自发性脑出血患者需要重症治疗预后模型的建立与验证

The Establishment and Verification of a Prognosis Model for Patients with Spontaneous Intracerebral Hemorrhage Need Critical Treatment

摘要: 背景和目的：自发性脑出血(sICH)患者在住院期间出现病情恶化较为常见，目前还没有建立对脑卒中患者住院期间需要接受重症治疗的预测模型。本研究旨在建立和验证一种新的模型来预测卒中患者需要接受重症治疗的风险。方法：回顾性队列研究纳入卒中前功能独立的自发性脑出血患者。基于1000次自举法抽样的多变量逻辑回归分析，我们建立了列线图模型并推导出评分模型。1000次自举法抽样和外部验证作为最终预测模型的验证。结果：在纳入训练队列的234例患者中，其中有42例接受了重症治疗。在多因素分析中，入院时的GCS评分、中性粒细胞/淋巴细胞比值、24小时内最小SpO₂和血浆D-二聚体与不良预后显著相关。评分模型的内部和外部验证的C-指数分别为0.908 (95% CI = 0.903~0.913)和0.919 (95% CI = 0.844~0.993)。评分模型内部和外部验证的Brier得分分别为0.080和0.068，Hosmer-Lemeshow拟合优度检验不显著。结论：新建立的评分模型可用于预测自发性脑出血患者住院期间接受重症治疗的发生。对于可能需要重症治疗的高危个体，有必要提前优化医疗资源配置。

Abstract: Background and Objective: It is common for patients with spontaneous intracerebral hemorrhage (sICH) to deteriorate during hospitalization. No predictive model has been established for patients with stroke who need to receive intensive treatment during hospitalization. This study aimed to establish and validate a new model to predict the risk of stroke patients requiring severe treat-ment. Methods: A retrospective cohort study included patients with spontaneous intracerebral hemorrhage before stroke who were functionally independent. Based on the multivariate logistic regression analysis of 1000 bootstrap sampling, we build a nomogram model and derive a scoring model. 1000 bootstrap sampling and external validation serve as validation of the final prediction model. Results: Of the 234 patients enrolled in the training cohort, 42 received severe treatment. In multivariate analyses, GCS score at admission, neutrophil/lymphocyte ratio, minimum SpO₂ within 24 hours, and plasma D-dimer were significantly associated with poor prognosis. The C-indices for internal and external validation of the scoring model were 0.908 (95%CI = 0.903~0.913) and 0.919 (95% CI = 0.844~0.993), respectively. The Brier scores for internal and external validation of the scoring model were 0.080 and 0.068, respectively, and the Hosmer-Lemeshow goodness of fit test was not significant. Conclusion: The newly established scoring model can be used to predict the oc-currence of severe treatment in patients with spontaneous cerebral hemorrhage during hospitali-zation. For high-risk individuals who may require severe treatment, it is necessary to optimize the allocation of medical resources in advance.