (38.6)	9879 (41.4)
65–79	10 438 (41.4)	627 (44.8)	9811 (41.2)
≥80	1226 (4.9)	92 (6.6)	1134 (4.8)
Female (%)	7621 (30.2)	497 (35.5)	7124 (29.9)	<0.001
Median length of stay (IQR)	1.0 (1.0–2.0)	2.0 (1.0–3.5)	1.0 (1.0–2.0)	<0.001
Cardiac history
Prior AF hospitalizations	15 839 (62.8)	884 (5.6)	516 (5.5)	0.761
Prior AF ablation	3088 (12.2)	142 (10.1)	2946 (12.4)	0.014
Hypertension	2842 (11.3)	240 (17.1)	2602 (10.9)	<0.001
Heart failure	2239 (8.9)	162 (11.6)	2077 (8.7)	<0.001
Valvular and rheumatic heart disease	919 (3.6)	74 (5.3)	845 (3.5)	0.001
Coronary artery disease	2401 (9.5)	185 (13.2)	2216 (9.3)	<0.001
Vascular disease	382 (1.5)	28 (2.0)	354 (1.5)	0.125
Noncardiac comorbidities
Diabetes	2849 (11.3)	153 (10.9)	2696 (11.3)	0.661
Chronic obstructive lung disease	304 (1.2)	34 (2.4)	270 (1.1)	<0.001
Chronic kidney disease	819 (3.3)	82 (5.9)	737 (3.1)	<0.001
Stroke or TIA	318 (1.3)	18 (1.3)	300 (1.3)	0.929
Hematologic disorders	1070 (4.2)	154 (11.0)	916 (3.8)	<0.001
Pneumonia	508 (2.0)	74 (5.3)	434 (1.8)	<0.001
Musculoskeletal and connective tissue disorders	1846 (7.3)	155 (11.0)	1691 (7.1)	<0.001
Dementia and senility	38 (0.2)	5 (0.4)	33 (0.1)	0.040

AF indicates atrial fibrillation; IQR, interquartile range; and TIA, transient ischemic attack.

Incidence of Complications

Overall, procedural complications occurred in 1400 (5.55%) patients in‐hospital or within 30 days of discharge (Table 2). Patients who experienced a complication were older (64.1 versus 62.5 years; P<0.001), were more likely to be women (35.5% versus 29.9%; P<0.001), and had higher rates of comorbidities such as hypertension (17.1% versus 10.9%), heart failure (11.6% versus 8.7%), coronary artery disease (13.2% versus 9.3%), chronic obstructive lung disease (2.4% versus 1.1%), and chronic kidney disease (5.9% versus 3.1%) compared with those who did not experience a complication (all P<0.001).

Table 2

Incidence of Complications After Catheter Ablation of AF

Procedural complications	Overall N (%)	In‐hospital n (%)	Postdischarge n (%)
Primary outcome—any complication*	1400 (5.55)	1095 (4.34)	368 (1.46)
Death	21 (0.08)	6 (0.02)	15 (0.06)
Cardiorespiratory failure and shock	47 (0.19)	43 (0.17)	4 (0.02)
Stroke/TIA	60 (0.24)	28 (0.11)	34 (0.13)
Pericardial effusion	188 (0.74)	166 (0.66)	25 (0.10)
Pericardiocentesis	107 (0.42)	91 (0.36)	16 (0.06)
Hemothorax/pneumothorax	33 (0.13)	19 (0.08)	15 (0.06)
Bleeding	835 (3.31)	693 (2.75)	165 (0.65)
Postprocedural hemorrhage/hematoma	645 (2.56)	582 (2.31)	74 (0.29)
Internal organ bleeding †	143 (0.57)	105 (0.42)	40 (0.16)
Blood transfusion	128 (0.51)	72 (0.29)	61 (0.24)
Vascular injury or intervention	56 (0.22)	32 (0.13)	26 (0.10)
Postprocedural infection	112 (0.44)	50 (0.20)	62 (0.25)
Pericarditis	71 (0.28)	56 (0.22)	16 (0.06)
Procedure‐related AMI	27 (0.11)	10 (0.04)	17 (0.07)
Venous thromboembolism	18 (0.07)	7 (0.03)	11 (0.04)
Acute kidney injury	73 (0.29)	66 (0.26)	7 (0.03)
Complications requiring cardiac surgery	25 (0.10)	15 (0.06)	10 (0.04)
Complete atrioventricular block	55 (0.22)	53 (0.21)	4 (0.02)

AF indicates atrial fibrillation, AMI indicates acute myocardial infarction; and TIA, transient ischemic attack.

