The Heart and Stroke Foundation of Ontario had no involvement in the design or conduct of the study, data management or analysis, or manuscript preparation, review, or authorization for submission. This study was supported by the Institute for Clinical Evaluative Sciences (ICES), which is funded by an annual grant from the Ontario Ministry of Health and Long-Term Care (MOHLTC). It was also supported by the Department of Emergency Services, Sunnybrook Health Sciences Centre, the Canadian Cardiovascular Outcomes Research Team, and the Institute for Clinical Evaluative Sciences. The opinions, results, and conclusions reported in this paper are those of the authors and are independent from the funding sources. No endorsement by ICES or the Ontario MOHLTC is intended or should be inferred.
Atrial fibrillation is common in the emergency department (ED). Mortality rates at 30, 90, and 365 days for ED patients with a main diagnosis of atrial fibrillation are 4, 6, and 11%, respectively; there are no data on the characteristics and outcomes of ED patients with atrial fibrillation who have alternative primary ED diagnoses.
In this single-site, retrospective cohort study, all electrocardiograms (ECGs) with confirmed atrial fibrillation performed in the ED from April 2007 to March 2008 were identified. Repeat ED visits were excluded. ECGs associated with a primary ED diagnosis of atrial fibrillation were excluded, and from the remaining ECGs of patients with a different primary ED diagnosis, half were randomly selected for abstraction. The main outcome measure was all-cause mortality at 30, 90, and 365 days post-ED visit, derived from linkage to a provincewide mortality database. As a secondary analysis, logistic regression was used to compare 90-day mortality of these patients to those with primary ED diagnoses of atrial fibrillation seen during the same time period.
Of 768 qualifying index ED visits, 416 charts were abstracted. Mean (±standard deviation [SD]) age was 80.3 (±11.8) years, and 50.7% were female. Two-thirds had a previous history of atrial fibrillation/flutter, 300 (72.1%) had a CHADS2 score ≥ 2, one died in the ED, and 275 (66.1%) were admitted. The most common primary ED diagnoses were congestive heart failure (12%), pneumonia (6%), and chest pain not yet diagnosed (6%), while most common in-hospital diagnoses were congestive heart failure (15%), chronic obstructive pulmonary disease exacerbation (6%), atrial fibrillation (5%), and pneumonia (5%). Mortalities at 30, 90, and 365 days were 10.6% (95% confidence interval [CI] = 7.8% to 14.0%), 17.4% (95% CI = 13.9% to 21.5%), and 34.2% (95% CI = 29.6% to 39.0%), respectively. In the adjusted analysis, an alternative primary ED diagnosis was associated with an increased risk of death (odds ratio [OR] = 2.75; p = 0.01).
Patients seen in the ED with atrial fibrillation and different primary ED diagnoses are older and have high short- and long-term mortality rates: mortality was three times higher than in patients with primary ED diagnoses of atrial fibrillation. Future studies of atrial fibrillation in the ED should distinguish between these two populations and the potential contribution of atrial fibrillation to mortality in the setting of other primary ED diagnoses.
Pacientes con Fibrilación Auricular y un Diagnóstico Principal Alternativo en el Servicio de Urgencias: Una Descripción de Sus Características y Su Evolución
La fibrilación auricular es frecuente en el servicio de urgencias (SU). Los porcentajes de mortalidad a los 30, 90 y 365 días para los pacientes del SU con un diagnóstico principal de fibrilación auricular son del 4%, 6% y 11%, respectivamente. No existen datos de las características y los resultados de los pacientes del SU con fibrilación auricular que tienen un diagnóstico principal alternativo en el SU.
