Super learning to hedge against incorrect inference from arbitrary parametric assumptions in marginal structural modeling.
Super learning to hedge against incorrect inference from arbitrary parametric assumptions in marginal structural modeling.
复制标题
DOI：
10.1016/j.jclinepi.2013.01.016
复制
发表时间：
2013-08
影响因子：
7.2
通讯作者：
O'Connor, Patrick J.
中科院分区：
文献类型：
--
作者：
Neugebauer, Romain;Fireman, Bruce;Roy, Jason A.;Raebel, Marsha A.;Nichols, Gregory A.;O'Connor, Patrick J.
关键词：
对于许多比较有效性研究（CER）问题，临床试验不太可能启动。然而，假设的随机试验的推论可以通过使用观察数据的边际结构建模（MSM）来模拟，但成功调整时间依赖性混杂和选择偏倚通常依赖于参数建模假设。如果这些假设被违反，从MSM推断可能是不准确的。在这篇文章中，我们鼓励应用一种称为超级学习的数据自适应估计方法，以避免依赖CER中的任意参数假设。使用新发2型糖尿病成人患者的电子健康记录数据，我们采用逆概率加权估计的MSM来评估三种口服抗糖尿病治疗对肾小球滤过率恶化的影响。来自IPW估计的推断对关于暴露和删失过程与主要疑似混杂源（即，血红蛋白A1c的时间依赖性测量。超级学习成功地实现了利用现有机器学习算法的灵活混杂和选择偏差调整。由于任意和不正确的建模决策而导致的关于临床有效性的错误IPW推断可以通过超级学习来避免。
Clinical trials are unlikely to ever be launched for many Comparative Effectiveness Research (CER) questions. Inferences from hypothetical randomized trials may however be emulated with marginal structural modeling (MSM) using observational data but success in adjusting for time-dependent confounding and selection bias typically relies on parametric modeling assumptions. If these assumptions are violated, inferences from MSM may be inaccurate. In this article, we motivate the application of a data-adaptive estimation approach called Super Learning to avoid reliance on arbitrary parametric assumptions in CER. Using the electronic health records data from adults with new onset type 2 diabetes, we implemented MSM with inverse probability weighting estimation to evaluate the effect of three oral anti-diabetic therapies on the worsening of glomerular filtration rate. Inferences from IPW estimation were noticeably sensitive to the parametric assumptions about the associations between both the exposure and censoring processes and the main suspected source of confounding, i.e., time-dependent measurements of hemoglobin A1c. Super Learning was successfully implemented to harness flexible confounding and selection bias adjustment from existing machine learning algorithms. Erroneous IPW inference about clinical effectiveness due to arbitrary and incorrect modeling decisions may be avoided with Super Learning.
登录
查看更多内容
影响因子：
158.5
作者：
Nissen, Steven E.;Wolski, Kathy
通讯作者：
Wolski, Kathy
影响因子：
7
作者：
McCaffrey, DF;Ridgeway, G;Morral, AR
通讯作者：
Morral, AR
影响因子：
2.5
作者：
HOERL, AE;KENNARD, RW
通讯作者：
KENNARD, RW
影响因子：
5
作者：
Cole, Stephen R.;Hernan, Miguel A.
通讯作者：
Hernan, Miguel A.
影响因子：
2.6
作者：
Neugebauer, Romain;Fireman, Bruce;Selby, Joe V.
通讯作者：
Selby, Joe V.