摘要目的:探讨重症监护室(ICU)内患者发生心脏骤停(CA)的影响因素，旨在建立并验证一种辅助临床预测ICU内患者发生CA的可视化评价工具。方法:收集急诊重症监护室合作研究数据库(eICU-CRD)所有入住ICU重症患者的临床数据，使用随机数字法将纳入患者分为训练集(68 396例)和内部验证集(29 313例)。通过单因素及多因素Logistic回归筛选CA的独立影响因素并构建列线图预测模型，以急性生理和慢性健康评分-Ⅳ(APACHE-Ⅳ)作对照。通过校正曲线、受试者特征曲线(ROC)和决策曲线分析(DCA)对模型的预测效能进行检验。结果:本研究共纳入97 709例患者，ICU内成人CA发生率为0.87%。研究结果显示，急性冠脉综合征、呼吸衰竭、24 h内PCI等11个变量是重症患者发生IHCA的独立影响因素。列线图在训练集和验证集的C指数分别为0.835和0.823。而APACHE-Ⅳ评分在训练集和验证集的C指数分别为0.734和0.740，列线图预测模型的曲线下面积(AUC)均大于0.80，对重症患者发生院内心脏骤停(IHCA)具有良好的诊断价值。校正曲线结果显示列线图模型在训练集和验证集中的预测概率和IHCA实际概率的平均绝对误差分别为0.036和0.045，具有良好的一致性，并优于APACHE-Ⅳ评分( P<0.05)。 结论:建立并验证一种较APACHE-Ⅳ评分预测效能强的列线图预测模型，有助于早期识别和筛选ICU内发生CA的高危患者。

abstractsObjective:To explore the influencing factors of cardiac arrest (CA) in intensive care unit (ICU) patients, and to establish and validate a visual evaluation tool for clinical prediction of CA in ICU patients.Methods:Clinical data of all critically ill ICU patients admitted to the Emergency Intensive Care Unit Collaborative Research Database (IECU-CRD) were collected and the patients were divided into a training set (68 396 cases) and an internal validation set (29 313 cases) using a random number method. Independent influencing factors of CA were screened by univariate and multivariate Logistic regression, and a nomogram prediction model was constructed. Acute physiological and chronic Health Score Ⅳ (APACHE-Ⅳ) was used as controls. The predictive efficiency of the model was tested by calibration curve, receiver operating characteristic (ROC) curve and decision curve analysis (DCA).Results:A total of 97 709 patients were included in this study, and the incidence of adult CA in ICU was 0.87%. The results showed that 11 variables, including acute coronary syndrome, respiratory failure and PCI within 24 h, were independent factors affecting the occurrence of IHCA in severe patients. The C-index of the nomogram in the training set and validation set was 0.835 and 0.823, respectively. The C index of APACHE-Ⅳ in the training set and the verification set was 0.734 and 0.740, respectively. The area under the curve (AUC) of the model predicted by the nomogram was all greater than 0.80, which had a good diagnostic value for severe patients with hospital CA. Correction curve results showed that the mean absolute error of the predicted probability and the actual probability of IHCA in the training set and validation set was 0.036 and 0.045, respectively, which had good consistency and was better than the APACHE-Ⅳ score ( P<0.05). Conclusion:This study successfully established and verified a nomogram prediction model with higher predictive efficacy than APACHE-Ⅳ score. It is helpful for early identification and screening of high-risk patients with CA in ICU.