When estimating the primary outcome, patients with multiple complications were counted only once. For all other outcomes, patients may have >1 complication. Therefore, the incidence across rows or columns may not sum to group totals.

Bleeding from the gastrointestinal, pulmonary, or urinary system. Intracranial bleeding was counted as stroke.

When specific complications were considered, bleeding was the most common complication, occurring in 3.31% of procedures. Of the bleeding events, 77.3% were attributable to postprocedural hemorrhage or hematoma, 17.1% was bleeding from internal organs (gastrointestinal, pulmonary, or urinary), and 15.3% required blood transfusion. Pericardial effusion was the second most common complication, which occurred in 0.74% of patients and 56.9% of these cases underwent pericardiocentesis. Death (0.08%), complications that required cardiac surgery (0.10%), and stroke or transient ischemic attack (0.24%) occurred infrequently.

Among patients who experienced a complication, 1095 (4.34%) had the complication during their hospital stay. Bleeding remained the most common complication (2.75%), with 10.4% of these patients requiring a blood transfusion. Pericardial effusion (0.66%) was the second most common, with 54.8% of these cases requiring drainage. Another 368 (1.46%) patients had procedural complications within 30 days of hospital discharge, with 0.25% of patients experiencing both in‐hospital and postdischarge complications. Bleeding was the most common cause of a postdischarge complication (0.65%), followed by postprocedural infection (0.25%) and stroke/transient ischemic attack (0.13%). Procedure‐related death occurred more frequently after discharge than during the index hospitalization (15 versus 6 deaths) and 5 of the 15 postdischarge deaths occurred in the community.

Risk‐Adjustment Model

Patient age, female sex, history of ablation, ablation of both atria, year of ablation, and 5 comorbidities were independently associated with the risk of complications (Table S4) and were used for the hospital‐level risk‐adjustment. The logistic regression model had moderate discrimination (C‐statistic of 0.604) and could predict a range of patient risk from 3.72% to 11.93% that closely approximated the observed risk, suggesting good model calibration (Hosmer‐Lemeshow c2=11.83; P=0.159) (Figure S1). Internal validation by bootstrapping with 100 replications revealed an average optimism of 0.005, corresponding to a corrected C‐statistic of 0.599.

Hospital Variation in the RSCRs

Among the 46 hospitals that performed at least 25 procedures during the study period, the crude median complication rate was 5.74% and ranged from 0.00% to 21.43% (Table S5). After risk‐standardization, the median RSCR was 5.50%, although the rate varied from 2.89% to 10.31% among hospitals (Figure 2A). Of these hospitals, 10 had complication rates significantly different from the cohort average, with 4 having the entire 95% CI below the average rate (indicating a better‐than‐average complication rate) and 6 with the entire estimated 95% CI above the average (indicating worse‐than‐average complication rate). There was no correlation between RSCR and the hospital’s annual ablation volume (Spearman correlation coefficient, −0.02; P=0.892; Figure 2B).

Figure 2

Institutional variation in the risk‐standardized complication rate (RSCR).

A, shows the RSCR with the corresponding 95% CI of the 46 hospitals. B, presents RSCR based on hospital’s annual ablation volume. C, shows the RSCR with the corresponding 95% CI when the outcome was limited to in‐hospital complications only. Analysis was limited to hospitals that performed ≥25 procedures during the study period with hospitals presented by ascending order of the RSCR in A, of hospital’s annual ablation volume in B, and of risk‐standardized in‐hospital complication rate in C.

