论文部分内容阅读
目的探讨GRASPS、SEDAN、HAT模型在预测非溶栓性脑梗死出血转化中的临床应用价值。方法选择570例未经溶栓的急性脑梗死患者,其中男性375例,女性195例;年龄41~90岁,平均年龄68.41岁。根据头颅CT或MRI检查是否出血分为出血转化组和非出血转化组,其中出血转化组123例,非出血转化组447例。两组同时给予GRASPS、SEDAN、HAT模型评分。采用受试者工作特性曲线(ROC)获得HAT模型、SEDAN模型和GRASPS模型的灵敏度和特异度,计算曲线下面积。结果 HAT模型预测出血转化的灵敏度为63.4%,特异度为70.5%,曲线下面积0.717[95%可信区间(CI)0.661~0.772],最佳诊断界值为1.5。SEDAN模型预测出血转化的灵敏度为48.3%,特异度为51.7%,曲线下面积0.601(95%CI 0.546~0.656),最佳诊断界值为1.5。GRASPS模型预测出血转化的灵敏度为58.5%,特异度为63.1%,曲线下面积0.620(95%CI 0.564~0.676),最佳诊断界值为77.5。结论HAT、GRASPS、SEDAN模型用于非溶栓性脑梗死出血转化有一定的预测价值,但以HAT模型预测能力最强。
Objective To investigate the clinical value of GRASPS, SEDAN and HAT models in predicting hemorrhage and conversion of non-thrombolytic cerebral infarction. Methods 570 patients without thrombolysis in acute cerebral infarction, 375 males and 195 females; aged 41 to 90 years, mean age 68.41 years. According to cranial CT or MRI, whether hemorrhage was divided into hemorrhagic transformation group and non-hemorrhagic transformation group, including 123 cases of hemorrhage conversion group and 447 cases of non-hemorrhage conversion group. The two groups were given GRASPS, SEDAN, HAT model score. The receiver operating characteristic curve (ROC) was used to obtain the sensitivity and specificity of the HAT, SEDAN and GRASPS models to calculate the area under the curve. Results The sensitivity of HAT model was 63.4%, specificity was 70.5%, and the area under the curve was 0.717 [95% confidence interval (CI) 0.661-0.772). The best diagnostic cutoff was 1.5. The SEDAN model predicts the sensitivity of hemorrhagic transformation to be 48.3%, the specificity is 51.7%, the area under the curve is 0.601 (95% CI 0.546-0.656), and the best diagnostic cutoff is 1.5. GRASPS model predicts hemorrhage conversion sensitivity of 58.5%, specificity of 63.1%, area under the curve 0.620 (95% CI 0.564 ~ 0.676), the best diagnostic cutoff value of 77.5. Conclusion The HAT, GRASPS and SEDAN models have some predictive value for the hemorrhagic transformation of non-thrombolytic cerebral infarction, but the HAT model has the best predictive ability.