key: cord-0993341-zttcvqul
authors: Musikantow, Daniel R.; Turagam, Mohit K.; Sartori, Samantha; Chu, Edward; Kawamura, Iwanari; Shivamurthy, Poojita; Bokhari, Mahmoud; Oates, Connor; Zhang, Chi; Pumill, Christopher; Malick, Waqas; Hashemi, Helen; Ruiz-Maya, Tania; Hadley, Michael B.; Gandhi, Jonathan; Sperling, Dylan; Whang, William; Koruth, Jacob S.; Langan, Marie-Noelle; Sofi, Aamir; Gomes, Anthony; Harcum, Stephanie; Cammack, Sam; Ellsworth, Betsy; Dukkipati, Srinivas R.; Goldman, Martin E.; Halperin, Jonathan L.; Fuster, Valentin; Reddy, Vivek Y.
title: Atrial Fibrillation in Patients Hospitalized with COVID-19: Incidence, Predictors, Outcomes and Comparison to Influenza
date: 2021-02-24
journal: JACC Clin Electrophysiol
DOI: 10.1016/j.jacep.2021.02.009
sha: 5aec8b12c83b31703a6038294f1c4ee0cfd087e5
doc_id: 993341
cord_uid: zttcvqul

Background Coronavirus Disease 2019 (COVID-19) results in increased inflammatory markers previously associated with atrial arrhythmias. However, little is known about their incidence or specificity in COVID-19, or their association with outcomes. We determined the incidence, predictors and outcomes of atrial fibrillation or flutter (AF/AFL) in patients hospitalized with COVID-19, or hospitalized with Influenza. Methods This is a retrospective analysis of 3,970 patients admitted with PCR-positive COVID-19 between 2/4/2020-4/22/2020 with manual review performed of 1,110. The comparator arm included 1,420 patients with influenza hospitalized between 1/1/2017-1/1/2020. Results Among 3970 inpatients with COVID-19, the incidence of AF/AFL was 10% (N=375) and in patients without a history of atrial arrhythmias, 4% (N=146). Patients with new-onset AF/AFL were older with increased inflammatory markers including Interleukin-6 (93 vs 68 pg/ml, P<0.01), and more myocardial injury (Troponin-I: 0.2 vs 0.06ng/ml, P<0.01). AF/AFL were associated with increased mortality (46% vs 26%, P<0.01). Manual review captured a somewhat higher incidence of AF/AFL (13%, N=140). Compared to inpatients with COVID-19, patients with Influenza (N=1420) had similar rates of AF/AFL (12%, n=163) but lower mortality. The presence of AF/AFL correlated with similarly increased mortality in both COVID-19 (RR 1.77) and Influenza (RR 1.78). Conclusions AF/AFL occurs in a subset of patients hospitalized with either COVID-19 or Influenza, and is associated with inflammation and disease severity in both infections. The incidence and associated increase in mortality in both cohorts suggests that AF/AFL in not specific to COVID-19, but is rather a generalized response to the systemic inflammation of severe viral illnesses.

Vivek Y. Reddy: No disclosures relevant to . Disclosures unrelated to this manuscript are viewable in the Supplement.

J o u r n a l P r e -p r o o f Introduction As of September 10th, 2020, there have been 28 million patients with COVID-19 infections worldwide and over 900,000 deaths 1 The pathophysiology of the SARS-CoV-2 viral infection appears driven by an inflammatory immune response with several markers of inflammation, such as C-reactive protein and the cytokine interleukin-6, correlating with disease severity and mortality. 2, 3 Even before the COVID-19 pandemic, atrial fibrillation and atrial flutter (AF/AFL) had been linked to conditions characterized by elevated inflammatory markers. 4, 5 Hence, it is not surprising that a high incidence of AF/AFL has been reported with COVID-19. [6] [7] [8] However, as available studies have been limited in scope and specificity, the true incidence of AF/AFL in this population is unknown. Also uncertain is whether the inflammatory milieu of COVID-19 is uniquely responsible for AF/AFL, or whether these arrhythmias reflect part of a nonspecific byproduct of severe viral respiratory illness.

