{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,10,13]],"date-time":"2026-10-13T18:15:05Z","timestamp":1791915305696,"version":"4.3.4"},"update-to":[{"DOI":"10.1371\/journal.pcbi.1013080","type":"new_version","label":"New version","source":"publisher","updated":{"date-parts":[[2025,5,27]],"date-time":"2025-05-27T00:00:00Z","timestamp":1748304000000}}],"reference-count":67,"publisher":"Public Library of Science (PLoS)","issue":"5","license":[{"start":{"date-parts":[[2025,5,19]],"date-time":"2025-05-19T00:00:00Z","timestamp":1747612800000},"content-version":"vor","delay-in-days":0,"URL":"http:\/\/creativecommons.org\/licenses\/by\/4.0\/"}],"content-domain":{"domain":["www.ploscompbiol.org"],"crossmark-restriction":false},"short-container-title":["PLoS Comput Biol"],"abstract":"<jats:p>Representation bias in health data can lead to unfair decisions and compromise the generalisability of research findings. As a consequence, underrepresented subpopulations, such as those from specific ethnic backgrounds or genders, do not benefit equally from clinical discoveries. Several approaches have been developed to mitigate representation bias, ranging from simple resampling methods, such as SMOTE, to recent approaches based on generative adversarial networks (GAN). However, generating high-dimensional time-series synthetic health data remains a significant challenge. In response, we devised a novel architecture (CA-GAN) that synthesises authentic, high-dimensional time series data. CA-GAN outperforms state-of-the-art methods in a qualitative and a quantitative evaluation while avoiding mode collapse, a serious GAN failure. We perform evaluation using 7535 patients with hypotension and sepsis from two diverse, real-world clinical datasets. We show that synthetic data generated by our CA-GAN improves model fairness in Black patients as well as female patients when evaluated separately for each subpopulation. Furthermore, CA-GAN generates authentic data of the minority class while faithfully maintaining the original distribution of data, resulting in improved performance in a downstream predictive task.<\/jats:p>","DOI":"10.1371\/journal.pcbi.1013080","type":"journal-article","created":{"date-parts":[[2025,5,19]],"date-time":"2025-05-19T14:08:55Z","timestamp":1747663735000},"page":"e1013080","update-policy":"https:\/\/doi.org\/10.1371\/journal.pcbi.corrections_policy","source":"Crossref","is-referenced-by-count":26,"title":["Generative AI mitigates representation bias and improves model fairness through synthetic health data"],"prefix":"10.1371","volume":"21","author":[{"ORCID":"https:\/\/orcid.org\/0009-0006-6595-5344","authenticated-orcid":true,"given":"Raffaele","family":"Marchesi","sequence":"first","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Nicolo","family":"Micheletti","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Nicholas","family":"I-Hsien Kuo","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0002-5919-372X","authenticated-orcid":true,"given":"Sebastiano","family":"Barbieri","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0002-2705-5728","authenticated-orcid":true,"given":"Giuseppe","family":"Jurman","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0001-7306-2972","authenticated-orcid":true,"given":"Venet","family":"Osmani","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"340","published-online":{"date-parts":[[2025,5,19]]},"reference":[{"issue":"1","key":"pcbi.1013080.ref001","doi-asserted-by":"crossref","first-page":"44","DOI":"10.1038\/s41591-018-0300-7","article-title":"High-performance medicine: the convergence of human and artificial intelligence","volume":"25","author":"EJ Topol","year":"2019","journal-title":"Nat Med"},{"key":"pcbi.1013080.ref002","volume-title":"Global strategy on digital health 2020\u20132025","author":"World Health Organization","year":"2021"},{"issue":"6464","key":"pcbi.1013080.ref003","doi-asserted-by":"crossref","first-page":"447","DOI":"10.1126\/science.aax2342","article-title":"Dissecting racial bias in an algorithm used to manage the health of populations","volume":"366","author":"Z Obermeyer","year":"2019","journal-title":"Science"},{"issue":"4","key":"pcbi.1013080.ref004","article-title":"Health data poverty: an assailable barrier to equitable digital health care","volume":"3","author":"H Ibrahim","year":"2021","journal-title":"Lancet Digit Health"},{"key":"pcbi.1013080.ref005","doi-asserted-by":"crossref","first-page":"321","DOI":"10.1613\/jair.953","article-title":"SMOTE: synthetic minority over-sampling technique","volume":"16","author":"NV Chawla","year":"2002","journal-title":"jair"},{"issue":"1","key":"pcbi.1013080.ref006","doi-asserted-by":"crossref","first-page":"21","DOI":"10.1109\/TIT.1967.1053964","article-title":"Nearest neighbor pattern classification","volume":"13","author":"T Cover","year":"1967","journal-title":"IEEE Trans Inform Theory"},{"key":"pcbi.1013080.ref007","first-page":"878","article-title":"Borderline-smote: a new over-sampling method in imbalanced data sets learning","volume-title":"Lect Notes Comput Science","author":"H Han","year":"2005"},{"key":"pcbi.1013080.ref008","first-page":"309","article-title":"Farthest smote: a modified smote approach","volume-title":"Advances in intelligent systems and computing","author":"A Gosain","year":"2019"},{"issue":"1","key":"pcbi.1013080.ref009","doi-asserted-by":"crossref","first-page":"4","DOI":"10.1504\/IJKESDP.2011.039875","article-title":"Borderline over-sampling for imbalanced data classification","volume":"3","author":"HM Nguyen","year":"2011","journal-title":"IJKESDP"},{"key":"pcbi.1013080.ref010","doi-asserted-by":"crossref","first-page":"106","DOI":"10.1186\/1471-2105-14-106","article-title":"SMOTE for high-dimensional class-imbalanced data","volume":"14","author":"R Blagus","year":"2013","journal-title":"BMC Bioinformatics"},{"key":"pcbi.1013080.ref011","doi-asserted-by":"crossref","first-page":"863","DOI":"10.1613\/jair.1.11192","article-title":"SMOTE for learning from imbalanced data: progress and challenges, marking the 15-year anniversary","volume":"61","author":"A Fernandez","year":"2018","journal-title":"jair"},{"issue":"9","key":"pcbi.1013080.ref012","doi-asserted-by":"crossref","first-page":"1263","DOI":"10.1109\/TKDE.2008.239","article-title":"Learning from imbalanced data","volume":"21","author":"He Haibo","year":"2009","journal-title":"IEEE Trans Knowl Data Eng"},{"key":"pcbi.1013080.ref013","unstructured":"Lu C, Reddy C, Wang P, Nie D, Ning Y. Multi-label clinical time-series generation via conditional GAN. arXiv preprint 2022. https:\/\/arxiv.org\/abs\/2204.04797"},{"key":"pcbi.1013080.ref014","doi-asserted-by":"crossref","first-page":"114582","DOI":"10.1016\/j.eswa.2021.114582","article-title":"Conditional Wasserstein GAN-based oversampling of tabular data for imbalanced learning","volume":"174","author":"J Engelmann","year":"2021","journal-title":"Exp Syst Appl"},{"key":"pcbi.1013080.ref015","doi-asserted-by":"crossref","first-page":"1009","DOI":"10.1016\/j.ins.2019.10.014","article-title":"Conditional Wasserstein generative adversarial network-gradient penalty-based approach to alleviating imbalanced data classification","volume":"512","author":"M Zheng","year":"2020","journal-title":"Inf