Atrial fibrillation (AF) may present as irregular, rapid, electrical and mechanical activation of the atria, resulting in asynchronous contraction of the atria that may promote thromboembolism formation [1]. AF is the most common arrhythmia in clinical settings, especially in intensive care unit (ICU) patients [2]. AF is mainly divided into paroxysmal AF, persistent AF, long-term persistent AF, and permanent AF [3]. Preexisting AF is very common among ICU patients, while new-onset AF is also a frequent complication in the ICU with an incidence of approximately 5% [4‐6].

Causes of new-onset AF might include electrolyte disturbances, fluid imbalances, neurohormonal disturbances, arrhythmic drug use, and inflammatory reactions [4]. Chronic heart failure, hypertension, valvular disease, and myocardial infarction trigger a variety of common inflammatory pathways, activation of the renin-angiotensin system and production of reactive oxygen species that lead to atrial fibrosis and further promote the occurrence of AF [7]. In patients who have undergone surgery and had a recent myocardial infarction, AF was often associated with poor outcomes, such as an increased risk of stroke [8‐12]. However, evidence of the effects of new-onset AF, preexisting and non-AF with prognosis in ICU patients is limited and contradictory.

Preexisting and new-onset AF have been shown to be associated with all-cause mortality [13, 14]. However, several studies have shown that preexisting and new-onset AF may not be independently associated with hospital mortality [15, 16]. A cohort study including more than 1300 critically ill patients with persistent arrhythmias found that AF was not associated with increased mortality [17]. Many previous studies had limited sample sizes, and their conclusions were controversial. The aim of this study was to identify the associations of new-onset AF, preexisting AF, and non-AF with hospital and 90-day mortality among ICU. This retrospective observational study was performed in accordance with the STROBE reporting checklist.

The incidence of new-onset AF is high in current critically ill patients and is associated with an increased mortality rate. However, strategies for treatment and management of these patients remain controversial. In this large retrospective study, we provide an important reference for the prognosis and management of new-onset AF. After adjusting for confounding factors in two models using Cox and logistic regression analyses, new-onset AF was found to be associated with significantly higher 90-day and hospital mortality than non-AF and preexisting AF. The prevalence of new-onset AF was approximately 9.4%. Furthermore, after adjusting for confounding factors in model 2 using multivariate Cox analyses, we also demonstrated that age, congestive heart failure, cerebral infarction, pulmonary embolism, sepsis, SOFA score, WBC count, amiodarone use, non-dihydropyridine CCB use, and digoxin use were associated with an increased risk of 90-day mortality in patients with AF, while hypertension, hyperlipidemia, β blockers use, statin use, warfarin use were protective factors for 90-day mortality in patients with AF. Patients with AF in the SICU had an increased risk of 90-day mortality compared with those in the CCU after adjusting for confounding variables.

Several previous studies have indicated that preexisting AF is associated with worse outcomes including higher rates of hospital and long-term mortality than non-AF patients [4, 9, 13]. In a previous retrospective study, preexisting AF was associated with an approximately 4 times increased risk of mortality compared with non-AF [20]. However, the effect of new-onset AF on mortality among critically ill patients remains controversial. Some prior studies did not support a significant correlation between new-onset AF and a high mortality risk [15, 21]. Another prospective cohort study showed that new-onset or preexisting AF was independently associated with increased mortality [14]. In our study, the univariate Kaplan-Meier method showed that the 90-day mortality rate among patients with preexisting AF was higher than that among those with new-onset and non-AF. However, consistent with previous research, after adjustment using two different models, new-onset AF was associated with higher 90-day mortality than preexisting and non-AF. This may be because multiple potential factors affect mortality associated with new-onset AF. Nevertheless, prior studies have failed to compare the mortality risk between patients with new-onset AF and those with preexisting AF [6, 22, 23]. The prevalence of new-onset AF in ICU patients in the current study is comparable to that reported in previous research [4, 6, 22, 24]. The reason why new-onset AF is associated with a poor prognosis is unclear. Infection and inflammation in critically ill patients may cause structural changes in the heart and accelerate electrical conduction [25, 26]. In addition, patients with electrolyte imbalances and those being treated with vasopressin therapy are also more likely to develop AF [27, 28]. An observational study revealed that the incidence of hemodynamic instability in patients with new-onset AF was significantly higher than that in patients with preexisting AF [24]. During AF, coordinated depolarization and contraction of the heart are disturbed by countless and disordered atrial electrical pulses, resulting in unstable contractions and loss of “atrial rhythmic beating”, thereby impairing cardiac output.

