重症颅脑损伤患者早期癫痫发作的风险预测列线图模型构建研究

Construction of nomogram model for risk prediction of early seizures in patients with severe craniocerebral injury

目的分析重症颅脑损伤患者早期癫痫发作的危险因素,并依此构建风险预测列线图模型。方法回顾性分析本院于2017年5月至2022年6月收治的325例重症颅脑损伤患者的临床资料,根据患者是否存在早期癫痫发作将其分为发作组(32例)和未发作组(293例)。采用Logistic回归分析重症颅脑损伤患者早期癫痫发作的危险因素并绘制列线图模型,列线图的预测效能和校准度分别用ROC曲线评估和Bootstrap法检验。结果颞叶损伤、开放性颅脑损伤、脑挫裂伤、颅内血肿、中线移位≥10 mm、颅骨骨折、入院时GCS评分、入院时EEG异常、蛛网膜下腔出血、未使用预防性抗癫痫药物均为重症颅脑损伤患者早期癫痫发作的独立危险因素(OR=3.071,OR=3.093,OR=2.843,OR=3.010,OR=3.377,OR=3.142,OR=3.391,OR=3.572,OR=3.364,OR=3.056;均P<0.05)。ROC曲线分析显示,列线图模型预测重症颅脑损伤患者早期癫痫发作的曲线下面积为0.865(95%CI:0.823~0.900),灵敏度为81.25%,特异度为83.28%;经Bootstrap法对模型进行内部验证,其一致性指数(C-index值)为0.835。结论基于风险因素构建的列线图预测模型效能较好,有助于临床及时筛选出重症颅脑损伤患者中早期癫痫发作的患者。

Objective To analyze the risk factors of early seizures in patients with severe craniocerebral injury,and to construct a nomogram model for risk prediction.Methods The clinical data of 325 patients with severe craniocerebral injury admitted to our hospital from May 2017 to June 2022 were retrospectively analyzed.According to whether the patients with early seizures or not,they were divided into seizure group(32 cases)and non seizure group(293 cases).Logistic regression was used to analyze the risk factors of early epileptic seizures in patients with severe brain injury,and a nomogram model was drawn.The predictive efficacy and calibration of the nomogram were evaluated by ROC curve and Bootstrap method respectively.Results Logistic regression analysis showed that temporal lobe injury,open craniocerebral injury,cerebral contusion and laceration,intracranial hematoma,midline shift≥10 mm,skull fracture,GCS score at admission,abnormal electroencephalogram at admission,subarachnoid hemorrhage and non use of preventive antiepileptic drugs were all independent risk factors for early seizures in patients with severe brain injury(OR=3.071,OR=3.093,OR=2.843,OR=3.010,OR=3.377,OR=3.142,OR=3.391,OR=3.572,OR=3.364,OR=3.056;all P<0.05).Taking the above risk factors as predictors,a nomogram model for predicting the risk of early seizures in patients with severe craniocerebral injury was constructed.The ROC curve analysis showed that the area under the curve of the nomogram model for predicting early seizures in patients with severe brain injury was 0.865(95%CI:0.823-0.900),and the sensitivity was 81.25%and specificity was 83.28%.After internal verification by Bootstrap method,the calculated consistency index(C-index value)was 0.835.Conclusion The efficacy of nomogram prediction model based on risk factors is better,and it is helpful to screen the patients with early seizures in the patients with severe craniocerebral injury.