En este estudio de cohorte retrospectivo de un único centro, se identificaron todos los electrocardiogramas realizados con fibrilación auricular confirmada en el SU desde abril del 2007 a marzo 2008. Se excluyeron las visitas repetidas al SU. Se excluyeron los electrocardiogramas asociados con un diagnóstico principal de fibrilación auricular en el SU, y del resto de electrocardiogramas de los pacientes con un diagnóstico principal diferente en el SU, se seleccionó la mitad de forma aleatorizada. El resultado principal fue la mortalidad por cualquier causa a los 30, 90 y 365 días tras la visita al SU, obtenida a través de un enlace a la base de datos de mortalidad de toda la provincia. Se realizó un análisis secundario mediante una regresión logística para comparar la mortalidad a 90 días de estos pacientes con aquéllos con diagnóstico principal de fibrilación auricular en el SU durante el mismo periodo del estudio.
De las 768 visitas al SU clasificadas, se escogieron 416 historias clínicas. La media de edad fue de 80,3 años (DE 11,8 años) y el 50,7% fueron mujeres. Dos tercios tenían una historia previa de fibrilación/fluter auricular, 300 (72,1%) tuvieron una puntuación de CHADS2 ≥ 2, uno falleció en el SU y 275 (66,1%) ingresaron. El diagnóstico principal más frecuente en el SU fue la insuficiencia cardiaca congestiva (12%), la neumonía (6%) y el dolor torácico indeterminado (6%); mientras que los diagnósticos más comunes en los ingresados fueron la insuficiencia cardiaca congestiva (15%), la exacerbación de la enfermedad pulmonar obstructiva crónica (6%), la fibrilación auricular (5%) y la neumonía (5%). La mortalidad a los 30, 90 y 365 días fue del 10,6% (IC 95% = 7,8% a 14,0%), 17,4% (IC 95% = 13,9% a 21,5%), y 34,2% (IC 95% = 29,6% a 39,0%), respectivamente. Tras el análisis ajustado, un diagnóstico principal alternativo en el SU se asoció con un incremento del riesgo de muerte (OR 2,75; p = 0,01).
Los pacientes atendidos en el SU con fibrilación auricular y un diagnóstico principal diferente son de mayor es edad y tienen mayor porcentajes de mortalidad a corto y largo plazo: la mortalidad fue tres veces mayor que en los pacientes con un diagnóstico principal de fibrilación auricular en el SU. Futuros estudios de la fibrilación auricular en el SU deberían distinguir entre estas dos poblaciones y la potencial contribución de la fibrilación auricular a la mortalidad en este escenario con otros diagnósticos principales.
Atrial fibrillation is common. In the United States it is estimated that 2.3 million people have atrial fibrillation, including 9% of persons over the age of 80 years. Many of these patients will visit the emergency department (ED): in 2006 there were 564,000 ED visits in the United States with primary diagnoses of atrial fibrillation, up 88% from 1994. Despite the high prevalence of atrial fibrillation in the ED, there is relatively little research on the emergency component of care of these patients,[2-4] with no published literature on patients with atrial fibrillation who are seen in the ED with different primary ED diagnoses. As the population ages, the number of patients seen in the ED with secondary (or tertiary, quaternary, etc.) diagnoses of atrial fibrillation is projected to markedly increase.
Emergency department patients with secondary diagnoses of atrial fibrillation may constitute a separate cohort from those with primary ED diagnoses of atrial fibrillation, with potentially different outcomes. While atrial fibrillation is considered a relatively benign disease if appropriately treated, in the emergency setting, a secondary diagnosis may have significantly altered prognostic implications. Some studies suggest that hospitalized congestive heart failure (CHF) patients have worse prognoses in the setting of atrial fibrillation, that atrial fibrillation is a marker for increased risk of death in patients with acute myocardial infarction (MI), and that new-onset atrial fibrillation is associated with markedly higher mortality in severe sepsis patients. Patients with secondary diagnoses of atrial fibrillation in the ED setting are often encountered, but have yet to be characterized. The purpose of this study was to report the characteristics, primary ED diagnoses, hospital admission rate, and mortality rates in the previously undescribed group of ED patients with secondary ED diagnoses of atrial fibrillation.
This was a single-site retrospective cohort study. We obtained ethics approval from Sunnybrook Health Sciences Centre.
