Abstract: Objective  To systematically evaluate the prediction models of symptomatic intracerebral hemorrhage after intravenous thrombolysis in acute ischemic stroke and to provide a reference for clinical decision-making of intravenous thrombolysis.
Methods  CNKI, WanFang Data, VIP, PubMed, Embase, Web of Science, and The Cochrane Library databases were searched to collect relevant studies, with a timeframe for searching the database from the establishment of the database to December 18, 2022. Two researchers independently screened the literature, extracted the data, assessed the risk of bias, and then systematically evaluated the basic characteristics and methodology of the included models. 
Results  Twenty studies were included, with a total of 30 prediction models. The AUC of the included models ranged from 0.42 to 0.94. The overall predictive performance of the 24 (80%) prediction models was good, with significant differences in model differentiation across outcome definitions and algorithms. The most common predictors were NIHSS score, age, imaging signs or scores of infarction, blood glucose, systolic blood pressure, and antiplatelet drugs.
Conclusions  In recent years, the prediction models of symptomatic intracerebral hemorrhage after intravenous thrombolysis in acute ischemic stroke have shown the characteristics of diversified modeling algorithms, better model performance, and diversified predictors. However, the overall risk of bias was high, and future studies need to calibrate the models further. More attention should be paid to updating and externally validating the models to improve extrapolation and clinical utility and utilize the models’ clinical practice implications.

Key words: Acute ischemic stroke; Intravenous thrombolysis; Symptomatic intracerebral hemorrhage; Prediction model; Systematic review

摘要： 目的 系统评价急性缺血性卒中静脉溶栓后症状性颅内出血预测模型的特征，为静脉溶栓临床决策提供参考。
方法 检索中国知网、万方数据知识服务平台、维普资讯、PubMed、Embase、Web of Science和The Cochrane Library数据库，收集急性缺血性卒中静脉溶栓后症状性颅内出血预测模型相关研究信息，检索时限为建库至2022年12月18日。由2位研究员独立筛选文献、提取资料并评价偏倚风险后对纳入模型的基本特征和方法论进行系统评价。 
结果 纳入20项研究，共30个预测模型。纳入模型ROC曲线的AUC值范围为0.42～0.94。24个（80%）预测模型的整体预测性能较好，模型对不同结局的定义和算法的区分度有明显差异。最常见的预测因子包括NIHSS评分、年龄、梗死的影像学征象或评分、血糖、收缩压和抗血小板药物。
结论 急性缺血性卒中静脉溶栓后症状性颅内出血预测模型呈现出建模算法多样化、模型性能更佳化、预测因素多元化等特点，但总体偏倚风险较高，未来研究还需要进一步校准模型。此外，应更加关注模型的更新与外部验证，提高其外推性及临床效用，从而发挥模型更大的临床价值。

关键词: 急性缺血性卒中; 静脉溶栓; 症状性颅内出血; 预测模型; 系统评价