Goaldriven management of interacting clinical guidelines for multimorbidity
Goal-driven management of interacting clinical guidelines for multimorbidity patients Alexandra Kogan , University of Haifa, Alexandra. Kogan. Oem@gmail. com Mor Peleg, University of Haifa, Peleg. Mor@gmail. com Samson W Tu, Stanford University, SWT@stanford. edu AMIA 2018 Annual Symposium, San Francisco, November 2018 Israel Science Foundation
Why does guideline-based CDS need to address multi-morbid patients? • Patients with multiple morbidities are increasingly prevalent • High complexity and Large quantities of knowledge involved in treating such patients • Computerizing guidelines (CIGs) provides patient-specific evidence-based advice at the point of care • Most CPGs are designed to address single diseases • CIGs need to utilize multiple CPGs CVD Prevention DU treatment Start Aspirin Stop Aspirin
Background • Others considered detecting interactions or finding nonconflicting solutions (not both) • Only one other solution considered goals • None used ontologies and standards Mitigation of adverse interactions in pairs of clinical practice guidelines using constraint logic programming Analyzing recommendations interactions in clinical guidelines
Aims • Development of a modular and reusable goal-based method to combine CIGs with general medical knowledge • Use of standard Ontologies and terminologies (reuse and scalability) • Utilization of the knowledge of drugs’ physiological effects and therapeutic usage • Incremental detection and management of interactions
A case study: Cardiovascular disease vs Duodenal ulcer aspirin, or other antiplatelet drug if not tolerated/contraindicated CVD 2 prevention DU 2 prevention
Methods Patterns General Medical Knowledge ALIUM(PROforma CIGs) Medical Knowledge Base (SNOMED, NDF-RT) Controller Treatment Specific patient information Patient Database Communication with the user
PROforma CIGs to Goal Trees • CPGs are encoded in PROforma to accommodate the structure of a hierarchical goal tree • The CIGs contain metaproperties that enable the transition of information from PROforma to the controller • The Formulation of categories of goals in NDF-RT terms • The Controller coordinates interactions among CIGs while considering new diagnoses, adverse events, medications <may-treat> <may-prevent> <Decrease-Platelet-Aggregation>
The Controller Incrementally Coordinates Possibly Contradictory Recommendations based on a Collection of Behavioral Patterns • Illustrative Patterns: • P 1: responding to a new goal • P 2 a: Responding to a Medication Request proposal • P 3: Response to an unsatisfied goal • Include users in decision making
Patterns: P 1: responding to a new goal Patient is on Aspirin for CVD 2 prevention DU developed due to NSAIDS Caregiver Interface Controller with Knowledge Base and EHR DU CIG Tree DU CIG G 2. 2 treatment of DU due to NSAIDs No goal inconsistencies (increase/decrease); G 2. 2 accepted CVD CIG Tree CVD CIG
Patterns: P 2 a: Responding to a Medication Request proposal Caregiver Interface Controller with Knowledge Base and EHR DU CIG Tree DU CIG CVD CIG Tree CVD CIG MR 2 Stop NSAIDs G 2. 2 planned; Action Inconsistency MR 2, MR 1 Inconsistent-Meds: Stop Aspirin or Discard G 2. 2 Accept MR 2 Stop Aspirin accepted; Tree. Log MR 2; G 2. 2 in progress
Patterns: P 3: Response to unsatisfied goal Caregiver Interface Controller with Knowledge Base and EHR DU CIG Tree Inconsistency detected: Goal G 1. 6. 1 unsatisfied (antiplatelet for CVD) Discard or Rerun G 1. 6. 1 DU CIG CVD CIG Tree CVD CIG
Patterns: P 4: Run CIG and Update Goal Tree Caregiver Interface Controller with Knowledge Base and EHR DU CIG Tree DU CIG CVD CIG Tree CVD CIG PROforma performs a decision regarding Medication Request MR 3: Clopidogrel (preferred); G 1. 6. 1 planned No more recommendations; Tree. Log MR 3(Clopidogrel) for G 1. 6. 1
Patterns: P 2 b: Respond to Medication Request proposal Caregiver Interface Controller with Knowledge Base and EHR DU CIG Tree DU CIG MR 3: Clopidogrel proposed for goal G 1. 6. 1 No inconsistencies with preferred action G 1. 6. 1: MR 3: Clopidogrel (preferred) MR 3(Clopidogrel) MR 3 Clopidogrel accepted; Tree. Log MR 3; G 1. 6. 1 in progress CVD CIG Tree CVD CIG
Summary • Goal-based approach for integrating CIGs with general medical ontologies • Controller implementing patterns for finding inconsistencies and mitigating them • Future work: • Detecting Adverse Events • Addressing temporal constraints in medications, procedures and recommendations
Thanks! Questions? Mor Peleg, U. Of Haifa Samson W Tu , Irit Hochberg, Stanford University RAMBAM We thank Deontics and John Fox Alexandra Kogan, U. Of Haifa
- Slides: 17