Deep Learning Identifies Novel Subphenotypes of Multiple Organ Dysfunction Syndrome in Critical Care

Deep Learning Identifies Novel Subphenotypes of Multiple Organ Dysfunction Syndrome in Critical Care | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Article Deep Learning Identifies Novel Subphenotypes of Multiple Organ Dysfunction Syndrome in Critical Care Huang Weijin, Qinyan Liang, Wei Lin, Xiaohuan Li, Jianrong Zhang, and 3 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-8901759/v1 This work is licensed under a CC BY 4.0 License Status: Posted Version 1 posted You are reading this latest preprint version Abstract Background : Multiple organ dysfunction syndrome (MODS) is associated with high mortality in critically ill patients, but traditional severity scores do not capture dynamic patterns of organ failure over time. Objectives : To identify novel MODS subphenotypes using unsupervised deep learning on high-resolution time-series data. Methods : Retrospective observational cohort study using the Medical Infor- mation Mart for Intensive Care (MIMIC)-IV version 3.1 database. We included 30,950 adult ICU stays with SOFA score ≥2 involving at least two organ systems and ICU length of stay > 48 hours. Hourly data from seven organ systems were extracted for the first 72 hours after ICU admission. A long short-term memory variational autoencoder was trained to generate latent representations, followed by UMAP visualization and K-means clustering. Results : Unsupervised clustering of VAE-derived latent representations iden- tified three distinct subphenotypes comprising 9,355 (30.2%), 9,590 (31.0%), and 12,005 (38.8%) ICU stays. Hospital mortality rates were 22.7%, 21.4%, and 11.2%, respectively (log-rank p < 0.001). In multivariable Cox regression adjusted for age, sex, baseline SOFA score, Charlson Comorbidity Index, APS III score, and surgical admission status, the two higher-mortality subphenotypes remained independently associated with increased risk compared with the ref- erence subphenotype (adjusted hazard ratios 1.60 [95% CI 1.49–1.71] and 1.38 [95% CI 1.29–1.48], respectively; both p < 0.001). The model showed good discrimination (C-index 0.725). Conclusions : Application of a variational autoencoder to high-resolution ICU time-series data identifies three clinically meaningful MODS subphenotypes with significantly different mortality risk, providing a foundation for precision critical care interventions. Biological sciences/Computational biology and bioinformatics Health sciences/Diseases Health sciences/Health care Health sciences/Medical research variational autoencoder multiple organ dysfunction syndrome intensive care unit unsupervised learning MIMIC-IV subphenotyping precision medicine Full Text Additional Declarations No competing interests reported. Supplementary Files SupplementalTable.docx Cite Share Download PDF Status: Posted Version 1 posted You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-8901759","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Article","associatedPublications":[],"authors":[{"id":606038642,"identity":"f6c5b4d9-ba36-46d4-9708-3cef6ad32165","order_by":0,"name":"Huang Weijin","email":"","orcid":"","institution":"Guangdong Medical University， China","correspondingAuthor":false,"prefix":"","firstName":"Huang","middleName":"","lastName":"Weijin","suffix":""},{"id":606038643,"identity":"c45d85d0-c85d-46d5-8458-f38f5ce4fa72","order_by":1,"name":"Qinyan Liang","email":"","orcid":"","institution":"Guangdong Medical University， China","correspondingAuthor":false,"prefix":"","firstName":"Qinyan","middleName":"","lastName":"Liang","suffix":""},{"id":606038644,"identity":"e5881eb6-32c5-4ae7-959c-32e9e3b4f52c","order_by":2,"name":"Wei Lin","email":"","orcid":"","institution":"Guangdong Medical University， China","correspondingAuthor":false,"prefix":"","firstName":"Wei","middleName":"","lastName":"Lin","suffix":""},{"id":606038645,"identity":"47c5fd65-4220-4feb-a63d-224a111cbe28","order_by":3,"name":"Xiaohuan Li","email":"","orcid":"","institution":"Guangdong Medical University， China","correspondingAuthor":false,"prefix":"","firstName":"Xiaohuan","middleName":"","lastName":"Li","suffix":""},{"id":606038646,"identity":"df0dfcee-cfba-4c68-b44e-70dda29dd98f","order_by":4,"name":"Jianrong Zhang","email":"","orcid":"","institution":"Guangdong Medical University， China","correspondingAuthor":false,"prefix":"","firstName":"Jianrong","middleName":"","lastName":"Zhang","suffix":""},{"id":606038647,"identity":"e28f87cc-22ac-405c-9391-f648e9eba0d8","order_by":5,"name":"Cuiqin Chen","email":"","orcid":"","institution":"Houjie Hospital of Dongguan","correspondingAuthor":false,"prefix":"","firstName":"Cuiqin","middleName":"","lastName":"Chen","suffix":""},{"id":606038648,"identity":"e52490c3-c15e-444e-89ac-6a2d5950214e","order_by":6,"name":"Hong Zhang","email":"","orcid":"","institution":"Guangdong Medical University， China","correspondingAuthor":false,"prefix":"","firstName":"Hong","middleName":"","lastName":"Zhang","suffix":""},{"id":606038649,"identity":"99d83992-1229-4ead-8309-7ea7c01abc8e","order_by":7,"name":"Qingping Huang","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAAt0lEQVRIiWNgGAWjYDCCA0CcwCDBw8be2PjwAwlabOT4eA43G0sQrYWBIc1YTiK9TYCHGB18N5KPPXjYdjixTfJhG4MEg52cbgMBLZI30tINEkFapBPbHhQwJBubHSCgxeBGjpkEVEu7gQTDgcRtxGuRPNgmwUOCljRjNglGIrVInnmWJpFwzkaOjScRGMgGRPiF73jyMckfZRI88u3HHz78UGEnR1ALujtJUz4KRsEoGAWjAAcAAIa3QrVZHvFWAAAAAElFTkSuQmCC","orcid":"","institution":"Guangdong Medical