Design, Rationale and Initial Findings From HERA‐FIB on 10 222 Patients With Atrial Fibrillation Presenting to an Emergency Department Over An 11‐Year Period

Abstract

Background

For the majority of patients with atrial fibrillation (AF), disease management has improved in recent years. However, there are still populations underrepresented or excluded in current registries and randomized controlled trials. HERA‐FIB (Heidelberg Registry of Atrial Fibrillation) was planned to assess real‐world evidence for the prevalence, demographic characteristics and management of patients with the diagnosis of AF presenting consecutively to a chest pain unit.

Methods and Results

HERA‐FIB is a retrospective, observational, single‐center study on patients with a diagnosis of AF presenting to a chest pain unit from June 2009 until March 2020. This article describes the structure, governance, outcome assessment, quality and data collection processes of the registry. Additionally, characteristics of populations of special interest are described. The study consecutively enrolled 10 222 patients presenting with AF to the chest pain unit of the University Hospital of Heidelberg. Clinical parameters and patient characteristics were assessed retrospectively. Outcome parameters included rates for all‐cause death, stroke, myocardial infarction and major bleedings. We were able to investigate patient cohorts of special interest such as advanced chronic kidney disease, octogenarians, and those with acute coronary syndrome who are often underrepresented in current studies and randomized controlled trials.

Conclusions

HERA‐FIB is one of the largest real‐world single‐center retrospective registries on patients with AF, which captures the era of transition from vitamin K antagonists to non–vitamin K oral anticoagulation regimens in clinical practice and offers the possibility to investigate patient populations usually underrepresented or excluded in current available randomized controlled trials and registries.

Registration

URL: https://www.clinicaltrials.gov; unique identifier: NCT05995561.

Nonstandard Abbreviations and Acronyms

CPU

chest pain unit

EHS

Euro Heart Survey

ENGAGE AF TIMI 48

Effective Anticoagulation With Factor Xa Next Generation in Atrial Fibrillation–Thrombolysis in Myocardial Infarction 48

EORP‐AF

EURObservational Research Program Atrial Fibrillation General Pilot Registry

GARFIELD‐AF

Global Anticoagulant Registry in the Field‐Atrial Fibrillation

GLORIA AF

Global Registry on Long‐Term Oral Antithrombotic Treatment in Patients With Atrial Fibrillation

HERA‐FIB

Heidelberg Registry of Atrial Fibrillation

NOAC

non–vitamin K oral anticoagulant

OAC

oral anticoagulant

ORBIT‐AF

Outcomes Registry for Better Informed Treatment of Atrial Fibrillation

RE‐LY

Randomized Evaluation of Long‐Term Anticoagulation Therapy

VKA

vitamin K antagonist

Clinical Perspective.

What Is New?

• HERA‐FIB (Heidelberg Registry of Atrial Fibrillation) was designed to assess real‐world evidence for patients with atrial fibrillation presenting to an emergency department.

What Are the Clinical Implications?

• HERA‐FIB offers the possibility to investigate patient populations of special interest, which are usually underrepresented or excluded in current available randomized controlled trials and registries.