Sensitivity Analysis

We performed several analyses to test the robustness of our findings. As most existing studies report in‐hospital complications exclusively and because these events may be more closely related to procedural technique and care quality, we repeated the RSCR estimation, limiting to in‐hospital complications only. We found persisting variation in RSCR (median, 4.15%; range, 2.07%–10.20%), with 7 hospitals having higher‐than‐average and 3 having lower‐than‐average in‐hospital complication rates (Figure 2C). Seven of these 10 hospitals were also outliers based on the 30‐day outcome. To determine the potential for unmeasured confounders to influence the results, we assessed the minimum strength of association that an unmeasured confounder would need to shift the interval estimate of the most outlying hospital (hospital 45 in Figure 2A—RSCR, 10.11%; 95% CI, 7.31%–13.29%) to cross the cohort average rate by calculating the E‐value for the lower 95% CI (7.31%), which yielded 1.97. This means that an unmeasured confounder would need to be 1.97 times more common in the outlier hospital compared with the national average and be associated with a 1.97‐times higher rate of complications to explain away the difference so that the hospital is no longer an outlier, while a weaker confounder could not. Moreover, to assess whether the observed variation could have occurred by chance, we repeated the analysis applying the Bonferroni correction, which tests the global null hypothesis that all hospitals have a risk‐standardized outcome rate similar to the national average. When a corrected P value of 0.001 (≈0.05/46) was applied (equivalent to 99.9% CIs) and 10 000 bootstrapped samples were used, 2 hospitals remained significantly different than average (all above the national average), making it unlikely that the observed variation was attributable to chance. Finally, the funnel plot of RSCRs (Figure S2), an alternative methodology for displaying variation in performance, also showed 7 hospitals with RSCRs exceeding the 95% limit of the average complication rate. These hospitals were also classified as having a higher‐ (4 hospitals) and lower‐than‐average (3 hospitals) complication rate using the bootstrapping method. The calculated φ and were 1.51 and 1.13, respectively, suggesting that we could assume φ=1 and that adjustment for possible overdispersion was not needed. But even if 10% winsorization is applied to adjust for possible overdispersion, the winsorized plot still shows 3 hospitals with RSCRs higher than the upper border of the 95% control limits (Figure S3). Collectively, these findings suggest that statistically significant and clinically meaningful institutional variation likely existed.

Discussion

In this population‐wide study of 25 237 patients undergoing AF ablation, we found that about 1 in 18 patients experienced a procedure‐related complication within 30 days of hospital discharge. However, the incidence of complications was highly dependent on the ablation center, with complication rates varying more than 3‐fold among hospitals even after adjusting for differences in patient and procedure characteristics, implying institutional disparities in care processes and quality control measures. Of all complications, 76.3% were attributable to bleeding, pericardial effusion, and infection—complications that can be reduced or avoided with established interventions such as vascular ultrasound, intracardiac echocardiography, or prophylactic antibiotics. Collectively, these findings call for concerted clinical and policy intervention to inform patients, improve procedural safety, and standardize care among hospitals.

Population studies with unselected cohorts (all age, all payer) that capture the full range of ablation facilities are sparse. Most existing studies report in‐hospital complications only , , and often fail to capture outpatient procedures, , , even though they could account for 37% to >90% of all AF ablations. We extend the literature by providing estimates from a national cohort that includes both inpatient and outpatient procedures and captures all complications including those that occurred following discharge. Although comparisons of complication rates among studies are often challenging because of differences in designs, data sources, and definitions of complications, our overall complication rate of 5.55% is consistent with the 3.5% to 7.4% range reported in population studies, including the Get With The Guidelines AF Registry. , , , , Our result is also comparable to the ≈6.9% rate reported in the multicenter CABANA (Catheter Ablation Versus Antiarrhythmic Drug Therapy for Atrial Fibrillation) trial. Our rate, however, is higher than the 2.9% (95% CI, 2.6%–3.2%) rate reported in a prior systematic review, in which most included studies assessed in‐hospital events only and may explain the discrepancy. Indeed, we found that nearly 30% of complications presented after discharge, highlighting the need for continued vigilance for complications after discharge.

We also extend the literature by demonstrating the institutional heterogeneity in AF ablation outcomes by showing clinically meaningful and statistically significant variation in complication rates among hospitals. Prior studies have compared procedural safety among hospitals by volume‐based grouping of ablation centers and found higher complications rates in low‐volume strata, , yet such studies do not provide insights into the performance of individual hospitals. Our study quantified hospital performance individually and did not find a significant relationship between a hospital’s ablation volume and its RSCR. The >3‐fold variation in overall complication rates suggests disparities in procedural quality among ablation centers. Such variation is perhaps unsurprising given the rapid dissemination of AF ablation and the highly heterogeneous nature of the procedure that relies on a wide variety of techniques, equipment, and resources. Indeed, marked institutional differences have been found in compliance with quality measures for AF management. Thus, it is conceivable that similar variation in care quality may occur for AF ablation. Recently, the fifth Atrial Fibrillation Network/European Heart Rhythm Association conference recommends defining and monitoring quality standards in AF care, including implementing a range of process and outcomes measures for AF ablation. The Heart Rhythm Society has also recently developed harmonized outcomes measures for use in AF, including AF ablation. Our findings firmly support implementing such measures across all ablation centers to standardize care and to guide targeted quality improvement efforts.