Sci"},{"key":"pcbi.1013080.ref016","doi-asserted-by":"crossref","unstructured":"Seibold M, Hoch A, Farshad M, Navab N, F\u00fcrnstahl P. Conditional generative data augmentation for clinical audio datasets. arXiv preprint 2022. https:\/\/arxiv.org\/abs\/2203.11570","DOI":"10.1007\/978-3-031-16449-1_33"},{"key":"pcbi.1013080.ref017","doi-asserted-by":"crossref","first-page":"487","DOI":"10.1016\/j.neucom.2018.10.109","article-title":"Data augmentation in fault diagnosis based on the Wasserstein generative adversarial network with gradient penalty","volume":"396","author":"X Gao","year":"2020","journal-title":"Neurocomputing"},{"key":"pcbi.1013080.ref018","doi-asserted-by":"crossref","unstructured":"Karras T, Laine S, Aila T. A style-based generator architecture for generative adversarial networks. In: 2019 IEEE\/CVF Conference on Computer Vision and Pattern Recognition (CVPR). 2019. p. 4396\u2013405.","DOI":"10.1109\/CVPR.2019.00453"},{"key":"pcbi.1013080.ref019","doi-asserted-by":"crossref","first-page":"108098","DOI":"10.1016\/j.patcog.2021.108098","article-title":"A survey on text generation using generative adversarial networks","volume":"119","author":"GH de Rosa","year":"2021","journal-title":"Pattern Recognit"},{"key":"pcbi.1013080.ref020","first-page":"17022","article-title":"Hifi-gan: generative adversarial networks for efficient and high fidelity speech synthesis","volume":"33","author":"J Kong","year":"2020","journal-title":"Adv Neural Inf Process Syst"},{"key":"pcbi.1013080.ref021","doi-asserted-by":"crossref","unstructured":"Savage N. Synthetic data could be better than real data. Nature. 2023. doi: 10.1038\/d41586-023-01445-8 37106108","DOI":"10.1038\/d41586-023-01445-8"},{"issue":"10","key":"pcbi.1013080.ref022","doi-asserted-by":"crossref","first-page":"1","DOI":"10.1145\/3559540","article-title":"Generative adversarial networks in time series: a systematic literature review","volume":"55","author":"E Brophy","year":"2023","journal-title":"ACM Comput Surv"},{"key":"pcbi.1013080.ref023","unstructured":"Alaa A, Van Breugel B, Saveliev ES, van der Schaar M. How faithful is your synthetic data? Sample-level metrics for evaluating and auditing generative models. In: Proceedings of the International Conference on Machine Learning. PMLR. p. 290\u2013306."},{"key":"pcbi.1013080.ref024","unstructured":"Ghosheh G, Li J, Zhu T. A review of generative adversarial networks for electronic health records: applications, evaluation measures and data sources. arXiv preprint 2022. https:\/\/arxiv.org\/abs\/2203.07018"},{"key":"pcbi.1013080.ref025","first-page":"214","article-title":"Wasserstein generative adversarial networks","volume-title":"Proceedings of the International Conference on Machine Learning","author":"M Arjovsky"},{"key":"pcbi.1013080.ref026","unstructured":"Gulrajani I, Ahmed F, Arjovsky M, Dumoulin V, Courville A. Improved training of wasserstein gans. Adv Neural Inf Process Syst. 2017."},{"issue":"1","key":"pcbi.1013080.ref027","doi-asserted-by":"crossref","first-page":"693","DOI":"10.1038\/s41597-022-01784-7","article-title":"The health gym: synthetic health-related datasets for the development of reinforcement learning algorithms","volume":"9","author":"NI-H Kuo","year":"2022","journal-title":"Sci Data"},{"issue":"4","key":"pcbi.1013080.ref028","doi-asserted-by":"crossref","first-page":"100946","DOI":"10.1016\/j.patter.2024.100946","article-title":"An evaluation of synthetic data augmentation for mitigating covariate bias in health data","volume":"5","author":"L Juwara","year":"2024","journal-title":"Patterns (N Y)"},{"key":"pcbi.1013080.ref029","doi-asserted-by":"crossref","first-page":"160035","DOI":"10.1038\/sdata.2016.35","article-title":"MIMIC-III, a