Atrial fibrillation (AF) is the most common arrhythmia in adults presenting to an emergency department (ED). ¹ Disease management, such as improvement of diagnostics, treatment and prevention of adverse outcomes has substantially improved in recent years. ¹ Being a major risk factor for stroke, ² AF is associated with increased mortality and morbidity, ³ leading to high mortality rates in the absence of adequate treatment. ⁴ When used appropriately in patients with AF, vitamin K antagonists (VKAs) were shown to be effective for stroke prevention. ⁵ However, VKAs have several problematic limitations including drug interactions and continuous drug dosage adjustments and may therefore have several disadvantages for many patients with AF. ⁶ Particularly in elderly patients, anticoagulation regimens containing VKA are a rising health care–related and economic issue. In US adults aged ≥65 years who are admitted to a hospital for adverse drug reactions, VKAs are assumed to account for 33%. ⁷ Addressing limitations and disadvantages of oral anticoagulant (OAC) regimens containing VKAs, non–vitamin K oral anticoagulant (NOAC) regimens including thrombin antagonists and factor Xa inhibitors have been developed over the past decade. ⁸ This fundamental shift in anticoagulation management has prompted the initiation of various registries and randomized controlled trials (RCTs) to investigate implications, advantages, and suitability of NOAC‐ versus VKA‐containing anticoagulation regimens. However, most studies are either biased due to incomplete enrollment of consecutive patients, such as large pharmaceutical industry–sponsored registries ⁹ , ¹⁰ or limited due to short recruitment periods or incomplete assessment of important variables in national registries. Thus, whether findings from RCTs could be generalized to the clinical real world is an ongoing matter of debate. There is a considerable variety in currently available registries design, objectives, and patient selection, as well as quality performance measures and standards. ¹¹ , ¹² Here, national registries ¹³ , ¹⁴ , ¹⁵ that collect a huge amount of available data from an entire country population, consecutively enrolling patients without exception, did not meet these limitations, since selection bias may be assumed and data quality, accuracy, and management of missing data could be problematic. Large international multicenter registries including verified pharmaceutical industry–sponsored or –funded registries differ significantly regarding highly specialized eligibility criteria and do not mandate enrollment of consecutive subjects, ¹⁶ , ¹⁷ leaving a gap of evidence for excluded or underrepresented patients. To address these issues, HERA‐FIB (Heidelberg Registry of Atrial Fibrillation) was designed to create real world data from a large number of all‐comer patients with AF, consecutively presenting to the chest pain unit (CPU) of the University Hospital of Heidelberg. The objectives of HERA‐FIB are (1) to characterize patients with AF presenting to a CPU regarding demographics, medical history, risk factors, treatment approaches, and outcomes; (2) to distinguish, characterize, and investigate patient populations of special interest within the registry, which may be underrepresented in current RCTs, such as octogenarians or patients with chronic kidney disease or acute coronary syndrome; and (3) to obtain data and outcomes regarding the shift of anticoagulation regimens from VKAs to NOACs.

Methods

The data that support the findings of this study are available from the corresponding author upon reasonable request.

Design

This single‐center retrospective all‐comer patient registry uses observational study methods ¹⁸ to collect clinical data and evaluate outcomes for patients with AF presenting to the CPU of University Hospital of Heidelberg from June 2009 to March 2020. Since we aimed to create an all‐comer registry, study enrollment of patients was consecutive. Enrollment consisted of 3 phases, and the number of available variables increased with time. The recruitment periods were 2009 to 2013, 2013 to 2017, and 2017 to 2020. We retrospectively pooled the recruitment periods. Accordingly, the pooled analysis contained a minimal data set for the entire cohort (Table 1). As electronic medical record software was unavailable from November 11, 2011, to December 9, 2011, patients presenting within this period were not included in HERA‐FIB. This retrospective observational study had no influence on patient treatment. Treatments, diagnostics, and management of the patients were left at the discretion of the treating physicians. This study was approved by the local ethics committee of the Medical Faculty of Heidelberg. Informed consent was waived for this retrospective analysis of patients from our institution, using data from routine care. This study was conducted according to ethical principles stated in the Declaration of Helsinki (2008). Patient identifiable data were pseudonymized to ensure data confidentiality and were not passed on to third parties. This study is registered at ClinicalTrials.gov. (NCT05995561).

Table 1.

Minimal Data Set Within HERA‐FIB

Variables of the minimal data set
Inclusion and exclusion criteria
Date and time of index presentation and discharge
Admission diagnosis
Demographic data including age, sex and vital parameters (blood pressure, heart rate, weight, height, ECG rhythm)
Laboratory values, if available such as high‐sensitivity cardiac troponin T, creatinine, hemoglobin, C‐reactive protein, NT‐proBNP, glutamic oxaloacetic transaminase, and international normalized ratio
Comorbidities and medical history such as arterial hypertension, diabetes, prior transient ischemic attack/stroke, prior coronary artery disease, coronary artery bypass graft, peripheral artery disease, and myocardial infarction
Antithrombotic and antiplatelet treatment at admission and discharge
Echocardiographic data, if available such as left ventricular ejection fraction
Information regarding AF, such as type of AF
Antiarrhythmic treatment and strategy
Coronary angiogram and revascularization strategy at index or history
Outcome variables such as all‐cause mortality, stroke, myocardial infarction, and major bleeding according to International Society on Thrombosis and Hemostasis criteria

AF indicates atrial fibrillation; HERA‐FIB, Heidelberg Registry of Atrial Fibrillation; and NTproBNP, N‐terminal pro‐B‐type natriuretic peptide.

