Four distinct clinical subtypes of new-onset atrial fibrillation were identified in critically ill patients, demonstrating a stepwise increase in 28-day mortality risk (HR 4.24-5.98; P<0.001).
Cohort (n=8,472)
Does unsupervised clustering identify distinct clinical subtypes with different prognoses in critically ill patients with new-onset atrial fibrillation?
Unsupervised clustering identified four distinct clinical subtypes of new-onset atrial fibrillation in ICU patients with varying 28-day mortality rates, enabling precision risk assessment.
Effect estimate: HR 4.24-5.98
p-value: p=< 0.001
Background New-onset atrial fibrillation (NOAF) is a common cardiovascular complication in critically ill patients and is consistently associated with adverse outcomes. However, substantial heterogeneity exists in its clinical presentation and prognosis. This study aimed to identify distinct clinical subtypes of NOAF and evaluate their prognostic and management-related implications. Methods Adult NOAF patients were extracted from the MIMIC-IV database. Demographic and laboratory data within 24 hours of ICU admission were analyzed. Consensus k-means clustering was used for identifying subtypes. Survival differences were compared using Kaplan-Meier and log-rank tests, and multivariable Cox models assessed mortality risk and pharmacologic treatment associations. Key variables identified by SHAP analysis were incorporated into a simplified six-variable model, validated externally in MIMIC-III. Results Among 8472 NOAF patients, four distinct subtypes were identified from the MIMIC-IV cohort (n=5554), showing progressively increased severity and mortality. Subtype A (30.28%) included mainly post-cardiac surgery patients with preserved homeostasis and the lowest 28-day mortality (4.9%). Subtype B (34.52%) was characterized by marked hypomagnesemia and a moderate burden of comorbid malignancy (28-day mortality 15.5%). Subtype C (19.70%) featured anemia, hypoxemia, and inflammation (28-day mortality 30.2%). Subtype D (15.50%) presented with organ failure and the highest 28-day mortality (42.7%). 28-day mortality risk increased stepwise across subtypes (HR 4.24-5.98; all P < 0.001). Pharmacologic responses, including heart rate control, sedation, and electrolyte therapy, varied across different subtypes. The simplified six-variable model demonstrated high predictive performance (AUC 0.89-0.96) in external validation. Conclusion Unsupervised clustering revealed four distinct NOAF subtypes in ICU patients, characterized by heterogeneous clinical trajectories. The simplified six-variable model enabled practical bedside classification, supporting precision risk assessment and potentially informing phenotype-oriented management of NOAF in the ICU.
Sheng et al. (2026) conducted a cohort in New-onset atrial fibrillation in critically ill patients (n=8,472). Clinical subtypes of new-onset atrial fibrillation was evaluated on 28-day mortality (HR 4.24-5.98, p=< 0.001). Four distinct clinical subtypes of new-onset atrial fibrillation were identified in critically ill patients, demonstrating a stepwise increase in 28-day mortality risk (HR 4.24-5.98; P<0.001).