Vasileios Ntinopoulos, Hector Rodriguez Cetina Biefer, Igor Tudorache, Nestoras Papadopoulos, Dragan Odavic, Petar Risteski, Achim Haeussler, Omer Dzemali
{"title":"用于从非结构化和半结构化电子健康记录中提取数据的大型语言模型:多模型性能评估。","authors":"Vasileios Ntinopoulos, Hector Rodriguez Cetina Biefer, Igor Tudorache, Nestoras Papadopoulos, Dragan Odavic, Petar Risteski, Achim Haeussler, Omer Dzemali","doi":"10.1136/bmjhci-2024-101139","DOIUrl":null,"url":null,"abstract":"<p><strong>Objectives: </strong>We aimed to evaluate the performance of multiple large language models (LLMs) in data extraction from unstructured and semi-structured electronic health records.</p><p><strong>Methods: </strong>50 synthetic medical notes in English, containing a structured and an unstructured part, were drafted and evaluated by domain experts, and subsequently used for LLM-prompting. 18 LLMs were evaluated against a baseline transformer-based model. Performance assessment comprised four entity extraction and five binary classification tasks with a total of 450 predictions for each LLM. LLM-response consistency assessment was performed over three same-prompt iterations.</p><p><strong>Results: </strong>Claude 3.0 Opus, Claude 3.0 Sonnet, Claude 2.0, GPT 4, Claude 2.1, Gemini Advanced, PaLM 2 chat-bison and Llama 3-70b exhibited an excellent overall accuracy >0.98 (0.995, 0.988, 0.988, 0.988, 0.986, 0.982, 0.982, and 0.982, respectively), significantly higher than the baseline RoBERTa model (0.742). Claude 2.0, Claude 2.1, Claude 3.0 Opus, PaLM 2 chat-bison, GPT 4, Claude 3.0 Sonnet and Llama 3-70b showed a marginally higher and Gemini Advanced a marginally lower multiple-run consistency than the baseline model RoBERTa (Krippendorff's alpha value 1, 0.998, 0.996, 0.996, 0.992, 0.991, 0.989, 0.988, and 0.985, respectively).</p><p><strong>Discussion: </strong>Claude 3.0 Opus, Claude 3.0 Sonnet, Claude 2.0, GPT 4, Claude 2.1, Gemini Advanced, PaLM 2 chat bison and Llama 3-70b performed the best, exhibiting outstanding performance in both entity extraction and binary classification, with highly consistent responses over multiple same-prompt iterations. Their use could leverage data for research and unburden healthcare professionals. Real-data analyses are warranted to confirm their performance in a real-world setting.</p><p><strong>Conclusion: </strong>Claude 3.0 Opus, Claude 3.0 Sonnet, Claude 2.0, GPT 4, Claude 2.1, Gemini Advanced, PaLM 2 chat-bison and Llama 3-70b seem to be able to reliably extract data from unstructured and semi-structured electronic health records. Further analyses using real data are warranted to confirm their performance in a real-world setting.</p>","PeriodicalId":9050,"journal":{"name":"BMJ Health & Care Informatics","volume":"32 1","pages":""},"PeriodicalIF":4.1000,"publicationDate":"2025-01-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC11751965/pdf/","citationCount":"0","resultStr":"{\"title\":\"Large language models for data extraction from unstructured and semi-structured electronic health records: a multiple model performance evaluation.\",\"authors\":\"Vasileios Ntinopoulos, Hector Rodriguez Cetina Biefer, Igor Tudorache, Nestoras Papadopoulos, Dragan Odavic, Petar Risteski, Achim Haeussler, Omer Dzemali\",\"doi\":\"10.1136/bmjhci-2024-101139\",\"DOIUrl\":null,\"url\":null,\"abstract\":\"<p><strong>Objectives: </strong>We aimed to evaluate the performance of multiple large language models (LLMs) in data extraction from unstructured and semi-structured electronic health records.</p><p><strong>Methods: </strong>50 synthetic medical notes in English, containing a structured and an unstructured part, were drafted and evaluated by domain experts, and subsequently used for LLM-prompting. 18 LLMs were evaluated against a baseline transformer-based model. Performance assessment comprised four entity extraction and five binary classification tasks with a total of 450 predictions for each LLM. LLM-response consistency assessment was performed over three same-prompt iterations.</p><p><strong>Results: </strong>Claude 3.0 Opus, Claude 3.0 Sonnet, Claude 2.0, GPT 4, Claude 2.1, Gemini Advanced, PaLM 2 chat-bison and Llama 3-70b exhibited an excellent overall accuracy >0.98 (0.995, 0.988, 0.988, 0.988, 0.986, 0.982, 0.982, and 0.982, respectively), significantly higher than the baseline RoBERTa model (0.742). Claude 2.0, Claude 2.1, Claude 3.0 Opus, PaLM 2 chat-bison, GPT 4, Claude 3.0 Sonnet and Llama 3-70b showed a marginally higher and Gemini Advanced a marginally lower multiple-run consistency than the baseline model RoBERTa (Krippendorff's alpha value 1, 0.998, 0.996, 0.996, 0.992, 0.991, 0.989, 0.988, and 0.985, respectively).