Study Population and Setting

The purpose of this registry was to provide data on patients with AF as the primary reason for admission or as a comorbidity admitted to the CPU of the University Hospital of Heidelberg with acute clinical presentations and symptoms. Patients were managed in the CPU, which represents a specialized ED that is led by a cardiologist and requires certification by the German Cardiac Society (Deutsche Gesellschaft für Kardiologie). CPUs represent the preferred facilities for cardiac emergencies, enabling a structured, guideline‐based treatment of potentially critical ill patients. Patients presenting to the CPU include but are not limited to patients with any kind of rhythm disorders, acute coronary syndrome, acute or worsening of dyspnea and other potentially cardiac‐related diagnosis or symptoms. Patients presenting to the CPU exhibited a broad range of symptoms including dyspnea, diaphoresis, palpitations, light‐headedness and angina pectoris. Reasons for admission were classified by treating physicians in the ED on the basis of symptoms and clinical cause at presentation to the CPU. Details on the CPU organization and certification requirements were published earlier. ¹⁹

Inclusion and Exclusion Criteria for HERA‐FIB

Inclusion criteria for HERA‐FIB were (1) a diagnosis of AF, either as the primary reason for admission to the CPU or as a comorbidity; (2) age ≥18 years. Patients without at least 1 available laboratory value of high‐sensitive cardiac troponin T at index and patients without complete follow‐up on all‐cause death were excluded. Since enrollment was executed over an 11‐year time period, data were adjusted for repeated visits. The first accessible visit for a patient within the whole recruitment period was chosen for inclusion. Other visits for the same patient were excluded. Figure 1 shows a flow diagram for included and excluded patients within the whole registry.

Figure 1. Flow diagram for included and excluded patients with AF in HERA‐FIB.

AF indicates atrial fibrillation; CPU, chest pain unit; and HERA‐FIB, Heidelberg Registry of Atrial Fibrillation.

Outcomes of Major Interest and Definitions

The following clinical outcomes were assessed during follow‐up: (1) all‐cause death, (2) stroke, (3) myocardial infarction (MI), and (4) major bleeding events. Stroke definition excluded primary hemorrhagic stroke but included ischemic or unknown causes of stroke. Patients who developed a hemorrhagic stroke were allocated as a major bleeding event according to the International Society on Thrombosis and Hemostasis major bleeding definition. Whenever available, computed tomography, magnetic resonance images or medical reports were used to verify the diagnosis. A major bleeding event was defined by International Society on Thrombosis and Hemostasis major criteria. The definition included (1) a fall of hemoglobin level by at least 2 g/dL or a need for red blood cell transfusions or (2) involvement of a critical anatomic site of bleeding such as intracranial or intraocular bleeding. ²⁰ Patients without a complete follow‐up on stroke, major bleeding, or MI were not included in analyses regarding these endpoints. AF was classified as paroxysmal when conversion to sinus rhythm occurred spontaneously within 7 days of presumed onset. ²¹ When terminated using cardioversion (either pharmacological or electrical) or AF was sustained for >7 days, AF type was classified as persistent. When persistent AF was sustained for >12 months, type of AF was classified as long‐standing persistent AF. When AF was accepted without attempts to restore sinus rhythm and the main purpose was heart rate control, AF type was classified as permanent AF. Valvular AF was defined as patients with mechanical prosthetic heart valve(s) or moderate to severe mitral stenosis. ²²

Data Management, Collection, and Follow‐Up

For data collection within the registry, the local electronical archive and all available medical files were screened using the research data warehouse system of the University Hospital of Heidelberg. All variables were cross‐checked by a study nurse and physician, not involved in patient care. Follow‐up was performed by review of in‐ and outpatient visits to the University Hospital of Heidelberg and affiliated hospitals. Here, the files from the electronic hospital information system and review of results from routine medical aftercare were used. For optimizing the integrity of routine follow‐up, a sequential follow‐up method was applied. First, a review of data from our hospital information system and hospital files within other hospitals affiliated with the University Hospital of Heidelberg were screened for information on outcome variables. Afterwards, structured patient phone calls were executed; if not possible, postal queries with standardized questionnaires were conveyed. If patients were still unattainable, registration offices were contacted. However, registration offices could only provide data on vital status and residency. Reported events were adjudicated by a physician, if unequivocal, in a team consisting of 3 cardiologists. All clinical data were accumulated in an electronic data capture system that ensures confidentially. Data were entered as described and checked by physicians and study nurses for quality review.

