Enhanced Language Models for Predicting and Understanding HIV Care Disengagement: A Case Study in Tanzania

Enhanced Language Models for Predicting and Understanding HIV Care Disengagement: A Case Study in Tanzania | 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 Method Article Enhanced Language Models for Predicting and Understanding HIV Care Disengagement: A Case Study in Tanzania Waverly Wei, Junzhe Shao, Rita Qiuran Lyu, Rebecca Hemono, Xinwei Ma, and 12 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-6608559/v2 This work is licensed under a CC BY 4.0 License Status: Posted Version 2 posted You are reading this latest preprint version Show more versions Abstract Sustained engagement in HIV care and adherence to ART are crucial for meeting the UNAIDS "95-95-95" targets. Disengagement from care remains a significant issue, especially in sub-Saharan Africa. Traditional machine learning (ML) models have had moderate success in predicting disengagement, enabling early intervention. We developed an enhanced large language model (LLM) fine-tuned with electronic medical records (EMRs) to predict individuals at risk of disengaging from HIV care in Tanzania. Using 4.8 million EMR records from the National HIV Care and Treatment Program (2018–2023), we identified risks of ART non-adherence, non-suppressed viral load, and loss to follow-up. Our enhanced LLM may outperform traditional machine learning models and zero-shot LLMs. HIV physicians in Tanzania evaluated the model’s predictions and justifications, finding 65% alignment with expert assessments, and 92.3% of the aligned cases were considered clinically relevant. This model can support data-driven decisions and may improve patient outcomes and reduce HIV transmission. Biostatistics Epidemiology Full Text Additional Declarations The authors declare no competing interests. Cite Share Download PDF Status: Posted Version 2 posted You are reading this latest preprint version Show more versions 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-6608559","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Method Article","associatedPublications":[],"authors":[{"id":570906007,"identity":"6c36342a-3edf-489d-8fd0-a95895ebc2fd","order_by":0,"name":"Waverly Wei","email":"","orcid":"","institution":"University of Southern California","correspondingAuthor":false,"prefix":"","firstName":"Waverly","middleName":"","lastName":"Wei","suffix":""},{"id":570906008,"identity":"5ffbedcf-da00-49ef-a210-2149390651bb","order_by":1,"name":"Junzhe Shao","email":"","orcid":"","institution":"University of California, Berkeley","correspondingAuthor":false,"prefix":"","firstName":"Junzhe","middleName":"","lastName":"Shao","suffix":""},{"id":570906009,"identity":"3cfc0b72-8052-4ce4-b9b7-bc2cda386ef4","order_by":2,"name":"Rita Qiuran Lyu","email":"","orcid":"","institution":"University of California, Berkeley","correspondingAuthor":false,"prefix":"","firstName":"Rita","middleName":"Qiuran","lastName":"Lyu","suffix":""},{"id":570906010,"identity":"60d66a2e-c3d3-44e7-9430-213136dc9dae","order_by":3,"name":"Rebecca Hemono","email":"","orcid":"","institution":"University of California, Berkeley","correspondingAuthor":false,"prefix":"","firstName":"Rebecca","middleName":"","lastName":"Hemono","suffix":""},{"id":570906011,"identity":"f06261d9-ebb6-4495-a0da-4aa2ec2f9b2b","order_by":4,"name":"Xinwei Ma","email":"","orcid":"","institution":"University of California, San Diego","correspondingAuthor":false,"prefix":"","firstName":"Xinwei","middleName":"","lastName":"Ma","suffix":""},{"id":570906012,"identity":"2a05b8b0-954a-4492-a5db-b5a57be0dc00","order_by":5,"name":"Joseph Giorgio","email":"","orcid":"","institution":"University of California, Berkeley","correspondingAuthor":false,"prefix":"","firstName":"Joseph","middleName":"","lastName":"Giorgio","suffix":""},{"id":570906013,"identity":"51d2a9bd-d476-4576-a5d9-64cd7a003bba","order_by":6,"name":"Zeyu