Wang Xiaoqian, Li Shuai, Cao Quanxia, Chang Jingjing, Pan Jingjing, Wang Qingtong, Wang Nan
Department of Critical Care Medicine, The First Affiliated Hospital of Anhui Medical University, Hefei, Anhui, China.
Anhui Public Health Clinical Center, Hefei, Anhui, China.
Heliyon. 2024 Aug 5;10(16):e35641. doi: 10.1016/j.heliyon.2024.e35641. eCollection 2024 Aug 30.
This study aimed to develop and validate a nomogram model for predicting 28-day mortality in patients with sepsis in the intensive care unit (ICU).
We retrospectively analyzed data from 331 patients with sepsis admitted to the ICU as a training set and collected a validation set of 120 patients. Both groups were followed for 28 days. Logistic regression analyses were performed to identify the potential prognostic factors for sepsis-related 28-day mortality. A nomogram model was generated to predict 28-day mortality in patients with sepsis in the ICU. Receiver operating characteristic (ROC) curve analysis, calibration curves, and decision curve analysis (DCA) were used to evaluate the model's prediction performance and clinical application. In addition, we used ROC curve analysis and DCA to compare this model with the sequential organ failure assessment (SOFA) and Acute Physiology and Chronic Health Evaluation (APACHE II) scores and further assessed the clinical value of our model.
Logistic multivariate regression analysis revealed that mechanical ventilation, oxygenation index, and lactate and blood urea nitrogen (BUN) levels were independent predictors of 28-day mortality in patients with sepsis in the ICU. We developed a nomogram model based on these results to further predict 28-day mortality. The model demonstrated satisfactory calibration curves for both training and validation sets. Additionally, in the training set, the area under the ROC curve (AUC) for this model was 0.80. In the validation set, the AUC was 0.82. DCA showed that the high-risk thresholds ranged between 0 and 0.86 in the training set and between 0 and 0.75 in the validation set. We compared the ROC curve and DCA of this model with those of SOFA and APACHE II scores in both the training and validation sets. In the training set, the AUC of this model was significantly higher than those of the SOFA ( = 0.032) and APACHE II ( = 0.004) scores. Although the validation set showed a similar trend, the differences were not statistically significant for the SOFA ( = 0.273) and APACHE II ( = 0.320) scores. Additionally, the DCA showed comparable clinical utility in all three assessments.
The present study used four common clinical variables, including mechanical ventilation, oxygenation index and lactate and BUN levels, to develop a nomogram model to predict 28-day mortality in patients with sepsis in the ICU. Our model demonstrated robust prediction performance and clinical application after validation and comparison.
本研究旨在开发并验证一种用于预测重症监护病房（ICU）脓毒症患者28天死亡率的列线图模型。
我们回顾性分析了331例入住ICU的脓毒症患者的数据作为训练集，并收集了120例患者的验证集。两组均随访28天。进行逻辑回归分析以确定脓毒症相关28天死亡率的潜在预后因素。生成列线图模型以预测ICU中脓毒症患者的28天死亡率。采用受试者工作特征（ROC）曲线分析、校准曲线和决策曲线分析（DCA）来评估模型的预测性能和临床应用。此外，我们使用ROC曲线分析和DCA将该模型与序贯器官衰竭评估（SOFA）和急性生理与慢性健康评估（APACHE II）评分进行比较，并进一步评估我们模型的临床价值。
逻辑多因素回归分析显示，机械通气、氧合指数以及乳酸和血尿素氮（BUN）水平是ICU中脓毒症患者28天死亡率的独立预测因素。基于这些结果，我们开发了一个列线图模型以进一步预测28天死亡率。该模型在训练集和验证集中均显示出令人满意的校准曲线。此外，在训练集中，该模型的ROC曲线下面积（AUC）为0.80。在验证集中，AUC为0.82。DCA显示，训练集中的高风险阈值在0至0.86之间，验证集中在0至0.75之间。我们在训练集和验证集中将该模型的ROC曲线和DCA与SOFA和APACHE II评分的曲线和DCA进行了比较。在训练集中，该模型的AUC显著高于SOFA（P = 0.032）和APACHE II（P = 0.004）评分的AUC。尽管验证集显示出类似趋势，但对于SOFA（P = 0.273）和APACHE II（P = 0.320）评分，差异无统计学意义。此外，DCA显示在所有三项评估中临床效用相当。
本研究使用了四个常见的临床变量，包括机械通气、氧合指数以及乳酸和BUN水平，来开发一个列线图模型以预测ICU中脓毒症患者的28天死亡率。我们的模型在验证和比较后显示出强大的预测性能和临床应用价值。
