Clinical foundation model for structured clinical data

结构化临床数据的临床基础模型

• 批准号: 10639397
• 负责人: Laila Rasmy Gindy Bekhet
• 金额: $ 35.1万
• 依托单位: UNIVERSITY OF TEXAS HLTH SCI CTR HOUSTON
• 依托单位国家: 美国
• 财政年份: 2023
• 资助国家: 美国
• 起止时间: 2023-09-01 至 2027-05-31
• 项目状态: 未结题
• 来源: https://reporter.nih.gov/project-details/10639397
• 关键词: Architecture; Attention; COVID-19 patient; Chemicals; Clinical; Clinical Data; Code; Data; Data Element; Data Protection; Data Sources; Development; Diagnosis; Electronic Health Record; Engineering; Evaluation; Event; Fostering; Foundations; Goals; Health; Health Care Costs; Healthcare Systems; Heart failure; Intake; Journals; Knowledge; Learning; Long COVID; Malignant neoplasm of pancreas; Methodology; Methods; Modeling; Molecular; Natural Language Processing; Patient risk; Patient-Focused Outcomes; Patients; Peer Review; Performance; Pharmaceutical Preparations; Policies; Population; Predictive Cancer Model; Preparation; Recommendation; Recording of previous events; Research Personnel; Risk; Source; Structure; Terminology; Time; Training; Unified Medical Language System; Variant; Work; base; cohort; comorbidity Index; coronavirus disease; design; diabetic patient; experience; flexibility; improved; knowledge integration; model building; predictive modeling; public health relevance; structured data; symposium

Abstract In the era of big clinical data, the availability of rich real-world clinical data sources (RWcD) enables the development of predictive models for different clinical events, bringing the potential to improve efficiency and lower the cost of health care. However, the currently in-use models in practice are mostly trained on local data, introducing issues of bias and lack of generalizability. We will develop comprehensive methods to efficiently train high-quality clinical foundation model (CFM) that learn informative representations from patients' structured clinical data either in the form of EHR or claims. Specifically, how to train CFM that can maximize the performance boost for any downstream prediction tasks regardless of the predictive model architecture and the size of the available training data. In this application we propose to 1) Develop a flexible framework to intake the temporal structured clinical data elements from heterogenous sources and enrich it with existing knowledge, 2) Optimize the foundation model architecture and pre-training strategy, 3) Develop prompting strategies for zero/few shot learning, and 4) Evaluating CFM on multiple clinical downstream tasks.

摘要 在大临床数据时代，丰富的真实世界临床数据源（RWcD）的可用性使 开发针对不同临床事件的预测模型，带来提高效率和 降低医疗费用。然而，目前在实践中使用的模型大多是在本地数据上训练的， 引入了偏见和缺乏普遍性的问题。我们将制定全面的方法， 训练高质量的临床基础模型（CFM），该模型从患者的 以EHR或索赔形式的结构化临床数据。具体来说，如何培养能够最大限度地 无论预测模型架构如何，任何下游预测任务的性能提升， 可用训练数据的大小。在本申请中，我们建议：1）开发一个灵活的框架， 从异构源中获取时间结构化临床数据元素，并使用现有的 知识，2）优化基础模型架构和预培训策略，3）开发提示 用于零/少射击学习的策略，以及4）在多个临床下游任务上评估CFM。