These findings have several additional implications for quality improvement efforts. Current consensus guidelines emphasize minimizing clinically important but rare complications such as atrio‐esophageal fistula formation. While this is important, our results imply that efforts to improve patient safety should also focus on reducing more common complications such as bleeding, pericardial effusion, and infection, which constitute 76.3% of all complications. Moreover, these complications are potentially preventable with existing interventions. For example, uninterrupted dabigatran has been shown to be associated with significantly fewer major bleeding events compared with warfarin. Similarly, good visualization during transseptal puncture by multiple fluoroscopic views or intracardiac ultrasound may help to avoid cardiac perforation. Among patients undergoing AF ablation under general anesthesia and routine urinary catheter placement, rates of urinary tract infection, which is significantly associated with risk of sepsis, could be reduced by 80% with prophylactic antibiotics. Moreover, the routine use of a urinary catheter could be safely avoided, as need‐based catheterization is shown to be associated with nearly 8 times lower odds of experiencing adverse outcomes, including cystitis, hematuria, dysuria, and urethral damage, compared with routine use. More broadly, implementing procedural safety checklists can reduce complications from cardiac catheterization procedures, including ablations. From a policy perspective, our observations support reporting of hospital‐specific complication rates to better inform decision making, guide quality improvement efforts, and standardize care among hospitals. Reporting hospital‐specific rates may be particularly important for true informed consent, as the average complication rate may have little meaning when discussing procedural risk with patients in the context of marked variation among hospitals.

Several limitations should be considered when interpreting our results. Administrative data are less granular than data collected specifically for research, however, validation studies have reported good accuracy (>85%) of diagnoses and procedures coding. We also focused on coding definitions used by prior studies to minimize the risk of erroneous coding influencing our results. , , , Data were available from all regions at up to 2017 only, after which new advances in techniques and technology may have occurred and impacted the contemporary complication rates. Similar to other population studies, atrio‐esophageal fistula, pulmonary vein stenosis, and phrenic nerve injury could not be reliably identified using administrative data because of a lack of specific codes. , , These complications can also present beyond 30 days; thus, our study is likely to underestimate the true complication rate. Nevertheless, atrio‐esophageal fistula is rare and may be captured under other categories such as sepsis, stroke, or death, while only a few cases with pulmonary vein stenosis and phrenic nerve injury cause symptoms or require treatment. , We also could not distinguish between different types of ablation used, although no technique is proven to have a superior safety profile. , Data regarding ablation lesions were also not available, and patients may have had additional ablations other than pulmonary vein isolation. Nevertheless, our sensitivity analyses suggest that the observed variation is unlikely to be explained either by chance or an unmeasured confounder.

Conclusions

Complications following catheter ablation of AF occur in ≈1 in 18 patients undergoing ablation, although the rate of complications is highly variable among hospitals, suggesting that clinically meaningful differences may exist in procedural quality and after‐care practices. Concerted clinical and policy efforts are needed to better inform patients, to improve care practices, and to standardize outcomes across ablation centers.

Sources of Funding

This work was supported by the National Heart Foundation of Australia (ID 101186).

Disclosures

Dr Ngo was a recipient of the Hospital Research Foundation Postgraduate Scholarship and a Research Training Program Scholarship from the University of Queensland during the course of this study. Dr Ali is a recipient of a Divisional Scholarship from the University of Adelaide. Drs Ganesan and Ranasinghe are recipients of a National Heart Foundation of Australia Future Leader Fellowship (IDs 101188 and 101186, respectively). In the past 3 years, Harlan Krumholz received expenses and/or personal fees from UnitedHealth, Element Science, Aetna, Reality Labs, Tesseract/4Catalyst, the Siegfried and Jensen Law Firm, Arnold and Porter Law Firm, Martin/Baughman Law Firm, and F‐Prime. He is a co‐founder of Refactor Health and HugoHealth, and is associated with contracts, through Yale New Haven Hospital, from the Centers for Medicare & Medicaid Services and through Yale University from Johnson & Johnson. The remaining authors have no disclosures to report.

Data S1

Tables S1–S5

Figures S1–S3

Click here for additional data file.