freely accessible critical care database","volume":"3","author":"AEW Johnson","year":"2016","journal-title":"Sci Data"},{"key":"pcbi.1013080.ref030","unstructured":"Theis L, van den Oord A, Bethge M. A note on the evaluation of generative models. In: International Conference on Learning Representations; 2016. http:\/\/arxiv.org\/abs\/1511.01844"},{"issue":"86","key":"pcbi.1013080.ref031","first-page":"2579","article-title":"Visualizing data using t-SNE","volume":"9","author":"L vanderMaaten","year":"2008","journal-title":"J Mach Learn Res"},{"key":"pcbi.1013080.ref032","unstructured":"McInnes L, Healy J, Melville J. UMAP: uniform manifold approximation and projection for dimension reduction. 2020."},{"key":"pcbi.1013080.ref033","unstructured":"Goodfellow I. Nips 2016 tutorial: Generative adversarial networks. CoRR. 2017. https:\/\/arxiv.org\/abs\/1701.00160"},{"issue":"1","key":"pcbi.1013080.ref034","doi-asserted-by":"crossref","first-page":"79","DOI":"10.1214\/aoms\/1177729694","article-title":"On information and sufficiency","volume":"22","author":"S Kullback","year":"1951","journal-title":"Ann Math Stat"},{"issue":"25","key":"pcbi.1013080.ref035","first-page":"723","article-title":"A kernel two-sample test","volume":"13","author":"A Gretton","year":"2012","journal-title":"J Mach Learn Res"},{"key":"pcbi.1013080.ref036","doi-asserted-by":"crossref","first-page":"239","DOI":"10.1093\/biomet\/33.3.239","article-title":"The treatment of ties in ranking problems","volume":"33","author":"MG Kendall","year":"1945","journal-title":"Biometrika"},{"key":"pcbi.1013080.ref037","doi-asserted-by":"crossref","first-page":"1651","DOI":"10.1109\/ICIP40778.2020.9191083","article-title":"Empirical analysis of overfitting and mode drop in gan training","volume-title":"2020 IEEE International Conference on Image Processing (ICIP)","author":"Y Yazici","year":"2020"},{"issue":"7","key":"pcbi.1013080.ref038","doi-asserted-by":"crossref","first-page":"1","DOI":"10.1145\/3616865","article-title":"Fairness in machine learning: a survey","volume":"56","author":"S Caton","year":"2024","journal-title":"ACM Comput Surv"},{"issue":"1","key":"pcbi.1013080.ref039","doi-asserted-by":"crossref","first-page":"4581","DOI":"10.1038\/s41467-022-32186-3","article-title":"Addressing fairness in artificial intelligence for medical imaging","volume":"13","author":"MA Ricci Lara","year":"2022","journal-title":"Nat Commun"},{"issue":"4","key":"pcbi.1013080.ref040","doi-asserted-by":"crossref","first-page":"1087","DOI":"10.1007\/s10877-021-00739-4","article-title":"Prediction of blood lactate values in critically ill patients: a retrospective multi-center cohort study","volume":"36","author":"B Mamandipoor","year":"2022","journal-title":"J Clin Monit Comput"},{"issue":"3","key":"pcbi.1013080.ref041","doi-asserted-by":"crossref","first-page":"80","DOI":"10.1007\/s00508-022-02130-y","article-title":"Hyperlactatemia and altered lactate kinetics are associated with excess mortality in sepsis: a multicenter retrospective observational study","volume":"135","author":"R Rezar","year":"2023","journal-title":"Wiener klinische Wochenschrift"},{"issue":"12","key":"pcbi.1013080.ref042","doi-asserted-by":"crossref","first-page":"1105","DOI":"10.3390\/diagnostics10121105","article-title":"Failure of lactate clearance predicts the outcome of critically ill septic patients","volume":"10","author":"RR Bruno","year":"2020","journal-title":"Diagnostics (Basel)"},{"key":"pcbi.1013080.ref043","unstructured":"Snell J, Swersky K, Zemel R. Prototypical networks for few-shot learning. Adv Neural Inf Process Syst. 2017."