Toward automated, scalable knowledge engineering for real world data using clinical terminologies
Active Project Grant Awarded Last action
- Award ID
- F31LM015316
- Recipient
-
The Trustees Of Columbia University In The City Of New York UEI QHF5ZZ114M72
- Awarding agency
-
National Institute of Environmental Health Sciences
National Institutes of Health, Department of Health and Human Services
- Obligated
- $50,114
Award facts
- Assistance listing
- 93.879 Medical Library Assistance
- Completion date 0% complete
- Place of performance
- New York, NY 10032, USA
Award details come from USAspending.gov.
Description
- Abstract
- High-quality real-world evidence depends on two intertwined capabilities: (1) the existence of harmonized clinical vocabularies that give consistent meaning to the codes embedded in electronic health records (ehrs) and (2) computable phenotypes that translate concept codes, temporal rules, and clinical context into reproducible patient cohorts. To date, no automated system has reliably harmonized clinical vocabularies or generated high-fidelity phenotype definitions at scale. While the advancement of transformer-based models, such as large language models (llms) offers some promise, they may hallucinate code mappings or concept relationships, and they lack explicit semantic grounding required for clinical applications. This work aims to address these limitations by creating an open-source pipeline that fuses symbolic artificial intelligence with llms to automate knowledge engineering. The methods will be anchored in the observational health data sciences and informatics (ohdsi) standardized vocabulary, which consists of over 198,000 concepts and 86 million expert-curated relationships. This semantic anchoring will constrain LLM outputs to validated clinical knowledge, alleviate hallucination, and enable rigorous benchmarking against gold standards. Our aims are as follows: (1) develop a vocabulary harmonization toolkit (including term similarity, subsumption, and grouping tasks) with LLM augmented algorithms; and (2) augment existing knowledge engineering pipelines with machine learning methods for phenotype specification and refinement. The proposed research will deliver an operational, end-to-end pipeline for automated clinical vocabulary harmonization and phenotype specification. By leveraging the scale and expert knowledge from the ohdsi standardized vocabulary and modern machine learning capabilities, this hybrid approach will surpass existing manual and purely data-driven methods in both interpretability and scalability. Ultimately, this work will reduce the need for manual knowledge curation, and the tools developed will be broadly applicable to other tasks such as terminology quality assurance, cohort definition review, and clinical documentation support.
Quick answers
Who received award F31LM015316?
F31LM015316 was awarded to The Trustees Of Columbia University In The City Of New York by National Institute of Environmental Health Sciences (National Institutes of Health) on February 28, 2027.
How much is award F31LM015316 worth?
$50,114 has been obligated on F31LM015316.
On GovTribe
See this award's full record
- Every transaction and subaward
- The recipient's other awards and teaming partners
- Similar grant awards
- Ask GovTribe AI about this award