Modified CAVE score for predicting late seizures after intracerebral hemorrhage

Abstract

BACKGROUND AND PURPOSE: Seizures commonly occur in patients with intracerebral hemorrhage (ICH). Anticonvulsants are commonly used for preventing seizures in patients with ICH. Thus, patients with ICH at high risk of seizures must be identified. The study aims to elucidate whether double the score of cortex involvement in ICH patients can increase accuracy of CAVE score for predicting late seizures. METHOD: This retrospective analysis of the medical records of surviving patients admitted between June 1, 2013, and December 31, 2019. Validated the CAVE score and modified it (CAVE2). The main outcome of patients with ICH was seizures. The first seizures occurring within 7 days after a stroke were defined as early seizures. Seizures occurring after 1 week of stroke onset, including patients who had experienced early seizures or patients who had not, were defined as late seizures. CAVE and CAVE2 scores were validated using the cohort. The accuracy and discrimination of those two scores were accessed by the area under the operating characteristic curve. Akaike information criterion, integrated discrimination improvement, and continuous net reclassification improvement were used to assess the performance of the CAVE and CAVE2 scores. RESULTS: In the cohort showed that late seizures occurred in 12.7% (52/408) of patients with ICH. Male sex, age > 65 years, cortex involvement, and early seizures were associated with the occurrence of late seizures, with odds ratios of 2.09, 2.04, 4.12, and 3.78, respectively. The risk rate of late seizures was 6.66% (17/255), 14.8% (17/115), and 47.4% (18/38) for CAVE scores ≤ 1, 2, and ≥ 3, and 4.6% (12/258), 18.3% (13/71), and 54.4 (20/37) for CAVE2 scores ≤ 1, 2, and ≥ 3 respectively. The C-statistics for the CAVE and CAVE2 scores were 0.73 and 0.74 respectively. CONCLUSION: The CAVE score can identify patients with ICH and high risk for late seizures. The CAVE can be modified by changing the score of cortex involvement to 2 points to improve accuracy in predicting late seizures in patients with ICH.