Liu Fan, Chen Jinlan, Huang Qinglan, Yan Xiaoning, Li Zhen, Yan Zheng
Department of Pediatrics/Neonatology, FuZhou First General Hospital Affiliated with Fujian Medical University, Fuzhou, Fujian, China.
Medicine (Baltimore). 2025 Jun 6;104(23):e42746. doi: 10.1097/MD.0000000000042746.
This study aimed to develop and validate a predictive model using multimodal data for early identification of septic shock in neonates with systemic inflammatory response syndrome (SIRS) or sepsis. This retrospective cohort study included neonates diagnosed with SIRS, sepsis, or septic shock at Fuzhou First General Hospital between January 2021 and December 2023. Univariate and multivariate logistic regression analyses were performed to identify independent predictors, and receiver operating characteristic curves were used to evaluate the model's predictive performance. Multivariate analysis identified 7 independent predictors of septic shock, namely, the age of neonatal (OR = 2.293, 95% CI = 2.129-3.482), critical illness score (OR = 1.835, 95% CI = 1.582-2.354), cerebral oxygen saturation (ScO2) (OR = 0.289, 95% CI = 0.281-0.359), pulsatility index of right middle cerebral artery (OR = 0.837, 95% CI = 0.828-1.022), peak velocity (PSV) of left renal hilum (OR = 0.952, 95% CI = 0.868-1.157), procalcitonin (OR = 1.875, 95% CI = 1.725-2.061), and lactate (OR = 9.654, 95% CI = 8.612-10.572) were independent influencing factors for the occurrence of septic shock. Based on the result of the regression analysis, we constructed a model for predicting septic shock, namely, multimodal model = (-0.378 × neonatal age) - (0.145 × critical illness score) - (0.366 × ScO2) + (0.416 × pulsatility index of right middle cerebral artery) + (0.825 × PSV of left renal hilum) + (12.288 × procalcitonin) + (5.167 × lactate) + 1.804. And results of receiver operating characteristic curve showed that the area under curve of the multimodal model for predicting septic shock in the SIRS + septic shock population, sepsis + septic shock population, and SIRS + sepsis + septic shock population is 0.862, 0.746, and 0.820, respectively. A multimodal prediction model incorporating clinical, hemodynamic, and biochemical parameters demonstrated robust performance in early identification of neonatal septic shock. Further validation through multicenter studies is warranted.
本研究旨在开发并验证一种使用多模态数据的预测模型,用于早期识别患有全身炎症反应综合征(SIRS)或脓毒症的新生儿败血症休克。这项回顾性队列研究纳入了2021年1月至2023年12月期间在福州市第一总医院被诊断为SIRS、脓毒症或败血症休克的新生儿。进行单因素和多因素逻辑回归分析以确定独立预测因素,并使用受试者工作特征曲线来评估模型的预测性能。多因素分析确定了7个败血症休克的独立预测因素,即新生儿年龄(OR = 2.293,95% CI = 2.129 - 3.482)、危重病评分(OR = 1.835,95% CI = 1.582 - 2.354)、脑氧饱和度(ScO2)(OR = 0.289,95% CI = 0.281 - 0.359)、右大脑中动脉搏动指数(OR = 0.837,95% CI = 0.828 - 1.022)、左肾门峰值流速(PSV)(OR = 0.952,95% CI = 0.868 - 1.157)、降钙素原(OR = 1.875,95% CI = 1.725 - 2.061)和乳酸(OR = 9.654,95% CI = 8.612 - 10.572)是败血症休克发生的独立影响因素。基于回归分析结果,我们构建了一个预测败血症休克的模型,即多模态模型 = (-0.378 × 新生儿年龄) - (0.145 × 危重病评分) - (0.366 × ScO2) + (0.416 × 右大脑中动脉搏动指数) + (0.825 × 左肾门PSV) + (12.288 × 降钙素原) + (5.167 × 乳酸) + 1.804。受试者工作特征曲线结果显示,多模态模型在SIRS + 败血症休克人群、脓毒症 + 败血症休克人群以及SIRS + 脓毒症 + 败血症休克人群中预测败血症休克的曲线下面积分别为0.862、0.746和0.820。一个纳入临床、血流动力学和生化参数的多模态预测模型在早期识别新生儿败血症休克方面表现出强大的性能。有必要通过多中心研究进行进一步验证。