广东工业大学学报 ›› 2024, Vol. 41 ›› Issue (01): 41-46.doi: 10.12052/gdutxb.230157

• 智慧医疗 • 上一篇    下一篇

基于工具变量的丁苯酞-急性缺血性卒中的因果效应评估

林容基1, 陈薇1, 黄志新2, 蔡瑞初1   

  1. 1. 广东工业大学 计算机学院, 广东 广州 510006;
    2. 广东省第二人民医院 神经内科, 广东 广州 510317
  • 收稿日期:2023-10-12 出版日期:2024-01-25 发布日期:2024-02-01
  • 通信作者: 蔡瑞初(1983–)，男，教授，博士生导师，主要研究方向为因果关系、机器学习、数据挖掘，E-mail：cairuichu@gmail.com
  • 作者简介:林容基(1998–)，男，硕士研究生，主要研究方向为因果关系
  • 基金资助:
    国家自然科学基金资助项目(61876043，61976052)；科技创新2030——“新一代人工智能”重大项目(2021ZD0111501)；国家优秀青年科学基金资助项目(62122022)；广州市科技项目(202201020359)

Assessment of Causal Effects of Butylphthalide-acute Ischemic Stroke Based on Instrumental Variables

Lin Rong-ji1, Chen Wei1, Huang Zhi-xin2, Cai Rui-chu1   

  1. 1. School of Computer Science and Technology, Guangdong University of Technology, Guangzhou 510006, China;
    2. Department of Neurology, Guangdong Second Provincial Hospital, Guangzhou 510317, China
  • Received:2023-10-12 Online:2024-01-25 Published:2024-02-01

HTML

PDF (PC)

393

摘要： 因果效应分析在临床统计中是一种常见的研究方法，其通常基于观察数据进行分析。然而，在使用观察数据进行因果效应分析时，常受到未观测变量的影响，从而使因果效应评估出现偏差。当无法忽略未观测变量带来的偏差或无法找到适当的代理变量来削弱这种偏差时，传统方法无法提供可靠的因果效应估计。为了解决这一问题，本文采用工具变量法，在临床统计的药效分析领域提出一种比传统方法更加准确的计算方法，将未观测变量的影响纳入误差项，以实现准确的因果效应估计。通过将观察数据中满足特定假设的变量作为工具变量，计算了丁苯酞(一种药物)对急性缺血性卒中(Acute Ischemic Stroke，AIS)患者在存在未观测变量的情况下，其3个月预后的因果效应，并评估了该因果估计量的置信区间。研究结果揭示了丁苯酞对急性缺血性卒中患者的预后恢复具有明显的积极作用。

关键词: 未观测变量, 工具变量, 丁苯酞, 因果效应估计

Abstract: Causal effect analysis is an important and popular method in clinical statistics, which typically conducted based on the observational data. However, the analysis based on observed data may be affected by unobserved variables, which may produce bias, leading to estimate the causal effects inaccurately. Existing methods ignore the unobserved variables or are unable to find appropriate proxy variables to weaken this bias, which fail to provide reliable estimation of the causal effects. To address this problem, this paper proposes the instrumental variable method, a more accurate computational method than traditional approaches in the realm of clinical statistic for drug efficacy analysis. This method incorporates the effect of unobserved into the error term, to estimate the accurate causal effect. Under some mild assumptions, the variables in the observational data is considered as instrumental variables. Then, the proposed method calculates the effects of butylphthalide (i.e., a drug) on patients with acute ischemic stroke (AIS) in the presence of unobserved variables. The causal effect of monthly prognosis and the confidence interval of this causal estimator is estimated. The study results show that the butylphthalide has a significant positive effect on the prognostic recovery of patients with acute ischemic stroke.

Key words: unobserved variable, instrumental variable, butylphthalide, causal effect estimation

中图分类号: 