Beyond inflammation, COVID-19 has been associated with both an elevated incidence of myocardial injury, and an increased risk of thrombotic events such as venous thromboembolism and ischemic stroke [9] [10] [11] [12] . Accordingly, it is possible both that AF/AFL may correlate with cardiac injury, and, in the context of a prothrombotic state, contribute to the increased risk of thromboembolic events such as ischemic stroke.

We performed a retrospective analysis of a large cohort of hospitalized patients afflicted with COVID-19 (n=3,970) to assess the incidence, predictors and outcomes of AF/AFL. To address the unusual clinical environment occurring during the New York City COVID-19 pandemic, in a subset of this cohort, we also performed a manual chart review of primary patient data including electrocardiograms (EKGs) and telemetry to assess for under-representation of J o u r n a l P r e -p r o o f arrhythmias in clinical coding. Finally, we compared these observations to patients hospitalized with Influenza to assess whether these atrial arrhythmias uniquely result from COVID-19, or whether they reflect a response to acute respiratory illness.

This multi-center retrospective cohort study included consecutive adult patients (≥18 years) with laboratory-confirmed COVID-19 infection, admitted to five hospitals within the Mount Sinai Health System. We studied 3 patient cohorts -of which two overlapped. i) The principal automated electronic record abstraction cohort (COVID-19 Primary ) included all patients with laboratory-confirmed COVID-19 admitted to the hospitals between February 4 th and April 22 th , 2020. ii) The manually-adjudicated patient cohort (COVID-19 Manual ) was drawn from the same population of patients, but only included consecutive patients admitted until March 28 th , 2020, and excluded patients who tested positive for COVID-19 more than one week into hospitalization. This exclusion as done as the manual cohort included patients diagnosed at onset of the pandemic including several who had prolonged hospitalizations with unrelated conditions and contracted COVID-19 while inpatient. iii) The automated electronic record abstraction influenza cohort (Influenza Primary ) included all patients with PCR-positive Influenza A or B from January 1 st , 2017 until January 1 st , 2020; there was no temporal overlap with the COVID-19 population. All patient data were de-identified prior to analysis, and data abstraction was approved by the Mount Sinai Institutional Review Board.

J o u r n a l P r e -p r o o f Data Collection Data were abstracted from the Electronic Health Records (EHR) including baseline demographics, laboratory measurements, inpatient medications, and outcomes. Using International Classification of Disease (ICD) 9/10 billing codes, comorbidities were identified; these included congestive heart failure (CHF), hypertension, diabetes, prior stroke/transient ischemic attack (TIA), chronic kidney and liver disease, human immunodeficiency virus (HIV), chronic obstructive pulmonary disease (COPD), asthma and obstructive sleep apnea (OSA). An analysis was then performed using ICD 9/10 codes for the occurrence of in-hospital ischemic stroke or TIA, and AF/AFL.

The COVID-19 Manual cohort included laboratory data, baseline demographics and hospital medications abstracted from the EHR and then manually reviewed. Baseline comorbidities, prehospital medications and in-hospital events, including neurological events, were obtained from available clinical records. All available EKGs were independently reviewed by a cardiologist or electrophysiologist and chart documentation was assessed for atrial arrhythmias.

Continuous variables were summarized as median and interquartile range (IQR) or means and standard deviations, as appropriate. Categorical variables were summarized as counts or percentages. No imputation was made for missing data. Median/Mann-Whitney U test, Fisher's exact test or χ² test was used to compare data where appropriate. A p-value ≤0.05 (2-tailed) was considered statistically significant. In the comparison of patients with influenza vs. COVID with new AF/AFL (Table S4) , we included only variables which were available in at least 75% of patients in both groups. We then plotted Kaplan-Meier curves for in-hospital mortality stratified by the presence of in-hospital AF/AFL. As our follow-up only included the duration of the J o u r n a l P r e -p r o o f hospitalization, patients discharged from the hospital were considered to have survived for the purposes of these curves. Separate Kaplan-Meier curves were also created in which discharged patients were censored (Supplementary Figure S1 

In the COVID-19 Primary cohort, 3970 patients admitted with PCR-confirmed COVID-19

were identified and incorporated into the analysis. The overall incidence of AF/AFL occurring during hospitalization was 10% (n=375 patients). As demonstrated in Table 1 , patients with AF/AFL were older (median 77 vs 65 years, p<0.01), with more baseline comorbidities, including hypertension (56% vs 32%, p<0.01), diabetes (33% vs 24%, p<0.01) and HF (25% vs 5%, p<0.01). Most patients with inpatient AF/AFL (61%) had a history of atrial arrhythmias, and of those with a history of atrial arrhythmias, 71% manifested AF/AFL during hospitalization.

