{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,3,13]],"date-time":"2026-03-13T09:09:52Z","timestamp":1773392992951,"version":"3.50.1"},"reference-count":48,"publisher":"Oxford University Press (OUP)","issue":"6","license":[{"start":{"date-parts":[[2021,2,16]],"date-time":"2021-02-16T00:00:00Z","timestamp":1613433600000},"content-version":"vor","delay-in-days":0,"URL":"https:\/\/academic.oup.com\/journals\/pages\/open_access\/funder_policies\/chorus\/standard_publication_model"}],"funder":[{"DOI":"10.13039\/100000002","name":"NIH","doi-asserted-by":"publisher","award":["R01AI145552"],"award-info":[{"award-number":["R01AI145552"]}],"id":[{"id":"10.13039\/100000002","id-type":"DOI","asserted-by":"publisher"}]},{"DOI":"10.13039\/100000002","name":"NIH","doi-asserted-by":"publisher","award":["R01CA246418"],"award-info":[{"award-number":["R01CA246418"]}],"id":[{"id":"10.13039\/100000002","id-type":"DOI","asserted-by":"publisher"}]},{"DOI":"10.13039\/100000002","name":"NIH","doi-asserted-by":"publisher","award":["U18DP006512"],"award-info":[{"award-number":["U18DP006512"]}],"id":[{"id":"10.13039\/100000002","id-type":"DOI","asserted-by":"publisher"}]},{"DOI":"10.13039\/100000002","name":"NIH","doi-asserted-by":"publisher","award":["R21AG068717"],"award-info":[{"award-number":["R21AG068717"]}],"id":[{"id":"10.13039\/100000002","id-type":"DOI","asserted-by":"publisher"}]},{"DOI":"10.13039\/100000002","name":"NIH","doi-asserted-by":"publisher","award":["R21CA245858"],"award-info":[{"award-number":["R21CA245858"]}],"id":[{"id":"10.13039\/100000002","id-type":"DOI","asserted-by":"publisher"}]}],"content-domain":{"domain":[],"crossmark-restriction":false},"short-container-title":[],"published-print":{"date-parts":[[2021,6,12]]},"abstract":"<jats:title>Abstract<\/jats:title><jats:sec><jats:title>Objective<\/jats:title><jats:p>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\u2014the so-called \u201ccounterfactuals.\u201d We propose a novel deep learning architecture for propensity score matching and counterfactual prediction\u2014the deep propensity network using a sparse autoencoder (DPN-SA)\u2014to tackle the problems of high dimensionality, nonlinear\/nonparallel treatment assignment, and residual confounding when estimating treatment effects.<\/jats:p><\/jats:sec><jats:sec><jats:title>Materials and Methods<\/jats:title><jats:p>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\u2019s 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\u2019 performances were assessed in terms of average treatment effects, mean squared error in precision on effect\u2019s heterogeneity, and average treatment effect on the treated, over multiple training\/test runs.<\/jats:p><\/jats:sec><jats:sec><jats:title>Results<\/jats:title><jats:p>The DPN-SA outperformed logistic regression and LASSO by 36%\u201363%, and DCN-PD by 6%\u201310% 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.<\/jats:p><\/jats:sec><jats:sec><jats:title>Discussion and Conclusion<\/jats:title><jats:p>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.<\/jats:p><\/jats:sec>","DOI":"10.1093\/jamia\/ocaa346","type":"journal-article","created":{"date-parts":[[2020,12,29]],"date-time":"2020-12-29T04:12:24Z","timestamp":1609215144000},"page":"1197-1206","source":"Crossref","is-referenced-by-count":13,"title":["Deep propensity network using a sparse autoencoder for estimation of treatment effects"],"prefix":"10.1093","volume":"28","author":[{"given":"Shantanu","family":"Ghosh","sequence":"first","affiliation":[{"name":"Department of Computer and Information Science and Engineering, University of Florida, Gainesville, Florida, USA"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"ORCID":"https:\/\/orcid.org\/0000-0002-2238-5429","authenticated-orcid":false,"given":"Jiang","family":"Bian","sequence":"additional","affiliation":[{"name":"Department of Health Outcomes and Biomedical Informatics, College of Medicine, University of Florida, Gainesville, Florida, USA"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Yi","family":"Guo","sequence":"additional","affiliation":[{"name":"Department