This is a subanalysis of the Health Outcome Predictive Evaluation (HOPE) COVID-19 registry, with an overall study sample of 6217 patients with a definitive diagnosis or high suspicion of SARS-CoV-2 infection.11 In brief, the HOPE registry is a retrospective cohort registry (ie, a “real-world” all comers type, with voluntary participation and with no financial compensation. All patients discharged (deceased or alive) after hospital admissions for COVID-19 were suitable for the study. There were no exclusion criteria, except for patients’ explicit refusal to participate. From March 23, 2020 to June 1, 2020 all patients fulfilling the inclusion criteria from 24 centers in Spain were assessed in the present study. Clinical and demographic data were collected at inclusion and during hospitalization. The study was performed according to the ethical principles of the Declaration of Helsinki and Good Clinical Practice Guidelines and was approved by the local Ethics Research Committee of the Hospital Clínico San Carlos (Madrid, Spain) (20/241-E). Written informed consent was waived because of the characteristics of the anonymized registry and the severity of the situation. However, at least verbal authorization from the patient (or familiar or caregiver when unavailable) was required. A list of participating hospitals, investigators, collaborators and the protocol are available in the appendix of the supplementary data.

Enrolled patients were divided into 2 groups according to AF history. Study definitions are available elsewhere on the registry's website.12 The primary endpoint was all-cause mortality at 60-days in either cohort. We evaluated whether the use of the CHA2DS2-VASc risk score was useful to assess the risk of death or embolism in patients with COVID-19. The CHA2DS2-VASc score was not recorded in the original dataset, but was obtained retrospectively for this research study. Enrolled patients were divided into 3 groups according to their CHA2DS2-VASc score (group 1: CHA2DS2-VASc=0 in men and ≤ 1 in women; group 2: CHA2DS2-VASc=1 in men or 2 in women; and group 3: CHA2DS2-VASc> 1 in men or> 2 in women). Other clinically relevant events were recorded as secondary endpoints: invasive mechanical ventilation, noninvasive mechanical ventilation, respiratory insufficiency, heart failure, renal failure, sepsis, systemic inflammatory response syndrome, clinically relevant bleeding, and embolic events. Events were classified following local researchers’ criteria according to the definitions of the HOPE COVID-19 registry. The vital status at 60 days of the patients discharged alive was confirmed by telephone interview.

Data are presented as mean±standard deviation for continuous variables with a normal distribution, median (interquartile range [IQR]) for continuous variables with a nonnormal distribution, and as frequency (%) for categorical variables. The student t test and the Mann-Whitney U test were used to compare continuous variables with normal and nonnormal distributions, when needed. The chi-square test or Fisher exact test was used to compare categorical variables. Univariate analysis was performed for qualitative variables and reported as odds ratios with 95% confidence interval (95%CI). Given the multiplicity of variables, only factors associated with all-cause mortality with P <.01 on univariate analysis (age, sex, hypertension, dyslipidemia, diabetes mellitus, smoke, chronic kidney failure, ischemic heart disease, heart failure, lung disease, cerebrovascular disease, cancer, renin-angiotensin-aldosterone system inhibitor treatment, AF, aspirin treatment, anticoagulation treatment, saturation O2 <92% on admission, D-dimer elevation, C-reactive protein elevation, lactate dehydrogenase elevation) were entered into the Cox multivariate regression analysis to define independent risk factors for the main outcome in the matched population. The assumption of proportionality of risks was verified by analyzing Schoenfeld residuals. The C-index and Gronesby Borgan test were calculated to determine discrimination and calibration, respectively. Kaplan-Meier curves were used to estimate survival function and compare subgroups with the log-rank test. Possible collinearity and interactions were evaluated with the introduction of multiplicative terms calculating the tolerance and the variance inflation factor. Propensity score matching (PSM) was estimated with AF as the dependent variable and the main clinical profiles at admission (PSM 1:1; 0.01 tolerance, without replacement, nearest neighbor) to obtain balanced pairs. The MatchIt package (Ho, Imai, King, & Stuart, 2007) was used. Variables included in the PSM were age, dyslipidemia, smoking, cerebrovascular disease, lung disease and were exactly added (sex, hypertension, diabetes, chronic kidney failure, and any type of cancer). Quality adjustment generated by the PSM model is shown in the figure 1 of the supplementary data. The area under the receiver operating characteristic curve (ROC curve) was used to measure how well the models discriminated the CHA2DS2-VASc score for 60-day all-cause mortality and risk of in-hospital embolic event. All tests were 2-sided. The statistical analysis was performed with the IBM SPSS 24.0 software package, STATA software, version 15 and R Core Team (2020).

