Experiences from a Large- scale Implementation of SNOMED CT in Argentina. Dr. Alejandro Lopez Osornio Dr. Guillermo Reynoso termMed IT Buenos Aires, Argentina
Hospital Italiano de Buenos Aires Located in Buenos Aires, Argentina Founded on 1853 500 bed hospital 10,000 visits a day in the ambulatory system 140,000 patients private health insurance system
Argentina, December 2001 Protests on the streets demanding the resignation of the president President De la Rua resigned Argentine Peso devalues from 1 U$S to 0.30 U$S Restrictions to cash withdrawal Credit crisis
Hospital Italiano, December 2001 “Management based on accurate clinical information will let us survive this crisis” (Dr. FGB Quiros) Great boost to the IT program Extension of the Hospital Information System to all areas: Ambulatory visits Inpatient episodes Emergency Department Diagnostic procedures Pharmacy and Medical Devices Home care
Terminology use cases Health insurance services management board Chronic disease management program Epidemiology department Clinical trials Primary care Physicians Incentives Program Research and teaching (University Hospital) Clinical decision support systems
Available vocabularies on 2001 Local Pharmacy terminology Local Devices terminology Local Procedures terminology Argentine Reimbursement vocabulary Diagnosis: Free text secondary coded with: ICPC and ICD-10 (outpatient) ICD-9CM and DRG (inpatient)
Requirements A new terminology System: Unobtrusive : clinical users do not like things that step in the middle of their usual workflow Immediate : instant codification is the essential to use any kind of CDSS Precise : allow the user to record exactly what he wants, appropriate detail level
Design Principles HIBA will use a local list of terms as user interface, constructed from the free- text entries repository The interface vocabulary will be concept oriented and will support “near natural language” expressiveness (lot of qualifiers) This Interface Vocabulary will be mapped to a Reference Terminology and the to different classifications
Terminologies levels and interactions DRG CIE-O CIAP Aggregate Terminology CrossMaps Mechanism ATC CIE-10 CIE-9-CM SCT Core Reference Terminology SNOMED CT Extension Interface Terminology Mechanism Local Vocabulary Natural Language
SNOMED CT content Concepts: great coverage, but detail is limited to avoid the ‘combinatorial explosion’. There is some regional bias. Descriptions: acceptable, valid and appropriate descriptions, but do not fully covers local jargon and acronyms (Spanish edition) SNOMED CT is extendable
• Implementation Strategy: – Standard SNOMED CT • User Experience: • Cost: – Standard SCT + New descriptions • User Experience: • Cost: – Standard SCT + New descriptions + New concepts • User Experience: • Cost:
Plan Extend SNOMED CT to cover our needs Hide the complexity of the system implementing simple user interfaces: Templates, pick-lists, checkboxes Search and refine
Terminology Team 1 Clinical terminologist (MD) (80% time) 1 Full time administrative supervisor 12 part-time medical students 1 Lead Software Analyst and Developer (75% time) and a second for 20% of his time Medical Informatics Department Advisory Board
SNOMED CT Mechanisms Extensions Cross-Maps Subsets
Extensions Namespace:
SNOMED CT and Local Extension SNOMED CT Relationships Concepts Descriptions
SNOMED CT and Local Extension SNOMED CT HI Extension Relationships Concepts Relationships Concepts Descriptions Descriptions
SNOMED CT and Local Extension SNOMED CT HI Extension New concepts definitions Relationships Concepts Relationships Concepts Descriptions Descriptions New Descriptions
SNOMED CT and Local Extension SNOMED CT HI Extension Osteoarthritis <- Is a Left knee Relationships Concepts Relationships Concepts osteoarthritis Left knee <- Finding site Descriptions Descriptions Left knee osteoarthritis Osteoarthritis of left knee
SNOMED CT and Local Extension SNOMED CT HI Extension Relationships Concepts Relationships Concepts Descriptions Descriptions Reference Terminology
SNOMED CT and Local Extension SNOMED CT HI Extension Relationships Concepts Relationships Concepts Descriptions Interface Terminology Descriptions
