Abstract:

Objective: To explore the risk factors for cluster seizures in children with epilepsy and to construct a risk prediction Nomogram model based on these factors. Methods: Total 200 children with epilepsy who were treated at The Affiliated Hospital of Xuzhou Medical University between March 2022 and September 2023 were enrolled. Based on the presence or absence of cluster seizures, the pediatric patients were divided into cluster seizures group (n=98) and no cluster seizures group (n=102). Univariate and multivariate stepwise Logistic regression analyses were employed to identify risk factors for cluster seizures. A Nomogram model was then constructed based on these factors. The model's performance was evaluated using receiver operating characteristic (ROC) curve analysis, calibration curves, and the Hosmer-Lemeshow goodness-of-fit test to validate its discriminative capacity, calibration accuracy and stability. Results: Logistic regression analysis identified the following independent risk factors for cluster seizures: structural etiology (OR=3.403, 95%CI: 1.442-8.027; P=0.005), infantile-onset seizures (OR=4.720, 95%CI: 2.150-10.365; P=0.000), multiple seizure types (OR=6.446, 95%CI: 2.085-19.933; P=0.001), and generalized-multifocal discharges (OR=13.257, 95%CI: 4.669-37.641; P=0.000). The Nomogram model incorporating these risk factors demonstrated excellent predictive performance, with an area under the curve (AUC) of 0.768 (95%CI: 0.703-0.834, P=0.000). The optimal cutoff value for predicting cluster seizures was 0.513. Calibration curves showed good agreement between predicted and observed probabilities, and the Hosmer-Lemeshow goodness-of-fit test indicated that the model had good stability (P=0.988). Conclusions: Epileptic children exhibiting structural etiology, infantile-onset seizures, multiple seizure types, and generalized-multifocal discharges are more susceptible to cluster seizures. The developed Nomogram model based on these factors provides a valuable clinical tool for predicting the risk of cluster seizures in children with epilepsy.

Key words: Epilepsy, Child, Cluster seizures (not in MeSH), Risk factors, Logistic models, Nomograms

摘要：

目的: 筛查儿童癫痫丛集性发作的危险因素并构建列线图（Nomogram）风险预测模型。方法: 纳入2022年3月至2023年9月徐州医科大学附属医院收治的200例癫痫患儿，根据有无丛集性发作分为丛集性发作组（98例）和无丛集性发作组（102例），单因素和多因素逐步法Logistic回归分析筛查儿童癫痫丛集性发作的危险因素，并基于危险因素构建Nomogram模型，绘制受试者工作特征（ROC）曲线和校准曲线并行Hosmer-Lemeshow拟合优度检验以验证模型的区分度、校准度和稳定性。结果: 多因素Logistic回归分析显示，结构性病因（OR=3.403，95% CI：1.442~8.027；P=0.005）、婴幼儿期发作（OR=4.720，95% CI：2.150~10.365；P=0.000）、多种发作并存（OR=6.446，95% CI：2.085~19.933；P=0.001）及广泛多灶病变（OR=13.257，95% CI：4.669~37.641；P=0.000）是儿童癫痫丛集性发作的危险因素。基于上述危险因素构建Nomogram模型，ROC曲线下面积为0.768（95% CI：0.703~0.834，P=0.000），预测儿童癫痫丛集性发作的最佳截断值为0.513；校准曲线显示该模型的预测概率与实际概率之间具有良好的一致性，Hosmer-Lemeshow拟合优度检验提示该模型具有良好的稳定性（P=0.988）。结论: 存在结构性病因、发作起始于婴幼儿期、多种发作并存及广泛多灶放电的癫痫患儿易发生丛集性发作，基于上述因素构建的Nomogram模型有助于临床预测儿童癫痫丛集性发作风险。

关键词: 癫痫, 儿童, 丛集性发作（非MeSH词）, 危险因素, Logistic模型, 列线图