Atrial fibrillation (AF) is the most prevalent cardiac arrhythmia, characterized by rapid and disorganized electrical activity in the atria, resulting in an irregular heartbeat (1, 2). Despite significant progress in its diagnosis and treatment (3), the global prevalence and incidence of AF have markedly increased over the past three decades. In 2021 alone, the global incidence of AF reached 4.48 million (4). In Europe and the United States, approximately 25% of individuals over 55 years of age are projected to develop AF during their lifetime (5). AF poses a substantial public health burden due to its high rates of comorbidity, increased mortality risks, and escalating healthcare costs (6). Early identification of individuals at heightened risk for AF development is critical for effective clinical management (7, 8).

Blood-based biomarkers have shown promise in improving prognostic accuracy in AF (9). However, current tools such as the CHADS₂ and CHA₂DS₂-VASc scores are limited in individualized risk prediction. Traditional biomarkers—including troponin and natriuretic peptides—lack sufficient specificity for AF, as they are elevated in various cardiovascular conditions (10). Moreover, the cost and complexity of certain tests, such as echocardiography, can hinder widespread clinical adoption (11). Despite extensive efforts, a definitive, reliable prognostic biomarker for AF remains elusive (12).

Blood urea nitrogen (BUN) and serum albumin are well-established indicators of renal function, nutritional status, and systemic inflammation (13). Elevated BUN has been linked to poor outcomes in cardiovascular conditions (14), while hypoalbuminemia is a known predictor of mortality in critical illness, including AF (15).

The blood urea nitrogen to serum albumin ratio (BAR), reflecting both renal function and nutritional status, has emerged as a prognostic indicator in various conditions, including heart failure (16), acute myocardial infarction (17), chronic obstructive pulmonary disease (18), and sepsis (19). However, its role in predicting short-term outcomes in AF patients remains unclear. This study aims to explore the association between BAR and the prognosis of patients with AF.

This study demonstrates that the BAR is independently associated with 28-day mortality in critically ill patients with AF. The association remained significant after adjusting for demographic, clinical, and laboratory factors. Compared to BUN or albumin alone, BAR showed superior prognostic performance, supporting its utility as a simple and accessible biomarker for early risk stratification.

In critically ill patients with AF, BUN and albumin levels are important prognostic indicators. Elevated BUN typically indicates renal impairment or fluid retention (20, 21), both of which are closely associated with the onset and persistence of AF. Increased BUN levels lead to fluid accumulation, raising cardiac workload, promoting atrial dilation, and contributing to atrial fibrosis—all of which create a substrate for AF (22, 23). Moreover, elevated BUN is often associated with electrolyte imbalances, such as hyperkalemia, which can disrupt the heart's electrical function, increase the risk of arrhythmias, and prolong AF duration, thus worsening cardiac function and prognosis (20, 24). On the other hand, low albumin levels generally indicate chronic illness, malnutrition, or impaired liver function (25). As a plasma protein, albumin plays a critical role in maintaining colloid osmotic pressure and transporting drugs and nutrients. Low albumin levels suggest nutritional depletion or severe illness, leading to increased vascular permeability, edema, and fluid shifts. Additionally, hypoalbuminemia may indicate an underlying inflammatory state (26), which can accelerate cardiac and organ failure, reducing survival rates.

BAR serves as an effective marker of this complex interplay, providing a comprehensive measure of disease severity. Elevated BAR levels in AF patients often signal multiorgan dysfunction, including renal and hepatic impairment, which is common in critically ill patients (27, 28). Moreover, BAR has been associated with increased thromboembolic risk and atrial remodeling. High BUN levels reflect systemic stress and a prothrombotic state (13), while low albumin correlates with reduced anticoagulant efficacy and increased embolic risk (29). Unlike isolated biomarkers, BAR offers a holistic assessment of the metabolic, hemodynamic, and inflammatory shifts that characterize AF, making it a practical tool for risk stratification and treatment planning, especially in high-risk patients.

