脑穿支动脉粥样硬化疾病早期神经功能恶化模型的构建与验证

Construction and validation of early neurological deterioration model in branch atherosclerosis disease

目的构建与验证脑穿支动脉粥样硬化疾病(BAD)患者发生早期神经功能恶化(END)的预测模型。方法选取2017年1月—2022年12月于黄河三门峡医院神经内科住院的出现END的246例BAD患者为研究对象。所有患者按照7∶3被随机分为建模组(n=172)和验证组(n=74)。收集患者的一般资料和临床资料。采用多因素Logistic回归分析BAD患者发生END的影响因素,采用一致性指数和受试者工作特征(ROC)曲线分析预测模型在建模组和验证组的区分度,绘制校正曲线评估预测模型在建模组和验证组中的准确性。结果建模组患者中,男109例(63.37%),女63例(36.63%);年龄(65.47±13.00)岁;中位发病时间4.00(2.00,7.00)h。验证组患者中,男39例(52.70%),女35例(47.30%);年龄(66.09±12.14)岁;中位发病时间2.50(1.90,6.00)h。多因素Logistic回归分析显示,是否吸烟、白质病变程度、中性粒细胞计数、低密度脂蛋白胆固醇水平、是否接受静脉溶栓治疗是BAD患者发生END的影响因素(P<0.05)。预测模型的敏感度为82.29%,特异度为72.37%;验证模型的敏感度为73.17%,特异度为72.73%。校准曲线均接近于直线为1的曲线。结论基于吸烟、白质病变程度、中性粒细胞计数、低密度脂蛋白胆固醇水平、静脉溶栓,用列线图建立的脑穿支动脉粥样硬化疾病早期神经功能恶化患者预后的预测模型可辅助预测患者的短期预后。

Objective To construct and validate the prediction model of early neurological deterioration(END)in patients with branch atheromatous disease(BAD).Methods From January 2017 to December 2022,246 END patients with BAD who were admitted to the Department of Neurology of Yellow River Sanmenxia Hospital were selected as participants.All patients were randomly divided into a modeling group(n=172)and a validation group(n=74)according to a 7∶3 ratio.General and clinical information of patients were collected.Multivariate Logistic regression was used to analyze the influencing factors of END in patients with BAD.Consistency index and receiver operating characteristic(ROC)curve were used to analyze the discriminative power of the predictive model in the modeling and validation groups,and a calibration curve was drawn to evaluate the accuracy of the predictive model in two groups.Results Among the patients in the modeling group,there were 109 males(63.37%)and 63 females(36.63%),with an age of(65.47±13.00)years and a median onset time of 4.00(2.00,7.00)hours.Among the patients in the validation group,there were 39 males(52.70%)and 35 females(47.30%),with an age of(66.09±12.14)years and a median onset time of 2.50(1.90,6.00)hours.Multivariate Logistic regression showed that smoking,degree of white matter lesions,neutrophil count,low-density lipoprotein cholesterol,and intravenous thrombolysis was the influencing factors for END in patients with BAD,and the differences were statistically significant(P<0.05).The sensitivity and specificity of the prediction model were 82.29%and 72.37%,respectively.The sensitivity and specificity of the validation model were 73.17%and 72.73%,respectively.The calibration curves were close to the curve with a straight line of 1.Conclusions Based on smoking,white matter lesion degree,neutrophil count,low-density lipoprotein cholesterol,intravenous thrombolysis,the prediction model of END in patients with BAD established by nomogram can assist in predicting the short-term prognosis of patients.