Zheng","email":"","orcid":"","institution":"University of California, Berkeley","correspondingAuthor":false,"prefix":"","firstName":"Zeyu","middleName":"","lastName":"Zheng","suffix":""},{"id":570906014,"identity":"9fa158f5-a757-4346-9a64-76ee73043fd8","order_by":7,"name":"Feng Ji","email":"","orcid":"","institution":"University of Toronto","correspondingAuthor":false,"prefix":"","firstName":"Feng","middleName":"","lastName":"Ji","suffix":""},{"id":570906015,"identity":"d7c3966a-c687-4286-8878-f25f7bb549e1","order_by":8,"name":"Xiaoya Zhang","email":"","orcid":"","institution":"University of Florida","correspondingAuthor":false,"prefix":"","firstName":"Xiaoya","middleName":"","lastName":"Zhang","suffix":""},{"id":570906016,"identity":"ee265dcb-1eb1-4153-b48c-780e2ecfa576","order_by":9,"name":"Emmanuel Katabaro","email":"","orcid":"","institution":"Health for a Prosperous Nation, Dar es Salaam, Tanzania","correspondingAuthor":false,"prefix":"","firstName":"Emmanuel","middleName":"","lastName":"Katabaro","suffix":""},{"id":570906017,"identity":"fa21d5bf-0562-40f4-b619-79240c46cd86","order_by":10,"name":"Matilda Mlowe","email":"","orcid":"","institution":"Health for a Prosperous Nation, Dar es Salaam, Tanzania","correspondingAuthor":false,"prefix":"","firstName":"Matilda","middleName":"","lastName":"Mlowe","suffix":""},{"id":570906018,"identity":"f50c25e5-5bcb-4e45-945b-7ca381f45d54","order_by":11,"name":"Amon Sabasaba","email":"","orcid":"","institution":"Health for a Prosperous Nation, Dar es Salaam, Tanzania","correspondingAuthor":false,"prefix":"","firstName":"Amon","middleName":"","lastName":"Sabasaba","suffix":""},{"id":570906019,"identity":"cb46e68a-d5ed-42e5-b558-056dece55bc9","order_by":12,"name":"Caroline Lister","email":"","orcid":"","institution":"Dodoma Referral Hospital, Dodoma, Tanzania","correspondingAuthor":false,"prefix":"","firstName":"Caroline","middleName":"","lastName":"Lister","suffix":""},{"id":570906020,"identity":"4f66692c-6621-4153-93e7-9beebf8d148e","order_by":13,"name":"Siraji Shabani","email":"","orcid":"","institution":"Ministry of Health, Dodoma, Tanzania","correspondingAuthor":false,"prefix":"","firstName":"Siraji","middleName":"","lastName":"Shabani","suffix":""},{"id":570906021,"identity":"914dbf0d-f411-46b2-aa07-75cc831bb676","order_by":14,"name":"Prosper Njau","email":"","orcid":"","institution":"Ministry of Health, Dodoma, Tanzania","correspondingAuthor":false,"prefix":"","firstName":"Prosper","middleName":"","lastName":"Njau","suffix":""},{"id":570906022,"identity":"fced9f04-ba8d-4bcf-92a3-0e7d6a0fa9ff","order_by":15,"name":"Sandra I. McCoy","email":"","orcid":"","institution":"University of California, Berkeley","correspondingAuthor":false,"prefix":"","firstName":"Sandra","middleName":"I.","lastName":"McCoy","suffix":""},{"id":570906023,"identity":"b338a3eb-8ed3-4709-85d9-daa925a43223","order_by":16,"name":"Jingshen Wang","email":"data:image/png;base64,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","orcid":"","institution":"University of California, Berkeley","correspondingAuthor":true,"prefix":"","firstName":"Jingshen","middleName":"","lastName":"Wang","suffix":""}],"badges":[],"createdAt":"2025-05-07 06:10:10","currentVersionCode":2,"declarations":{"humanSubjects":true,"vertebrateSubjects":false,"conflictsOfInterestStatement":false,"humanSubjectEthicalGuidelines":true,"humanSubjectConsent":true,"humanSubjectClinicalTrial":true,"humanSubjectCaseReport":true,"vertebrateSubjectEthicalGuidelines":false},"doi":"10.21203/rs.3.rs-6608559/v2","doiUrl":"https://doi.org/10.21203/rs.3.rs-6608559/v2","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":100373088,"identity":"0d6f7ab4-3d56-46d8-bc95-8e1066b9dd30","added_by":"auto","created_at":"2026-01-16 