Mercy Heart and Vascular Hospital St. Louis, Missouri
Disclosure(s): No relevant financial relationship(s) to disclose.
Introduction: Timely electroencephalography (EEG) initiation and interpretation is necessary for fast management of nonconvulsive seizures (NCS), as delays in EEG access have been associated with poor outcomes. Artificial intelligence (AI)-enabled point-of-care (POC) EEG systems provide rapid access to EEG with constant monitoring. We pose that the introduction of AI-enabled POC EEG (Ceribell, Inc.) at our community hospitals positively impacted neurological care and patient outcomes.
Methods: We analyzed retrospective data from adult patients at three Mercy MO hospitals with limited conventional EEG (convEEG) prior to POC EEG availability. In patients who received a ≥30-minute EEG within 7 days of admission. Two cohorts were enrolled: pre-POC EEG patients monitored with convEEG (Jun-Dec 2019) and patients monitored with POC EEG after its adoption (Jun-Dec 2023) as part of standard of care. We included only patients’ first eligible visit. We report door-to-EEG as the time from first patient contact to EEG start. We extracted EEG findings from neurology reports and the maximum AI-based 5-min seizure burden displayed at the bedside.
Results: A total of 148 patients were included (convEEG: 38; POC EEG: 110). Mean age (63y (16.7) vs. 61y (15.1); p = 0.4) and median admission GCS score (11 [3,15] vs. 12 [6,15]; p = 0.56) were similar between cohorts. The POC EEG cohort had shorter median door-to-EEG time compared to the convEEG arm (26.0 h [16.8, 47.1] vs. 7.2 h [3.4,16.4]; p < 0.001). Of patients presenting to the hospital afterhours (on weekends or weekdays 5pm-8am), a greater proportion in the POC EEG arm vs. convEEG arm started their EEGs afterhours (16.7% vs. 91.1%, p< 0.001). Across the convEEG cohort, zero patients had NCS, while in the POC EEG cohort, 5 patients had NCS (all registering a maximum AI-based seizure burden ≥10%) and 4 with nonconvulsive status epilepticus (NCSE), which was in all cases detected by the AI algorithm, alerting the bedside team. Five of these seizure cases were recorded afterhours.
Conclusions: AI-enabled POC EEG significantly improved timely access to EEG monitoring, particularly during afterhours, which represents a majority of the hospital working hours. POC EEG also facilitated management of NCS/NCSE and potentially impacted patient outcomes.
