目的 分析ICU脓毒症患者亚谵妄综合征的危险因素，构建并验证亚谵妄风险预测模型，以期为临床提供有效的预测工具。 方法 选取2021年4月—2023年10月厦门市某三级甲等医院重症医学科收治的脓毒症患者作为研究对象。利用重症监护谵妄筛查量表及ICU意识模糊评估量表综合评估患者的谵妄状态，区分出亚谵妄组与无谵妄组。通过单因素Logistic回归分析及LASSO回归进行变量筛选，多因素Logistic回归分析建立亚谵妄风险预测模型并构建静态及动态列线图。利用校准曲线图、受试者操作特征曲线及决策曲线分析综合评估模型的预测价值。结果 Logistic回归分析结果显示，年龄（OR=1.045）、吸烟史（OR=3.100）、身体约束（OR=3.611）、瑞芬太尼使用（OR=0.040）以及Richmond躁动-镇静评分（OR=0.244）被确定为ICU脓毒症患者发生亚谵妄的独立风险因素。模型内部评价及验证结果显示，建模组与验证组的受试者操作特征曲线下面积分别为0.877（95%CI：0.839~0.914）和0.812（95%CI：0.718~0.906），表明模型具有较高的预测准确性；Hosmer-Lemeshow检验P值分别为0.548和0.623，显示模型拟合良好。结论 该研究构建的ICU脓毒症患者亚谵妄风险预测模型，有助于临床医护人员早期识别并采取预防性措施，降低ICU脓毒症患者发生亚谵妄综合征的风险。

Objective To analyze the risk factors of subsyndromal delirium（SSD） in Intensive Care Unit（ICU） patients with sepsis，construct and validate a risk prediction model for SSD，and provide an effective predictive tool for clinical practice. Methods The sepsis patients admitted to the intensive care unit（ICU） of a tertiary Grade A hospital in Xiamen，China were selected as research objects from April 2021 to October 2023. The delirium status of the patients was comprehensively evaluated using the Intensive Care Delirium Screening Checklist and the Confusion Assessment Method for the ICU，and the patients were divided into the SSD group and the non-delirium group. Univariate logistic regression analysis and LASSO regression were used for variable selection，and multivariate logistic regression analysis was conducted to establish a risk prediction model for SSD and construct static and dynamic nomograms. The predictive value of the model was comprehensively evaluated using calibration curves，receiver operating characteristic curves（ROC curves），and decision curve analysis（DCA）. Results The results of logistic regression analysis showed that age（OR=1.045），smoking history（OR=3.100），physical restraint（OR=3.611），remifentanil use（OR=0.040），and Richmond Agitation-Sedation Scale score（OR=0.244） were identified as independent risk factors for SSD in ICU patients with sepsis. The internal evaluation and validation results of the model showed that the areas under the ROC curves of the modeling group and the validation group were 0.877（95%CI：0.839~0.914） and 0.812（95%CI：0.718~0.906），respectively，indicating that the model had high predictive accuracy. The P values of the Hosmer-Lemeshow test were 0.548 and 0.623，respectively，indicating that the model fit well. Conclusion This study constructed a risk prediction model for SSD in ICU patients with sepsis，which is helpful for clinical medical staff to identify SSD early and take preventive measures，thereby reducing the risk of SSD in ICU patients with sepsis.