Authors :
Stacey Kim, MEng, BA – Kern Medical
Yohan Ko, BS – Kern Medical
Aaron Fernandez, BS – Kern Medical
Presenting Author: Alan Salim, BS – Kern Medical
Katayoun Sabetian, MD – Kern Medical
Charles Liu, MD, PhD – University of Southern California
Hari Veedu, MD, FACNS – Kern Medical, United Neuroscience Institute
Rationale:
Post-stroke epilepsy (PSE) causes significant long-term morbidity in stroke patients, yet early risk detection remains a challenge. The SeLECT and CAVE scores predict PSE following ischemic stroke (IS) and intracerebral hemorrhage (ICH), respectively. SeLECT evaluates 5 clinical factors: seizure at stroke onset, stroke location, cortical involvement, stroke type, and stroke severity [1]. CAVE assesses 4 variables: cortical involvement, age, hematoma volume, and early seizure [2]. A modified version, CAVE-2, doubles the weight of cortical involvement to enhance predictive accuracy. While validated at academic centers, their performance in underserved community hospitals remains largely unexplored. This study validates SeLECT and CAVE retrospectively in a Central Valley, CA community hospital.
Methods:
This retrospective study analyzes 106 patients 18+ years old, 93 IS and 13 ICH, admitted 2019-2024. All patients admitted to the emergency with an initial stroke were verified by CT head and/or MRI brain. If a patient was unable to undergo CT angiography, MR angiography was performed to assess the severity of large-artery atherosclerosis and to evaluate occlusion or intraluminal thrombus. Patients with history of stroke or seizures and/or without ≥ 12 months follow-up after initial stroke were excluded. SeLECT score (0-9) classifies patients as low-risk (0–2), moderate-risk (3–4), and high-risk (≥ 5). Similarly, CAVE (0-4) and CAVE-2 (0-5) categorize patients as low-risk (0–1), moderate-risk (2–3), and high-risk (≥ 4).
Results:
Based on SeLECT risk stratification, 46.2% of patients classified as low-risk, 31.2% as moderate-risk, and 22.6% as high-risk. Imaging showed acute IS primarily in the middle cerebral artery (MCA) and deep perforating territories. Large-artery atherosclerosis was present in 52.7%, cortical involvement 50.5%, MCA infarcts 45.2%, and 3.2% had a seizure within seven days of stroke onset. SeLECT demonstrated an accuracy of 48.4% in predicting PSE in IS.
CAVE classified 46.2% as low-risk, 53.9% as moderate-risk, and no high-risk, while CAVE-2 redistributed the cohort into 46.2% low-, 23.1% moderate-, and 30.8% high-risk. Hematomas were classified as cortical, subcortical (thalamic and basal ganglia), or brainstem and cerebellum. Both CAVE and CAVE-2 presented 61.5% accuracy predicting PSE in ICH.
Conclusions:
Majority IS patients had low SeLECT scores, potentially due to referral patterns in our region, where severe strokes are transported to tertiary centers. SeLECT appears applicable in community hospitals but needs broader representation of moderate- to high-risk patients for further validation. Though our ICH cohort was limited to 13 cases, CAVE demonstrated reasonable predictive performance. These preliminary results underscore the need for larger community-based studies to assess these scores’ reliability in non-academic healthcare environments, where patient characteristics and stroke care may differ substantially from tertiary care settings.Funding:
N/A