基于血压及心率变异度预测重症患者院内死亡风险

Prediction of in-hospital death risk for critically ill patients based on the coefficient of variation of blood pressure and heart rate

目的 利用入住重症监护病房(intensive care unit, ICU)时的血压变异度(CV-MAP)及心率变异度(CV-HR)构建预测模型,预测ICU患者院内死亡的风险。方法 回顾性分析在美国重症监护医学信息数据库Ⅲ(medical information mart for intensive care, MIMICⅢ)中年龄≥18岁,且首次入住ICU患者的临床资料。通过多因素Logistic分析筛选危险因素并构建评分系统,采用受试者工作特征(receiver operator characteristic, ROC)曲线和校准曲线评估模型区分度和校准度,采用临床决策曲线评估模型实际应用价值。结果 共筛选符合标准的患者38 824例,院内死亡患者4075例(住院病死率为10.5%)。从危险因素中选择年龄、是否合并肝脏疾病、是否合并血液系统恶性肿瘤、是否合并转移癌、住院类型、24 h心率变异系数、24 h血压变异系数、是否使用血管活性药、是否接受镇痛治疗、是否接受镇静治疗、是否接受有创机械通气构建简化预测模型。模型预测院内死亡的ROC曲线下面积(AUC)为0.743(95%CI 0.735~0.750,P<0.001),Hosmer-Lemeshow检验χ^(2)=4.978,P=0.083。使用Bootstrap法进行1000次重复采样进行内部验证,校正曲线判断预测值与实际值一致性较好。决策曲线分析提示,在高阈值风险0.1~0.6时,预测模型具有较高的实用价值。结论 基于CV-MAP及CV-HR建立ICU患者院内死亡风险预测模型具有较好的临床预测价值,有助于识别高危患者。

Objective To establish a prediction model based on the coefficient of variation of mean arterial pressure(CV-MAP) and heart rate(CV-HR) of critically ill patients admitted to the intensive care unit(ICU) to predict the risk of in-hospital death. Methods Clinical data of critically ill patients aged ≥ 18 years and admitted to the ICU for the first time in the medical information mart for intensive care Ⅲ(MIMIC Ⅲ) were analyzed retrospectively. Multivariate Logistic regression analysis was used to determine the risk factors for in-hospital death and a prediction model was developed. The receiver operating characteristic(ROC) curve and calibration curve were applied to evaluate the differentiation and calibration degree of the prediction model. The clinical decision curve was used to evaluate the application value. Results A total of 38 824 patients were screened out, 4075 patients died in hospital, and the in-hospital mortality was 10.5%. Age, history of liver disease, history of hematological malignancies(HEM), history of metastatic cancer(METS), the type of hospitalization, CV-MAP, CV-HR, the usage of vasopressor, the usage of sedatives, the usage of analgesic, and invasive mechanical ventilation were used to establish the prediction model. The area under ROC(AUC) of the prediction model for in-hospital death risk was 0.743(95%CI 0.735-0.750, P<0.001). Hosmer-Lemeshow test found a good calibration ability of the prediction model(χ^(2)=4.978,P=0.083). Bootstrap was used to conduct 1000 times repeated sampling for verification. The calibration curve showed the predicted values were in good agreement with actual values. The clinical decision curve showed that the prediction model had certain clinical practicability in the high-risk threshold range(0.1-0.6). Conclusions The establishment of a prediction model for in-hospital death risk of ICU patients based on CV-MAP and CV-HR has good clinical prediction value, which is conducive to identifying high-risk patients.