Statistical Analysis

Continuous variables were tested for normal distribution using the Kolmogorov–Smirnov test. If distributed normally, data were processed as means (SD); if not, data were processed as medians (interquartile range [IQR], 25th–75th percentiles). Statistical analyses were performed using MedCalc version 20.105.

Results

In a total of 76 528 screened visits to the CPU, 24.3% were excluded by repeated visits and 59.6% because of diagnoses other than AF, resulting in a total of 12 297 visits within the inclusion period. Next, 1.5% were excluded for nonsufficient laboratory values, 5.4% after readjustment of the AF diagnosis, and 10% were lost to follow‐up, resulting in a total of 10 222 included patients (Figure 1). Baseline characteristics for the entire cohort are shown in Table 2. Median age was 75 years (IQR, 66–81), 58.6% were male. The median heart rate at presentation was 90 bpm (IQR, 73–117). The median high‐sensitive cardiac troponin T serum level was 19 ng/L (IQR, 10–38). Arterial hypertension was the most frequent risk factor (82.5%). Within the 21 baseline variables shown in Table 2, the mean completeness in data collection was 95.4% (SD, 14.4). With an overall completeness rate of 43.4%, NT‐proBNP (N‐terminal pro‐B‐type natriuretic peptide) was the most sparsely collected variable within the minimal data set of the whole registry.

Table 2.

Baseline Characteristics of HERA‐FIB

Variables		Completeness rate (%)
Age, median (IQR)	75 (66–81)	100
Sex, male, n%all	5957 (58.3)	100
Blood pressure, mm Hg, systolic, median (IQR)	146 (130–160)	99.5
Blood pressure, mm Hg, diastolic, median (IQR)	85 (25–97)	99.5
Heart rate, bpm, median (IQR)	90 (73–117)	100
Body mass index, kg/m2, median (IQR)	26.8 (24.0–30.5)	62.5
hs‐cTnT ng/L, median (IQR)	19 (10–38)	100
Hemoglobin, g/dL, median (IQR)	13.2 (11.8–14.5)	100
C‐reactive protein, mg/L, median (IQR)	6.5 (1.99–21.4)	99.5
Creatinine, mg/dL, median (IQR)	0.99 (0.81–1.29)	100
NT‐proBNP, ng/L, median (IQR)	3158 (1217–8071)	43.4
Glutamic oxaloacetic transaminase, U/L, median (IQR)	27 (21–37)	98.6
International normalized ratio, median (IQR)	1.1 (1.0–1.7)	98.8
Arterial hypertension, n%all	8436 (82.5)	100
Diabetes, n%all	2043 (20.0)	100
Prior coronary artery bypass graft, n%all	942 (9.2)	100
Prior MI, n%all	1672 (16.2)	100
Prior peripheral artery disease, n%all	796 (7.8)	100
Prior coronary artery disease, n%all	4414 (43.2)	100
Prior transient ischemic attack/stroke, (n%all)	1293 (12.7)	100
Valvular AF, (n%all)	227 (2.2)	100

AF indicates atrial fibrillation; HERA‐FIB, Heidelberg Registry of Atrial Fibrillation; hs‐cTnT, high‐sensitivity cardiac troponin T; IQR, interquartile range; MI, myocardial infarction; and NT‐proBNP, N‐terminal pro‐B‐type natriuretic peptide.

Outcome Data Within HERA‐FIB

Taking into consideration that follow‐up for all‐cause death was an inclusion criterion, a complete follow‐up on all‐cause death was achieved. Completeness of follow‐up rates for other outcome parameters such as stroke, MI, and major bleeding were 94.7%, 94.6%, and 94.6%, respectively (Table 3). Kaplan–Meier analyses for all‐cause death, stroke, major bleeding and MI are shown in Figure 2. Within a median follow‐up time of 23 months (IQR, 12–35), a total of 21.3% patients died, 3.3% developed a stroke, 4.4% suffered an MI, and 5.9% suffered a major bleeding event according to International Society on Thrombosis and Hemostasis major criteria. Outcome variables stratified by 1‐year event rates were 14.7% for all‐cause death, 1.7% for stroke, 1.9% for MI, and 1.2% for major bleeding events.

Table 3.

Outcome Variables of HERA‐FIB

Variables	n (%)	Completeness of follow‐up (%)	1‐year outcome, n (%)
All‐cause death	2173 (21.3)	100	1502 (14.7)
Stroke	287 (3.3)	94.7	160 (1.7)
Major bleeding	514 (5.9)	94.6	318 (3.3)
Myocardial infarction	382 (4.4)	94.6	179 (1.9)

HERA‐FIB indicates Heidelberg Registry of Atrial Fibrillation.