Study Setting and Population
The ED at Sunnybrook Health Sciences Centre is an academic site in Toronto that serves 48,000 patients annually. It is a provincial trauma center and has in-house access to all services except obstetrics (including neurosurgery, vascular surgery, etc). Acuity level is high, with a 24.2% admission rate. It is one of the base hospitals for the University of Toronto emergency medicine residency training program. All patients age 18 to 105 who were seen in the Sunnybrook Health Sciences Centre ED from April 1, 2007, to March 31, 2008 with a “secondary diagnosis” of atrial fibrillation were eligible for inclusion.
The National Ambulatory Care Reporting System (NACRS) is a province-wide administrative database that contains abstracted data on all visits to Ontario EDs. Each hospital in Ontario is mandated to record NACRS data and submit the data to the Canadian Institutes of Health Information. At each hospital, trained NACRS abstractors collect over 200 data points for inclusion in the NACRS database, including the primary ED diagnosis, and up to nine additional ED diagnoses, herein referred to collectively as secondary diagnoses.
To determine out-of-hospital mortality, data were linked to the Registered Persons Database, a population database in Ontario that provides accurate mortality data for all Ontario residents. To determine in-hospital information on admitted patients, we linked the study cohort to the Discharge Abstract Database, which contains information on all Ontario hospitalizations. The Ontario Health Insurance Plan includes outpatient visits and was used in conjunction with NACRS and the Discharge Abstract Database, to determine Adjusted Diagnosis Group (ADG) score.[9-11] Because Ontario provides universal health care coverage, the vast majority of inpatient and outpatient visits are documented in these databases.
Tracemaster software (Philips Healthcare, Andover, MA) was used to gather all electrocardiograms (ECGs) that were performed in the ED during our study period (>15,000 ECGs). The software allowed us to run an inquiry on these ECGs for those with a final interpretation (by a cardiologist) of atrial fibrillation. This method of identification of study patients was necessary because while NACRS does collect up to nine “secondary” ED diagnoses, these fields may be of limited value given that emergency physicians (EPs) infrequently record more than one ED diagnosis on the ED chart. Thus the definition of a secondary diagnosis of atrial fibrillation had to be based on ECG criteria, to create a cohort that avoided the low sensitivity of reported secondary ED diagnoses.
We identified eligible patients using Tracemaster software to collect all ED ECGs with cardiologist interpretations of atrial fibrillation; ECGs with atrial flutter were excluded. By pulling all patient charts, we separated patients with a primary diagnosis of atrial fibrillation from those with other primary ED diagnoses, using the EPs' first diagnosis written on the ED chart. From the latter ECGs, we randomly selected half of the index ED visits (first ED visit within the study period) for full chart abstraction; all ECGs associated with a primary ED diagnosis of atrial fibrillation were also abstracted. If a patient had ED visits with both a primary and a secondary ED diagnosis of atrial fibrillation during the study period, he or she could be included for both the index primary and the secondary diagnosis ED atrial fibrillation visits.
Data Collection and Processing
A data dictionary was created a priori. Where data were not recorded, it was marked as negative, as per previous chart abstractions. For example, if hypertension was not recorded on the emergency chart (physician notes, nursing notes, triage note, consultation note, hospital discharge summary when admitted), then it was recorded as “no history of hypertension.” We validated the accuracy of the emergency chart when using this method on a separate sample of 50 charts with primary diagnoses of atrial fibrillation. Using the consultation reports, discharge summaries, and, where these were not available, subsequent office visits as the criterion standard, we assessed agreement on ED chart variables of past medical history and medication use: simple agreement ranged from 89.4% (hypertension) to 100% (CHADS2 ≥ 2, history of diabetes, history of ischemic stroke, hydrochlorathiazide use, amiodarone use, and glyburide use), with the majority over 95%.
