GLINDA: Automated Reasoning for Application of Clinical Guidelines BMIR Research-in-Progress Presentation Csongor Nyulas Samson Tu
Acknowledgement Funder: National Library of Medicine Project Members Mark Musen Mary Goldstein Susana Martins Hyunggu Jung Pamela Kum
Problem Statement Populations are aging worldwide Older adults tend to have multiple chronic conditions Data?? Management of multiple comorbidities presents a challenging problem Almost all clinical practice guidelines focus on the management of single diseases May take comorbidities into account Simultaneous application of multiple guidelines leads to suboptimal care
Research Goals Develop a modular and extensible platform for exploring informatics and clinical issues Integrate and reuse best-of-breed knowledge resources and applications Create methods for detecting, repairing and integrating treatment recommendations from multiple guideliens
Method Adapt BioSTORM agent architecture Task decomposition Problem-solving method Reuse ATHENA CDSS Clinical domains: Hypertension, diabetes mellitus, heart failure, hyperlipidemia, chronic kidney disease Develop new agents for detecting, repairing, and integrating treatment recommendations Apply methods on anonymized patient cases from the Stanford STRIDE database
Outline of Method Section STRIDE patient selection and preparation BioSTORM agent architecture and its application to GLINDA ATHENA DSS agents New agents for detecting, repairing and integrating guideline recommendations Presentation for review
STRIDE Patient Extraction
Test Patients Selection
Data Preparation
BioSTORM Agent Architecture
GLINDA Agent Configurations
ATHENA CDSS
Integrating Recommendations from Multiple Guidelines
Detecting Interactions
Current Status
Future Work