当前位置:首页 > 医学进展 > 心血管 > 正文

[ESC2017]心梗患者的风险预测,原来还可以这么简单!

——CAMI最新数据:中国人专属的STEMI死亡风险预测模型


整理报道,未经授权请勿转载。


中国急性心肌梗死注册登记研究(CAMI)是由国家心血管病中心、中国医学科学院阜外医院牵头的一项真实世界的前瞻性多中心研究,是目前国内最大规模的急性心肌梗死注册研究,属十二五科技部支撑项目。在ESC2017年会上,共有3项CAMI研究的最新数据亮相,阜外医院杨进刚教授提交的壁报“基于机器学习方法的中国ST段抬高型心肌梗死(STEMI)住院死亡率的简单风险预测模型”是其中之一。


杨进刚教授指出,基于真实世界的患者资料建立死亡风险预测模型对改善STEMI治疗至关重要,但目前临床上常用的TIMI评分、GRACE评分等评估工具并不适合真实世界的中国STEMI患者。本研究采用机器学习方法开发和验证了一种新型、简单实用的STEMI住院死亡率风险预测模型,该模型与TIMI评分的预测效果相似。


研究纳入了2013年1月至2016年1月期间,CAMI研究中的28 705名STEMI住院患者(推导集18 744例,验证集9961例)。此外,还在China PEACE研究的5849名STEMI患者中进行了外部验证。


CAMI-STEMI评分得到了7个日常临床实践中容易获得的变量:女性(1分),心率(HR)≥100bmp(2分),年龄≥70岁(2分),收缩压≤115mmHg(2分),Killip分级> 1(2分),心跳骤停(4分)和前壁梗死(1分)。


CAMI-STEMI评分的预测精度与TIMI评分相似:CAMI-STEMI评分在推到队列、验证队列和外部验证队列中的C统计量分别为0.789、0.771和0.772,而TIMI评分分别为0.778、0.758和0.780。此外,采用CAMI-STEMI评分将患者分为低、中、高危对死亡风险的预测能力优于GRACE评分。


图 不同CAMI-STEMI评分患者的住院死亡风险


与同类的其他评分相比,CAMI-STEMI评分简单实用,不需要抽血化验及详细询问病史,在中国STEMI患者住院死亡率的预测精度与TIMI评分、GRACE评分相似。


杨进刚教授介绍,目前该评分正在进一步优化,去掉心跳骤停和前壁梗死,已缩减到5个变量。我们期待详细研究结果的早日发表。


热点专题>>>2017年欧洲心脏病学会年会(ESC2017)


【Abstract: P5582】


Simple risk prediction model to assess hospital mortality in Chinese patients with ST elevation myocardial infarction based on a machine learning approach: from China acute myocardial infarction (CAMI)


Authors:J.G. Yang1, H.F. Liu2, H.Y. Xu1, X. Li2, W. Li1, Y. Wang1, Y.J. Yang1, 1Cardiovascular Institute & Fuwai Hospital - Beijing - China People's Republic of, 2IBM Research - China, Cognitive HealthCare - Beijing - China People's Republic of,


On behalf: China Acute Myocardial Infarction (CAMI) Registry research group

Background: Establishment of risk prediction model is crucial for improving care and should be population dependent and reflect true life populations. The commonly used tools for assessing hospital mortality for patients with ST Elevation Myocardial Infarction (STEMI) may not be suitable to real world Chinese STEMI patients.


Purpose: The objective of this study is to develop and validate a practical and simple risk prediction model for in-hospital mortality in Chinese STEMI patients using a machine learning approach and have comparable risk prediction performance with TIMI risk prediction score.


Method: 28 705 patients admitted with STEMI (18 744 in derivation set, 9961 in validation set) at hospitals participating in the China Acute Myocardial Infarction (CAMI) registry were included between Jan. 2013 and Jan. 2016. A machine learning approach was adopted to risk stratify subjects, identify discriminating rules between strata, and adapt these rules to develop a risk prediction score for in-hospital mortality among Chinese STEMI patients. A decision tree method (C5.0) was used to identify discriminating rules between the risk strata.


Results: The new risk assessment tool were composed of seven variables that were readily available in routine clinical practice: female (1 point), heart rate (HR) ≥100 bmp (2 points), age ≥70 years (2 points), systolic blood pressure ≤115 mmHg (2 points), Killip class >1 (2 points), cardiac arrest (4 points), and anterior wall infarction (1 point). The prediction accuracy of the resulting score, or CAMI scores (with AUC of 0.783 and 0.766 respectively in the derivation and validation cohorts) was similar with TIMI (AUC of 0.769 and 0.750 respectively).


Conclusions: CAMI scores can predict in-hospital mortality among Chinese STEMI patients with similar performance to the well-established TIMI score, while relying solely on simple and practical variables, that is, do not need to ask the history and draw blood to checkup biochemistry.

 

阅读次数:  

发表评论