of Health Outcomes and Biomedical Informatics, College of Medicine, University of Florida, Gainesville, Florida, USA"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"ORCID":"https:\/\/orcid.org\/0000-0002-9021-5595","authenticated-orcid":false,"given":"Mattia","family":"Prosperi","sequence":"additional","affiliation":[{"name":"Department of Epidemiology, College of Public Health and Health Professions & College of Medicine, University of Florida, Gainesville, Florida, USA"}],"role":[{"role":"author","vocabulary":"crossref"}]}],"member":"286","published-online":{"date-parts":[[2021,2,16]]},"reference":[{"issue":"7","key":"2021061318595285400_ocaa346-B1","doi-asserted-by":"crossref","first-page":"369","DOI":"10.1038\/s42256-020-0197-y","article-title":"Causal inference and counterfactual prediction in machine learning for actionable healthcare","volume":"2","author":"Prosperi","year":"2020","journal-title":"Nat Mach Intell"},{"key":"2021061318595285400_ocaa346-B2","author":"Sibbald B, Roland","year":"1998; 316 (7126): 201"},{"issue":"25","key":"2021061318595285400_ocaa346-B3","doi-asserted-by":"crossref","first-page":"1887","DOI":"10.1056\/NEJM200006223422507","article-title":"Randomized, controlled trials, observational studies, and the hierarchy of research designs","volume":"342","author":"Concato","year":"2000","journal-title":"N Engl J Med"},{"issue":"1","key":"2021061318595285400_ocaa346-B4","doi-asserted-by":"crossref","first-page":"41","DOI":"10.1093\/biomet\/70.1.41","article-title":"The central role of the propensity score in observational studies for causal effects","volume":"70","author":"Rosenbaum","year":"1983","journal-title":"Biometrika"},{"key":"2021061318595285400_ocaa346-B5","author":"Pearl","year":"2016"},{"key":"2021061318595285400_ocaa346-B6","article-title":"Causal Inference: What if. Boca Raton, FL: Chapman & Hill\/CRC;","author":"Hern\u00e1n","year":"2010"},{"issue":"1","key":"2021061318595285400_ocaa346-B7","doi-asserted-by":"crossref","first-page":"189","DOI":"10.1146\/annurev.publhealth.22.1.189","article-title":"Confounding in health research","volume":"22","author":"Greenland","year":"2001","journal-title":"Annu Rev Public Health"},{"issue":"387","key":"2021061318595285400_ocaa346-B8","doi-asserted-by":"crossref","first-page":"516","DOI":"10.1080\/01621459.1984.10478078","article-title":"Reducing bias in observational studies using subclassification on the propensity score","volume":"79","author":"Rosenbaum","year":"1984","journal-title":"J Am Stat Assoc"},{"key":"2021061318595285400_ocaa346-B9","author":"Garrido"},{"key":"2021061318595285400_ocaa346-B10","doi-asserted-by":"crossref","DOI":"10.1093\/aje\/kwj047","article-title":"Results of multivariable logistic regression, propensity matching, propensity adjustment, and propensity-based weighting under conditions of nonuniform effect","author":"Kurth","year":"2006","journal-title":"Am J Epidemiol"},{"issue":"1","key":"2021061318595285400_ocaa346-B11","doi-asserted-by":"crossref","DOI":"10.1186\/s12911-018-0719-2","article-title":"Big data hurdles in precision medicine and precision public health","volume":"18","author":"Prosperi","year":"2018","journal-title":"BMC Med Inform Decis Mak"},{"key":"2021061318595285400_ocaa346-B12","doi-asserted-by":"crossref","DOI":"10.1097\/EDE.0b013e3181a663cc","article-title":"High-dimensional propensity score adjustment in studies of treatment effects using health care claims data","author":"Schneeweiss","year":"2009","journal-title":"Epidemiology"},{"key":"2021061318595285400_ocaa346-B13","doi-asserted-by":"crossref","DOI":"10.1093\/ije\/dyy120","article-title":"Evaluating large-scale propensity score performance through real-world and synthetic data experiments","author":"Tian","year":"2018","journal-title":"Int J Epidemiol"},{"issue":"1","key":"2021061318595285400_ocaa346-B14","doi-asserted-by":"crossref","first-page":"115","DOI":"10.1007\/BF00993164","article-title":"Approximation and estimation bounds for artificial neural networks","volume":"14","author":"Barron","year":"1994","journal-title":"Mach