Notably, our study also demonstrated that age, congestive heart failure, cerebral infarction, pulmonary embolism, sepsis, SOFA score, WBC count, amiodarone use, non-dihydropyridine CCB use, and digoxin use were risk factors for 90-day mortality in critically ill patients with AF, while hypertension, hyperlipidemia, β blockers, statin and warfarin were protective factors with 90-day mortality (all p < 0.05). The incidence of previous cerebral infarction was significantly higher among patients with new-onset and preexisting AF than among patients with non-AF. Therefore, these patients may have had arrhythmic events prior to admission. However, we have no clear evidence to prove support this hypothesis. We focused on the effect of new-onset AF during hospitalization on 90-day prognosis. The final conclusion may be biased. Development of AF during a critically ill period is related to the presence of more serious disease before the onset of AF and to clinical deterioration after AF; therefore, it is difficult to identify a causal role of AF in affecting patient prognosis. However, comparison of our findings with those of other studies confirms that elderly age, congestive heart failure, cerebral infarction, pulmonary embolism, and sepsis were associated with an obviously increased mortality risk in critically ill patients with AF [27, 29]. After evaluating the data, we found that the SOFA score in more than 90% of patients was less than 10 points. However, the SOFA scores we included were those obtained on admission, and not all patients were in a very critical condition at admission. Condition of patients may undergo a series of changes during their hospital stays. The distribution of data is shown in the Additional file 1: Fig. 1. We believe that these data are still representative of patients with severe illness. The SOFA score was independently associated with 90-day mortality in patients with AF, which was also consistent with the founding of a previous study [30].

At present, there are still many controversies about the treatment strategy for patients with AF. Commonly used rate control drugs include non-dihydropyridine CCBs, β blockers and digoxin. Rhythm control drugs usually include magnesium and amiodarone, both of which have rhythm and rate control properties. In the current study, we found that β blockers, statins and warfarin were significantly associated with a decreased 90-day mortality risk, while amiodarone, non-dihydropyridine CCBs, and digoxin were associated with an increased risk. We evaluated the effect of warfarin on 90-day mortality in all patients with AF, including both those with new-onset and preexisting AF. Warfarin was associated with a decreased 90-day mortality risk in patients with new-onset and preexisting AF compared with those non-AF. β blockers have rate control, negative muscle strength and vasodilatory effects. They exert sympathetic effects by antagonizing β-1 receptors, resulting in decreased conductivity and a reduced effect of catecholamines on the myocardium [31]. In accordance with the present results, a previous study demonstrated that β blockers might be associated with lower hospital mortality in AF patients than amiodarone, non-dihydropyridine CCBs, and digoxin [32]. Digoxin slows the heart rate by increasing vagal nerve tension and may be related to hypotension. There is an association between digoxin and increased mortality, especially in patients with serum digoxin concentrations greater than 1.2ng/ml [33]. In addition, the vagal nerve effects of digoxin may be less effective in severe diseases characterized by a high catecholamine status [34]. Amiodarone inhibits adrenergic stimulation, blocks delayed current, and slows atrioventricular conduction. However, this drug still has many shortcomings, such as hypotension, pro-arrhythmic effects, and pulmonary toxicity. In this study, we found that the most widely used antiarrhythmic drugs in patients with AF were β blockers (75.2%), amiodarone (34.9%), CCBs (21.2%), and digoxin (12.5%). Among severely ill patients with AF in the United Kingdom, the most commonly used drug was amiodarone (> 80%), followed by β blockers (12%) [35]. Interestingly, in our study, statin use was also an independent protective factor for 90-day mortality in AF patients. This finding might be due to the anti-inflammatory effects of statins [36]. Patients with AF in SICU had increased 90-day mortality compared with those in the CCU after adjusting for confounding variables; thus, these patients may require more aggressive management.

In conclusion, after adjusting for confounding factors using Cox and logistic regression analyses, critically ill patients with new-onset AF had a significantly increased risk of hospital and 90-day mortality compared with those with preexisting AF and non-AF. Patients with AF in the CSRU had a decreased risk of 90-day mortality, while patients with AF in the SICU had an increased risk of 90-day mortality, compared with those hospitalized in the CCU. Management of patients with new-onset AF patients in the ICU, especially in the SICU, requires robust measures. β blockers may be used as a first-line treatment for patients with AF in the ICU.