| OCR Text |
Show Automating FHIR Element Recognition to Enhance Semantic Mapping in Oncology Informatics Addison Stuart, BSN, RN Key Findings: Strategies were developed for protecting URIs in a FHIR model from human error. Displaying a concept map FHIR URI in the same screen as it is required by workflow steps reduced click count. The incorporation of dropdown lists reduced risk for errors in FHIR URI creation. Iterative feedback loops significantly enhanced the design and functionality of implemented features. Internally-deployed LLMs streamlined the development of complex queries, facilitating efficient workflow. Background Results • Fast Healthcare Interoperability Resources (FHIR) standards are essential in health informatics • Concept maps in a FHIR model match EHR terms to standards (LOINC, RxNorm, SNOMED CT, etc) • FHIR Uniform Resource Identifier (URI): a unique address for data objects in a FHIR model • FHIR URIs are needed to save concept maps in the correct location with a FHIR model • Human error in accessing or creating a FHIR URI is a threat to FHIR database integrity • Automated tests for FHIR Device profile yielded no critical errors • SQL query executed successfully across various contexts, retrieving FHIR URIs for concept maps • FHIR URI Display stakeholder feedback led to improved design and implementation • FHIR URI Display reduced Click Count for finding concept map FHIR URI from 9 clicks to 1 click • Collaborative workflow analysis of FHIR URI Builder demonstrated reduction in risk for errors • Both interface improvements worked successfully when tested across various contexts FHIR URI Display Methods Conclusions • Created FHIR profile for Device resource o Used Forge and Simplifier by Firely o Allowed discovery of FHIR URI location in database • Wrote SQL query to find a concept map’s FHIR URI o Internally-deployed LLM used for SQL help o Tested query with various concept maps and tenants • Improved use of FHIR URI in Informatics tooling o FHIR URI Display in concept map publish screen o Provides information in context required by workflow o FHIR URI Builder uses dropdowns to prevent errors o Queries Simplifier API for resources in FHIR model • All project objectives completed successfully • Iterative demo and feedback cycles led to improved design and functionality • Needed info in right context reduces click count • Critical data objects can be protected from errors by limiting naming options with dropdown lists • Internally-deployed LLMs reduce barriers to development of complex queries • Publications are needed detailing best practices for FHIR URI creation and management • Future projects should seek to automate the providence of a FHIR URI into the FHIR model FHIR URI Builder COLLEGE OF NURSING Automating FHIR Element Recognition to Enhance Semantic Mapping in Oncology Informatics Background • Fast Healthcare Interoperability Resources (FHIR) standards are essential in health informatics • Concept maps in a FHIR model match EHR terms to standards (LOINC, RxNorm, SNOMED CT, etc) • FHIR Uniform Resource Identifier (URI): a unique address for data objects in a FHIR model • FHIR URIs are needed to save concept maps in the correct location with a FHIR model • Human error in accessing or creating a FHIR URI is a threat to FHIR database integrity Source: https://www.hl7.org/fhir/ ©UNIVERSITY OF UTAH HEALTH COLLEGE OF NURSING Automating FHIR Element Recognition to Enhance Semantic Mapping in Oncology Informatics Methods • Created FHIR profile for Device resource o Used Forge and Simplifier by Firely o Allowed discovery of FHIR URI location in database FHIR URI Display • Wrote SQL query to find a concept map’s FHIR URI o Internally-deployed LLM used for SQL help o Tested query with various concept maps and tenants • Improved use of FHIR URI in Informatics tooling o FHIR URI Display in concept map publish screen o Provides information in context required by workflow FHIR URI Builder o FHIR URI Builder uses dropdowns to prevent errors o Queries Simplifier API for resources in FHIR model ©UNIVERSITY OF UTAH HEALTH COLLEGE OF NURSING Automating FHIR Element Recognition to Enhance Semantic Mapping in Oncology Informatics Results • Automated tests for FHIR Device profile yielded no critical errors • SQL query executed successfully across various contexts, retrieving FHIR URIs for concept maps • FHIR URI Display stakeholder feedback led to improved design and implementation • FHIR URI Display reduced Click Count for finding concept map FHIR URI from 9 clicks to 1 click • Collaborative workflow analysis of FHIR URI Builder demonstrated reduction in risk for errors • Both interface improvements worked successfully when tested across various contexts Workflow analysis to identify risk for FHIR URI pain points and errors. ©UNIVERSITY OF UTAH HEALTH COLLEGE OF NURSING Automating FHIR Element Recognition to Enhance Semantic Mapping in Oncology Informatics Conclusions • All project objectives completed successfully • Iterative demo and feedback cycles led to improved design and functionality • Needed info in right context reduces click count • Critical data objects can be protected from errors by limiting naming options with dropdown lists • Internally-deployed LLMs reduce barriers to development of complex queries • Publications are needed detailing best practices for FHIR URI creation and management • Future projects should seek to automate the providence of a FHIR URI into the FHIR model A diagram of the steps in the query needed to fetch a concept map’s source FHIR URI ©UNIVERSITY OF UTAH HEALTH COLLEGE OF NURSING Automating FHIR Element Recognition to Enhance Semantic Mapping in Oncology Informatics Epilogue I finished the last stage of this project on February 29th. st *On March 1 , the company was abruptly shut down. Although my interventions will not enjoy the long tenure in Prod that I was hoping for, I will apply the lessons I have learned about leveraging FHIR URIs throughout my informatics career. *Unrelated to project completion or implementation ©UNIVERSITY OF UTAH HEALTH COLLEGE OF NURSING Automating FHIR Element Recognition to Enhance Semantic Mapping in Oncology Informatics Questions? Thank you! ©UNIVERSITY OF UTAH HEALTH COLLEGE OF NURSING |