08:13:37","extension":"docx","order_by":0,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":3960102,"visible":true,"origin":"","legend":"","description":"","filename":"round3manuscriptcleaned.docx","url":"https://assets-eu.researchsquare.com/files/rs-6608559/v2/c9b3fe2882cfc656fca57dfb.docx"},{"id":100371522,"identity":"6c64d179-c67e-432c-a2c8-4722c8d38bce","added_by":"auto","created_at":"2026-01-16 08:10:28","extension":"json","order_by":1,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":342,"visible":true,"origin":"","legend":"","description":"","filename":"rs6608559.json","url":"https://assets-eu.researchsquare.com/files/rs-6608559/v2/e13e4cee3df55ba5753ccf85.json"},{"id":100372368,"identity":"74de1786-9fa4-43ba-96b6-cfc87f094f1f","added_by":"auto","created_at":"2026-01-16 08:12:08","extension":"xml","order_by":2,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":98308,"visible":true,"origin":"","legend":"","description":"","filename":"rs66085592enriched.xml","url":"https://assets-eu.researchsquare.com/files/rs-6608559/v2/6defbd4161a0eb695044d1ca.xml"},{"id":100273376,"identity":"f7f7308f-4da7-48af-9f0c-0ad63f17f17c","added_by":"auto","created_at":"2026-01-14 21:27:46","extension":"jpeg","order_by":3,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":677872,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage1.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-6608559/v2/5dcc9091ce7ba7b3c86f8293.jpeg"},{"id":100373083,"identity":"e30b7c86-0b4b-417f-b312-b6c40066df92","added_by":"auto","created_at":"2026-01-16 08:13:36","extension":"jpeg","order_by":4,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":458415,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage2.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-6608559/v2/3f843d0c96f2c76e67ab6f2b.jpeg"},{"id":100373308,"identity":"21daefb5-63d2-4807-95c4-ee2ce656a875","added_by":"auto","created_at":"2026-01-16 08:14:03","extension":"jpeg","order_by":5,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":496049,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage3.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-6608559/v2/4baed5d6d966be57ead2c808.jpeg"},{"id":100373087,"identity":"99bedc59-1121-4851-80a4-b3898a5b84a7","added_by":"auto","created_at":"2026-01-16 08:13:36","extension":"jpeg","order_by":6,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":412299,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage4.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-6608559/v2/4fc3034b79c19baace442999.jpeg"},{"id":100273384,"identity":"5a7f611f-cd72-47b2-9a8c-58d438c5e179","added_by":"auto","created_at":"2026-01-14 21:27:46","extension":"png","order_by":7,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":159618,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage1.png","url":"https://assets-eu.researchsquare.com/files/rs-6608559/v2/ad84eb1c87c1c344905847db.png"},{"id":100371652,"identity":"3228c669-c0a1-4279-ab78-b1284227add2","added_by":"auto","created_at":"2026-01-16 08:10:40","extension":"png","order_by":8,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":89314,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage2.png","url":"https://assets-eu.researchsquare.com/files/rs-6608559/v2/d82d58c985d3aa550ff49427.png"},{"id":100273381,"identity":"1b9e9a81-c0d6-4560-918f-8f0c280e2a18","added_by":"auto","created_at":"2026-01-14 21:27:46","extension":"png","order_by":9,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":121507,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage3.png","url":"https://assets-eu.researchsquare.com/files/rs-6608559/v2/ac009343fe20ea01a4f5984f.png"},{"id":100373311,"identity":"7ca673fd-1950-462d-8bdc-863f3c961c50","added_by":"auto","created_at":"2026-01-16 08:14:03","extension":"png","order_by":10,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":71386,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage4.png","url":"https://assets-eu.researchsquare.com/files/rs-6608559/v2/87be9126f0d7bbb6ee5fdfc9.png"},{"id":100273386,"identity":"04f5ef6b-d9ba-4922-ac6e-b5969cedc732","added_by":"auto","created_at":"2026-01-14 