©2013 MFMER | slide-1
A Sematic Web-Based Framework for Quality Assurance of Electronic Medical Records Data for Secondary Use
Guoqian Jiang, Harold Solbrig, Christopher Chute Mayo Clinic W3C RDF Validation Workshop September 10, 2013
©2013 MFMER | slide-2
Secondary use of EHR data • The secondary use of electronic health records
(EHR) sourced data is a broad domain. • It includes patient safety and clinical quality metrics
and development programs • The entire categories of clinical and translational
research are fundamentally dependent on effective secondary use of clinical information, including
• clinical trials, • observational cohorts, • outcomes research, • comparative effectiveness, • and best evidence discovery.
©2013 MFMER | slide-3
The dimensions of data quality and methods of data quality assessment
Source from: Weiskopf NG, Weng C. J Am Med Inform Assoc. 2013
©2013 MFMER | slide-4
Use of Data Standards • A central part of good data practice that
improves data quality • Meaningful Use
• A national strategy in the US • A roadmap for the advancement of health
information exchange • Using certified EHR technology to improve
quality, safety, efficiency and reduce health disparities
• http://www.cms.gov/Regulations-and-Guidance/Legislation/EHRIncentivePrograms/Meaningful_Use.html
©2013 MFMER | slide-5
Clinical Information Modeling Initiative (CIMI)
• an international collaboration effort, • has been actively working on developing a
common format for detailed specifications for the representation of health information content
• http://informatics.mayo.edu/CIMI/index.php/Main_Page
©2013 MFMER | slide-6
CIMI Information Model Ideas
Source from: Dr. Stan Huff, CIMI
©2013 MFMER | slide-7
Standardized clinical information models for secondary use
• There is an emerging interest in the use of clinical information models in the context of secondary use of Electronic Medical Records (EMR) data
• The Office of the National Coordinator (ONC)’s Strategic Health IT Advanced Research Projects Area 4 (SHARPn) project has adopted Intermountain Healthcare clinical element models (CEMs) for normalizing patient data extracted from EMR systems
• http://sharpn.org
©2013 MFMER | slide-8
SHARPn Project Organization (Secondary Use of EHR Data)
©2013 MFMER | slide-9
Clinical Element Models • Developed by the Intermountain Healthcare • A model for describing and representing
detailed clinical information through defining a standard data structure
• Presently, over 4,000 CEMs, such as blood pressure measurement or specific laboratory tests, are defined.
• Authored in the Constrain Definition Language (CDL)
• Distributed in XML schema (XSD)
©2013 MFMER | slide-10
CEMs in CDL
CEMs in XSD
Patient Data in EMRs
Instance Data in XML
CEMs in OWL Instance Data in RDF
SPARQL Inference Notation (SPIN) Framework
Semantic Web Layer
Current Normalization
Pipeline
System Architecture
©2013 MFMER | slide-11
©2013 MFMER | slide-12
CEMs in XSD -> OWL
©2013 MFMER | slide-13
Instance Data in XML -> RDF
©2013 MFMER | slide-14
Check constraints and perform data validation Define SPIN constraints
©2013 MFMER | slide-15
Check whether the units are identical cem:SecondaryUseNotedDrug
©2013 MFMER | slide-16
Check whether lowerLimit is greater than upperLimit
cem:SecondaryUseNotedDrug
©2013 MFMER | slide-17
Constraint Violation Warning When (takenDoseLowerLimit.PQ.unit.value “TAB” = takenDoseUpperLimit.PQ.unit.value “TAB”) and (takenDoseLowerLimit.PQ.value “3” > takeDoseUpperLimit.PQ.value “2”)
©2013 MFMER | slide-18
Standardized Semantic Web Services • We develop a CIMI-compliant semantic web
representation and services for clinical information models
• Convert the CIMI reference model in the UML to OWL
• Generate domain templates from the BRIDG model (which is a domain analysis model in the domain of clinical research)
©2013 MFMER | slide-19
The Parts of the CIMI Reference Model • Core Model • Party Model – with HL7 V3 RIM mappings • Data Value Types • Primitive Types
©2013 MFMER | slide-20
CIMI Reference Model -‐ Core class CIMI Core Model
COMPOSITION
category :CODED_TEXTlanguage :CODED_TEXTjurisdiction :CODED_TEXT
CONTENT_ITEM
ENTRY
language :CODED_TEXT
SECTION
ARCHETYPED
archetype_id :Stringrm_version :String = 1.0.11 {readOnly}
LOCATABLE
archetype_node_id :Stringname :String
LINK
meaning :TEXT
ITEM
ELEMENT
null_flavor :CODED_TEXT [0..1]value :DATA_VALUE [0..1]
CLUSTER
structure_type :CODED_TEXT [0..1]
PARTICIPATION
function :CODED_TEXTdetails :ITEM [0..*]
CORE_LOCATABLEPARTY
details :ITEM [0..*]
Name: CIMI Core ModelAuthor: CIMI MTFVersion: 1.0.10Created: 30/11/2012 8:32:29 PMUpdated: 9/04/2013 1:29:40 PM
data 1..*
item
0..*
archetype_details
0..1
sourcelink
0..*
content0..*
item
1..*
participation
0..*
party
1..1
target
1
©2013 MFMER | slide-21
CIMI Reference Model – Data Values
©2013 MFMER | slide-22
CIMI Reference Model in OWL • http://informatics.mayo.edu/ontologies/cimi/cimi-
rm-1.0.12b.owl
©2013 MFMER | slide-23
Domain Templates in CIMI-compliant Representation Using Semantic Web
©2013 MFMER | slide-24
©2013 MFMER | slide-25
BRIDG Associations
BRIDG Attributes
Mindmap View
©2013 MFMER | slide-26
cimi:ENTRY
cimi:CLUSTER
cimi:ELEMENT
©2013 MFMER | slide-27
Discussion • RDF validation mechanism against CIMI models
• Model level (CIMI models) • Instance level (Patient data)
• Challenging issues • Data types • Value set binding
• XML Semantics Reuse technology • XSD -> OWL • XML -> RDF
©2013 MFMER | slide-28
BRIDGing Technology
? Object Constraint Language (OCL)
UML
SWRL/SPARQL Rules
RDF/OWL
XML Rules
XSD/XML
©2013 MFMER | slide-29
Questions & Discussion