Using a standardized data collection instrument, patient demographics, visit characteristics, ED disposition data, and hospital discharge information was abstracted by two trained chart abstractors (KL and CY). When the abstraction proceeded more quickly than expected, an extra 32 secondary diagnosis charts were abstracted using the randomization schedule. A third abstractor (CA) reabstracted 20% of the same charts for two a priori selected variables: CHADS2 score13 and ED discharge diagnoses. Data were cleaned using range and logic checks.
Using the unique encrypted health card number for each patient, the data were linked to the Registered Persons Database, the Discharge Abstract Database, and the Ontario Health Insurance Plan. Descriptive analyses were performed using 95% confidence intervals (CIs), means with standard deviations (SDs), and medians with interquartile ranges (IQRs), where appropriate. A secondary analysis was performed using logistic regression to compare the adjusted 90-day mortality to the group of patients seen in the same time period with a primary ED diagnosis of atrial fibrillation. Model covariates included age, sex, CHA2DS2-VASc criteria (history of CHF, hypertension, diabetes mellitus, stroke, and coronary artery disease), history of atrial fibrillation or flutter, presenting heart rate, triage score (CTAS 1/2/3, or 4/5, which corresponds to high acuity [should be seen by a physician immediately, at 15 minutes, or at 30 minutes, respectively] or low acuity [should be seen within 1 or 2 hours, respectively]), international normalized ratio (INR) level (<2, 2 to 3, or >3), admission to hospital, a secondary diagnosis of acute coronary syndrome, discharge (from hospital or ED) on warfarin, and ADG score.[9-11] ADG risk scores are used in ambulatory patient populations in place of a risk score tool for hospitalized patients; given that many patients were not hospitalized, a Charlson score would not be available for them. Collinearity and goodness-of-fit testing were performed. All analyses were performed using SAS software (Version 9.2, SAS Institute Inc., Cary, NC).
During the study period there were 1,498 ED visits with ECGs that had cardiologist interpretations of atrial fibrillation. By index ED visits, there were 1,034 ECGs. There were 266 associated with ED visits with primary ED diagnoses of atrial fibrillation. From the remaining 768 eligible ECGs, we abstracted data from a random sample of 416 charts. Raw agreement of reabstracted variables was 100%.
Patient demographics, presenting symptoms, and past medical history are shown in Table 1, with primary atrial fibrillation patients provided for comparison. The most common presenting chief complaints (experienced by 5% or more) are listed in Table 2, along with the top 10 primary ED diagnoses. Among the 416 patients, two or more patients had a total of 26 different primary ED diagnoses.
Table 1. Patient Demographics in 416 ED Patients With AF and an Alternative Primary ED Diagnosis
Secondary AF Diagnosis, n = 416 (%)
Primary AF Diagnosis, n = 266 (%)
AF = atrial fibrillation; CHADS2 ≥2 = congestive heart failure, hypertension, age ≥ 75, diabetes mellitus, stroke; COPD = chronic obstructive pulmonary disease; CTAS = Canadian Triage Acuity Scale; MI = myocardial infarction.
Ten secondary AF diagnosis patients missing a triage score.
Small cell size, <5, cannot be reported for privacy reasons.
There were five (1.2%) patients who were on antiarrhythmic agents on arrival in the ED, and 229 (55.4%) were on warfarin. Of the 300 patients with CHADS2 ≥ 2 (plus four patients with valvular disease), 182 (59.9%) arrived on warfarin (another 47 were on warfarin and had CHADS2 < 2). Taking into account possible contraindications, including a history of hemorrhagic stroke, gastrointestinal bleeding, alcoholism, falls, and dementia (total 99 patients), 39 of the patients with CHADS2 ≥ 2 and/or valvular disease had one of these relative contraindications, leaving 79 (26.0%) eligible patients not on warfarin when they arrived in the ED. Of the 229 patients who arrived on warfarin, 197 had ED INRs available: 52 (26.4%) were subtherapeutic, while 68 were supratherapeutic (34.5%). Vitamin K was administered in the ED to 22 patients.
