目的 探讨凝血纤溶标志物在脑出血（ICH）患者预后中的预测价值，并构建多因素logistic回归模型以实现个体化风险预测。 方法 回顾性纳入2020年1月至2023年12月就诊于南方医科大学南方医院的101例ICH患者，依据出院与入院GCS评分差值（ΔGCS）将患者分为预后不良组（ΔGCS ≤ 0）与预后良好组（ΔGCS > 0）。收集患者入院时的凝血纤溶指标及基础资料，采用LASSO回归筛选变量，70%的样本构建logistic回归模型，30%的样本用于验证。通过ROC曲线、校准曲线、Hosmer-Lemeshow检验及决策曲线分析评估模型的预测效能。 结果 单因素分析提示凝血酶-抗凝血酶复合物（TAT）、D-二聚体（D-D）及年龄在不同预后组间存在显著差异。LASSO回归筛选出TAT、D-D和年龄为关键变量并纳入logistic回归模型。模型表达式为：logit（P） = -6.234 + 1.132 × TAT + 0.867 × D-D + 0.699 ×年龄。在训练集的AUC为0.795，校准曲线显示良好的拟合度，Hosmer-Lemeshow检验P = 0.8568，DCA显示模型在0.1 ~ 0.8风险阈值范围内具有稳定净获益。 结论 TAT、D-D和年龄是ICH患者不良预后的独立危险因素。基于上述变量构建的预测模型具有良好的判别性与临床实用性，列线图可直观实现个体化风险评估，为ICH患者的早期风险分层与临床干预提供支持。

Objective To explore the prognostic implications of coagulation-fibrinolysis biomarkers in intracerebral hemorrhage （ICH） and to construct a multivariable logistic regression model for individualized risk prediction. Methods A total of 101 ICH patients who were admitted to Nanfang Hospital of Southern Medical University from January 2020 to December 2023 were retrospectively enrolled. These patients were stratified into a poor outcome group （ΔGCS ≤ 0） and a good outcome group （ΔGCS > 0） according to the difference in Glasgow Coma Scale （GCS） scores between discharge and admission. Coagulation and fibrinolysis markers collected upon admission were analyzed. The Least Absolute Shrinkage and Selection Operator （LASSO） regression was employed to screen variables. A logistic regression model was constructed using 70% of the cases （the training set）， while the remaining 30% were utilized for validation. The performance of the model was evaluated through receiver operating characteristic （ROC） curves， calibration plots， Hosmer-Lemeshow goodness-of-fit test， and decision curve analysis （DCA）. Results Univariate analysis indicated that thrombin-antithrombin complex （TAT）， D-dimer， and age exhibited significant differences between the two outcome groups （P < 0.05）. These three variables were selected via LASSO regression and incorporated into the logistic model. The final model equation was expressed as： logit（P） = -6.234 + 1.132 × TAT + 0.867 × D-dimer + 0.699 × Age. In the training set， the area under the ROC curve （AUC） was 0.795. The calibration curve demonstrated excellent agreement between the predicted and observed outcomes， with a Hosmer-Lemeshow test P-value of 0.8568. DCA revealed that the model achieved net clinical benefit across a broad range of risk thresholds （0.1 ~ 0.8）. Conclusions TAT， D-dimer， and age are independent predictors of poor prognosis in patients with ICH. The logistic regression model based on these variables demonstrates favorable discriminatory ability and clinical utility. The nomogram derived from this model enables individualized risk assessment and may aid clinicians in early prognostic evaluation and treatment planning.