Jonathan Moran Sierra
Program: Biomedical Informatics and Data Science
Current advisor: Philip Payne, PhD, FACMI
Undergraduate university: University of Puerto Rico-Rio Piedras
Research summary
Most valuable information in the Electronic Health Record is embedded in free text.
Thus, data that can be crucial for development of personalized health plans can be buried under an overload of information in a patient chart. By using Large Language Models we can optimize extraction of the most relevant information required by clinicians to make a health management change. Ideally these methods can be applied to the surgical space; by integrating information from clinical notes in addition to pre-operative lab values we can build better personalized care plans for surgical patients. This would reduce intra-/post-operative complications, decrease recovery time, and reduce morbidity and mortality in many cases.
Graduate publications
Bhattarai K, Oh IY, Sierra JM, Tang J, Payne PRO, Abrams Z, Lai AM. 2024 Leveraging GPT-4 for identifying cancer phenotypes in electronic health records: a performance comparison between GPT-4, GPT-3.5-turbo, Flan-T5, Llama-3-8B, and spaCy’s rule-based and machine learning-based methods. JAMIA Open, 7(3):ooae060.