Linkage to the Registered Persons Database was successful for 407 study patients (97.8%). Mortality rates, admission rates, in-hospital deaths (only one patient died in the ED), and final hospital diagnoses are shown in Table 3. The five most common primary ED diagnoses of patients who died in-hospital were CHF (15%), pneumonia (9%), intracranial hemorrhage (9%), sepsis (9%), and ischemic stroke (6%), while the top five presenting chief complaints were altered level of consciousness (18%), abdominal pain (15%), shortness of breath (15%), weakness/fatigue (9%), and chest pain (6%). Median presenting heart rate was similar in admitted versus discharged patients (89.0 beats/min, IQR = 74.0 to 110.0 beats/min vs. 87.0 beats/min, IQR = 71.0 to 98.0 beats/min) and in those who had died at 30 days compared to those who survived (94.0 beats/min, IQR = 70.0 to 110.0 beats/min vs. 88.0 beats/min, IQR = 73.0 to 104.0 beats/min). By chronicity of the atrial fibrillation, there was no difference in mortality at 30 days in those with a history of atrial fibrillation versus new atrial fibrillation (11.4%, 95% CI = 7.8% to 15.8% vs. 9.6%, 95% CI = 5.2% to 15.9%), nor at 90 days (17.8% each). In the adjusted analysis, a secondary diagnosis of atrial fibrillation was associated with a higher risk of 90-day death (hazard ratio = 2.75, 95% CI = 1.24 to 6.11; p = 0.01), relative to a primary ED diagnosis of atrial fibrillation (Figure 1).
Table 3. Outcomes of the 416 Patients With AF and an Alternative Primary ED Diagnosis
Secondary AF Diagnosis (n = 416)
Primary AF Diagnosis (n = 266)
Values are reported as n (%; 95% CI)
AF = atrial fibrillation.
A total 407 secondary AF diagnosis patients able to be linked to Registered Persons Database
To the best of our knowledge, this is the first study to report the clinical characteristics and outcomes of the commonly seen ED patient who has atrial fibrillation and a different primary ED diagnosis. We found that these patients had remarkably high mortality rates, both overall and relative to those with a primary ED diagnosis of atrial fibrillation: almost 8% went on to die in hospital, with 11% dead at 30 days, 17% dead at 90 days, and a full third of them dead by 1 year after the ED visit. In comparison, primary ED atrial fibrillation patients have 30-day mortality rates of 3.6% at 30 days, 6.1% at 90 days, and 11% at 1 year. It appears that a patient with atrial fibrillation as a secondary ED diagnosis is markedly different, in terms of prognosis, from one with a primary diagnosis of atrial fibrillation.
The top three primary ED diagnoses in the study group were the relatively common CHF, pneumonia, and chest pain not yet diagnosed; however, even the most common primary diagnosis, CHF, constituted only 12% of study patients. In all ED CHF patients, the documented 30-day mortality rate is slightly lower than what was reported in our cohort, at 9.7%. Because some of these patients likely have atrial fibrillation, the rate is likely lower in a cohort of CHF patients who do not have atrial fibrillation. Our results suggest the possibility that ED patients with both acute CHF and atrial fibrillation are a group of patients who are at high risk of poor outcomes, a finding that has been demonstrated in the non-ED setting.[5, 18] It may be the same for patients with a history of atrial fibrillation who develop other diseases, such as pneumonia (which has a 30-day mortality rate of approximately 5%). Our results are consistent with studies on acute MI, where concomitant atrial fibrillation results in worse outcomes, regardless of whether the atrial fibrillation is new or chronic. A recent study on new-onset atrial fibrillation in the setting of sepsis showed a markedly worse prognosis for these patients, compared to patients with sepsis without atrial fibrillation, again suggesting that atrial fibrillation in the setting of another acute disease process has a pronounced effect on prognosis. However, as the current study has no matched control group without atrial fibrillation to allow a direct comparison of mortality between these secondary atrial fibrillation patients and similar patients without atrial fibrillation, these data can only raise additional questions regarding the potential effect of atrial fibrillation on mortality in patients with other emergent conditions: these data cannot determine whether new atrial fibrillation that occurs secondary to an emergent condition portends a worse prognosis.
