Successfully reported this slideshow.
We use your LinkedIn profile and activity data to personalize ads and to show you more relevant ads. You can change your ad preferences anytime.
Comparable and Consistent Data from     Heterogeneous Systems  Health IT Summit Conference iHT2   San Francisco, March 27,...
From Practice-based Evidence              to Evidence-based Practice                Clinical                              ...
The ChallengeMost clinical data in the United States is heterogeneous – non-standard  – Within Institutions  – Between In...
© 2012 Mayo Clinic
SHARP Area 4:             Secondary Use of EHR Data• Agilex Technologies              • Harvard Univ.• CDISC (Clinical Dat...
Cross-integrated suite of                                                                                        projects ...
MissionTo enable the use of       Leverage health informatics to:EHR data for secondary     • generate new knowledgepurpos...
Normalization OptionsNormalize data at source  – Fiat, regulation, patient expectationsTransformation and mapping  – Sou...
Modes of NormalizationGenerally true for both structured and un-structured dataSyntactic transformation  – Clean up mess...
Transformation Target?Normalization begs a “normal form”Extant national and international standards do not fully specify...
Clinical Data Normalization                     Dr. Huff on Data Normalization                           Stanley M. Huff,...
Clinical Data NormalizationData Normalization– Comparable and consistent data is foundational to  secondary useClinical ...
© 2012 Mayo Clinic
Data Element Harmonization     http://informatics.mayo.edu/CIMI/Stan Huff – CIMI– Clinical Information Model InitiativeN...
Core CEMsRecognize that use-case specific work-flow enters into CEM-like objects  – Clinical EHR implementation  – CLIA o...
That Semantic Bit…Canonical semantics reduce to Value-set Binding to CEM objectsValue-sets should obviously be drawn fro...
On MappingSemantic mapping from local codes to “standard” codes is required  – Not magic, humanly curatedReality of idio...
Normalization PipelinesInput heterogeneous clinical data  – HL7, CDA/CCD, structured feedsOutput Normalized CEMs  – Crea...
Normalization Flow   © 2012 Mayo Clinic
NLP Deliverables and Tools      http://informatics.mayo.edu/sharp/index.php/Tools cTAKES Releases  – Smoking Status Class...
High-Throughput PhenotypingPhenotype - a set of patient characteristics :  – Diagnoses, Procedures  – Demographics  – Lab...
Drools-based Architecture                                        Jyotishman Pathak, PhD.                                  ...
SHARP and Beacon SynergiesSHARP will facilitate the mapping of comparable and consistent data into information and knowle...
Medication Reconciliation         a Pan-SHARP ProjectTest data between and within enterpriseMedications lists with demog...
More Information      SHARP Area 4: (SHARPn)     Secondary Use of EHR Data          www.sharpn.orgSoutheast Minnesota Beac...
Upcoming SlideShare
Loading in …5
×

Presentation“Comparable and Consistent Patient Data from Heterogeneous Systems”

514 views

Published on

Dr. Chute will overview progress around data normalization and high throughput clinical phenotyping (recognizing groups of patients for quality, practice, or research use-cases from electronic medical records (EMRs). These techniques were demonstrated to generate comparable and consistent information from multiple academic medical centers with heterogeneous EMR systems and record structures in the NHGRI funded eMERGE consortium (gwas.net). Tools and techniques for data normalization and phenotyping have been generalized and partially commoditized as open-source archetypes software in the ongoing SHARPn (SHARPn.org).