Figure 2. Kaplan–Meier analyses for all‐cause death (A), stroke (B), major bleeding events (C), and MI (D).

MI indicates myocardial infarction.

Populations of Special Interest and OAC Regimens Within HERA‐FIB

Within HERA‐FIB, 36.9% of the patients presented with AF as primary reason for admission. Among the 63.1% of patients in which AF was a comorbidity, cardiac decompensation (15.1%), acute coronary syndrome (11.9%), and infection (7.7%) were the most frequent reasons for admission to the CPU. A list of the top reasons for admission is shown in Table 4. Within the registry, we assessed the prevalence of patient populations of special interests, such as patients with elevated high‐sensitive cardiac troponin T >99th reference limit (60.7%), elderly (52.0%), octogenarians (27.5%), or patients with an uncontrolled heart rate (>110 bpm; 30.2%) at admission (Table 5). Risk for all‐cause death differed among populations of patients with AF of special interest. All‐cause death in patients with AF presenting to the CPU with an acute coronary syndrome was lower compared with patients with AF presenting with decompensated heart failure (Figure 3). Kaplan–Meier analyses for other outcome parameters such as stroke, major bleeding, and MI are shown in Figure 2. Temporal trends for the use of any anticoagulant between 2009 and 2020 are shown in Figure 4. Here, rates of patients receiving any OAC increased from 48.7% in 2009 and 45.3% in 2010 to 84.9% in 2019 and 83.8% in 2020, respectively. Since this registry covers the period of transition in OAC regimens from VKAs to NOACs, the trend for distribution of OAC regimens at discharge changed within the observed years. Temporal trends for the transition from VKAs to NOACs between 2009 and 2020 are shown in Figure 5. Rates of VKAs decreased, as rates for NOAC regimens at discharge increased from 0% and 0.6% in 2009 and 2010 to 87.6% and 86.3% in 2019 and 2020, respectively. For the entire inclusion period, the rate of newly initiated OAC regimens for anticoagulation‐naive patients was 23.9%. Of those, 27.8% were regimens containing VKAs, and 72.2% were NOAC regimens. Rates of newly initiated OAC regimens of VKAs and NOACs varied over time. In 2009 and 2010, newly initiated OAC regimens containing VKAs were initiated in 100% and 99.4%, whereas in 2019 and 2020, VKA‐containing regimens were initiated in 2.1% and 0%, respectively. As rates of newly initiated VKAs declined, rates of NOACs increased over time, with a maximum of 97.9% in 2019 and 100% in 2020 (Figure 6).

Table 4.

Reasons for Admission to the ED in HERA‐FIB

Variables	Prevalence, n (%)
AF as primary reason for admission, n%all	3776 (36.9)
Decompensated heart failure with AF, n%all	1538 (15.1)
Acute coronary syndrome with AF, n%all	1213 (11.9)
Any infection with AF, n%all	791 (7.7)
Noncardiac chest pain with AF, n%all	391 (3.8)
Hypertension with AF, n%all	345 (3.4)
Syncope with AF, n%all	178 (1.7)
Others,* n%all	1990 (19.5)

AF indicates atrial fibrillation; ED, emergency department; and HERA‐FIB, Heidelberg Registry of Atrial Fibrillation.

^*Other reasons for admission with AF included but were not limited to: exacerbated chronic obstructive pulmonary disease, acute asthma exacerbation, pulmonary embolism, acute aortic dissection, and myocarditis.

Table 5.

Prevalence of Study Populations of Special Interest

Variables	Prevalence, n (%)
Patients without OAC at discharge	3096 (30.3)
Patients with OAC at discharge	7126 (69.7)
NOAC at discharge	4241 (59.5)
VKA at discharge	2885 (40.5)
Newly initiated OAC	2448 (23.9)
Elevated hs‐cTnT >99th reference limit	6229 (60.9)
Elderly (≥75 y)	5318 (52.0)
Octogenarians*	2816 (27.5)
Uncontrolled heart rate>110 bpm	3086 (30.2)
Noninvasive rhythm control strategy†	1552 (15.2)
CKD‡	3888 (38.0)

Patients might be counted twice, as they might appear in multiple different patient populations of interest. CKD indicates chronic kidney disease; hs‐cTnT, high‐sensitivity cardiac troponin T; NOAC, non–vitamin K anticoagulant; OAC, oral anticoagulation; and VKA, vitamin K antagonist.

