{"id":434,"date":"2025-07-06T18:18:22","date_gmt":"2025-07-06T22:18:22","guid":{"rendered":"https:\/\/medmultilingua.com\/english\/?p=434"},"modified":"2025-07-06T18:18:22","modified_gmt":"2025-07-06T22:18:22","slug":"bridging-data-worlds-multimodal-fine-tuning-for-predicting-covid-19-outcomes","status":"publish","type":"post","link":"https:\/\/medmultilingua.com\/english\/bridging-data-worlds-multimodal-fine-tuning-for-predicting-covid-19-outcomes\/","title":{"rendered":"Bridging Data Worlds: Multimodal Fine-Tuning for Predicting COVID-19 Outcomes"},"content":{"rendered":"\n<p><strong>By Dr. Marco V. Benavides S\u00e1nchez.<\/strong><\/p>\n\n\n\n<p>As the <a href=\"https:\/\/www.publichealth.columbia.edu\/news\/epidemic-endemic-pandemic-what-are-differences\">pandemic<\/a> tested the limits of healthcare systems worldwide, a new study by Henriksson et al. (2023) offers a compelling leap forward in clinical prediction: combining structured and unstructured data through <strong><a href=\"https:\/\/deepwiki.com\/unslothai\/notebooks\/3.3-multimodal-fine-tuning\">multimodal fine-tuning<\/a><\/strong> of language models.<\/p>\n\n\n\n<p>Traditional models tend to lean heavily on structured data\u2014things like lab values, vitals, and demographic variables\u2014while neglecting the nuanced information embedded in free-text clinical notes. But these notes, rich with physician observations, patient context, and care nuances, are a goldmine of untapped insights. That\u2019s where this study breaks new ground.<\/p>\n\n\n\n<p>\ud83e\uddea <strong>The Approach<\/strong><br>The researchers updated pre-trained language models, like <strong><a href=\"https:\/\/arxiv.org\/abs\/1904.05342\">ClinicalBERT<\/a><\/strong>, with both structured and unstructured inputs from six emergency departments, encompassing thousands of <strong><a href=\"https:\/\/www.who.int\/health-topics\/coronavirus#tab=tab_1\">COVID-19<\/a><\/strong> patients. The models were trained to predict three key outcomes:<\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li><strong>30-day mortality<\/strong><\/li>\n\n\n\n<li><strong>Safe discharge<\/strong><\/li>\n\n\n\n<li><strong>Readmission<\/strong><\/li>\n<\/ul>\n\n\n\n<p>By fusing these data modalities into a single model, the authors created a system that <strong><a href=\"https:\/\/arxiv.org\/abs\/2505.02467#\">outperformed unimodal baselines<\/a><\/strong> across all three predictions.<\/p>\n\n\n\n<p>\ud83d\udd0d <strong>Why It Matters<\/strong><br>Multimodal models, trained end-to-end, showed significant gains in accuracy and generalizability. Sensitivity analyses revealed how the models adapted across patient subgroups\u2014suggesting real-world applicability in diverse hospital settings. An <strong><a href=\"https:\/\/en.wikipedia.org\/wiki\/Ablation_%28artificial_intelligence%29\">ablation study<\/a><\/strong> further revealed that not all notes contribute equally; physician impressions and assessment plans had outsized value.<\/p>\n\n\n\n<p>\ud83c\udfaf <strong>Implications for Practice<\/strong><br>With healthcare systems grappling with staff shortages and capacity constraints, predictive tools that leverage <em>all<\/em> available data could help triage and manage patient care more efficiently. By integrating free-text notes, we move closer to more human-like, context-aware AI in medicine.<\/p>\n\n\n\n<p>In an age of <strong>information abundance<\/strong>, it\u2019s not just about having data\u2014it\u2019s about teaching machines to understand it holistically. This study proves that when structured metrics meet clinical narratives, the result is a sharper lens on patient outcomes.<\/p>\n\n\n\n<p><strong>For further reading:<\/strong><\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li>Henriksson, A., Pawar, Y., Hedberg, P., &amp; Naucl\u00e9r, P. (2023). Multimodal fine-tuning of clinical language models for predicting COVID-19 outcomes. <em>Artificial Intelligence in Medicine, 146<\/em>, 102695. <a href=\"https:\/\/doi.org\/10.1016\/j.artmed.2023.102695\">https:\/\/doi.org\/10.1016\/j.artmed.2023.102695<\/a><\/li>\n<\/ul>\n\n\n\n<p><em>Dedicated to my friend from Medical School, Dr. Carlos Eduardo Moye de Alba, on his birthday. Happy birthday!<\/em><\/p>\n\n\n\n<p><strong>#ArtificialIntelligence #Medicine #Surgery #Medmultilingua<\/strong><\/p>\n","protected":false},"excerpt":{"rendered":"<p>By Dr. Marco V. Benavides S\u00e1nchez. As the pandemic tested the limits of healthcare systems worldwide, a new study by Henriksson et al. (2023) offers a compelling leap forward in clinical prediction: combining structured and unstructured data through multimodal fine-tuning of language models. Traditional models tend to lean heavily on structured data\u2014things like lab values,&#8230;<\/p>\n","protected":false},"author":1,"featured_media":444,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[1],"tags":[],"class_list":["post-434","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-blog"],"_links":{"self":[{"href":"https:\/\/medmultilingua.com\/english\/wp-json\/wp\/v2\/posts\/434","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/medmultilingua.com\/english\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/medmultilingua.com\/english\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/medmultilingua.com\/english\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/medmultilingua.com\/english\/wp-json\/wp\/v2\/comments?post=434"}],"version-history":[{"count":10,"href":"https:\/\/medmultilingua.com\/english\/wp-json\/wp\/v2\/posts\/434\/revisions"}],"predecessor-version":[{"id":445,"href":"https:\/\/medmultilingua.com\/english\/wp-json\/wp\/v2\/posts\/434\/revisions\/445"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/medmultilingua.com\/english\/wp-json\/wp\/v2\/media\/444"}],"wp:attachment":[{"href":"https:\/\/medmultilingua.com\/english\/wp-json\/wp\/v2\/media?parent=434"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/medmultilingua.com\/english\/wp-json\/wp\/v2\/categories?post=434"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/medmultilingua.com\/english\/wp-json\/wp\/v2\/tags?post=434"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}