©2011 MFMER | slide-1
Developing, Implementing, and Sharing Pharmacogenomics CDS
Robert R. Freimuth, Ph.D.Biomedical Statistics and Informatics,Applied Clinical Informatics
AMIA TBI – March 18, 2013
©2011 MFMER | slide-2
Pharmacogenomics (PGx) CDS
• Mayo Clinic PGx CDS
• Case Study: Abacavir
• Challenges
• Challenges for Genomics-Based CDS
• Sharing Knowledge
• Standardization
• Challenges of the Future
©2011 MFMER | slide-3
EHR at Mayo Clinic, Rochester
• >12 different systems that store primary data
• GE
• Started from a partnership with HP in late 1980s
• Mainframe design
• Blaze Advisor rule engine
• Built for large-scale financial applications (FICO)
• Designed for complex, multi-faceted problems
• Full forward and backward chaining logic
• Tightly coupled to data sources
©2011 MFMER | slide-4
CDS at Mayo Clinic, Rochester
• Developing CDS solutions for >20 years
• Initially developed over home-grown EHR
• Batch-oriented rules for events and conditions
• Standalone systems for different departments• e.g., Pharmacy, Infectious Disease
• Started using Blaze ~2005
• Provided real-time CDS
• >100 high-level rules composed of ~1000 atomic rules
©2011 MFMER | slide-5
CDS Rule Development Process
• Synthesis of clinical knowledge
• Committee approvals
• Technical specifications
• UI screen design
• Web service calls, database queries
• Implementation and deployment
• Compiling
• Loading to rule servers on mainframe
• Run modes
• Silent
• "Live"
©2011 MFMER | slide-6
Abacavir and HLA-B*5701
• Abacavir
• Nucleoside RT-inhibitor for HIV infection
• HLA-B
• >1500 known variants
• *5701 allele is associated with life-threatening hypersensitivity reactions to abacavir
• Results are either "positive" or "negative"
• At Mayo Clinic, *5701 screening is recommended:
• Prior to initiating treatment with abacavir
• Prior to reinitiation of abacavir in patients of unknown HLA-B*5701 status who have previously tolerated abacavir
©2011 MFMER | slide-7
Abacavir-HLA B*5701: Governance
• Executive sponsors:
• Center for Individualized Medicine
• Pharmacogenomics Task Force
• Mayo Pharmacy Formulary Committee
• Disease-Oriented Task Force: ID
• Other committee approvals:
• Practice Convergence Council (MC-CPC)
• MC Clinical Decision Support Subcommittee
• Implementation
• Site-based P&T committees, coordinate with CIM
• Local CDS teams
©2011 MFMER | slide-8
Abacavir-HLA B*5701 Workflow
Internal lab result?
External lab result?
What is it?
Capture result
Inbox message
Prompt for test
©2011 MFMER | slide-9
Abacavir-HLA B*5701 Results
• Text report
• MCF, MCA, MCHS
• Structured data
• MCR
• Storage of results
• Internal: lab test
• External: custom table
• Added as an allergy
©2011 MFMER | slide-10
Challenges: Abacavir
• Currently using the Allergy model
• Can be deleted or altered
• Policy for positive test/deleted allergy
• Not a true allergy
• Reaction has not occurred, but is possible
• Storing external results in a custom table
• Inaccessible by other systems (e.g., pharmacy)
• Providers can't see entries from prior admissions
©2011 MFMER | slide-11
Challenges: External Results
• Capture
• Streamline current alert-heavy process
• Representation
• Test orders and results change over time
• Storage
• "Clinical event" (MCF/MCA)
• "Observation/finding" (MCR)
• Exchange
• CDA (partial solution for providers?)