The overall incidence of AF/AFL in patients without a history of atrial arrhythmias (new-onset AF/AFL) was 4% (n=146 patients).

Aside from age (median 74 vs 66 years, p<0.01) and race, patients with newly detected AF/AFL did not different significantly in terms of baseline characteristics from those who did not develop AF/AFL. However, there were differences in certain laboratory values, including an increase in peak inflammatory markers: C-reactive protein (median 232 mg/dl vs 175 mg/dl, p p<0.01) and Interleukin-6 (median 93.5 mg/dl vs 67.8 mg/dl, p<0.01). There were also increases in other previously described markers of disease severity, including peak troponin (median 0.2ng/ml vs 0.07 ng/ml, p<0.01), peak D-dimer (median 3.7 ug/ml vs 2.3 ug/ml, p<0.01) and B-type-natriuretic peptide (BNP; median 125 pg/ml vs 56 pg/ml, p<0.01)

A multivariate logistic regression model was constructed including individual laboratory values and in-hospital treatment along with comorbidities found to be predictive of developing new AF (Table 2) . No admission laboratory value demonstrated significant predictive value in this analysis; however, in-hospital markers of peak inflammation including C-reactive protein and platelet nadir, along with evidence of myocardial injury (Troponin ≥ 0.03 ng/mL) maintained predictive value. Use of steroids as well as mechanical ventilation were also associated with a higher incidence of new AF in this analysis.

To perform a manual review of the COVID-19 Primary dataset, a consecutive subset of patients, the COVID-19 Manual cohort (n=1,110 patients) was screened for both baseline and outcome characteristics, including atrial arrhythmias. There were substantially higher rates of certain common comorbidities identified in the COVID-19 Manual group compared to the corresponding COVID-19 Primary group such as hypertension (63% vs 35%, p<0.01) and diabetes (38% vs 24%, p<0.01; Supplementary Table S1 ). Of those with a pre-existing history of atrial arrhythmias, 43% were considered paroxysmal, 25% persistent and the remaining could not be determined. Most importantly, including both EKG-confirmed and reported AF/AFL, the overall incidence was higher than the 10% captured in the COVID-19 Primary analysis -the AF/AFL in COVID-19 Manual was instead 12.6% (n=140 patients) with 6.6% of patients demonstrating new AF/AFL. Similar to the larger COVID-19 Primary cohort, AF/AFL in this COVID-19 Manual cohort was associated with increases in baseline comorbidities such as heart failure (HF), hypertension, and age (Supplementary Table S2 ). In this COVID-19 Manual cohort, the pre-admission medications of the AF/AFL patients included more frequent use of anticoagulant, lipid-lowering, and antihypertensive medications, but no significant difference in use of ACE-inhibitors (20% vs 14%, p=0.07), or ARBs (16% vs 17%, p=0.81).

In the COVID-19 Primary cohort, patients with AF/AFL were slightly less often treated with hydroxychloroquine (68% vs 76%, p=.03); Interleukin-6 inhibitors was similar compared to those without AF/AFL (Table 3) . Corticosteroid use differed significantly between groups (40% vs 28%, p<0.01), and this association was stronger in patients with new-onset AF/AFL (47% vs 28%, p<0.01). In-hospital treatment of AF/AFL frequently included therapeutic anticoagulation with either parenteral heparin or oral anticoagulants (78%). While there was not a significant difference in peak hospitalization heart rates, AF patients frequently received antiarrhythmic drugs (25%), predominately amiodarone (86 of 95 patients, 91%).

Overall, the presence of AF/AFL was associated with worse outcomes, including higher rates of intubation (27% vs 15%, RR 1.8 p<0.01), ischemic stroke (1.6% vs 0.6%, RR 2.7 p=0.05), and mortality (46% vs 26%, RR 1.78 p<0.01).

