老年脓毒症患者住院期间死亡风险预测模型的建立与验证

邢冬梅; 隋冰冰; 王磊

doi:10.7619/jcmp.20233722

老年脓毒症患者住院期间死亡风险预测模型的建立与验证

Establishment and validation of risk prediction model for mortality in elderly patients with sepsis during hospitalization

摘要

摘要:
目的建立并验证可预测老年脓毒症患者住院死亡风险的模型。
方法回顾性纳入2019年1月—2022年12月哈尔滨医科大学附属第一医院重症医学科住院治疗的238例脓毒症患者，以住院期间转归情况为主要结局指标，分为死亡组68例(28.57%)和生存组170例(71.43%)。采用多因素Logistic回归法筛选脓毒症患者住院死亡的独立危险因素，并根据影响因素构建预测脓毒症患者住院死亡风险的模型。采用受试者工作特征(ROC)曲线对预测模型的性能进行评定，结果以曲线下面积(AUC)表示; 基于2016年1月—2018年12月的176例脓毒症患者的临床资料进行外部验证。
结果单因素分析显示，与生存组比较，死亡组年龄>70岁的比率、急性肾损伤(AKI) Ⅲ期比率及红细胞分布宽度(RDW)、纤维蛋白原、乳酸、血肌酐、英国早期预警评分(NEWS)、快速序贯器官衰竭评分(qSOFA)升高，氧合指数、白蛋白降低，差异均有统计学意义(P＜0.05)。多因素Logistic回归分析结果显示，年龄>70岁(OR=1.426, 95%CI: 1.055~1.928)、乳酸>6 mmol/L(OR=1.436, 95%CI: 1.105~1.867)、RDW>16%(OR=1.354, 95%CI: 1.080~1.698)、AKI Ⅲ期(OR=1.982, 95%CI: 1.407~2.791)、qSOFA>2分(OR=1.853, 95%CI: 1.255~2.738)是脓毒症患者住院期间死亡的独立危险因素(P＜0.05)。根据回归分析结果，建立脓毒症患者死亡风险方程，一致性指数(C_index)=-1.694+0.355×年龄+0.303×RDW+0.362×乳酸+0.684×AKI Ⅲ期+0.617×qSOFA。ROC曲线显示，内部验证时C_index预测脓毒症患者住院期间死亡的AUC为0.882(95%CI: 0.834~0.929), 灵敏度为83.82%, 特异度为77.06%; 外部验证时C_index预测脓毒症患者住院期间死亡的AUC为0.823(95%CI: 0.757~0.889), 灵敏度为74.13%, 特异度为81.36%。
结论年龄、乳酸、RDW、AKI分期、qSOFA与老年脓毒症死亡风险具有相关性，基于这些参数构建的模型可能有助于预测老年脓毒症住院期间全因死亡风险。

Abstract:
Objective To establish and validate a model that can predict the risk of death during hospitalization in elderly patients with sepsis.
Methods A total of 238 hospitalized patients with sepsis in the Intensive Care Unit of the First Hospital Affiliated to Harbin Medical University from January 2019 to December 2022 were retrospectively included, and they were divided into death group with 68 cases (28.57%) and survival group with 170 cases (71.43%) according to the prognosis during hospitalization as the primary outcome indicator. Multivariate Logistic regression was used to screen the independent risk factors for death during hospitalization in sepsis patients, and a model for predicting the risk of death during hospitalization in sepsis patients was established based on these factors. The performance of the prediction model was evaluated by the receiver operating characteristic (ROC) curve, and the results were expressed by the area under the curve (AUC); external validation was performed based on the clinical data of 176 sepsis patients from January 2016 to December 2018.
Results Univariate analysis showed that when compared with the survival group, the death group had higher ratios of patients aged over 70 years, ratio of injury (AKI) in stage Ⅲ, higher levels of red cell distribution width (RDW), fibrinogen, lactate, blood creatinine, the National Early Warning Score (NEWS) and the quick Sequential Organ Failure Assessment (qSOFA) score, but lower levels of oxygenation index and albumin, and the differences were statistically significant (P < 0.05). Multivariate Logistic regression analysis showed that age over 70 years old (OR=1.426, 95%CI, 1.055 to 1.928), lactate>6 mmol/L (OR=1.436, 95%CI, 1.105 to 1.867), RDW>16% (OR=1.354, 95%CI, 1.080 to 1.698), AKI in stage Ⅲ (OR=1.982, 95%CI, 1.407 to 2.791), and qSOFA score>2 points (OR=1.853, 95%CI, 1.255 to 2.738) were the independent risk factors for death during hospitalization in patients with sepsis (P < 0.05). Based on the regression analysis results, a risk equation for death in patients with sepsis was established, and consistency index (C_index)=-1.694+0.355×age+0.303×RDW+0.362×lactate+0.684×AKI in stage Ⅲ+0.617×qSOFA score. The ROC curve showed that the AUC of C_index for predicting death during hospitalization in patients with sepsis was 0.882 (95%CI, 0.834 to 0.929) with a sensitivity of 83.82% and a specificity of 77.06% by internal validation, and 0.823 (95%CI, 0.757 to 0.889) with a sensitivity of 74.13% and a specificity of 81.36% by external validation.
Conclusion Age, lactate, RDW, AKI staging and qSOFA score are correlated with the risk of death in elderly patients with sepsis, and the model constructed based on these parameters may help predict the risk of all-cause death during hospitalization in elderly patients with sepsis.

HTML全文

参考文献(25)

施引文献

资源附件(0)