^*Defined as patients aged between 80 and 89 y.

^† Noninvasive rhythm control strategy includes patients with electric or pharmacologic cardioversion.

^‡ Defined as estimated glomerular filtration rate <60 mL/min per 1.73 m².

Figure 3. Kaplan–Meier analysis for all‐cause death and populations of special interest within HERA‐FIB.

Patients might be counted twice as they might appear in multiple different patient populations of interest. ACS indicates acute coronary syndrome; CKD, chronic kidney disease; HERA‐FIB, Heidelberg Registry of Atrial Fibrillation; and HF, heart failure.

Figure 4. Usage rates of oral anticoagulation regimens versus no oral anticoagulation within HERA‐FIB.

HERA‐FIB indicates Heidelberg Registry of Atrial Fibrillation.

Figure 5. Trend for distribution of OAC regimens at discharge within HERA‐FIB.

HERA‐FIB indicates Heidelberg Registry of Atrial Fibrillation; OAC, oral anticoagulation; and VKA, vitamin K antagonist.

Figure 6. Trend for new initiated OAC regimens within HERA‐FIB.

HERA‐FIB indicates Heidelberg Registry of Atrial Fibrillation; OAC, oral anticoagulation; and VKA, vitamin K antagonist.

Discussion

HERA‐FIB is a large single‐center observational study, consecutively including patients with AF presenting to a CPU. Its purpose was to collect data and create evidence for patients with the diagnosis of AF in the clinical setting of an ED. HERA‐FIB is enabling insights to the clinical real world for this particular patient population and creates evidence for subgroups of patients who have been routinely excluded or are underrepresented in RCTs or registries.

Observational studies offer insights into factors, outcome parameters, and subgroups of patients usually not included or underrepresented in RCTs. They are valuable instruments for monitoring the course and progression of a disease and have proven to be effective in evaluating treatments and treatment effectiveness in recent years. For patients with AF, large observational studies with consecutive enrollment are rare. Furthermore, owing to missing randomization, interpretation of results from such studies might be challenging. HERA‐FIB is an observational study with a comprehensive inclusion of patients with AF presenting over an 11‐year period and one of the largest single‐center all‐comer registries on patients with AF.

As a result of the study design, this registry also includes patients with preexisting AF. Thus, the structure and design of HERA‐FIB differs significantly from other studies or registries such as large industry‐sponsored registries like the GLORIA AF (Global Registry on Long‐Term Oral Antithrombotic Treatment in Patients With Atrial Fibrillation). ²³ , ²⁴ Besides the inclusive study design, the studied patient population also differs from other AF registries. Patients in HERA‐FIB were managed in a CPU, which represents an organized structure for cardiac emergencies. These patients were presenting in an acute setting with a potentially critical clinical situation. Within HERA‐FIB, a total of 36.9% of the patients presented with AF as the admission diagnosis, whereas in 63.1%, AF was a comorbidity. These differences in study design and study population allow HERA‐FIB to create evidence for an important patient population in clinical practice and offers unique insights to the characteristics of patients with AF in a CPU.

In our study, we report a median age of 75 years and found coronary artery disease as a comorbidity in 43% of the patients presenting in the setting of an ED. In contrast, rates of coronary artery disease were 24% in a large Swedish nationwide cohort with data extracted from a nationwide registry, regardless of the clinical context. In the Swedish registry, the median age was 71 years, and thus a median of 4 years younger than in HERA‐FIB, suggesting that both age and clinical context may influence overall rates of comorbidities such as coronary artery disease. ²⁵

The all‐cause mortality rate in HERA‐FIB was 21.3% for the entire follow‐up period with a median of 23 months, and 14.7% for the 1‐year follow‐up, respectively. Here, we report a noticeably higher all‐cause mortality rate than multicenter studies such as the EHS (Euro Heart Survey) or the EORP‐AF (EURObservational Research Program Atrial Fibrillation General Pilot Registry) in their 1‐year reports who report all‐cause mortality rates of 5.3% and 5.7%, respectively. ²⁶ , ²⁷ However, HERA‐FIB addresses patients with AF presenting in an acute setting also including critical ill patients, whereas EHS or EORP‐AF predominantly included stable in‐ and outpatients with AF. ²⁶ , ²⁷ Therefore, all‐cause mortality rates in our registry are not unexpectedly higher than in EHS or EORP‐AF. Thus, the 1‐year stroke rate of HERA‐FIB was 1.7% and is comparable to the stroke rate of the EORP‐AF 1‐year pilot registry, which reports a stroke rate of 1.8%. Again, this fact might be attributed to the study setting. A CPU is led by a cardiologist and offers management for patients with acute cardiovascular and cardiopulmonary emergencies. Therefore, a selection bias for critical ill patients with AF with stroke may be assumed, since these patients are triaged to stroke units or neurological EDs in the clinical real world. ²⁶