</p><p><strong>Discussion: </strong>Claude 3.0 Opus, Claude 3.0 Sonnet, Claude 2.0, GPT 4, Claude 2.1, Gemini Advanced, PaLM 2 chat bison and Llama 3-70b performed the best, exhibiting outstanding performance in both entity extraction and binary classification, with highly consistent responses over multiple same-prompt iterations. Their use could leverage data for research and unburden healthcare professionals. Real-data analyses are warranted to confirm their performance in a real-world setting.</p><p><strong>Conclusion: </strong>Claude 3.0 Opus, Claude 3.0 Sonnet, Claude 2.0, GPT 4, Claude 2.1, Gemini Advanced, PaLM 2 chat-bison and Llama 3-70b seem to be able to reliably extract data from unstructured and semi-structured electronic health records. Further analyses using real data are warranted to confirm their performance in a real-world setting.</p>\",\"PeriodicalId\":9050,\"journal\":{\"name\":\"BMJ Health & Care Informatics\",\"volume\":\"32 1\",\"pages\":\"\"},\"PeriodicalIF\":4.1000,\"publicationDate\":\"2025-01-19\",\"publicationTypes\":\"Journal Article\",\"fieldsOfStudy\":null,\"isOpenAccess\":false,\"openAccessPdf\":\"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC11751965/pdf/\",\"citationCount\":\"0\",\"resultStr\":null,\"platform\":\"Semanticscholar\",\"paperid\":null,\"PeriodicalName\":\"BMJ Health & Care Informatics\",\"FirstCategoryId\":\"1085\",\"ListUrlMain\":\"https://doi.org/10.1136/bmjhci-2024-101139\",\"RegionNum\":0,\"RegionCategory\":null,\"ArticlePicture\":[],\"TitleCN\":null,\"AbstractTextCN\":null,\"PMCID\":null,\"EPubDate\":\"\",\"PubModel\":\"\",\"JCR\":\"Q1\",\"JCRName\":\"HEALTH CARE SCIENCES & SERVICES\",\"Score\":null,\"Total\":0}","platform":"Semanticscholar","paperid":null,"PeriodicalName":"BMJ Health & Care Informatics","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/10.1136/bmjhci-2024-101139","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"Q1","JCRName":"HEALTH CARE SCIENCES & SERVICES","Score":null,"Total":0}
Large language models for data extraction from unstructured and semi-structured electronic health records: a multiple model performance evaluation.
Objectives: We aimed to evaluate the performance of multiple large language models (LLMs) in data extraction from unstructured and semi-structured electronic health records.
Methods: 50 synthetic medical notes in English, containing a structured and an unstructured part, were drafted and evaluated by domain experts, and subsequently used for LLM-prompting. 18 LLMs were evaluated against a baseline transformer-based model. Performance assessment comprised four entity extraction and five binary classification tasks with a total of 450 predictions for each LLM. LLM-response consistency assessment was performed over three same-prompt iterations.
Results: Claude 3.0 Opus, Claude 3.0 Sonnet, Claude 2.0, GPT 4, Claude 2.1, Gemini Advanced, PaLM 2 chat-bison and Llama 3-70b exhibited an excellent overall accuracy >0.98 (0.995, 0.988, 0.988, 0.988, 0.986, 0.982, 0.982, and 0.982, respectively), significantly higher than the baseline RoBERTa model (0.742). Claude 2.0, Claude 2.1, Claude 3.0 Opus, PaLM 2 chat-bison, GPT 4, Claude 3.0 Sonnet and Llama 3-70b showed a marginally higher and Gemini Advanced a marginally lower multiple-run consistency than the baseline model RoBERTa (Krippendorff's alpha value 1, 0.998, 0.996, 0.996, 0.992, 0.991, 0.989, 0.988, and 0.985, respectively).
Discussion: Claude 3.0 Opus, Claude 3.0 Sonnet, Claude 2.0, GPT 4, Claude 2.1, Gemini Advanced, PaLM 2 chat bison and Llama 3-70b performed the best, exhibiting outstanding performance in both entity extraction and binary classification, with highly consistent responses over multiple same-prompt iterations. Their use could leverage data for research and unburden healthcare professionals. Real-data analyses are warranted to confirm their performance in a real-world setting.
Conclusion: Claude 3.0 Opus, Claude 3.0 Sonnet, Claude 2.0, GPT 4, Claude 2.1, Gemini Advanced, PaLM 2 chat-bison and Llama 3-70b seem to be able to reliably extract data from unstructured and semi-structured electronic health records. Further analyses using real data are warranted to confirm their performance in a real-world setting.