To understand the specificity of observed atrial arrhythmias in COVID-19, we studied a cohort of 1,420 Influenza patients (Influenza Primary group). Comorbidities occurred more frequently in patients hospitalized with influenza than COVID-19 ( Table 4 ). The incidence of in-hospital AF/AFL was higher in the Influenza Primary than the COVID-19 Primary cohort (12% vs 10%, p=0.03), but the incidence of new-onset AF/AFL was similar (4% vs 4%, p=0.93). Not surprisingly, despite more frequent comorbidities, the Influenza Primary cohort had a substantially lower incidence of in-hospital mortality (9% vs 29%, p<0.01).

Like the COVID-19 patients, influenza patients with in-hospital AF/AFL were older and had higher rates of comorbidities including HF and hypertension (Supplementary Table S3 ). The levels of inflammatory markers were not significantly different in AF/AFL patients in the Influenza Primary cohort; however, they had increased markers of cardiac injury (median 0.08 ng/ml vs 0.05 ng/ml, p<0.01). Use of corticosteroids was similar in those with in-hospital AF/AFL (39% vs 41%, p=0.73). As in COVID-19 patients, in-hospital AF/AFL in the Influenza Primary cohort was associated with more frequent intubation (14% vs 7%, p=0.004) and death (16% vs 10%, p=0.003).

We performed sensitivity analyses to address the potential impact of differences in acuity of illness and baseline characteristics in these two groups (Supplementary Table S5 ). After correcting for differences in age, sex, race and various co-morbidities, these analyses revealed similar rates of atrial arrhythmias in both the COVID-19 Primary and Influenza Primary cohorts for 

In this study, we examined the incidence, predictors and outcomes of patients This is the first manuscript involving a large cohort of patients to address the incidence of in-hospital AF in COVID-19 patients. A previous analysis involving a manual review of 115 inpatients for various cardiac arrhythmias reported an AF incidence of 16.5%, all of which occurred in intensive care units. 13 In a nationwide study in Denmark, there was a 47% decrease in the incidence of new-onset AF compared with the corresponding weeks in 2019. 14 While this most likely resulted from underreporting and lower healthcare utilization, it suggests that COVID-19 patients overall do not develop atrial arrhythmias at a greater frequency than other acutely ill hospitalized patients.

While the incidence of AF/AFL in patients with COVID-19 is not exceedingly high, the occurrence of in-hospital AF/AFL appears impactful to a patient's clinical course, as indicated by the frequent use of antiarrhythmic therapy. Despite the cumulative risk potentially anticipated by combining the prothrombotic state of COVID-19 with the stasis of blood flow during AF/AFL, there was only a modest (1%) absolute increase in ischemic stroke in patients with atrial arrhythmias, perhaps due to concurrent use of therapeutic anticoagulation (76%) during hospitalization. Not unexpectedly, as AFL/AFL was associated with comorbidities, markers of inflammation and disease severity, mortality was exceptionally high in this group (46%).

Previous studies have noted new-onset AF in patients with Influenza. 15 This had been attributed to higher rates of pro-inflammatory cytokines including IL-6, which is not specific to either Influenza or COVID-19 infection. 16 The observation that AF occurs in COVID-19 at a frequency similar to that in Influenza argues against a unique effect of either virus on atrial rhythm. This similarity may be confounded by a higher rate of baseline comorbidities in the Influenza group, more severe systemic illness in COVID-19, and longer duration of hospitalization during the COVID-19 pandemic. Importantly, as many patients with COVID-19

and Influenza are managed as outpatients, our study only reflects those patients whose severity of illness warranted hospitalization.

The height of the COVID-19 pandemic imposed a unique stress on the medical system in New York hospitals as providers were called to perform unfamiliar roles. As a result, we felt it important to ascertain whether the automated data abstraction strategy applied to such a large dataset (using ICD 9/10 codes) accurately reflected the incidence of atrial arrhythmias. Manual chart review of a large subset of this population found that 23% of patients with AF/AFL were not identified by the automated analysis -accordingly, the true rate of AF/AFL increased from 10% to 13%. Furthermore, manual review identified substantially higher rates of common comorbidities that were not discovered by automated indexing of ICD codes. On the other hand, there were minimal differences in laboratory values, treatments or outcomes between the two methods of data abstraction. This highlights the limitations and strengths of "big data" studies which have become commonplace during the COVID-19 pandemic.