Published in: Health & Medicine, Technology
  • Be the first to comment

  • Be the first to like this

Presentation“Comparable and Consistent Patient Data from Heterogeneous Systems”

  1. 1. Comparable and Consistent Data from Heterogeneous Systems Health IT Summit Conference iHT2 San Francisco, March 27, 2012 Christopher G. Chute, MD DrPH, Professor, Biomedical Informatics, Mayo Clinic Chair, ISO TC215 on Health Informatics Chair, International Classification of Disease, WHO
  2. 2. From Practice-based Evidence to Evidence-based Practice Clinical Registries et al.Data Databases Comparable and Consistent Inference Patient Standards Medical KnowledgeEncounters Vocabularies & TerminologiesDecision Expert Clinical Knowledge Guidelinessupport Systems Management “Secondary Use” © 2012 Mayo Clinic
  3. 3. The ChallengeMost clinical data in the United States is heterogeneous – non-standard – Within Institutions – Between InstitutionsMeaningful Use is mitigating, but has not yet “solved” the problem – Achieving standardization in Meaningful Use is sometimes minimized © 2012 Mayo Clinic
  4. 4. © 2012 Mayo Clinic
  5. 5. SHARP Area 4: Secondary Use of EHR Data• Agilex Technologies • Harvard Univ.• CDISC (Clinical Data Interchange • Intermountain Healthcare Standards Consortium) • Mayo Clinic• Centerphase Solutions • Mirth Corporation, Inc.• Deloitte • MIT• Group Health, Seattle • MITRE Corp.• IBM Watson Research Labs • Regenstrief Institute, Inc.• University of Utah • SUNY• University of Pittsburgh • University of Colorado © 2012 Mayo Clinic
  6. 6. Cross-integrated suite of projects and products Themes Projects Players Data Quality & Evaluation FrameworksData Normalization Phenotype Recognition Clinical Data Normalization IBM, Mayo, Utah, Agilex, Regenstrief Harvard, Group Health, IBM, Utah, Mayo, MIT, SUNY, i2b2, Pittsburgh, Natural Language Processing Colorado, MITRE High-Throughput Phenotyping CDISC, Centerphase, Mayo, Utah UIMA & Scaling Capacity IBM, Mayo, Agilex, Mirth Data Quality Mayo, Utah Evaluation Framework Agilex, Mayo, Utah © 2012 Mayo Clinic
  7. 7. MissionTo enable the use of Leverage health informatics to:EHR data for secondary • generate new knowledgepurposes, such as • improve careclinical research andpublic health. • address population needsTo support the community • open-source toolsof EHR data consumers by • servicesdeveloping: • scalable software © 2012 Mayo Clinic
  8. 8. Normalization OptionsNormalize data at source – Fiat, regulation, patient expectationsTransformation and mapping – Soul of “ETL” in data warehousingHybrid (graceful evolution to source) – “New” systems normalize at source – Transformation of legacy system data © 2012 Mayo Clinic
  9. 9. Modes of NormalizationGenerally true for both structured and un-structured dataSyntactic transformation – Clean up message formats – HL7 V2, CCD/CDA, tabular data, etc – Emulate Regenstrief HOSS pipelineSemantic normalization – Typically vocabulary mapping © 2012 Mayo Clinic
  10. 10. Transformation Target?Normalization begs a “normal form”Extant national and international standards do not fully specify – Focus on HIE or internal messaging – Canonical data representation wanting – Require fully machine manageable data © 2012 Mayo Clinic
  11. 11. Clinical Data Normalization  Dr. Huff on Data Normalization Stanley M. Huff, M.D.; SHARPn Co- Principal Investigator; Professor (Clinical) - Biomedical Informatics at University of Utah - College of Medicine and Chief Medical Informatics Officer Intermountain Healthcare. Dr. Huff discusses the need to provide patient care at the lowest cost with advanced decision support requires structured and coded data. © 2012 Mayo Clinic
  12. 12. Clinical Data NormalizationData Normalization– Comparable and consistent data is foundational to secondary useClinical Data Models – Clinical Element Models (CEMS)– Basis for retaining computable meaning when data is exchanged between heterogeneous computer systems.– Basis for shared computable meaning when clinical data is referenced in decision support logic. © 2012 Mayo Clinic
