2. Agenda
• About Clinical Decision Support Services
• Experience of the Clinical Decision Support
Consortium
• Standards Efforts Underway to Make Them
Widely Available
• Opportunities and Challenges
3. About Clinical Decision Support Services (CDSS)
Outside the Cloud Inside the Cloud
CDSS
Firewall
Data Normalization and
Classification
Services
Cloud-based
Clinical
Decision
Support
Services
CDSS
EHR
Consumer
PATIENT DATA
ASSESSMENTS and
RECOMMENDATIONS
1) Externalizes application of CDS Logic that can provide assessments and
guidance
2) Externalizes curation of clinical knowledge to a CDSS provider or their
respective content supplier
3) EHR vendor is responsible for making it possible to send data in the
appropriate workflow context and receive assessments and
recommendations from CDSS
4) EHR vendor is responsible for making it possible to insert CDSS guidance
into the appropriate EHR workflow context
5) Implementing consumer still needs to determine insertion, support
ongoing semantic harmonization
4. Clinical Decision Support Consortium (CDSC)
1. Knowledge Management Life Cycle
2. Knowledge
Specification
3. Knowledge Portal and
Repository
4. CDS Public Services
and Dashboard
5. Evaluation Process for each CDS Assessment and Research Area
6. Dissemination Process for each Assessment and Research Area
• Knowledge management lifecycle
• Knowledge specification
• Knowledge Portal and Repository
• CDS Knowledge Content and Public Web Services
• Evaluation
• Dissemination
Led by Dr. Blackford Middleton
AHRQ funded from 2008-2013
5.
6. CDSC Conceptual Approach
CDSC Evidence-based Guidelines (e.g., DM, HTN, CAD)
Level 1
Translation
Dissemination
Level 2 and Level 3
Specifications
CDS Services
Provider
Dashboard
Developer
Dashboard
KM Portal and
Repository
EMR End user
access
Refinement
Performance Measures
Collaboration Collaboration
NextGen
Centricity
Regenstrief
Partners LMR
8. Legal Agreements Developed to
Address Liability Points of Failure
• CDS manufacturing defect
– Software does not perform as designed
– i.e. alerts fail to notify due to gap in software or service Device
– CDS supplier must be able to audit/trace all guidance provided
• CDS implementation defect
– Customer implementation of software results in defective functioning
– i.e. alerts fail to fire because customer has incorrectly implemented services
insertion or failed to notify CDS supplier that their dictionary changed
• CDS user error
– Software performs as designed, customer has implemented correctly, however
user does not utilize correctly
– i.e. user ignores alert, turns off alerts, fails to notice alert
– Blurred distinctions here because users typically blame CDS manufacturer or
implementation team for creating unusable CDS.
– Legal precedent to date still renders the Provider accountable for determining if
the CDS guidance is appropriate because the Provider has the richer context of the
patient to interpret the relevance of the guidance
9. CDSC Services Rule Building Blocks
Problem Classes Drug Classes
Classes of Observations
And Test Results
Indication
State
Inferences
Goal
State
Inferences
Contraindication
State
Inferences
Infobutton
Knowledge
Access
Diagnostic
Testing
Care
Management
Observation
Dictionaries
LOINC or SNOMED
Order
Classes
Drug
Dictionaries
RX Norm
Order
Catalogues
Patient
Assessment
Problem Dictionaries
SNOMED
Risk
State
Inferences
Recommendations
Guideline
Context
Family
History
Patient
Preference
Phenotypic
State
Genotypic
State
Classes
Of
Raw Data
Inferred Patient
Context (Clinical State
Rules)
Patient
Education
10. Qualifying for
ACEI
ARB
Drug Class
ACEi
Drug Class
Contra
Indication
To ACEi/ARB
Allergy to
ACEi/ARB
Clinical State
Rule
Pregnant
Clinical State
Rule
Pregnant
Prob
Class Subset
Preg.
Complications
Prob Class Subset
Low BP
Clinical State
Rule
Low BP
Prob
Class Subset
Hyperkalemia
Disease
State Rule
Hyperkalemia
Prob
Class Subset
NOTNOT NOT
Non-Gest
DM Disease
State Rule
DM Disease
State Rule
DM
Prob Class Subset
DM Complications
Prob Class Subset
Gest DM
Disease
State Rule
Gest DM
Prob Class Subset
NOT
Simple Rules are NOT Simple:
IF in non-gestational DM pt with non-ESRD qualifying for ACEi
Key:
___ AND
. . . OR
Non-ESRD
CRF Disease
State Rule
CRF
Prob
Class Subset
Creat >2
w/in 12 mo
Calc Rule
GFR<50
Calc Rule
Proteinuria
Disease State
Rule
Malb/cre>30
Calc Rule
Proteinuria
Prob Class
Subset
ESRD
Disease
State Rule
ESRD Prob
Class Subset
Dialysis Comp
Prob
Class Subset
&NOTO
R
11. CDSC, ACDS, HL7 and Other Standards
ONC S&I Health eDecisions Use Case 1
– Data Model for CDS Artifact
Authoring
– Knowledge Representation for
CDS Artifacts
– Artifacts can be value sets
(groupers), rules, order sets,
documentation templates, etc
– Enable interoperable knowledge
sharing
CDS Artifact
Sharing Use Case
FR & Data
Elements
VMR
GEM
eRECS
CDSC
L3
HL7
Order
Set
Model
SHARP
ARDEN
Inputs
Use Case 1: CDS Artifact Sharing
HeD
Artifact
Sharing
Standard
Harmonization
and Modeling
for 3 Artifact
Types
CREF
12. CDSC, Open CDS, HL7
ONC S&I Health eDecisions Use Case 2
• Use Case 2: CDS Guidance Services
– CDS Services Insertion
– Patient Data Output by EHR/PHR system to
CDS Guidance Service
CDS Guidance
Services Use Case
FR & Data
Elements
HL7
Consolidated
CDA
Other
Standards
HL7 DSS
Services
Inputs
HeD CDS
Guidance
Services
Standards
Harmonization
and Modeling
HL7
Context Aware
Information
Retrieval
13. Opportunities and Challenges
• EHR vendors can’t expect their customers to curate
all this knowledge inside their EHR walls
Immunizations, Genetics, Genomics, Personalized
Medicine
• Few EHR CDS systems can actually execute the kind
of inferencing required for personalized medicine
• Hence, Memorial Sloan Kettering is training IBM
Watson Personalized Cancer Rx service
• We need to advocate for EHR vendors to move
beyond “walled gardens of simple CDS”