21:27:46","extension":"xml","order_by":11,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":94604,"visible":true,"origin":"","legend":"","description":"","filename":"rs66085592structuring.xml","url":"https://assets-eu.researchsquare.com/files/rs-6608559/v2/c04863effca317c6bdbb5a42.xml"},{"id":100273388,"identity":"e0f64d0f-04e8-4946-9029-a6dada808001","added_by":"auto","created_at":"2026-01-14 21:27:46","extension":"html","order_by":12,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":108991,"visible":true,"origin":"","legend":"","description":"","filename":"earlyproof.html","url":"https://assets-eu.researchsquare.com/files/rs-6608559/v2/ee19333ee53e1a9a2a4741aa.html"},{"id":100383907,"identity":"22f55277-02b4-4d70-bb02-e3393a01d675","added_by":"auto","created_at":"2026-01-16 10:48:45","extension":"pdf","order_by":1,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":635614,"visible":true,"origin":"","legend":"","description":"","filename":"round3manuscriptcleaned.pdf","url":"https://assets-eu.researchsquare.com/files/rs-6608559/v2_covered_88aac505-1613-45cb-9aef-afd2976b4f44.pdf"}],"financialInterests":"The authors declare no competing interests.","formattedTitle":"Enhanced Language Models for Predicting and Understanding HIV Care Disengagement: A Case Study in Tanzania","fulltext":[],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":false,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":true,"hideJournal":true,"highlight":"","institution":"University of California, Berkeley","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":"","lastPublishedDoi":"10.21203/rs.3.rs-6608559/v2","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-6608559/v2","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eSustained engagement in HIV care and adherence to ART are crucial for meeting the UNAIDS \"95-95-95\" targets. Disengagement from care remains a significant issue, especially in sub-Saharan Africa. Traditional machine learning (ML) models have had moderate success in predicting disengagement, enabling early intervention. We developed an enhanced large language model (LLM) fine-tuned with electronic medical records (EMRs) to predict individuals at risk of disengaging from HIV care in Tanzania. Using 4.8\u0026nbsp;million EMR records from the National HIV Care and Treatment Program (2018\u0026ndash;2023), we identified risks of ART non-adherence, non-suppressed viral load, and loss to follow-up. Our enhanced LLM may outperform traditional machine learning models and zero-shot LLMs. HIV physicians in Tanzania evaluated the model\u0026rsquo;s predictions and justifications, finding 65% alignment with expert assessments, and 92.3% of the aligned cases were considered clinically relevant. This model can support data-driven decisions and may improve patient outcomes and reduce HIV transmission.\u003c/p\u003e","manuscriptTitle":"Enhanced Language Models for Predicting and Understanding HIV Care Disengagement: A Case Study in Tanzania","msid":"","msnumber":"","nonDraftVersions":[{"code":2,"date":"2026-01-14 21:27:41","doi":"10.21203/rs.3.rs-6608559/v2","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}},{"code":1,"date":"2025-05-08 14:32:11","doi":"10.21203/rs.3.rs-6608559/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":"6351d622-6700-470c-aa09-ead0880ffe14","owner":[],"postedDate":"January 14th, 2026","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[{"id":60752421,"name":"Biostatistics"},{"id":60752422,"name":"Epidemiology"}],"tags":[],"updatedAt":"2025-05-08T14:32:11+00:00","versionOfRecord":[],"versionCreatedAt":"2026-01-14 21:27:41","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v2","identity":"rs-6608559","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-6608559","identity":"rs-6608559","version":["v2"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}