Liu Ronghua, Li Xiang, Yang Jie, Peng Yue, Liu Xiaolu, Tian Chanchan
Department of Laboratory Medicine, The Second People's Hospital of China Three Gorges University, The Second People's Hospital of Yichang, Third Floor, No. 21, Xiling 1st Road, Yichang, 443000, Hubei, China.
Respiratory and Critical Care Department, The Second People's Hospital of China Three Gorges University, The Second People's Hospital of Yichang, Yichang, 443000, Hubei, China.
Eur J Med Res. 2025 Mar 25;30(1):201. doi: 10.1186/s40001-025-02448-z.
It aimed to identify the key risk factors associated with carbapenem-resistant organism (CRO) infections in septic patients, and subsequently develop a nomogram and assess its predictive accuracy.
The study population comprised adult critically ill patients with sepsis, drawn from the MIMIC-IV 2.0 data set. The data were split into a training set and a validation set at a 7:3 ratio. Independent predictors were identified using both univariate and multivariate logistic regression models, followed by the construction of a nomogram. The predictive performance of the model was evaluated using the C-index, receiver operating characteristic (ROC) curve, area under the curve (AUC), calibration curve, and decision curve.
We enrolled 8814 patients, with 529 (6%) CRO-infected and 8285 (94%) non-CRO-infected. Using risk factors such as age, gender, laboratory values (WBC_max, Creatinine_max, BUN_max, Hemoglobin_min, Sodium_max), and medical conditions (COPD, hypoimmunity, diabetes), along with medications (meropenem, ceftriaxone), we developed a predictive model for CRO infection in septic patients. The model demonstrated good performance, with AUC values of 0.747 for the training set and 0.725 for the validation set. The calibration curve indicates that predicted outcomes align well with observed outcomes. The clinical decision curve results indicate that the nomogram prediction model has a high net benefit, which is clinically beneficial.
The nomogram we have developed for predicting the risk of CRO infection in sepsis patients is reasonably accurate and reliable.
Not applicable.
旨在确定脓毒症患者中与耐碳青霉烯类微生物（CRO）感染相关的关键危险因素，随后构建列线图并评估其预测准确性。
研究人群包括从MIMIC-IV 2.0数据集中选取的成年脓毒症重症患者。数据按7:3的比例分为训练集和验证集。使用单变量和多变量逻辑回归模型确定独立预测因素，随后构建列线图。使用C指数、受试者工作特征（ROC）曲线、曲线下面积（AUC）、校准曲线和决策曲线评估模型的预测性能。
我们纳入了8814例患者，其中529例（6%）发生CRO感染，8285例（94%）未发生CRO感染。利用年龄、性别、实验室值（白细胞计数最高值、肌酐最高值、尿素氮最高值、血红蛋白最低值、钠最高值）、疾病状况（慢性阻塞性肺疾病、免疫低下、糖尿病）以及药物（美罗培南、头孢曲松）等危险因素，我们建立了脓毒症患者CRO感染的预测模型。该模型表现良好，训练集的AUC值为0.747，验证集的AUC值为0.725。校准曲线表明预测结果与观察结果吻合良好。临床决策曲线结果表明列线图预测模型具有较高的净效益，具有临床实用性。
我们开发的用于预测脓毒症患者CRO感染风险的列线图相当准确且可靠。
不适用。