},{"key":"pcbi.1013080.ref044","doi-asserted-by":"crossref","first-page":"119898","DOI":"10.1016\/j.ins.2023.119898","article-title":"On generating trustworthy counterfactual explanations","volume":"655","author":"J Del Ser","year":"2024","journal-title":"Inf Sci"},{"key":"pcbi.1013080.ref045","unstructured":"Carrington AM, Manuel DG, Fieguth PW, Ramsay T, Osmani V, Wernly B. Deep ROC analysis and AUC as balanced average accuracy to improve model selection, understanding and interpretation. arXiv preprint 2021. https:\/\/arxiv.org\/abs\/2103.11357"},{"issue":"1","key":"pcbi.1013080.ref046","doi-asserted-by":"crossref","first-page":"329","DOI":"10.1109\/TPAMI.2022.3145392","article-title":"Deep ROC analysis and AUC as balanced average accuracy, for improved classifier selection, audit and explanation","volume":"45","author":"AM Carrington","year":"2023","journal-title":"IEEE Trans Pattern Anal Mach Intell"},{"key":"pcbi.1013080.ref047","doi-asserted-by":"crossref","unstructured":"Kuo N, Perez-Concha O, Hanly M, Mnatzaganian E, Hao B, Di Sipio M. Enriching data science and healthcare education: application and impact of synthetic datasets through the health gym project. JMIR Med Educ. 2023.","DOI":"10.2196\/preprints.51388"},{"key":"pcbi.1013080.ref048","unstructured":"Bai S, Kolter J, Koltun V. An empirical evaluation of generic convolutional and recurrent networks for sequence modeling. arXiv preprint 2018. https:\/\/arxiv.org\/abs\/1803.01271"},{"key":"pcbi.1013080.ref049","first-page":"1","article-title":"An input residual connection for simplifying gated recurrent neural networks","volume-title":"2020 International Joint Conference on Neural Networks (IJCNN)","author":"NIH Kuo","year":"2020"},{"key":"pcbi.1013080.ref050","doi-asserted-by":"crossref","unstructured":"Johansson P, Bright J, Krishna S, Fischer C, Leslie D. Exploring responsible applications of synthetic data to advance online safety research and development. arXiv preprint 2024. https:\/\/arxiv.org\/abs\/2402.04910","DOI":"10.2139\/ssrn.4699185"},{"key":"pcbi.1013080.ref051","doi-asserted-by":"crossref","DOI":"10.1101\/2023.12.11.23299819","volume-title":"AI learns racial information from the values of vital signs","author":"B Velichkovska","year":"2023"},{"key":"pcbi.1013080.ref052","doi-asserted-by":"crossref","DOI":"10.1101\/2022.02.03.22270291","volume-title":"Vital signs as a source of racial bias","author":"B Velichkovska","year":"2022"},{"key":"pcbi.1013080.ref053","unstructured":"Goodfellow I, Pouget-Abadie J, Mirza M, Xu B, Warde-Farley D, Ozair S. Generative adversarial networks. arXiv preprint 2014. https:\/\/arxiv.org\/abs\/1406.2661"},{"key":"pcbi.1013080.ref054","unstructured":"Mirza M, Osindero S. Conditional generative adversarial nets. 2014."},{"key":"pcbi.1013080.ref055","unstructured":"Yoon J, Jarrett D, VanderSchaar M. Time-series generative adversarial networks. Adv Neural Inf Process Syst. 2019."},{"key":"pcbi.1013080.ref056","doi-asserted-by":"crossref","unstructured":"Yu L, Zhang W, Wang J, Yu Y. Seqgan: sequence generative adversarial nets with policy gradient. CoRR. 2016. https:\/\/arxiv.org\/abs\/1609.05473","DOI":"10.1609\/aaai.v31i1.10804"},{"key":"pcbi.1013080.ref057","first-page":"251","article-title":"The Earth Mover\u2019s distance is the Mallows distance: some insights from statistics","volume-title":"Proceedings Eighth IEEE International Conference on Computer Vision. ICCV 2001","author":"E Levina"},{"key":"pcbi.1013080.ref058","unstructured":"Arjovsky M, Bottou L. Towards principled methods for training generative adversarial networks. arXiv preprint 