识别气管插管大于48 h且通过脱机试验的神经重症患者拔管的独立风险因素，并建立拔管风险预测模型。

收集2020年4月至2023年8月安徽医科大学第一附属医院神经重症监护病房中接受规范化监护及治疗的83例神经重症患者的病例资料。根据拔管后96 h内是否需要再次插管或气管切开将83例患者分为拔管成功组（61例）和拔管失败组（22例）。比较两组患者的一般资料、神经功能、气道保护能力情况以及生理生化指标，最终通过多因素回归分析建立预测模型。进一步绘制受试者工作特征（ROC）曲线，检验研究结果的预测价值、敏感度和特异度，并采用Z检验对各评分的曲线下面积（AUC）进行比较。

83例患者中有61例（73.5%）成功拔管，22例（26.5%）在拔管后96 h内呼吸障碍再次插管。两组患者格拉斯哥昏迷量表评分的运动子项（GCS-M）得分、肌力分级、拔管前意识水平、呛咳反应、吞咽功能、拔管后吸氧方式、拔管前体温、血钠离子浓度、Godet's评分、VISAGE评分和呼吸功能不全量表-插管（RIS-i）评分比较，差异均有统计学意义（P均< 0.05）。将GCS-M得分、拔管前意识水平、呛咳反应、吞咽功能及拔管前体温纳入多因素logistic回归分析，结果显示拔管前的GCS-M得分< 4分[比值比（OR）= 8.835，95%置信区间（CI）（1.638，47.659），P = 0.011]、呛咳反应> Ⅱ级[OR = 10.281，95%CI（2.366，44.129），P = 0.002]、体温≥ 37 ℃[OR = 4.506，95%CI（1.024，19.833），P = 0.046]是神经重症患者拔管失败的危险因素。进一步根据OR值对每个独立危险因素进行赋分，并建立神经危重拔管失败风险预测量表，总分为5分，ROC曲线分析显示本研究量表评分> 1分预测拔管失败（高失败风险）的敏感度为77.3%，特异度为85.2%，AUC为0.850[95%CI（0.775，0.919），P < 0.001]。Z检验结果显示，当前量表的预测价值高于Godet's评分（Z = 2.050，P = 0.040）、VISAGE评分（Z = 1.990，P = 0.047）。

该临床预测评分可为神经重症患者脱机后拔管提供指导价值。

To identify the independent risk factors for extubation in neurological critically ill patients who had been intubated for more than 48 hours and passed the weaning test, and to establish an extubation risk prediction model.

The data of 83 neurological critically ill patients who received standardized monitoring and treatment in the neurological intensive care unit of the First Affiliated Hospital of Anhui Medical University from April 2020 to August 2023 were collected. According to whether they needed re-intubation or tracheotomy within 96 hours after extubation, the 83 patients were divided into a successful extubation group (61 cases) and a failed extubation group (22 cases). The general data, neurological function, airway protection ability, and physiological and biochemical indicators of the two groups were compared, and finally a prediction model was established by multivariate regression analysis. The receiver operating characteristic (ROC) curve was further drawn to test the predictive value, sensitivity, and specificity of research results, and the area under the curve (AUC) of each score was compared by the Z test.

Among the 83 patients, 61 patients (73.5%) were successfully extubated, and 22 patients (26.5%) were reintubated due to respiratory dysfunction within 96 hours after extubation. There were significant differences in the Glasgow coma scale-motor (GCS-M) score, muscle strength grade, consciousness level before extubation, choking reaction, swallowing function, oxygen inhalation method after extubation, body temperature before extubation, blood sodium concentration, Godet's score, VISAGE score, and respiratory insufficiency scale-intubated (RIS-i) score between the two groups (all P < 0.05). The GCS-M score, consciousness level before extubation, choking reaction, swallowing function, and body temperature before extubation were included in multivariate logistic regression analysis. The results showed that the GCS-M score before extubation < 4 [odds ratio (OR) = 8.835, 95% confidence interval (CI) (1.638, 47.659), P = 0.011], choking reaction > grade Ⅱ [OR = 10.281, 95%CI (2.366, 44.129), P = 0.002], and body temperature ≥ 37 ℃ [OR = 4.506, 95%CI (1.024, 19.833), P = 0.046] were risk factors for extubation failure in neurological critically ill patients. Each independent risk factor was further scored according to the OR value, and a risk prediction scale for failure of extubation in critically ill patients with neurological diseases was established, with a total score of 5. ROC analysis showed that the sensitivity of the scale score > 1 in predicting failure of extubation (high risk of failure) was 77.3%, the specificity was 85.2%, and the AUC was 0.850 [95%CI (0.775, 0.919), P < 0.001]. The Z test showed that the predictive value of the current scale was higher than that of the Godet's score (Z = 2.050, P = 0.040) and VISAGE score (Z = 1.990, P = 0.047).

This clinical prediction score can provide guidance for extubation in critically ill patients with neurological diseases after weaning.