  13. 13. © 2012 Mayo Clinic
  14. 14. Data Element Harmonization http://informatics.mayo.edu/CIMI/Stan Huff – CIMI– Clinical Information Model InitiativeNHS Clinical StatementCEN TC251/OpenEHR ArchetypesHL7 TemplatesISO TC215 Detailed Clinical ModelsCDISC Common Clinical ElementsIntermountain/GE CEMs © 2012 Mayo Clinic
  15. 15. Core CEMsRecognize that use-case specific work-flow enters into CEM-like objects – Clinical EHR implementation – CLIA or FDA regulatory overheadSecondary Use tends to focus on dataCreate “core” CEMs – Labs, Rxs, Dxs, Pxs, demographics © 2012 Mayo Clinic
  16. 16. That Semantic Bit…Canonical semantics reduce to Value-set Binding to CEM objectsValue-sets should obviously be drawn from “standard” vocabularies – SNOMED-CT and ICD – LOINC – RxNorm – But others required: HUGO, GO, HL7 © 2012 Mayo Clinic
  17. 17. On MappingSemantic mapping from local codes to “standard” codes is required – Not magic, humanly curatedReality of idiosyncratic local codes is perverse – Why does every clinical lab in the country insist on making up its own lab codes? © 2012 Mayo Clinic
  18. 18. Normalization PipelinesInput heterogeneous clinical data – HL7, CDA/CCD, structured feedsOutput Normalized CEMs – Create logical structures within UIMA CASSerialize to a persistence layer – SQL, RDF, “PCAST like”, XMLRobust Prototypes exist – Early version production Q3 2012 © 2012 Mayo Clinic
  19. 19. Normalization Flow © 2012 Mayo Clinic
  20. 20. NLP Deliverables and Tools http://informatics.mayo.edu/sharp/index.php/Tools cTAKES Releases – Smoking Status Classifier – cTAKES Side Effects module – Medication Annotator – Modules for relation extraction Integrated cTAKES(icTAKES) – an effort to improve the usability of cTAKES for end users NLP evaluation workbench – the dissemination of an NLP algorithm requires performance benchmarking. The evaluation workbench allows NLP investigators and developers to compare and evaluate various NLP algorithms. SHARPn NLP Common Type – SHARPn NLP Common Type System is an effort for defining common NLP types used in SHARPn; UIMA framework. © 2012 Mayo Clinic
  21. 21. High-Throughput PhenotypingPhenotype - a set of patient characteristics : – Diagnoses, Procedures – Demographics – Lab Values, MedicationsPhenotyping – overload of terms – Originally for research cohorts from EMRs – Obvious extension to clinical trial eligibility – Quality metric Numerators and denominators – Clinical decision support - Trigger criteria © 2012 Mayo Clinic
  22. 22. Drools-based Architecture Jyotishman Pathak, PhD. discusses leveraging open-source tools such Clinical Data Access as Drools.Element LayerDatabase Business Logic Transformation Inference Layer Engine (Drools) Service for List of Transform physical representation Creating Output Diabetic  Normalized logical representation (File, Database, (Fact Model) Patients etc) © 2012 Mayo Clinic
  23. 23. SHARP and Beacon SynergiesSHARP will facilitate the mapping of comparable and consistent data into information and knowledgeSE MN Beacon will facilitate the population-based generation of best evidence and new knowledgeSE MN Beacon will allow the application of Health Information Technology to primary care practice – Informing practice with population-based data – Supporting practice with knowledge © 2012 Mayo Clinic 23
  24. 24. Medication Reconciliation a Pan-SHARP ProjectTest data between and within enterpriseMedications lists with demographicsDe-identified (age, dates)Create Normalized CEMs – RxNorm, strength, form, route, ISO dates,Move into SmartAccess RDF renderingReconciliation as “App” in SHARP 3 © 2012 Mayo Clinic
  25. 25. More Information SHARP Area 4: (SHARPn) Secondary Use of EHR Data www.sharpn.orgSoutheast Minnesota Beacon Community www.semnbeacon.org © 2012 Mayo Clinic

×