Literature DB >> 33594415

Deep propensity network using a sparse autoencoder for estimation of treatment effects.

Shantanu Ghosh1, Jiang Bian2, Yi Guo2, Mattia Prosperi3.   

Abstract

OBJECTIVE: Drawing causal estimates from observational data is problematic, because datasets often contain underlying bias (eg, discrimination in treatment assignment). To examine causal effects, it is important to evaluate what-if scenarios-the so-called "counterfactuals." We propose a novel deep learning architecture for propensity score matching and counterfactual prediction-the deep propensity network using a sparse autoencoder (DPN-SA)-to tackle the problems of high dimensionality, nonlinear/nonparallel treatment assignment, and residual confounding when estimating treatment effects.
MATERIALS AND METHODS: We used 2 randomized prospective datasets, a semisynthetic one with nonlinear/nonparallel treatment selection bias and simulated counterfactual outcomes from the Infant Health and Development Program and a real-world dataset from the LaLonde's employment training program. We compared different configurations of the DPN-SA against logistic regression and LASSO as well as deep counterfactual networks with propensity dropout (DCN-PD). Models' performances were assessed in terms of average treatment effects, mean squared error in precision on effect's heterogeneity, and average treatment effect on the treated, over multiple training/test runs.
RESULTS: The DPN-SA outperformed logistic regression and LASSO by 36%-63%, and DCN-PD by 6%-10% across all datasets. All deep learning architectures yielded average treatment effects close to the true ones with low variance. Results were also robust to noise-injection and addition of correlated variables. Code is publicly available at https://github.com/Shantanu48114860/DPN-SAz. DISCUSSION AND
CONCLUSION: Deep sparse autoencoders are particularly suited for treatment effect estimation studies using electronic health records because they can handle high-dimensional covariate sets, large sample sizes, and complex heterogeneity in treatment assignments.
© The Author(s) 2021. Published by Oxford University Press on behalf of the American Medical Informatics Association. All rights reserved. For permissions, please email: journals.permissions@oup.com.

Entities:  

Keywords:  big data; biomedical informatics; causal AI; causal inference; deep learning; electronic health record; propensity score; treatment effect

Mesh:

Year:  2021        PMID: 33594415      PMCID: PMC8661404          DOI: 10.1093/jamia/ocaa346

Source DB:  PubMed          Journal:  J Am Med Inform Assoc        ISSN: 1067-5027            Impact factor:   4.497


  18 in total

1.  Results of multivariable logistic regression, propensity matching, propensity adjustment, and propensity-based weighting under conditions of nonuniform effect.

Authors:  Tobias Kurth; Alexander M Walker; Robert J Glynn; K Arnold Chan; J Michael Gaziano; Klaus Berger; James M Robins
Journal:  Am J Epidemiol       Date:  2005-12-21       Impact factor: 4.897

2.  Ischemic stroke lesion segmentation using stacked sparse autoencoder.

Authors:  G B Praveen; Anita Agrawal; Ponraj Sundaram; Sanjay Sardesai
Journal:  Comput Biol Med       Date:  2018-06-02       Impact factor: 4.589

3.  Evaluating large-scale propensity score performance through real-world and synthetic data experiments.

Authors:  Yuxi Tian; Martijn J Schuemie; Marc A Suchard
Journal:  Int J Epidemiol       Date:  2018-12-01       Impact factor: 7.196

4.  Should Health Care Demand Interpretable Artificial Intelligence or Accept "Black Box" Medicine?

Authors:  Fei Wang; Rainu Kaushal; Dhruv Khullar
Journal:  Ann Intern Med       Date:  2019-12-17       Impact factor: 25.391

5.  Explainable artificial intelligence models using real-world electronic health record data: a systematic scoping review.

Authors:  Seyedeh Neelufar Payrovnaziri; Zhaoyi Chen; Pablo Rengifo-Moreno; Tim Miller; Jiang Bian; Jonathan H Chen; Xiuwen Liu; Zhe He
Journal:  J Am Med Inform Assoc       Date:  2020-07-01       Impact factor: 4.497

6.  Methods for constructing and assessing propensity scores.

Authors:  Melissa M Garrido; Amy S Kelley; Julia Paris; Katherine Roza; Diane E Meier; R Sean Morrison; Melissa D Aldridge
Journal:  Health Serv Res       Date:  2014-04-30       Impact factor: 3.402

7.  High-dimensional propensity score adjustment in studies of treatment effects using health care claims data.

Authors:  Sebastian Schneeweiss; Jeremy A Rassen; Robert J Glynn; Jerry Avorn; Helen Mogun; M Alan Brookhart
Journal:  Epidemiology       Date:  2009-07       Impact factor: 4.822

8.  Deep learning enables rapid identification of potent DDR1 kinase inhibitors.

Authors:  Alex Zhavoronkov; Yan A Ivanenkov; Alex Aliper; Mark S Veselov; Vladimir A Aladinskiy; Anastasiya V Aladinskaya; Victor A Terentiev; Daniil A Polykovskiy; Maksim D Kuznetsov; Arip Asadulaev; Yury Volkov; Artem Zholus; Rim R Shayakhmetov; Alexander Zhebrak; Lidiya I Minaeva; Bogdan A Zagribelnyy; Lennart H Lee; Richard Soll; David Madge; Li Xing; Tao Guo; Alán Aspuru-Guzik
Journal:  Nat Biotechnol       Date:  2019-09-02       Impact factor: 54.908

9.  A comparison of 12 algorithms for matching on the propensity score.

Authors:  Peter C Austin
Journal:  Stat Med       Date:  2013-10-07       Impact factor: 2.373

10.  Big data hurdles in precision medicine and precision public health.

Authors:  Mattia Prosperi; Jae S Min; Jiang Bian; François Modave
Journal:  BMC Med Inform Decis Mak       Date:  2018-12-29       Impact factor: 2.796

View more

北京卡尤迪生物科技股份有限公司 © 2022-2023.