Grouping of Emergency Department-based Cardiac Arrest Patients According to Clinical Features to Assess Patient Outcomes

Comprehensive Summary

Leow et al. used a machine learning algorithm to identify distinct clusters of emergency department-based cardiac arrests and examined how this clustering approach related to patient clinical outcomes. Primary outcomes included emergency department mortality and length of stay. Three risk clusters were identified based on time to cardiac arrest: cluster 1 (immediate risk), cluster 2 (early risk), and cluster 3 (late risk), with cluster 1 patients experiencing the shortest median time to cardiac arrest onset. Cluster 1 patients most commonly presented in near-arrest states at triage, while respiratory illnesses and sepsis were more common precipitating factors in clusters 2 and 3. Cluster 3 exhibited the highest mortality rate (58%), compared to cluster 1 (48%) and cluster 2 (35%). Cluster 1 had the shortest median length of stay (4 hours), while cluster 3 had the longest (81 hours).

Outcomes and Implications

The distinct characteristics of each cluster suggest opportunities for targeted interventions tailored to patients' risk profiles and underlying etiologies. Identifying which cluster a patient belongs to upon arrival could enable clinicians to implement cluster-specific treatment protocols and monitoring strategies to improve outcomes. Early cluster identification may also enable preventive interventions that reduce the likelihood of cardiac arrest in at-risk patients before arrest occurs.