Although BAR was independently associated with 28-day mortality, its discriminatory power was moderate, with an AUC of 0.65. However, BAR showed better predictive performance compared with its individual components, with an AUC of 0.63 for BUN and 0.59 for albumin. This suggests that BAR may provide a more integrated assessment of renal function and nutritional status. These findings support its potential role as a composite prognostic biomarker. Nevertheless, due to its limited sensitivity and specificity, BAR may not be sufficient as a standalone risk stratification tool. Further research is needed to assess whether incorporating BAR into established risk prediction models, such as CHA₂DS₂-VASc or HAS-BLED, could improve prognostic accuracy and inform clinical decision-making in patients with atrial fibrillation.

Notably, our findings revealed relatively low rates of anticoagulant use, particularly in patients with high BAR levels (Q4 group), where warfarin was used in only 16.77% of cases and NOACs in just 3.53%. Current guidelines recommend oral anticoagulation as standard therapy for patients with atrial fibrillation, especially those with persistent or permanent AF at elevated thromboembolic risk. However, our dataset lacked detailed information on AF duration or classification, precluding analysis of whether the patients had newly diagnosed or chronic AF. The underuse of guideline-directed medical therapy (GDMT), including anticoagulants, may partially explain the higher mortality observed in the high-BAR group. Inadequate anticoagulation could contribute to increased thromboembolic complications or progression of atrial remodeling, compounding the risk associated with renal dysfunction and hypoalbuminemia. Future studies should incorporate detailed rhythm history and therapeutic adherence data to better delineate these associations.

In our study based on the MIMIC-IV database, the median ICU length of stay among intubated patients was 5.8 days (IQR: 3.1–10.7), which appears relatively short. Several factors may account for this. First, early ICU mortality was common in our cohort and could contribute to shorter ICU stays. Second, MIMIC includes a substantial proportion of patients with treatment-limiting orders such as Do-Not-Resuscitate (DNR) or Do-Not-Intubate (DNI), which may lead to earlier withdrawal of intensive care. Third, hospital-specific discharge practices and ICU resource limitations may have led to early transitions to step-down or general wards. Similar ICU durations have also been reported in previous studies using the MIMIC-III and IV datasets (30, 31). Importantly, the relatively short ICU stay may have implications for our results. It could lead to underestimation of disease progression or complications that typically manifest later during prolonged ICU admissions. Moreover, the early timing of data collection may favor early predictors of mortality, while longer-term prognostic factors may be less represented. Therefore, while our findings remain valid within the context of early ICU outcomes, caution is warranted when generalizing them to patients with prolonged ICU courses.

This study has several strengths, including the use of a large, well-curated critical care database and rigorous statistical adjustment. However, limitations should be acknowledged. First, the retrospective design inherently introduces potential selection bias and limits the ability to draw causal inferences. Second, this study was based on data from the MIMIC-IV database, which represents a single-center ICU cohort in the United States. As such, the generalizability of our findings to other healthcare settings, geographic regions, or patient populations may be limited. Third, BAR may be affected by unmeasured confounders such as hydration status, liver function, and chronic malnutrition, which were not fully accounted for due to data limitations. Fourth, echocardiographic data were not available in the MIMIC-IV database, which limited our ability to assess cardiac structure and function, including left ventricular ejection fraction and atrial size. Lastly, our study did not compare BAR with established AF-specific risk scores (e.g., CHA₂DS₂-VASc, ATRIA, or HAS-BLED), which limits our ability to evaluate the incremental prognostic value of BAR beyond current tools. In addition, while MIMIC-IV is a valuable resource, analyses based on such databases require careful attention to case definitions, data quality, and residual confounding. For example, evaluating the effect of anticoagulation is challenging, as treated patients may differ systematically in comorbidity burden or disease severity. Without careful statistical adjustment, treatment effects may be confounded. Moreover, data on anticoagulation intensity and adherence are limited, and MIMIC-IV does not capture post-discharge outcomes. This constrains assessment of long-term endpoints such as stroke, bleeding, or readmission. External validation in prospective cohorts is needed to confirm these findings.

In conclusion, BAR is an independent and easily accessible predictor of short-term mortality in critically ill patients with AF. Its integration into routine clinical assessment may enhance early risk stratification and inform personalized treatment strategies in this vulnerable population.