SNOMED CT and Local Extension SNOMED CT HI Extension one to one relationships 'equivalent to' Relationships Concepts Relationships Concepts Descriptions Descriptions Interface Terminology
SNOMED CT and Local Extension SNOMED CT HI Extension Asthma <- Equivalent to Asthma Relationships Concepts Relationships Concepts Descriptions Descriptions Asthma Asthmatic Interface Terminology
Dealing with changes on SNOMED CT, new releases every 6 months SNOMED CT HI Extension New Concepts Unchanged Relationships Relationships Concepts Concepts Retired Concepts X Descriptions Descriptions Update time: One person, half day work, during one week
Local Interface Vocabulary for problems and procedures 2 million free text entries 110,000 selected descriptions Normalization Frequency count > 10 25,000 concepts 9,000 equivalencies 16,000 new definitions
Subsets Simple Subsets: HI Terminology ID: ID: ID: 328423100001131 4433100001131 9843100001131 Signs & Diagnosis Drugs & Symptoms substances Surgical Procedures Diagnostic Procedures Medical Devices ID: ID: 22223100001137 ID: 45565100001137 6743100001131 OB-Gyn Subset ID: 76656100001137
Subsets Language Subsets: Context: Concept A: Description X Description Y Concept A: Medical Description X (Preferred) department 1 Description Y Concept A: Medical Description X department 2 Description Y (Preferred) Navigation Subsets:
SNOMED CT Cross-maps ICD-9CM ICD-10 ICPC SNOMED CT HI Extension Relationships Concepts Relationships Concepts Descriptions Descriptions
Tools Apelon TDE - DTS Protege In house development
Terminology tools in Hospital Italiano SNOMED CT Release Domain Experts Terminology Administra1on tool (Java 2EE Web based Applica-on + CLUE Browser) Terminology Server (Web Services) Clinical So3ware Clinical Informa-on Applica-ons Retrieval
Requirements for the Terminology Administration tool SNOMED CT Release Domain Experts SNOMED CT periodic update Add, delete concepts and descrip-ons Terminology SNOMED CT modeling Administra1on tool (Java 2EE Web based Subset Administra-on Applica-on + CLUE Browser) Cross‐maps Administra-on Workflow Administra-on Terminology Server Quality Assurance
Requirements for the Terminology Server Interface Vocabulary Search and Refine Subset retrieval Terminology Server Users Sugges-ons (Web Services) Structured Post‐coordina-on Specific Knowledge Inference Cross map conversion Clinical So3ware Clinical Informa-on Applica-on Retrieval
Data retrieval Members of the terminology team analyze “Data retrieval support requests” and define appropriate subsets using an interactive Subset Definition tool: List of membership clauses: Include / Exclude descendants of concept... Include / Exclude concepts with role xx Include / Exclude concepts with role/target xx/yy The “Epidemiology department” uses the subsets on simple SQL sentences More than Data retrieval 1,200 subsets had been created
Implementation example:
Implementation example:
Results Concept Oriented Thesaurus compilation 2002-2004 Software development 2003-2006 Full implementation in mid-2006 Free submission of new terms on problems list interface Submission rate has stabilized for the problems list but not for procedures Very good acceptance by the doctors = No complaints
Other standards HL7 for messaging between different applications Map with ICD-9CM and DRG grouper DSM-IV Map with local specific reimbursement terminologies Participating on a pilot implementation of the draft version of the HL7 decision support services specification (Kawamoto - Lobach, Duke Univ.) Information Model No standardization of the information model yet
Critical success factors Full commitment from the to organization levels Respect for local vocabulary Successfully hiding the terminology complexity from the users
Which were the biggest hurdles to be overcome? Initial processing of back history of free text Training modelers for the extension Designing a QA process for the adherence to Concept Model guidelines Train clinical users to avoid ‘non-clinical’ statements Manage variability on Procedures domain to maintain expressiveness
Lessons learned There’s no such thing as too much effort to make things easier to you clinical users
Thanks! termMed IT Argentina
Recommend
More recommend
