蛛网膜下腔出血患者预后指标的开发与验证：基于MIMIC-IV数据库的分析

Development and Validation of Prognostic Indicators in Patients with Subarachnoid Hemorrhage: An Analysis of the MIMIC-IV Database

摘要: 背景：蛛网膜下腔出血是一种严重的神经系统事件，死亡率高。本研究旨在基于MIMIC-IV数据库，为重症监护病房内的蛛网膜下腔出血患者识别有效的死亡率预后指标。方法：对来自数据库的1005名SAH患者进行回顾性分析。首先，分析基线特征以比较生存组和死亡组之间的各项临床指标。随后，对临床指标进行单因素和多因素Cox回归分析以确定预后指标。基于这些预后指标构建了列线图模型，并使用校准曲线、受试者工作特征曲线和决策曲线分析验证了列线图模型的预测准确性。此外，还对SAH患者进行了Kaplan-Meier生存曲线分析。最后，为探究预后指标与SAH之间的关系，应用了限制性立方样条模型。结果：来自MIMIC-IV数据库的合格生存组和死亡组在年龄、种族、平均动脉压、心率等临床指标上表现出显著差异。为预测列线图确定的预后指标包括：白细胞计数、舒张压、阴离子间隙、体温和序贯器官衰竭评估评分。该列线图显示出良好的预测性能和临床实用性。此外，Kaplan-Meier生存分析显示，预后指标高/低亚组之间的生存概率存在显著差异；限制性立方样条模型表明预后指标与死亡风险之间存在非线性关系。结论：本研究确定了SAH患者的预后因素为白细胞计数、阴离子间隙、体温、舒张压和SOFA评分。基于这些预后因素构建的预测模型可作为临床医生有价值的决策工具。

Abstract: Background: Subarachnoid hemorrhage is a serious nervous system event with high mortality. The purpose of this study is to identify effective mortality prognostic indicators for patients with subarachnoid hemorrhage in intensive care unit based on MIMIC-IV database. Methods: 1005 SAH patients from MIMIC-IV database were analyzed retrospectively. Firstly, the baseline characteristics were analyzed to compare the clinical indexes between the survival group and the death group. Subsequently, the clinical indicators were analyzed by univariate and multivariate Cox regression to determine the prognostic indicators. Based on these prognostic indicators, a nomogram model is constructed, and the prediction accuracy of the nomogram model is verified by the analysis of calibration curve, receiver operating characteristic curve and decision curve. In addition, Kaplan-Meier survival curve of SAH patients was analyzed. Finally, in order to explore the relationship between prognostic indicators and SAH, the restricted cubic spline model is applied. Results: There were significant differences in age, race, mean arterial pressure, heart rate and other clinical indicators between the qualified survival group and the death group from the MIMIC-IV database. Prognostic indicators for predicting nomogram include: white blood cell count, diastolic blood pressure, anion gap, body temperature and evaluation score of sequential organ failure. The nomogram shows good predictive performance and clinical practicability. In addition, Kaplan-Meier survival analysis showed that there were significant differences in survival probability between high/low subgroups of prognostic indicators; Restrictive cubic spline model shows that there is a nonlinear relationship between prognosis index and death risk. Conclusion: The prognostic factors of SAH patients are white blood cell count, anion gap, body temperature, diastolic blood pressure and SOFA score. The prediction model based on these prognostic factors can be used as a valuable decision-making tool for clinicians.