The percentage of patients testing positive patients for SARS-CoV-2 infection by nasopharyngeal PCR was 89%. The baseline characteristics of COVID-19 patients are shown in table 1. Mean age was 66±17 years, 57.6% of patient were male and the median interval from disease onset to admission was 6 [IQR 5] days. Of the total reported patients, 3020 (50.3%) had hypertension, 2237 (37.5%) dyslipidemia, 1237 (19.9%) diabetes mellitus, 1243 (30.5%) previous pulmonary disease, 265 (4.2%) heart failure, and 416 (7.1%) chronic kidney failure.

Patients were categorized in 2 groups according to history of AF. When we compared these groups, we observed that patients with AF were older and had a greater number of comorbidities. Furthermore, this group had more frequently received prior treatment with antiplatelets, anticoagulants, and renin-angiotensin-aldosterone system inhibitors. These differences were controlled after statistical matching (table 1).

Management is depicted in table 2. The specific COVID-19 drugs most frequently used were hydroxychloroquine (86.1%), followed by antibiotics (77.9%) and lopinavir/ritonavir (54.4%). Corticoids were prescribed in approximately 31% of the patients. For respiratory support, prone positioning was used in 7%, and noninvasive mechanical ventilation in 10%. An invasive mechanical ventilation approach was required in more than 6%. When we compared these groups according to AF, antiviral drugs and tocilizumab were more frequently used in non-AF patients; in contrast, corticoids and antibiotics were more frequently used in AF patients. After PSM these differences were attenuated (table 2).

In-hospital events are shown in figure 1. The most common was bilateral pneumonia (75%) with concomitant respiratory insufficiency in 52.3% of all patients. In the overall COVID-19 population, acute renal injury, sepsis and systemic inflammatory response syndrome were reported in roughly 20% of the patients. Systemic inflammatory response syndrome, heart failure and respiratory failure were more frequent in the AF group. In parallel, we observed a higher incidence of hemorrhagic complications in the AF group (9.8% vs 2.6%; OR, 4.03; 95%CI, 2.56-6.33), but with comparable outcomes in embolic events between the 2 groups. Despite the statistical matching, the AF group continued to show a higher incidence of all these complications (table 2).

Figure 1.

Adverse events during hospitalization in patients with COVID-19 and comparative analysis according to atrial fibrillation. 95%CI, 95% confidence interval; AF, atrial fibrillation; OR, odds ratio; PSM, propensity score matching; SIRS, systemic inflammatory response syndrome.

(0.2MB).

During hospitalization, close to 80% of all patients received some type of anticoagulation therapy, with 62.7% of them receiving a prophylactic dose, while 17.1% received the full anticoagulant dose. In particular, in the AF group, only 135 (57%) patients received anticoagulation at an appropriate dose [102 (75.6%) of them using intravenous/subcutaneous anticoagulation with heparin/enoxaparin, 18 (13.3%) with acenocumarol and 15 (11.1%) with a direct-acting anticoagulant], 61 (25.7%) received a prophylactic dose, and 41 (17.3%) did not receive anticoagulant treatment. We did not observe differences in age (79.4 vs 81.5; P=.248) or in the CHA2DS2-VASc score (3.8 vs 3.8, P=.277) between patients who had received some dose of anticoagulant treatment and those who had not. Despite this low percentage of patients treated with appropriate doses of anticoagulant treatment, the incidence of relevant bleeding complications during admission was higher in the AF group (9.8% vs 2.6%; OR, 4.03; 95%CI, 2.56-6.33). When we compared the entire cohort of patients, we observed that those who received full anticoagulant doses had a higher risk of bleeding than those who received only a prophylactic dose or did not receive any (OR full dose vs prophylactic dose 4.17; 95%CI, 2.90-6.00; and OR full dose vs any dose 3.32; 95%CI, 2.05-5.35). However, these differences were not observed when we analyzed only the group of patients with AF (OR full dose vs prophylactic dose 1.42, 95%CI, 0.49-4.11; and OR full dose vs any dose 1.57, 95%CI, 0.43-5.72). When we analyzed embolic events, we observed no differences between the groups based on the type of anticoagulant (2 [6.1%] in nonvitamin K antagonist oral anticoagulant group vs 4 [6.6%] in vitamin K antagonist group vs 33 [5.1%] in unfractionated heparin group vs 115 [3.9%] in low-molecular-weight-heparin group; P=.298) and dose received (40 [4.7%] in the full-dose anticoagulation group vs 99 [3.2%] prophylaxis group vs 15 [1.5%] group without anticoagulation; P=.231).