2017. https:\/\/arxiv.org\/abs\/1701.04862"},{"key":"pcbi.1013080.ref059","doi-asserted-by":"crossref","first-page":"259","DOI":"10.1080\/01638539809545028","article-title":"An introduction to latent semantic analysis","volume":"25","author":"TK Landauer","year":"1998","journal-title":"Discourse Processes"},{"key":"pcbi.1013080.ref060","unstructured":"Mottini A, Lheritier A, Acuna-Agost R. Airline passenger name record generation using generative adversarial networks. CoRR. 2018. doi: abs\/1807.06657"},{"issue":"8","key":"pcbi.1013080.ref061","doi-asserted-by":"crossref","first-page":"1735","DOI":"10.1162\/neco.1997.9.8.1735","article-title":"Long short-term memory","volume":"9","author":"S Hochreiter","year":"1997","journal-title":"Neural Comput"},{"key":"pcbi.1013080.ref062","first-page":"799","article-title":"Bidirectional LSTM networks for improved phoneme classification and recognition","volume-title":"Proceedings of the 15th International Conference on Artificial Neural Networks: Formal Models and Their Applications - Volume Part II","author":"A Graves","year":"2005"},{"issue":"3","key":"pcbi.1013080.ref063","first-page":"69","article-title":"Statistics corner: a guide to appropriate use of correlation coefficient in medical research","volume":"24","author":"MM Mukaka","year":"2012","journal-title":"Malawi Med J"},{"key":"pcbi.1013080.ref064","doi-asserted-by":"crossref","unstructured":"Kuo N, Jorm L, Barbieri S. Generating synthetic clinical data that capture class imbalanced distributions with generative adversarial networks: Example using antiretroviral therapy for HIV. arXiv preprint 2022. https:\/\/arxiv.org\/abs\/2208.08655","DOI":"10.1016\/j.jbi.2023.104436"},{"key":"pcbi.1013080.ref065","unstructured":"Kuo N, Jorm L, Barbieri S. Synthetic health-related longitudinal data with mixed-type variables generated using diffusion models. arXiv preprint 2023. https:\/\/arxiv.org\/abs\/2303.12281"},{"key":"pcbi.1013080.ref066","first-page":"2256","article-title":"Deep unsupervised learning using nonequilibrium thermodynamics","volume-title":"Proceedings of the International Conference on Machine Learning","author":"J Sohl-Dickstein"},{"key":"pcbi.1013080.ref067","doi-asserted-by":"crossref","first-page":"151","DOI":"10.1109\/IACS.2018.8355458","article-title":"Evaluation of bidirectional LSTM for short-and long-term stock market prediction","volume-title":"2018 9th International Conference on Information and Communication Systems (ICICS)","author":"KA Althelaya","year":"2018"}],"updated-by":[{"DOI":"10.1371\/journal.pcbi.1013080","type":"new_version","label":"New version","source":"publisher","updated":{"date-parts":[[2025,5,27]],"date-time":"2025-05-27T00:00:00Z","timestamp":1748304000000}}],"container-title":["PLOS Computational Biology"],"original-title":[],"language":"en","link":[{"URL":"https:\/\/dx.plos.org\/10.1371\/journal.pcbi.1013080","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2025,5,27]],"date-time":"2025-05-27T14:33:34Z","timestamp":1748356414000},"score":1,"resource":{"primary":{"URL":"https:\/\/dx.plos.org\/10.1371\/journal.pcbi.1013080"}},"subtitle":[],"editor":[{"given":"Piero","family":"Fariselli","sequence":"first","affiliation":[],"role":[{"vocabulary":"crossref","role":"editor"}]}],"short-title":[],"issued":{"date-parts":[[2025,5,19]]},"references-count":67,"journal-issue":{"issue":"5","published-online":{"date-parts":[[2025,5,19]]}},"URL":"https:\/\/doi.org\/10.1371\/journal.pcbi.1013080","relation":{"has-preprint":[{"id-type":"doi","id":"10.1101\/2023.09.26.23296163","asserted-by":"object"}]},"ISSN":["1553-7358"],"issn-type":[{"value":"1553-7358","type":"electronic"}],"subject":[],"published":{"date-parts":[[2025,5,19]]}}}