摘要
目的探讨可能导致颅骨成形术后癫痫发作(PCS)的危险因素,并构建预测模型。方法回顾性分析163例行颅骨成形术患者的临床资料,根据患者术后有无癫痫发作分为癫痫发作组(35例)和无癫痫发作组(128例)。采用单因素和进一步的多因素Logistic回归分析,探讨可能导致PCS的危险因素,并构建预测模型,利用Hosmer-Lemeshow(H-L)检验及受试者工作特征(ROC)曲线评价该模型对PCS的诊断效能和预测价值。结果共163例患者被纳入本研究,出现PCS的患者共35例,发生率为21.5%。单因素分析显示:两组患者格拉斯哥预后量表(GOS)评分、原发疾病后发生癫痫发作、皮瓣情况比较存在差异(P<0.05)。进一步的多因素Logistic回归分析显示:原发疾病后发生癫痫发作、低GOS评分、皮瓣塌陷>0.5 cm是导致PCS的独立危险因素(P<0.05)。模型logit(P)=5.105+3.266x_(1)-1.775x_(2)-1.241x_(3)-1.576x_(4),经H-L检测显示χ^(2)=6.637(P>0.05),表明模型与观测值具有较好的拟合度;按照诊断几率logit(P)绘制患者PCS的ROC曲线,当logit(P)>5.515时,曲线下面积(AUC)为0.850,95%CI为(0.776,0.925),敏感度为71.4%,特异度为87.5%。结论原发疾病后发生癫痫发作、低GOS评分、皮瓣塌陷>0.5 cm可能增加患者PCS的发生率,构建的预测模型具有较好的区分度、一致性及临床适应性。
Objective To explore the risk factors for post-cranioplasty seizures(PCS)and construction of prediction model.Methods The clinical data of 163 patients who underwent cranioplasty were retrospectively analyzed,and the patients were divided into a seizure group(35 patients)and a non-seizure group(128 patients)according to the presence or absence of postoperative seizures.Univariate and further multifactorial Logistic regression analyses were used to explore the risk factors that may lead to PCS,and a predictive model was constructed to evaluate the diagnostic efficacy and predictive value of the model for PCS using the Hosmer-Lemeshow(H-L)test and the receiver operating characteristic(ROC)curve.Results A total of 163 patients were included in this study,and a total of 35 patients developed PCS,with an incidence of 21.5%.Univariate analysis showed that there were differences in Glasgow outcome scale(GOS)scores,occurrence of seizures after primary disease and skin flap between the two groups(P<0.05).Further multifactorial Logistic regression analysis showed that the occurrence of seizures after primary disease,low GOS score,and flap collapse>0.5 cm were independent risk factors for PCS(P<0.05).Model logitt(P)=5.105+3.266x_(1)-1.775x_(2)-1.241x_(3)-1.576x_(4),the H-L test showed that the model had a good fit with the observed values.The ROC curve of PCS in patients was plotted according to the diagnostic probability logit(P),and when logit(P)>5.515,the area under the curve(AUC)was 0.850,with a 95%CI of(0.776,0.925),and the sensitivity was 71.4%,and the specificity was 87.5%.Conclusion The occurrence of seizures after primary disease,low GOS score,and flap collapse>0.5 cm may increase the incidence of PCS in patients.The constructed prediction model has good differentiation,consistency,and clinical adaptability.
作者
于涛
徐震
王文成
李钊
刘儒斌
刘斌
YU Tao;XU Zhen;WANG Wen-cheng(Department of Neurosurgery,the First People's Hospital of Jining City,Jining 272000,China)
出处
《中国实用医药》
2024年第16期25-29,共5页
China Practical Medicine
关键词
颅骨成形
癫痫
危险因素
预测模型
Cranioplasty
Epilepsy
Risk factors
Predictive modeling