Learn"},{"key":"2021061318595285400_ocaa346-B15","author":"Bishop"},{"key":"2021061318595285400_ocaa346-B16","doi-asserted-by":"crossref","DOI":"10.1162\/neco.1996.8.1.164","article-title":"Neural Networks for Optimal Approximation of Smooth and Analytic Functions","author":"Mhaskar","year":"1996","journal-title":"Neural Comput"},{"key":"2021061318595285400_ocaa346-B17","author":"Alaa","year":"2017"},{"key":"2021061318595285400_ocaa346-B18","author":"Shi"},{"key":"2021061318595285400_ocaa346-B19","author":"Makhzani","year":"2014"},{"issue":"5","key":"2021061318595285400_ocaa346-B20","doi-asserted-by":"crossref","first-page":"688","DOI":"10.1037\/h0037350","article-title":"Estimating causal effects of treatments in randomized and nonrandomized studies","volume":"66","author":"Rubin","year":"1974","journal-title":"J Educ Psychol"},{"issue":"469","key":"2021061318595285400_ocaa346-B21","doi-asserted-by":"crossref","first-page":"322","DOI":"10.1198\/016214504000001880","article-title":"Causal inference using potential outcomes: design, modeling, decisions","volume":"100","author":"Rubin","year":"2005","journal-title":"J Am Stat Assoc"},{"issue":"3","key":"2021061318595285400_ocaa346-B22","doi-asserted-by":"crossref","first-page":"706","DOI":"10.1093\/biomet\/87.3.706","article-title":"The role of the propensity score in estimating dose-response functions","volume":"87","author":"Imbens","year":"2000","journal-title":"Biometrika"},{"key":"2021061318595285400_ocaa346-B23","doi-asserted-by":"publisher","author":"Lechner","year":"2001","DOI":"10.1007\/978-3-642-57615-7_3"},{"key":"2021061318595285400_ocaa346-B24","author":"Peters"},{"issue":"6","key":"2021061318595285400_ocaa346-B25","doi-asserted-by":"crossref","first-page":"1057","DOI":"10.1002\/sim.6004","article-title":"A comparison of 12 algorithms for matching on the propensity score","volume":"33","author":"Austin","year":"2014","journal-title":"Stat Med"},{"issue":"2","key":"2021061318595285400_ocaa346-B26","doi-asserted-by":"crossref","first-page":"781","DOI":"10.3982\/ECTA11293","article-title":"Matching on the estimated propensity Score","volume":"84","author":"Abadie","year":"2016","journal-title":"Econometrica"},{"key":"2021061318595285400_ocaa346-B27","first-page":"1050","author":"Gal"},{"key":"2021061318595285400_ocaa346-B28","article-title":"Dropout: a simple way to prevent neural networks from overfitting","author":"Srivastava","year":"2014","journal-title":"J Mach Learn Res"},{"key":"2021061318595285400_ocaa346-B29","first-page":"1601","article-title":"Computable eligibility criteria through ontology-driven data access: a case study of hepatitis C virus trials","volume":"2018","author":"Zhang","year":"2018","journal-title":"AMIA Annu Symp Proc"},{"key":"2021061318595285400_ocaa346-B30","first-page":"1601","article-title":"Assessing the Validity of a a priori patient-trial generalizability score using real-world data from a large clinical data research network: a colorectal cancer clinical trial case study","volume":"2019","author":"Li","year":"2020","journal-title":"AMIA. Annu Symp Proceedings AMIA Symp"},{"issue":"5786","key":"2021061318595285400_ocaa346-B31","doi-asserted-by":"crossref","first-page":"504","DOI":"10.1126\/science.1127647","article-title":"Reducing the dimensionality of data with neural networks","volume":"313","author":"Hinton","year":"2006","journal-title":"Science"},{"key":"2021061318595285400_ocaa346-B32","article-title":"Autoencoders, minimum description length and helmholtz free energy","author":"Hinton","year":"1993","journal-title":"Proceedings of the 6th International Conference on Neural Information Processing Systems"},{"key":"2021061318595285400_ocaa346-B33","author":"Zhou"},{"key":"2021061318595285400_ocaa346-B34","author":"Cho","year":"2015"},{"issue":"9","key":"2021061318595285400_ocaa346-B35","doi-asserted-by":"crossref","first-page":"1038","DOI":"10.1038\/s41587-019-0224-x","article-title":"Deep learning enables rapid identification of potent DDR1 kinase inhibitors","volume":"37","author":"Zhavoronkov","year":"2019","journal-title":"Nat