Clinical foundation model for structured clinical data

Document text

Principal Investigator: Laila Rasmy Gindy Bekhet
Organization: UNIVERSITY OF TEXAS HLTH SCI CTR HOUSTON
Fiscal Year: 2024
Award: $351,000
Funding agency: National Library of Medicine

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.

Terms: <Architecture><Attention><COVID infected patient><COVID patient><COVID positive patient><COVID-19><COVID-19 infected patient><COVID-19 patient><COVID-19 positive patient><COVID19 patient><COVID19 positive patient><CV-19><Chemicals><Clinical><Clinical Data><Code><Coding System><Coronavirus Infectious Disease 2019><Data><Data Element><Data Protection><Data Sources><Development><Diagnosis><Drugs><Electronic Health Record><Engineering><Engineering / Architecture><Evaluation><Event><Fostering><Foundations><Goals><Health><Health Care Costs><Health Care Systems><Health Costs><Healthcare Costs><Healthcare Systems><Heart failure><History><Intake><Investigators><Journals><Knowledge><Learning><Long COVID><Long COVID-19><Long coronavirus disease><Long coronavirus disease 2019><Magazine><Malignant Pancreatic Neoplasm><Malignant neoplasm of pancreas><Medication><Methodology><Methods><Modeling><Molecular><Natural Language Processing><Pancreas Cancer><Pancreatic Cancer><Patient outcome><Patient risk><Patient-Centered Outcomes><Patient-Focused Outcomes><Patients><Peer Review><Performance><Pharmaceutical Preparations><Policies><Population><Predictive Cancer Model><Preparation><Recommendation><Recording of previous events><Research Personnel><Researchers><Risk><SARS-CoV-2 infected patient><SARS-CoV-2 patient><SARS-CoV-2 positive patient><Source><Structure><Terminology><Time><Training><UMLS><Unified Medical Language System><Variant><Variation><Work><base><bases><cardiac failure><chronic COVID><chronic COVID-19><chronic novel coronavirus disease 2019><co-morbidity index><cohort><comorbidity Index><computer based prediction><conference><convention><coronavirus disease 2019><coronavirus disease 2019 infected patient><coronavirus disease 2019 patient><coronavirus disease 2019 positive patient><coronavirus disease infected patient><coronavirus disease patient><coronavirus disease positive patient><coronavirus disease-19><coronavirus disease-19 patient><coronavirus infectious disease-19><coronavirus patient><data to train><dataset to train><design><designing><developmental><diabetic patient><drug/agent><electronic health care record><electronic health medical record><electronic health plan record><electronic health registry><electronic medical health record><experience><flexibility><flexible><histories><improved><knowledge integration><long haul COVID><long haul COVID-19><long haul coronavirus disease><long haul coronavirus disease 2019><long-hauler COVID><long-hauler COVID-19><long-hauler coronavirus disease 2019><long-hauler syndrome><long-term COVID><long-term COVID-19><long-term coronavirus disease><long-term coronavirus disease 2019><longterm COVID><longterm COVID-19><longterm coronavirus disease><longterm coronavirus disease 2019><model building><natural language understanding><pancreatic malignancy><patient infected with COVID><patient infected with COVID-19><patient infected with SARS-CoV-2><patient infected with coronavirus disease><patient infected with coronavirus disease 2019><patient infected with severe acute respiratory syndrome coronavirus 2><patient oriented outcomes><patient with COVID><patient with COVID-19><patient with COVID19><patient with SARS-CoV-2><patient with coronavirus disease><patient with coronavirus disease 2019><patient with severe acute respiratory distress syndrome coronavirus 2><persistent COVID-19><post COVID syndrome><post COVID-19 syndrome><post acute COVID syndrome><post acute COVID-19><post acute COVID-19 syndrome><post acute SARS-CoV-2><post acute coronavirus disease 2019><post acute coronavirus disease 2019 syndrome><post acute coronavirus disease syndrome><post acute severe acute respiratory syndrome coronavirus 2><post coronavirus disease 2019 syndrome><post coronavirus disease syndrome><post-acute phases of COVID-19><pre-trained model><pre-trained transformer><predictive modeling><preparations><prolonged COVID-19 symptoms><public health relevance><severe acute respiratory syndrome coronavirus 2 infected patient><severe acute respiratory syndrome coronavirus 2 patient><severe acute respiratory syndrome coronavirus 2 positive patient><shot learning><structured data><summit><symposia><symposium><training data><transformer architecture><transformer based model><transformer model>