Limitations. This study was limited by several intrinsic challenges of the COVID-19 pandemic including limited access to telemetry monitoring in the non-ICU setting, a high incidence of sedated and non-communicative patients, and the potential for under-detection of J o u r n a l P r e -p r o o f ischemic stroke due to the difficulty of performing brain imaging tests in infected patients. From a methodological perspective, the comparison between the COVID-19 and Influenza patients was performed with a similar automated strategy; however, it remains possible that unlike during the "normal" Influenza season, the throes of a pandemic may have resulted in a differentially lower rate of detection of AF/AFL (or potential underreporting using ICD 9/10 codes) in the COVID-19 Primary cohort. However, given the rapid ventricular response common with AF/AFL in the hospitalized COVID-19 pneumonia patient, it seems unlikely that there were many instances of clinically-undetected AF/AFL. Additionally, the exact onset of AF/AFL cannot be accurately determined in part because of limitations in the available data and in part because of the potential delays in diagnosis of atrial arrhythmias during the COVID-19 pandemic. As such, it is difficult to discern the temporal relationship between the factors associated with the development of AF/AFL and their occurrence during the hospital course.

It also bears mentioning that these data only pertain to hospitalized patients: it is possible that non-hospitalized COVID-19 patients have different predictors of developing AF/AFL and different outcomes. Also, we cannot rule-out the possibly that the decision-making for hospitalizing a COVID-19 patient during a pandemic may differ than for an Influenza patient.

However, the directionality of any such variance is unclear -perhaps lower threshold for hospitalization with COVID-19 because of the apprehension surrounding a pandemic, or perhaps a higher threshold related to scarce resources or apprehension related to hospitalization. On the other hand, since AF/AFL was more likely to occur in the most critically-ill patients regardless of the viral etiology, it is likely that the patients most likely to develop AF/AFL were hospitalized. Finally, since our follow-up only extended to hospital discharge, the impact of atrial arrhythmias on the patient's clinical course post-hospitalization was not examined in this analysis.

In this study, we found that AF/AFL occurred in ~13% of hospitalized patients with COVID-19. However, new-onset AF/AFL occurred in only a small minority (4%), a rate that was similar to that observed in hospitalized Influenza patients. In both cohorts, new-onset AF/AFL correlated best with higher degrees of inflammation and disease severity, independent of patient baseline characteristics. These data suggest that these atrial arrhythmias are less likely specific to the COVID-19 viral infection but are rather a generalized response to the systemic inflammatory response of severe viral illness. New-onset atrial fibrillation: incidence, characteristics, and related events following a national COVID-19 lockdown of 5.6 million people. Eur Heart J. 2020.

Chang 

Prognostic value of interleukin-6, C-reactive protein, and procalcitonin in patients with COVID-19

Predictors of mortality in hospitalized COVID-19 patients: A systematic review and meta-analysis

C-reactive protein elevation in patients with atrial arrhythmias: inflammatory mechanisms and persistence of atrial fibrillation

Association of inflammatory factors with occurrence and recurrence of atrial fibrillation: a meta-analysis

Risk Factors Associated with Clinical Outcomes in 323 COVID-19 Hospitalized Patients in Wuhan, China

Clinical Characteristics of 138 Hospitalized Patients With 2019 Novel Coronavirus-Infected Pneumonia in Wuhan

COVID-19 and Cardiac Arrhythmias

Prevalence and Impact of Myocardial Injury in Patients Hospitalized with COVID-19 Infection

Association of Cardiac Injury With Mortality in Hospitalized Patients With COVID-19 in Wuhan, China

Incidence of thrombotic complications in critically ill ICU patients with COVID-19

D-dimer levels on admission to predict in-hospital mortality in patients with Covid-19

Arrhythmias in COVID-19 Patients

Brain natriuretic peptide, pg/mL 69.3 (27.2-214