AF and MI often coexist. However, outcome data focusing on incident MI in large all‐comer AF registries is rare. A meta‐analysis from 2016 reported an annual MI rate ranging from 0.4% to 2.5% depending on the examined patient population, trial design, and clinical setting. ²⁸ Within the RE‐LY (Randomized Evaluation of Long‐Term Anticoagulation Therapy) study ²⁹ reporting on a patient population with a median age of 71.5 years and a follow‐up of 2 years, the observed number of MIs was 1.5% to 1.0%, whereas in the ENGAGE AF TIMI 48 (Effective Anticoagulation With Factor Xa Next Generation in Atrial Fibrillation–Thrombolysis in Myocardial Infarction 48) trial, ³⁰ an MI rate of 1.9% to 2.0% in a population of patients with AF with a median age of 72.0 years during a follow‐up of 2.8 months was observed. In HERA‐FIB, we report an annual MI rate of 1.9% and an overall MI rate of 4.4%. However, HERA‐FIB consisted of a patient population with a median age of 75 years and was conducted in the setting of an ED. Thus, higher MI rates might be related to the study setting and a generally elderly patient population, as expected from a real world registry focusing on AF patients presenting to an ED. Within HERA‐FIB, outcomes of AF patients differed in respect to subpopulations depending on the diagnosis at presentation. Therefore, an individualized subpopulation focused assessment of risk factors and determinants of outcomes is necessary in order to expand our understanding of usually underrepresented subsets of patients with AF presenting to an ED.

HERA‐FIB was executed in 3 phases from 2009 to 2013, 2013 to 2017, and 2017 to 2020, enabling a single‐center summary of OAC use in clinical practice of patients with AF. Many multicenter studies are executed in multiple, particularly overlapping phases such as GLORIA AF, ²³ where patients with preexisting AF were excluded. Similar to the results from large multicenter trials such as GLORIA AF ²³ , ²⁴ or the GARFIELD‐AF (Global Anticoagulant Registry in the Field–Atrial Fibrillation), ³¹ OAC use in HERA‐FIB increased over time and rates of NOAC‐ versus VKA‐containing regimens also increased. Thus, our results concerning the shift of OAC regimens are in line in with the data from other registries, showing a shift from VKA‐ to NOAC‐containing regimens in last decade. ¹¹ Usage rates of OAC regimens in HERA‐FIB are lower compared with other registries. In the EORP‐AF registry collecting data from February 2012 to March 2013 including patients with primary diagnosed AF, the OAC usage rate was 80%. In our all‐comer real world registry, rates of OAC in 2012 and 2013 were 55.7% and 66.2%, respectively. The reported difference of OAC usage is potentially attributed to the difference in the study design between EORP‐AF, ²⁶ where patients with primary diagnosed AF were included and OAC therapy was initated, and HERA‐FIB, which included patients with preexisting AF, where OAC regimens or refusal of an OAC was preexisting. However, in the beginning of HERA‐FIB, the use of OACs was inappropriately low but increased over time. The fast uptake may be explained by higher guidelines adherence, stimulated by growing evidence from studies and registries showing low implementation rates of OACs in clinical routine. Additionally, the implementation of NOACs in clinical routine created a higher awareness for an appropriate use of OACs in patients with AF.