• Clinical notes (patient-reported data)
©2011 MFMER | slide-12
Challenges: Enterprise-Level
• Coordination across the Mayo enterprise (multi-site)
• Group formed to focus on challenges related to the implementation of PGx CDS
• Identify issues with our current infrastucture
• Identify bottlenecks in processes
• Propose short and long term solutions
• Consensus for functional requirements
• Abacavir vs. Carbamazepine
©2011 MFMER | slide-13
Challenges: Education
• Clinical knowledge
• Proactive: Videos, articles
• On-demand: AskMayoExpert (AME)
• Contextual: Alerts link to AME
• Rule implementation
• Consequences of going against recommendation
• "If the patient has been on abacavir for 10 years, do I still need to order the test?"
• Workflow
• "If I click yes (or no), how many more alerts will pop up?"
©2011 MFMER | slide-14
Generalizing
A
T C
G
N
©2011 MFMER | slide-15
Measuring Outcomes of PGx CDS
• Evaluating effectiveness
• What data should be captured?
• Case-by-case decision?
• Increased effectiveness of therapy
• Decreased adverse reactions or complications
• Fiscal responsibility
• Over-testing
• Time to resolution
• Increased efficiency of healthcare delivery
©2011 MFMER | slide-16
Refining PGx CDS
• Create highly specific triggers
• Limit by drug, genotype
• Ethnicity?
• Requires accurate data, not always available
• Probabilities?
• Allele, adverse reaction, effective treatment
• Context-sensitive
• IL28B, peginterferon, and hepatitis C
• Timing: diagnosis vs. CPOE
©2011 MFMER | slide-17
Increasing Complexity of Genomics CDS
• Evolution from "simple" CDS to guided workflows
• Context-sensitive expert systems
• Require provider input
• Prompt for data (e.g., family history)
• Technical limitations in popup alerts
• Important for complex decisions
• Need to capture clinical judgement
• Not limited to genomics…
©2011 MFMER | slide-18
Sharing Knowledge…
• Human-readable text
• PharmGKB
• Structured representations
• TPP/CPIC tables
• Semantically computable
• PGx Guideline Repository
• Implementation-independent algorithms
• Generalized syntax, metadata
• Executable code (local adoption)
Tuesday
Posters
©2011 MFMER | slide-19
Sharing Knowledge… Is Hard
• Who?
• Many stakeholders
• What?
• Institutional vs. national standards of care
• Not everything is generalizable
• Where?
• Distributed vs. Centralized
• How?
• Knowledgebases
• Code libraries
• Web services
Main
tenance
Governance
©2011 MFMER | slide-20
Standardization
• Guideline development and interpretation
• Clinical practice and standard of care
• Allele definitions (molecular)
• Nomenclature is only part of the solution
• Stability
• Inferrence
• Phenotype definitions
• Genotype-phenotype translation
• Test results
• Sequence vs. interpretation
• CDS rules
HealtheDecisions
©2011 MFMER | slide-21
Challenges of the Future
• Combinatorial expansion, complexity
• Gene-Drug / Gene-Drug Class
• Gene-Gene-Drug / Pathway-Drug
• Advances in knowledge
• Stable representation of results
• Sequence? Interpretation?
• Reinterpretation process
• Store updated results
• Educate providers
• Advances in technology
• Merge, reconcile results
©2011 MFMER | slide-22
Challenges of the Future
• Genome on a thumb drive
• The ultimate in patient-provided data
• Patient mobility
©2011 MFMER | slide-23
Acknowledgements – Abacavir Implementation
• Michelle Elliott, MD
• Joseph Paul, MD
• David Blair, MD
• John Black, MD (DLMP)
• Joseph Yao, MD (DLMP)
• Robert Freimuth, PhD (Knowledge Engineering)
• Jyotishman Pathak, PhD (Knowledge Engineering)
• Mark Siska, RPh
• Kelly Wix, DPh
• Pedro Caraballo, MD (GIM)
• Robert Bleimeyer (Blaze, MN)
• Mark Dobie (Blaze, MN)
• Padma Rao (Blaze, MN)
• Liz Clark (Cerner, AZ)
• Gaurav Jain (Cerner, MCHS)
• Charles Pugh (Cerner, FL)
• Cloann Schultz (PM, CIM)
• Caer Vitek (Education, CIM)
• Donald Gabrielson (PM)
• and many others