| 摘要: |
| 目的 急诊脓毒症患者预后发生死亡风险较高，基于快速序贯性器官功能衰竭（qSOFA）-序贯性器官功能衰竭（SOFA）评分及床旁快速检测（POCT）生物标志物分析影响其预后的风险因素，并以此构建列线图预测模型，为临床实践提供依据。方法 回顾性选取2022年1月至2024年12月本院接受治疗的急诊脓毒患者100例，依据病历系统记录患者入院后28d的转归情况将研究对象分为存活组、死亡组。收集对比患者一般资料、临床特征资料、POCT生物标志物，依次进行选择算子（LASSO）回归分析、多因素logistic分析筛选出急诊脓毒症患者预后的风险因素，利用R软件建立列线图预测模型并验证。构建随机森林生存曲线评估各因素对患者生存预测及重要性。结果 本研究共纳入急诊脓毒症患者100例，依据病历系统记录，19例患者入院后28d发生死亡，占比19.00%，纳入死亡组，81例患者入院后28d存活，占比81.00%，纳入存活组。死亡组、存活组患者qSOFA评分、SOFA评分、血乳酸（Lac）、降钙素原（PCT）、C反应蛋白（CRP）、D-二聚体、检测白细胞计数（WBC）对比具有显著统计学差异（P＜0.05），经LASSO回归分析可知，qSOFA评分、SOFA评分、Lac、PCT、CRP未存在多重共线性、过度拟合，纳入Logistic回归模型分析发现结果表明，qSOFA评分、SOFA评分、Lac、PCT、CRP均为急诊脓毒症预后危险因素（OR=10.901、1.640、7.776、2.735、1.101，P＜0.05）；基于Logistic回归分析结果构建急诊脓毒症预后的列线图风险预测模型，绘制ROC曲线显示AUC值为0.993，95%CI为0.983~1.00，校准曲线显示模型预测结果与急诊脓毒症预后的实际情况吻合良好，Brier Score为：0.028，模型拟合度P值为：0.996，统计量为：1.224，临床决策曲线基本高于两条极端曲线，显示列线图中包含的因素对急诊脓毒症预后预测具有较高净收益。使用随机森林模型进行生存分析，评估各预测因子对28天生存结局的影响。随机森林生存分析显示模型的C-index为：0.894，表明模型具有良好的预测准确性，变量重要性图显示各预测因子对模型预测生存时间的相对贡献，生存曲线直观展示了不同风险患者的生存概率差异，从而为临床决策提供更有力的支持。结论qSOFA评分、SOFA评分、Lac、PCT、CRP均为影响急诊脓毒症预后的相关因素，基于上述因素构建风险预测列线图模型对患者预后死亡风险及生存时间具备一定预测价值，临床需早期筛选出高风险人群并针对性优化治疗方案。 |
| 关键词: 脓毒症 快速序贯器官衰竭评分 生物标志物 预测模型 |
| DOI： |
| 分类号: |
| 基金项目: |
|
| Risk prediction of Emergency sepsis: Development and Validation of the Combined Model of qSOFA and POCT biomarkers |
|
Jiang Hao,Zhou Yuebo,Bai Lin,Li Yin
|
|
Fudan University Affiliated Huadong Hospital
|
| Abstract: |
| Objective Patients with emergency sepsis have a relatively high risk of death. Based on the Rapid sequential Organ Failure Assessment (qSOFA) - sequential Organ Failure assessment (SOFA) score and bedside rapid testing (POCT) biomarkers, the risk factors affecting their prognosis were analyzed, and a nomogram prediction model was constructed accordingly to provide a basis for clinical practice.Methods: A retrospective selection was made of 100 emergency sepsis patients treated in our hospital from January 2022 to December 2024. The research subjects were divided into the survival group and the death group based on the prognosis of the patients 28 days after admission recorded in the medical record system. The general data, clinical characteristic data and POCT biomarkers of the patients were collected and compared. Selection operator (LASSO) regression analysis and multivariate logistic analysis were conducted successively to screen out the risk factors for the prognosis of emergency sepsis patients. A nomogram prediction model was established and verified using R software. Construct the random forest survival curve to evaluate the prediction and importance of each factor for the survival of patients. Results: A total of 100 emergency sepsis patients were included in this study. According to the records in the medical record system, 19 patients died 28 days after admission, accounting for 19.00%, and were included in the death group. 81 patients survived 28 days after admission, accounting for 81.00%, and were included in the survival group. There were significant statistically significant differences (P < 0.05) in the qSOFA score, SOFA score, blood lactate (Lac), procalcitonin (PCT), C-reactive protein (CRP), D-dimer, and detected white blood cell count (WBC) between the death group and the survival group. It was known through LASSO regression analysis that There was no multicollinearity or overfitting in the qSOFA score, SOFA score, Lac, PCT and CRP. The results of the Logistic regression model analysis indicated that qSOFA score, SOFA score, Lac, PCT and CRP were all prognostic risk factors for emergency sepsis (OR=10.901, 1.640, 7.776, 2.735, 1.101, P < 0.05). Based on the results of Logistic regression analysis, a nomogram risk prediction model for the prognosis of emergency sepsis was constructed. The ROC curve plotted showed that the AUC value was 0.993, and the 95%CI was 0.983-1.00. The calibration curve showed that the model prediction results were in good agreement with the actual situation of the prognosis of emergency sepsis. The Brier Score was: 0.028, the P-value of the model fitting degree was 0.996, and the statistic was 1.224. The clinical decision curve was basically higher than the two extreme curves, indicating that the factors included in the nomogram have a relatively high net benefit for the prognosis prediction of emergency sepsis. Survival analysis was conducted using the random forest model to evaluate the impact of each predictor on the 28-day survival outcome. Random forest survival analysis shows that the C-index of the model is 0.894, indicating that the model has good predictive accuracy. The variable importance plot shows the relative contribution of each predictor to the predicted survival time of the model. The survival curve visually presents the differences in survival probabilities of patients with different risks, thereby providing more powerful support for clinical decision-making. Conclusion: qSOFA score, SOFA score, Lac, PCT and CRP are all related factors affecting the prognosis of emergency sepsis. Constructing a risk prediction nomogram model based on the above factors has certain predictive value for the prognosis, mortality risk and survival time of patients. Clinically, high-risk populations need to be screened out early and treatment plans should be optimized specifically. |
| Key words: Sepsis Rapid Sequential Organ Failure Score Biomarker Prediction model |