Biotechnol"},{"issue":"7","key":"2021061318595285400_ocaa346-B36","doi-asserted-by":"crossref","first-page":"e6098","DOI":"10.1371\/journal.pone.0006098","article-title":"Deconvolution of blood microarray data identifies cellular activation patterns in systemic lupus erythematosus","volume":"4","author":"Abbas","year":"2009","journal-title":"PLoS One"},{"key":"2021061318595285400_ocaa346-B37","article-title":"Multitask autoencoder model for recovering human poses","author":"Yu","year":"2018","journal-title":"IEEE Trans Ind Electron"},{"key":"2021061318595285400_ocaa346-B38","doi-asserted-by":"crossref","first-page":"1","DOI":"10.1155\/2018\/3078374","article-title":"Feature representation using deep autoencoder for lung nodule image classification","volume":"2018","author":"Mao","year":"2018","journal-title":"Complexity"},{"key":"2021061318595285400_ocaa346-B39","doi-asserted-by":"crossref","first-page":"38","DOI":"10.1016\/j.compbiomed.2018.05.027","article-title":"Ischemic stroke lesion segmentation using stacked sparse autoencoder","volume":"99","author":"Praveen","year":"2018","journal-title":"Comput Biol Med"},{"issue":"1","key":"2021061318595285400_ocaa346-B40","doi-asserted-by":"crossref","first-page":"146","DOI":"10.1186\/s12859-020-3465-2","article-title":"PathME: pathway based multi-modal sparse autoencoders for clustering of patient-level multi-omics data","volume":"21","author":"Lemsara","year":"2020","journal-title":"BMC Bioinformatics"},{"key":"2021061318595285400_ocaa346-B41","author":"Kingma","year":"6980"},{"key":"2021061318595285400_ocaa346-B42","doi-asserted-by":"crossref","DOI":"10.1198\/jcgs.2010.08162","article-title":"Bayesian nonparametric modeling for causal inference","author":"Hill","year":"2011","journal-title":"J Comput Graph Stat"},{"key":"2021061318595285400_ocaa346-B43","first-page":"604","author":"LaLonde","year":"1986"},{"issue":"3","key":"2021061318595285400_ocaa346-B44","doi-asserted-by":"crossref","first-page":"239","DOI":"10.1023\/A:1024068626366","article-title":"Inference for the generalization error","volume":"52","author":"Nadeau","year":"2003","journal-title":"Mach Learn"},{"key":"2021061318595285400_ocaa346-B45","doi-asserted-by":"crossref","DOI":"10.1093\/jamia\/ocaa053","article-title":"Explainable artificial intelligence models using real-world electronic health record data: a systematic scoping review","author":"Payrovnaziri","year":"2020","journal-title":"J Am Med Inform Assoc"},{"key":"2021061318595285400_ocaa346-B46","doi-asserted-by":"crossref","DOI":"10.7326\/M19-2548","article-title":"Should health care demand interpretable artificial intelligence or accept \u2018black Box\u2019 Medicine?","author":"Wang","year":"2020","journal-title":"Ann Intern Med"},{"key":"2021061318595285400_ocaa346-B47","author":"Yoon","year":"2018"},{"key":"2021061318595285400_ocaa346-B48","first-page":"129","volume-title":"Limits of estimating heterogeneous treatment effects: Guidelines for practical algorithm design. In: Proceedings of the 35th International Conference on Machine Learning, PMLR"}],"container-title":["Journal of the American Medical Informatics Association"],"original-title":[],"language":"en","link":[{"URL":"http:\/\/academic.oup.com\/jamia\/article-pdf\/28\/6\/1197\/38615544\/ocaa346.pdf","content-type":"application\/pdf","content-version":"vor","intended-application":"syndication"},{"URL":"http:\/\/academic.oup.com\/jamia\/article-pdf\/28\/6\/1197\/38615544\/ocaa346.pdf","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2024,8,20]],"date-time":"2024-08-20T08:36:34Z","timestamp":1724142994000},"score":1,"resource":{"primary":{"URL":"https:\/\/academic.oup.com\/jamia\/article\/28\/6\/1197\/6139936"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2021,2,16]]},"references-count":48,"journal-issue":{"issue":"6","published-online":{"date-parts":[[2021,2,16]]},"published-print":{"date-parts":[[2021,6,12]]}},"URL":"https:\/\/doi.org\/10.1093\/jamia\/ocaa346","relation":{},"ISSN":["1527-974X"],"issn-type":[{"value":"1527-974X","type":"electronic"}],"subject":[],"published-other":{"date-parts":[[2021,6,1]]},"published":{"date-parts":[[2021,2,16]]}}}