A cross-sectional study of early-onset epilepsy of intracerebral hemorrhage and construction of a risk prediction model
10.3760/cma.j.cn121430-20221008-00878
- VernacularTitle:影响脑出血早发癫痫的横断面研究及风险预测模型的构建
- Author:
Xiangyan BAI
1
;
Liang ZHANG
;
Hailin LI
;
Dengjun GUO
;
Guangchao YIN
Author Information
1. 浙江省立同德医院急诊医学科，杭州 310012
- Keywords:
Intracerebral hemorrhage;
Early-onset epilepsy;
Predictive model;
Validation;
Cross-sectional study
- From:
Chinese Critical Care Medicine
2022;34(12):1273-1279
- CountryChina
- Language:Chinese
-
Abstract:
Objective:To study the early-onset epilepsy of intracerebral hemorrhage and build a prediction model to evaluate its prediction efficiency.Methods:A cross-sectional investigation was conducted to construct a specialized optimized prediction model. The prediction model was converted into a visual optimized scoring scale, so as to quantify the probability of secondary epilepsy after intracerebral hemorrhage. Based on the current prediction model of acute cerebral infraction and post-stroke seizure (AIS-PSS), the evaluation efficacy of optimized score for secondary epilepsy after hemorrhagic stroke was explored.Results:① After sample size calculation and sufficient inclusion and exclusion, 159 patients with cerebral hemorrhage were continuously selected as the model group of this cross-sectional study. A total of 29 patients with early-onset epilepsy and 130 patients without secondary epilepsy were enrolled. The time span was from January 2021 to August 2021. In addition, 77 patients with acute cerebral hemorrhage from August 2021 to February 2022 were selected as the verification group, among which 12 patients had early-onset epilepsy and 65 patients had not any secondary epilepsy. ② There were significant differences in demographic characteristics such as diabetes history, cerebral infarction history, smoking history, National Institutes of Health Stroke Scale (NIHSS) score, intracerebral hemorrhage hematoma volume, serum creatinine (SCr), neuron-specific enolase (NSE), S-100 protein and intracerebral hemorrhage site between the two model groups with different prognosis (all P < 0.05). ③ The above indexes were included in univariate and multivariate Poisson regression analysis, and the results showed that the duration of diabetes [relative risk ( RR) = 1.229, 95% confidence interval (95% CI) was 1.065-1.896, P = 0.036], smoking history ( RR = 1.419, 95% CI was 1.133-2.160, P = 0.030), history of cerebral infarction ( RR = 1.634, 95% CI was 1.128-2.548, P = 0.041), hematoma volume of cerebral hemorrhage ( RR = 1.222, 95% CI was 1.024-2.052, P = 0.041), NES content ( RR = 1.146, 95% CI was 1.041-1.704, P = 0.032), were independent influencing factors to constitute the prediction model. The prediction model was converted into a visual optimized scoring scale in the form of a line diagram to obtain the prediction probability corresponding to the corresponding score. ④ Receiver operator characteristic curve (ROC curve) was used to test the evaluation efficiency of optimized score and AIS-PSS score for early-onset cerebral hemorrhage epilepsy. Relevant data of patients in the verification group were extracted according to the information of two scores, and the final score of each patient in the verification group was obtained. The score and prognosis were put into the ROC curve to evaluate the predictive ability of different prediction models. The results showed that the cut-off value of the optimized score and the AIS-PSS score were 144 points and 7 points, respectively, and the area under the ROC curve (AUC) and the Yoden index of the optimized score were slightly lower than the AIS-PSS score. However, compared with AIS-PSS score, there was no significant difference in the evaluation efficiency of optimized score for early-onset epilepsy ( Z = 1.874, P > 0.05). Conclusion:This study constructed a specific early-onset epilepsy prediction model for patients with hemorrhagic stroke, and transformed it into an optimized score that is easy for clinical use, and its evaluation efficiency is reliable.
