linked open data for medical guidelines interactions
TRANSCRIPT
@AIME, Pavia, 19th June 2015
Linked Open Data for Medical Guidelines Interactions
Veruska Zamborlini, Marcos da Silveira, Cedric Pruski, Annette ten Teije and Frank van Harmelen, Rinke Hoekstra
Epidemiology of multimorbidity and implications for health care, research, and medical education: a cross-sectional study (2012)
Karen Barnett, Stewart W Mercer, Michael Norbury, Prof Graham Watt, Prof Sally Wyke, Bruce Guthrie
Project Smart WardExpectation:
multi-morbidity holds for psychiatric patients as well
In practice:✤ Paper-based Clinical Guideline
(CG)
✤ One guideline per disease
✤ Common co-morbidities (2) are addressed during CG development
✤ Not suitable for detecting interactions
What about…✤ Computer-interpretable
Clinical Guideline (CIG)✤ How to combine CIGs?
✤ CIG Languages: ✤ mainly designed for execution✤ not suitable for detecting
interactions
What do we propose?✤ Address multimobidity at CG level:
✤ scalable in number of guidelines✤ reusable rules designed for diverse types of interactions✤ binary cumulative rules - allows for combination of n
recommendations
✤ represent recommendations based on transitions promoted by actions✤ do give aspirin x don’t give aspirin
=> different recommendations about the same action✤ hierarchy of actions✤ causation beliefs
✤ reuse of existent background knowledge available online (LOD)
Case Study
Case Study
Case Study
Case Study
Case Study
Case Study
Case Study
Case Study
Reusable Rules IF Positive recommendation R1 to action A1 & Negative recommendation R2 to action A2 & Actions A1 and A2 are the same or subsuming one
anotherTHEN R1 and R2 might contradict each other
FOL Rules
Systematic Analysis
Systematic Analysis
Case Study
Conclusion so far✤ Address multimobidity at CG level:
✤ reusable/domain-independent rules for detecting types of interactions
✤ scalable in number of guidelines
Next step✤ Re-use of existent background knowledge available
online (LOD)
Ibuprofen
Incompatible Drugs
Drugbank
Aspirin
Ibuprofen
Incompatible Drugs
Drugbank
Aspirin
Ibuprofen
Incompatible Drugs
Drugbank
Aspirin
Incompatible
Ibuprofen
Drugbank
Aspirin
Anti-platelets
Epoprostenol
Ibuprofen
Drugbank
Aspirin
Anti-platelets
Epoprostenol
Ibuprofen
Drugbank
Aspirin
Anti-platelets
Epoprostenol
Alternative Drug
Alternative Drug
Ibuprofen High Blood Pressure
Side-EffectSide-Effect
Sider
ThiazideHigh BloodSugar Level
Ibuprofen High Blood Pressure
Side-EffectSide-Effect
Sider
ThiazideHigh BloodSugar Level
Ibuprofen High Blood Pressure
Side-EffectSide-Effect
Sider
ThiazideHigh BloodSugar Level
Ibuprofen High Blood Pressure
Side-EffectSide-Effect
Sider
ThiazideHigh BloodSugar Level
Side-Effect
Side-Effect
Using LOD
Using LOD
Using LOD
Using LOD
Using LOD
Using LOD
Using LOD
Using LOD
Our guideline model in LOD using NanoPublications standard
NanopublicationProvenance about the publication:When, by whom, how this publication was produced…
“Atomic” piece of information:E.g. causation beliefs, recommendations.
Provenance about the assertion:Who/where it was originally asserted;In case the source is a text, the specific piece od text can also be pointed out.
Nanopublication
Nanopublication
Conclusion ✤ Guideline model: represent recommendations
based on transitions promoted by actions(hierarchy of actions & causation beliefs)
✤ Address multimobidity at guideline level✤ scalable in number of guidelines✤ reusable rules designed for diverse types of interactions
✤Semantic web technology✤ reuse of existent background knowledge available online (LOD)✤ use nano publications for guideline model
Smart Ward: use of guidelines
Smart Ward: interactions detection independent
of specific diseases
Relevant technology