  • TP399
[1] YAO L, CHU Z, LI S, et al. A survey on causal inference [J]. ACM Transactions on Knowledge Discovery from Data (TKDD) , 2021, 15(5): 1-46.
[2] IMBENS G W, RUBIN D B. Causal inference in statistics, social, and biomedical sciences[M]. Cambridge: Cambridge University Press, 2015.
[3] BRICK J M, KALTON G. Handling missing data in survey research [J]. Statistical Methods in Medical Research, 1996, 5(3): 215-238.
[4] ABADIE A, DRUKKER D, HERR J L, et al. Implementing matching estimators for average treatment effects in Stata [J]. The Stata Journal, 2004, 4(3): 290-311.
[5] LOH W Y. Classification and regression trees [J]. WIREs:Data Mining and Knowledge Discovery, 2011, 1(1): 14-23.
[6] BENGIO Y, COURVILLER A, VINCENT P. Representation learning: a review and new perspectives [J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2013, 35(8): 1798-1828.
[7] CARUANA R. Multitask learning [J]. Machine Learning, 1997, 28(1): 41-75.
[8] HOSPEDALES T, ANTONIOU A, MICAELLI P, et al. Meta-learning in neural networks: a survey [J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2021, 44(9): 5149-5169.
[9] BAIOCCHI M, CHENG J, SMALL D S. Instrumental variable methods for causal inference [J]. Statistics in Medicine, 2014, 33(13): 2297-2340.
[10] MARTENS E P, PESTMAN W R, DE BOER A, et al. Instrumental variables: application and limitations [J]. Epidemiology, 2006, 17(3): 260-267.
[11] OHLSSON H, KENDLER K S. Applying causal inference methods in psychiatric epidemiology: a review [J]. JAMA Psychiatry, 2020, 77(6): 637-644.
[12] BROOKHART M A, SCHNEEWEISS S. Preference-based instrumental variable methods for the estimation of treatment effects: assessing validity and interpreting results [J]. The International Journal of Biostatistics, 2007, 3(1): 14.
[13] MUKAMAL K J, STAMPFER M J, RIMM E B. Genetic instrumental variable analysis: time to call mendelian randomization what it is. The example of alcohol and cardiovascular disease [J]. European Journal of Epidemiology, 2020, 35: 93-97.
[14] DAVEY SMITH G, EBRAHIM S. ‘Mendelian randomization’: can genetic epidemiology contribute to understanding environmental determinants of disease? [J]. International Journal of Epidemiology, 2003, 32(1): 1-22.
[15] RASSEN J A, SCHNEEWEISS S, GLYNN R J, et al. Instrumental variable analysis for estimation of treatment effects with dichotomous outcomes [J]. American Journal of Epidemiology, 2009, 169(3): 273-284.
[16] 伍德里奇. 计量经济学导论: 现代方法[M]. 北京: 清华大学出版社, 2017.
[17] KIM H Y. Statistical notes for clinical researchers: risk difference, risk ratio, and odds ratio [J]. Restorative Dentistry & Endodontics, 2017, 42(1): 72-76.
[18] BROOKHART M A, WANG P, SOLOMON D H, et al. Evaluating short-term drug effects using a physician-specific prescribing preference as an instrumental variable [J]. Epidemiology, 2006, 17(3): 268.
[19] MCCOY C E. Understanding the intention-to-treat principle in randomized controlled trials [J]. Western Journal of Emergency Medicine, 2017, 18(6): 1075.
[20] MA H, ROBINS J M. Instruments for causal inference: an epidemiologist's dream? [J]. Epidemiology, 2006, 17(4): 360-372.
[1] 吴菊华, 李俊锋, 陶雷. 基于知识图谱嵌入与深度学习的药物不良反应预测[J]. 广东工业大学学报, 2024, 41(01): 19-26,40.
[2] 冯广, 潘庭锋, 伍文燕. 基于贝叶斯网络模型的在线学习行为分析[J]. 广东工业大学学报, 2022, 39(03): 41-48.
[3] 梁轰, 冯丽, 徐方鑫, 李光程, 周郭许. 算力网络中一种新颖的看门狗故障检测协议[J]. 广东工业大学学报, 2021, 38(06): 35-46.
[4] 刘瑞雪, 曾碧, 汪明慧, 卢智亮. 一种基于高效边界探索的机器人自主建图方法[J]. 广东工业大学学报, 2020, 37(05): 38-45.
[5] 李卫华, 李志猛. 基于大数据运输集团生产运营决策系统的构建及应用[J]. 广东工业大学学报, 2018, 35(03): 113-118.
[6] 范锐, 颜思伟, 彭中煌, 廖永乐, 陈月峰, 罗小行, 林恒, 谭治. 可拓策略生成软件架构及其应用研究[J]. 广东工业大学学报, 2017, 34(02): 1-5.
Viewed
Full text


Abstract

Cited
  Shared   
  Discussed   
No Suggested Reading articles found!
版权所有 © 2017 《广东工业大学学报》编辑部
地址：广州市东风东路729号广东工业大学 邮编：510090
电话：020-37626653,020-37626139,020-37626396
邮箱：xbzrb@gdut.edu.cn
本系统由北京玛格泰克科技发展有限公司设计开发