Through the retrospective comprehensive character of the registry, one could assume low data collection rates. However, for clinical end points within the registry, we were able to achieve a data collection rate of at least >94% for stroke, MI, and major bleeding events. To date, many national and countrywide studies and registries were prompted to create data on patients with AF such as the ORBIT‐AF (Outcomes Registry for Better Informed Treatment of Atrial Fibrillation), ³² the Atrial Fibrillation Network, ³³ and EORP‐AF. ³⁴ Through the design of an all‐comer registry with consecutive enrollment over an 11‐year period, HERA‐FIB allows creating evidence for subgroups of patients with AF, who are clinically important and frequently underrepresented in other studies, national registries, or RCTs. Since the discovery that systemic anticoagulation is superior to thrombocyte inhibition, ⁵ the introduction of NOAC regimens for stroke prevention represented the most important advancement in AF treatment in the past decades. Owing to the long inclusion period, without substantial exclusion criteria HERA‐FIB documents patients with AF, use of OAC regimens in the shift from VKAs to NOACs, and allows investigations within different treatment regimens and eras.

Strengths

HERA‐FIB provides information on >10 000 consecutive patients with AF recruited continuously from 2009 until 2020. This registry contains real world evidence on treatment practice; guideline adherence; and insights into populations of special interest that are underrepresented in RCTs, such as the elderly, octogenarians and patients with severe kidney disease, acute coronary syndromes and heart failure. The registry is unique, as it is one of the largest single‐center registries and covers the most interesting period of time for studying OAC regimens starting at the time when NOACs entered the market. Additionally, HERA‐FIB offers a follow‐up period over a median of 23 months with very high follow‐up rates for incident stroke, MI, and major bleeding complications according International Society on Thrombosis and Hemostasis criteria. Together with the close follow‐up, the collection of a large set of demographic and clinical data allows an evaluation of risk stratification tools including biomarkers, such as high‐sensitivity cardiac troponin T, as well as an assessment of outcomes by treatment with NOACs versus VKAs.

Limitations

This study is an observational single‐center study on patients with AF presenting to an ED. Thus, it is essential to interpret our findings in the context of an acute setting with a potentially critical clinical situation. As described above, the data for the registry had to be collected from electronic archives, discharge letters, structured phone interviews, questionnaires and registration offices. Therefore, we cannot fully exclude human errors leading to an incomplete or incorrect collection within this registry. Rates of patients lost to follow‐up were relatively high, owing to the design of an observational retrospective study on patients in an emergency situation without substantial exclusion criteria. Thus, survival status could be obtained in ≥90%, which is remarkable for an observational retrospective study over an 11‐year period. The design of the registry did not assess time in therapeutic range and thus does not allow any conclusions on the quality of anticoagulation therapies with coumadin or phenprocoumon. Additionally, HERA‐FIB did not systematically collect data on health care outcomes related to resource usage such as readmissions, nor data reflecting the community diversity such as economic data. However, the admission policy of a university hospital reduces the likelihood of inequity, as patients are admitted and treated regardless of insurance or economic status. Owing to the heterogenicity of outcomes among the subpopulations of patients with AF within HERA‐FIB, an individualized analysis is necessary for each subpopulation of patients with AF. However, this subanalysis is far beyond the scope of the present work but might allow a more comprehensive understanding of risk factors and treatment strategies affecting outcomes and risk factors of these patient populations.

Conclusions

HERA‐FIB represents a large, structured, real world collective of patients with AF presenting to an ED over an 11‐year period. The scope of the present study was to describe the composition and data collection process of HERA‐FIB, including brief insights for patient populations of special interest. However, a detailed investigation of different subpopulations will require in‐depth analyses for each subgroup of interest. This strategy would exceed the scope of the present work and is intended as future publications.

Sources of Funding

The study was supported by a research grant from Daiichi Sankyo and Bayer Diagnostics. The sponsor had no influence on the study concept, data collection, or interpretation. For the publication fee, we acknowledge financial support by Heidelberg University.

Disclosures

Dr Müller‐Hennessen received research funding from BRAHMS, Thermo Fisher Scientific, and Roche Diagnostics. Dr Biener received research support from AstraZeneca outside the submitted work. Dr Giannitsis received honoraria for lectures from Roche Diagnostics, AstraZeneca, Bayer, Daiichi‐Sankyo, and Lilly Eli Deutschland. He serves as a consultant for Roche Diagnostics, BRAHMS Thermo Fisher Scientific, and Boehringer Ingelheim and has received research funding from BRAHMS Thermo Fisher Scientific, Roche Diagnostics, Bayer Vital, and Daiichi Sankyo. Dr Frey has received speaker honoraria from Daiichi Sankyo, Astra Zeneca, Boehringer Ingelheim, and Bayer Vital. He serves as a consultant for CMS. Dr Milles received research funding from Daiichi Sankyo. The remaining authors have no disclosures to report regarding this manuscript.

Supporting information

Figure S1