Part I: Introduction to SHARPn Normalization Hongfang Liu, PhD, Mayo Clinic Tom Oniki, PhD, Intermountain Healthcare
Data Normalization Goals –To conduct the science for realizing semantic interoperability and integration of diverse data sources –To develop tools and resources enabling the generation of normalized EMR data for secondary uses
Data Normalization Information Models Target Value Sets Raw EMR Data Tooling Normalized EMR Data Normalization Targets Normalization Process
Normalization Targets Clinical Element Models –Intermountain Healthcare/GE Healthcare’s detailed clinical models Terminology/value sets associated with the models –using standards where possible
CEM Models Different models for different use cases “CORE” models
“Core” Models CORE Lab CEM model CORE Lab CEM model attribute 1 attribute 2 attribute 3 attribute 4 CEM A CEM C CEM B CEM D
“Core” Models CORE Lab CEM model CORE Lab CEM model attribute 1 attribute 2 attribute 3 attribute 4 Secondary Use Lab CEM model CEM A CEM C CEM B CEM D attribute 1 attribute 2 attribute 3 attribute 4 CEM A CEM C CEM B CEM D
“Core” Models CORE Lab CEM model CORE Lab CEM model attribute 1 attribute 2 attribute 3 attribute 4 Secondary Use Lab CEM model CEM A CEMC CEM B CEMD attribute 1 attribute 2 attribute 3 attribute 4 Clinical Trial Lab CEM model Clinical Trial Lab CEM model CEMA CEM C CEM B CEM D attribute 1 attribute 2 attribute 3 attribute 4 CEM A CEM C CEM B CEM D
“Core” Models CORE Lab CEM model CORE Lab CEM model attribute 1 attribute 2 attribute 3 attribute 4 Secondary Use Lab CEM model CEM A CEMC CEM B CEMD attribute 1 attribute 2 attribute 3 attribute 4 Clinical Trial Lab CEM model Clinical Trial Lab CEM model CEM A CEM C CEM B CEM D attribute 1 attribute 2 attribute 3 attribute 4 CEM A CEM C CEM B CEM D EMR Lab CEM model EMR Lab CEM model CEM A CEM C CEM B CEM D attribute 1 attribute 2 attribute 3 attribute 4
CEM Models Different models for different use cases “CORE” models –CORENotedDrug -> SecondaryUseNotedDrug –COREStandardLab -> SecondaryUseStandardLab (+ 6 data type- specific models) –COREPatient -> SecondaryUsePatient
Generating XSDs Secondary Use Lab CEM model CEM A CEM C CEM B CEM D attribute 1 attribute 2 attribute 3 attribute 4
Generating XSDs Secondary Use Lab CEM model CEM A CEM C CEM B CEM D attribute 1 attribute 2 attribute 3 attribute 4 SHARP “reference class” SHARP “reference class” attribute 5 attribute 6 attribute 7 attribute 8 CEM E CEM G CEM F CEM H
Generating XSDs Secondary Use Lab CEM model CEM A CEM C CEM B CEM D attribute 1 attribute 2 attribute 3 attribute 4 SHARP “reference class” SHARP “reference class” attribute 5 attribute 6 attribute 7 attribute 8 CEM E CEM G CEM F CEM H COMPILE Secondary Use Lab XSD attribute attribute attribute attribute attribute attribute
Terminology/Value Sets Terminology value sets define the valid values used in the models Terminology standards are used wherever possible
Terminology/Value Sets Terminology value sets define the valid values used in the models Terminology standards are used wherever possible Secondary Use Patient CEM model Secondary Use Patient CEM model CEM B CEM A CEM C administrativeGender attribute X attribute Y attribute Z Gender CEM Gender Value Set: HL7 AdminGender {M, F}
CEM Request Site and Browser
Normalization Process Prepare Mapping UIMA Pipeline to transform raw EMR data to normalized EMR data based on mappings
Mappings Two kinds of mappings needed: –Model Mappings –Terminology Mappings
Model Mappings HL7CEM Secondary Use Patient CEM model CEM A CEM C CEM B CEM D attribute 1 attribute 2 attribute 3 attribute 4 MSH PID 1 2 … OBR OBX … Secondary Use Lab CEM model CEM A CEM C CEM B CEM D attribute 1 attribute 2 attribute 3 attribute 4
Model Mappings
Terminology Mappings HL7 from MayoCEM Local Gender Codes 1 = MALE 2 = FEMALE HL7 AdministrativeGender M = MALE F = FEMALE
Terminology Mappings CEM FieldsLocalCodeTargetCodeTargetCodeSystem GenderMMHL7 Gender GenderFFHL7 Gender Race CDC Race RaceW2106-3CDC Race RouterMethodDeviceORALPOHL7 Route DoseFreqBID &0800, SNOMED DoseFreqBID &0800, SNOMED DoseFreqDAILY & SNOMED DoseFreqQ24HRS SNOMED DoseFreqONE TIME ORDER SNOMED DoseUNITPuff SNOMED DoseUNITTABLET SNOMED DoseUNITtsp SNOMED DoseUNITCAPSULE (HA SNOMED DoseUNITpatch SNOMED DoseUNITgr SNOMED DoseUNITmL SNOMED
Pipeline Implement in UIMA (Unstructured Information Management Architecture) Configurable –Data sources – HL7, CCD, CDA, and Table format –Model mappings (different EMR systems may have different formats) –Terminology mappings –Inference mappings – infer ingredients from clinical drugs