In comparison to ED patients with a primary diagnosis of atrial fibrillation, these patients were significantly older, with a similar sex distribution (51% female). After adjustment for a range of covariates, a secondary diagnosis of atrial fibrillation remained associated with increased mortality following an ED visit, compared to patients with a primary diagnosis of atrial fibrillation. Based on these findings, we propose that future clinical prediction instruments for atrial fibrillation in the ED should consider excluding patients with atrial fibrillation who have a clear alternative primary ED diagnosis, which may be cleaner than relying on statistical methods to fully adjust for illness severity. As well, clinical decision instruments for entities such as CHF, pneumonia, and others should consider evaluating the effect of adding a diagnosis of atrial fibrillation to their tool.
While researchers prefer to focus on patients with a single disease entity, often excluding patients with multiple comorbidities from clinical studies, in reality a large proportion of patients have a combination of chronic diseases, diseases that tend to be concurrent, such as atrial fibrillation and CHF, atrial fibrillation and COPD, etc. This patient population is understudied despite a high prevalence, and future studies are needed to examine patients with multiple active comorbidities and the contribution that the combination of diseases in the ED makes to the outcomes.
In our study, over a quarter of eligible patients who arrived in the ED were not on anticoagulation, even after accounting for possible contraindications. Even more surprising was that the majority of patients who were on warfarin were either sub- or supratherapeutic (61%); the latter led to vitamin K administration in a substantial portion of patients. While not a planned focus of our study, this snapshot of anticoagulation, in a real-world population of atrial fibrillation patients who are coming to the ED for other reasons, highlights what appears to be inadequate prescription, compliance, and ongoing monitoring of warfarin anticoagulation in this patient population.
External validity is a limitation of this single-center study, and thus the results may only apply to academic sites and potentially those that serve patients with similar demographics. Identification of these patients was very labor-intensive, as coding of secondary diagnoses is likely to be inaccurate at the level of the ED chart and subsequently within databases. Thus the feasibility of a multicenter study of these patients is very low. Due to resource limitations, we were only able to abstract half of the secondary atrial fibrillation patients, but we chose a random selection of 416 patients, and thus we should have a representative sample. We did not perform chart abstraction with two independent abstractors for all variables, but did do this for key variables in a random sample of 20% of patients, which showed 100% agreement. Thus we believe that our data are valid.
The definition of a secondary diagnosis of atrial fibrillation was defined here by the receipt of an ECG during the ED visit; our results do not apply to patients with atrial fibrillation who did not warrant an ED ECG. In a small proportion of cases the distinction between a “primary” and secondary diagnosis of atrial fibrillation is not clear, but this is a clinical reality in the ED, which is unavoidable; we dealt with this by using the managing physician's main diagnosis and including all persons with ECG proof of atrial fibrillation in the ED. It is possible, however, that the physician might have listed atrial fibrillation first for a patient who presented with a nonspecific diagnosis such as weakness, which may or may not be related to the preexisting atrial fibrillation. Such a patient would be categorized here as primary even if the physician was thinking that the weakness may be unrelated to the atrial fibrillation. Thus, these data are potentially subject to errors in patient classification. Last, given that this was an observational study, our regression results may be subject to bias due to unmeasured confounding. Inclusion of ADG risk scores and CHA2DS2-VASc criteria should limit potential bias, but due to this limitation, and because it was our primary goal to simply describe the characteristics and outcomes of a previously unreported, common ED patient group, we made the regression analysis a secondary analysis.
In this cohort of patients with atrial fibrillation and another primary ED diagnosis, subsequent mortality, both short-term and long-term, was very high. These data suggest that in this previously undescribed but commonly encountered group of ED patients, in the setting of an emergent condition, atrial fibrillation appears to be associated with very poor outcomes. Future studies are needed to evaluate the specific contribution of atrial fibrillation to mortality in the setting of certain primary ED diagnoses.