Univariable analysis of 60-day all-cause mortality from COVID-19 showed a linear relation between AF development and mortality (mortality in patients with AF 43.6% vs mortality in patients without AF 18%; OR, 3.51; 95%CI, 2.71-4.55). In the multivariate analysis (C-Index and 95% Jackknife CI, 0.750 (0.71-0.788), Groennesby and Borgan test P=.782) (table 3), the presence of AF was independently associated with 60-day all-cause mortality in these patients (HR, 1.234; 95%CI, 1.003-1.519) together with other variables such as age, lactate dehydrogenase elevated on admission, saturation on admission <92%, and chronic kidney disease. In addition, we observed high mortality in patients who were on anticoagulant treatment before admission (41.9% vs 16.9%; OR 3.55; 95%CI, 2.99-4.28), specifically, 43.9% within the AF group. The 60-day Kaplan-Meier survival analysis after PSM confirmed the higher mortality among AF patients (figure 2).

Figure 2.

Kaplan-Meier survival landmark analysis according to atrial fibrillation history. A: before propensity score matching. B: after propensity score matching.

(0.19MB).

A total of 1185 patients died during the 60-day follow-up. The main causes of mortality in our registry were respiratory failure (59.2%), combined cause (19.9%), infectious etiology (6.2%), and systemic inflammatory response (5.0%). Cardiovascular cause was the main cause of death in 1.4% of the patients. In addition, a total of 154 embolic events were recorded during hospital admission, with no differences between the groups (table 2).

Kaplan-Meier survival landmark analysis according to the CHA2DS2-VASc score is shown in figure 3. CHA2DS2-VASc score had a modest ability to predict 60-day all-cause mortality in the entire cohort (area ROC, 0.748; 95%CI, 0.733-0.764); however, it had a poor performance when the group of patients with AF was specifically evaluated (area ROC, 0.618; 95%CI, 0.546-0.689). Furthermore, CHA2DS2-VASc score was also unable to predict the incidence of embolism during admission in the overall cohort (area ROC, 0.519; 95%CI, 0.471-0.568) and AF group (area ROC, 0.411; 95%CI, 0.147-0.675).

Figure 3.

Kaplan-Meier survival landmark analysis according to the CHA2DS2-VASc score. Group 1: CHA2DS2-VASc=0 in men and ≤ 1 in women; Group 2: CHA2DS2-VASc=1 in men or 2 in women; and Group 3: CHA2DS2-VASc> 1 in men or> 2 in women. A: entire cohort. B: patients without AF. C: patients with AF.

(0.2MB).

The design of this study entails some constraints. Some incident events in the participating centers may not have been diagnosed and/or reported. The calculation of the incidence of the events is not precise since recruitment was performed in participating centers without other sampling procedures other than the broad inclusion criteria (hospital discharge). Regarding the management applied, at all times it was decided by the attending medical team, as well as in the comparison group.

Other considerations to take into account are that we did not have information on the type of AF (paroxysmal/permanent) so we cannot know if this classification influenced the prognosis or management of these patients; although we performed a PSM, some variables such as heart failure and ischemic heart disease were not balanced; although we attempted to adjust for many confounders, other unmeasured or unknown confounders might have played a role; we tried to report all the treatments used during admission, but the protocols differed between the centers, which may influence the results. In addition, we were not able to record possible in-hospital AF events, because the health situation experienced in some hospitals included in the registry did not allow electrocardiographic monitoring or daily electrocardiograms to be performed in many patients. The variable “previous anticoagulation” was not included in the matching process; therefore, we cannot rule out that this variable could have modified our results. Despite this, we did include this variable in the multivariate analysis and it was not predictive of a worse prognosis. Furthermore, our registry only included in-hospital complications (except for mortality) and, therefore, we cannot exclude the possibility that some of the patients may have had an embolic or hemorrhagic event at discharge. The precise impact of AF on COVID-19 warrant further investigation.