University， China","correspondingAuthor":true,"prefix":"","firstName":"Qingping","middleName":"","lastName":"Huang","suffix":""}],"badges":[],"createdAt":"2026-02-17 13:54:40","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-8901759/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-8901759/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":109439592,"identity":"cda2032f-6319-42fd-be56-c608adf16e61","added_by":"auto","created_at":"2026-05-18 06:56:21","extension":"pdf","order_by":1,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":4866043,"visible":true,"origin":"","legend":"","description":"","filename":"maintex3.0.pdf","url":"https://assets-eu.researchsquare.com/files/rs-8901759/v1_covered_ea37ae2e-76cd-40f1-a193-47dbf1c3f259.pdf"},{"id":104769433,"identity":"0b10bbc5-246b-4544-b244-743ca070f6a9","added_by":"auto","created_at":"2026-03-17 04:48:22","extension":"docx","order_by":0,"title":"","display":"","copyAsset":false,"role":"supplement","size":15614,"visible":true,"origin":"","legend":"","description":"","filename":"SupplementalTable.docx","url":"https://assets-eu.researchsquare.com/files/rs-8901759/v1/584acd262b5f67dc359255d3.docx"}],"financialInterests":"No competing interests reported.","formattedTitle":"Deep Learning Identifies Novel Subphenotypes of Multiple Organ Dysfunction Syndrome in Critical Care","fulltext":[],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":false,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":true,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":true,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":true,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true},"keywords":"variational autoencoder, multiple organ dysfunction syndrome, intensive care unit, unsupervised learning, MIMIC-IV, subphenotyping, precision medicine","lastPublishedDoi":"10.21203/rs.3.rs-8901759/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-8901759/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003e\u003cstrong\u003eBackground\u003c/strong\u003e: Multiple organ dysfunction syndrome (MODS) is associated with high mortality in critically ill patients, but traditional severity scores do not capture dynamic patterns of organ failure over time.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eObjectives\u003c/strong\u003e: To identify novel MODS subphenotypes using unsupervised deep learning on high-resolution time-series data.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eMethods\u003c/strong\u003e: Retrospective observational cohort study using the Medical Infor- mation Mart for Intensive Care (MIMIC)-IV version 3.1 database. We included 30,950 adult ICU stays with SOFA score ≥2 involving at least two organ systems and ICU length of stay \u0026gt; 48 hours. Hourly data from seven organ systems were extracted for the first 72 hours after ICU admission. A long short-term memory variational autoencoder was trained to generate latent representations, followed by UMAP visualization and K-means clustering.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eResults\u003c/strong\u003e: Unsupervised clustering of VAE-derived latent representations iden- tified three distinct subphenotypes comprising 9,355 (30.2%), 9,590 (31.0%), and 12,005 (38.8%) ICU stays. Hospital mortality rates were 22.7%, 21.4%, and 11.2%, respectively (log-rank p \u0026lt; 0.001). In multivariable Cox regression adjusted for age, sex, baseline SOFA score, Charlson Comorbidity Index, APS III score, and surgical admission status, the two higher-mortality subphenotypes remained independently associated with increased risk compared with the ref- erence subphenotype (adjusted hazard ratios 1.60 [95% CI 1.49–1.71] and 1.38 [95% CI 1.29–1.48], respectively; both p \u0026lt; 0.001). The model showed good discrimination (C-index 0.725).\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eConclusions\u003c/strong\u003e: Application of a variational autoencoder to high-resolution ICU time-series data identifies three clinically meaningful MODS subphenotypes with significantly different mortality risk, providing a foundation for precision critical care interventions.\u003c/p\u003e","manuscriptTitle":"Deep Learning Identifies Novel Subphenotypes of Multiple Organ Dysfunction Syndrome in Critical Care","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2026-03-17 04:48:18","doi":"10.21203/rs.3.rs-8901759/v1","editorialEvents":[{"type":"communityComments","content":0}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"1a272de2-1503-41af-9762-a0906286acca","owner":[],"postedDate":"March 17th, 2026","published":true,"recentEditorialEvents":[{"type":"decision","content":"Rejected","date":"2026-05-18T06:42:39+00:00","index":"","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2026-05-13T12:45:34+00:00","index":145,"fulltext":""},{"type":"reviewerAgreed","content":"45480718896017714610646955580263597546","date":"2026-05-13T10:36:25+00:00","index":144,"fulltext":""}],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[{"id":64497476,"name":"Biological sciences/Computational biology and bioinformatics"},{"id":64497477,"name":"Health sciences/Diseases"},{"id":64497478,"name":"Health sciences/Health care"},{"id":64497479,"name":"Health sciences/Medical research"}],"tags":[],"updatedAt":"2026-05-18T06:55:03+00:00","versionOfRecord":[],"versionCreatedAt":"2026-03-17 04:48:18","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-8901759","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-8901759","identity":"rs-8901759","version":["v1"]},"buildId":"XKTyCvWXoU3ODBz1xrDgd","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}