肺部感染并发脓毒症患者入院...8_d内死亡预测模型的构建_王子文.pdf

山东医药2023 年第 63 卷第 3 期肺部感染并发脓毒症患者入院28 d内死亡预测模型的构建王子文，张林娜，徐猛，赵文静，晁亚丽徐州医科大学附属医院重症医学科，江苏徐州 221000摘要：目的建立肺部感染并发脓毒症患者入院28 d内死亡的预测模型，以预测肺部感染并发脓毒症患者的院内预后。方法收集徐州医科大学附属医院2017年1月2021年12月入住ICU的肺部感染并发脓毒症患者（建模组）及2015年1月2016年12月入住ICU的肺部感染并发脓毒症患者（验证组）的基线特征（年龄、身高、体质量、合并基础疾病种类），生命体征（入ICU 24 h的平均动脉压、心率、呼吸频率、SpO2、体温），实验室检查指标（入ICU 24 h内血气分析、电解质、肝肾功能、凝血功能、血常规），器官功能状态评分（APACHE评分、SOFA评分），治疗措施（是否镇静、呼吸支持、血管活性药物）及疾病转归情况等资料。对建模组所有患者上述资料进行单因素及多因素Logistic回归分析，筛选出肺部感染并发脓毒症患者死亡的独立影响因素，建立回归方程即肺部感染并发脓毒症患者入院28 d内死亡预测模型（简称死亡预测模型），利用R4.0.3软件“rms”包将回归方程可视化为列线图。在验证组中，通过建立的死亡预测模型对患者入院28 d内死亡结局进行预测，并与临床常用的ICU预后评分如SOFA评分、APACHE评分的诊断效能相比较；采用受试者工作特征曲线下面积（AUR）以及校准曲线分别评价该预测模型的区分度和校准度。结果本研究共纳入970例肺部感染并发脓毒症患者，建模组739例，死亡117例，生存622例；验证组231例，死亡24例、生存207例。死亡组与生存组年龄、体质量、SOFA评分，合并肥胖占比、合并贫血占比、使用血管活性药物占比、机械通气占比、心率、呼吸频率、平均动脉压、体温、SpO2、红细胞计数、谷草转氨酶、白蛋白、血糖、肌酐、乳酸、国际标准化比值、血钾、动脉血氧分压比较，P均0.05。多因素Logistic回归分析结果显示，肺部感染并发脓毒症患者入院28 d内死亡的独立影响因素为年龄、体质量、SOFA评分、肌酐、乳酸及呼吸频率，分别将其赋值为X1X6，然后基于上述独立影响因素建立回归方程：Logit（P）=-7.673+0.047X1-0.022X2+0.202X3+0.013X4+0.130X5+0.137X6，其中P是肺部感染并发脓毒症患者28 d死亡预测概率。死亡预测模型对肺部感染并发脓毒症患者入院 28 d 内死亡预测的 AUR 在建模和验证组分别为 0.826（95%CI 0.7790.873）和 0.850（95%CI 0.7630.937），SOFA评分分别为0.721（95%CI 0.6610.780）、0.698（95%CI 0.5890.808），APACHE评分分别为0.623（95%CI 0.5530.693）和 0.554（95%CI 0.4350.673）。死亡预测模型的校准曲线在建模组和验证组均与平面直角坐标系中45的直线重合度较高。结论成功构建了肺部感染并发脓毒症患者入院28 d内死亡的预测模型，即Logit（P）=-7.673+0.047X1-0.022X2+0.202X3+0.013X4+0.130X5+0.137X6，其可以有效地对肺部感染并发脓毒症患者死亡进行预测。关键词：肺部感染并发症；脓毒症；肺部感染；死亡预测模型doi：10.3969/j.issn.1002-266X.2023.03.008 中图分类号：R631；R563 文献标志码：A 文章编号：1002-266X（2023）03-0037-07Establishment of death prediction model within 28-day admission for patients with sepsis induced by pulmonary infectionWANG Ziwen，ZHANG Linna，XU Meng，ZHAO Wenjing，CHAO YaliDepartment of Intensive Care Unit，Affiliated Hospital of Xuzhou Medical University，Xuzhou 221000，ChinaAbstract：Objective To establish the death prediction model within 28-day admission for patients with sepsis induced by pulmonary infection，and to predict their nosocomial prognosis.Methods The clinical data of patients with sepsis induced by pulmonary infection admitted to the ICU from January 2017 to December 2021（modeling group）and the clinical data of patients with sepsis induced by pulmonary infection admitted to the ICU from January 2015 to December 基金项目：徐州市科学技术项目（KC17172，KC20154）；徐州医科大学附属医院院课题项目（2021ZA33）。第一作者简介：王子文（1987-），女，副主任医师，主要研究方向为脓毒症。E-mail：通信作者简介：晁亚丽（1981-），女，主任医师，主要研究方向为脓毒症。E-mail：开放科学（资源服务）标识码（OSID）37山东医药2023 年第 63 卷第 3 期2016（verification group）were retrospectively collected，including the baseline characteristics（age，height，body weight，type of underlying diseases），vital signs（mean arterial pressure，heart rate，respiratory rate，SpO2，body temperature），laboratory tests（blood gas analysis within 24 hours of ICU admission，electrolytes，liver and kidney function，coagulation function，blood routine），organ function status score（APACHEII score，SOFA score），treatment measures（whether sedation，respiratory support，vasoactive drugs），and outcome.Univariate and multivariate Logistic regression analyses were performed on data in the modeling group.The influencing factors of death in patients with sepsis induced by pulmonary infection were screened out.Then Logistic regression equation was established and visualized as a nomogram using the R4.0.3 software rms package.The model was verified in the validation group.The patient outcome was predicted by 28-day death prediction regression equation and compared with the real outcome.The discrimination and calibration degree of the model were evaluated by the area under receiver operating characteristic curve（AUROC）and the calibration curve.It was compared with the diagnostic performance of ICU prognostic scores commonly used in clinical practice，such as SOFA score and APACHEII score.Results In this study，a total of 970 patients with sepsis induced by pulmonary infection were included.There were 739 cases in the modeling group，117 cases died，and 622 cases survived.There were 231 cases in the verification group，24 cases died，and 207 cases survived.Variables that were statistically different between the death group and the survival group included age，weight，SOFA score，obesity，anemia，use of vasoactive drugs，mechanical ventilation，heart rate，respiratory rate，mean arterial pressure，body temperature，SpO2，red blood cell count，aspartate aminotransferase，albumin，blood glucose，creatinine，lactic acid，international normalized ratio，blood potassium，and arterial oxygen partial pressure（all P0.05），and there were no statistical differences in the other indicators.Multivariate Logistic regression analysis showed that the independent influencing factors were age，body weight，SOFA score，creatinine，lactate and respiratory rate.We assigned them as X1-X6，respectively，and then，based on the above independent influencing factors，the Logistic regression equation was established：Logit（P）=-7.673+0.047X1-0.022X2+0.202X3+0.013X4+0.130X5+0.137X6.P was the predicted probability of 28-day death in patients with sepsis induced by