Country / Region
Americas
Tags
Artificial intelligence, Data analytics, Innovation, Mapping, Research
Hospitals in the United States often implement fundamentally incompatible schemas and coding systems, with majority of healthcare organizations unable to effectively share structured clinical data across institutional boundaries, severely limiting data uplift for policymaking, resource planning, and budgeting. We present HARMONY (Healthcare Alignment and Record Modernization via Ontologies and Neural sYstems), a method for standardizing healthcare records across hospitals and states leveraging SNOMED CT.
Description
Patient procedures are often recorded in inconsistent formats by hospitals in the United States, making it difficult to use this data effectively for policymaking, resource planning, and budgeting. We present HARMONY (Healthcare Alignment and Record Modernization via Ontologies and Neural sYstems), a method for standardizing healthcare records across hospitals and states. We evaluate HARMONY through embedding space metrics and expert validation. Results show that HARMONY consistently outperforms baseline methods, producing more accurate, consistent, and reusable standardized records-enabling robust applications like state-wide dashboards and national health data integration.
Scope
SNOMED CT consistently provides a baseline reference for standardized description and relationship of healthcare entities describing clinical procedures, in addition to diseases, symptoms, treatments and cause-and-effect functionalities. We are aiming to resolve the diversity of reporting standards adopted by various hospitals so that we can compare data reported by them at the national level; and present a clear picture of the sate of healthcare services in the United States. SNOMED CT is an obvious choice for this effort.
How SNOMED CT will be used
HARMONY leverages large language models (LLMs) with in-context prompting to structure free-text medical descriptions, and uses embedding-based similarity with the SNOMED CT knowledge graph to map them to standardized concepts. It covers a wide range of clinical data-including tests, procedures, medications, and diagnoses-from diverse hospital systems.
Why SNOMED CT will be used
Contact
