160 likes | 290 Views
The InterMed TM Collaboratory –the early years (1994-1996). Biomedical informatics researchers & systems developers at 5 sites: Harvard/Brigham and Women’s Hospital Harvard/Massachusetts General Hospital Columbia Presbyterian Medical Center University of Utah School of Medicine
E N D
The InterMedTM Collaboratory –the early years (1994-1996) • Biomedical informatics researchers & systems developers at 5 sites: • Harvard/Brigham and Women’s Hospital • Harvard/Massachusetts General Hospital • Columbia Presbyterian Medical Center • University of Utah School of Medicine • Stanford University School of Medicine
The initial InterMed concept • Biomedical application development too slow • Silos, limited reuse • Internet could provide common resources (middleware) • Domain-specific • Seven-tiered model • Built on infrastructure “service” layer
InterMed’s Seven-Tiered Model Policies & Procedures Applications Composition Methods Component Tools Data Bases & Knowledge Bases Vocabulary / Taxonomy Infrastructure Classes & Services
UMLS Collaborative Model Development Input/Output Specifications and Protocols Browsing & Maintenance Requirements Implemented using convenient representation language SNOMED Map into local vocabularies for existing applications MED (Columbia) InterMed Vocabulary Model InterMed Vocabulary Server LOINC IVORY (WARP) Local vocabs for new applications TheNetSys . . . • Define meta-terminology • Keep implementation-independent • Provide generic solution Populated with sample domain-specific content
Intended applications • Linking DXplain to a clinical information system • Internet-based clinical guidelines • Internet-based clinical research queries
Component development work • DXplain access (MGH) - tiers 3,4 • Guideline tools (MGH, BWH versions) - tiers 3,4 • Semantic net tool: TheNetSys - tier 2,4 • Semantic model for CXR reports - tier 3 • Form tool (a la HTML 3) using semantic model - tiers 3,4 • Development of import/export (TheNetSys/Ontolingua) • Image tool (with pan, zoom, grayscale adjust) - tier 4 • Tree walker for state maps & protocols - tier 4 • WWW access modules - tier 4
Application construction tools (Tier 5) • Interface builder • Use of visual embedding/ compound document architectures (Open Doc, OLE-2) • Script-based communication among components • Work-flow-management-oriented framework • An approach to integration based on clinical guidelines: the GEODE-CM project
Early findings ~ 1995 • Activities weren’t converging • Progress in the different labs was not aligned • Each was working on what interested it • Vision was too abstract • Hard to identify measures of success • We decided to select a single application focus to drive the project • Needed to be something all of the labs cared about
Common guideline representation as a second year focus • Multiple purposes of guidelines • education (reference) • training (how to do something) • problem-solving • responding to events (alerts) • protocolized care/data capture • care plans/critical paths -- UR and QA • consultation • workflow management (GEODE)
Multiple representation approaches • Flow charts • Narratives • Decision tables • Rule-base • MLM procedures • State model database
Collaborative Model Development Input/Output Specifications and Protocols Browsing & Maintenance Requirements AHCPR Guidelines Implemented using convenient representation language T-Helper Protocols Map into local representation for existing applications CPMC Clinical Pathways & Guidelines InterMed Guidelines Model InterMed Guidelines Server UTI and Other Guidelines at MGH Local versions for new applications BWH GEODE Clin. Mgt. Pathways • Define meta-terminology • Keep implementation-independent • Provide generic solution . . . Populated with sample domain-specific content
Goals to explore: • Whether a common representation could be devised that maintains all of the information needed to support all of the applications • How the guideline knowledge should interact with vocabulary knowledge • What components and tools would be needed
Work was then able to be partitioned • Requirements analysis • Model development • Tool building • Testing and refinement • Application exploration • Cognitive studies • SDO participation, conference planning, community development
Difficulties still occurred when: • Groups disagreed about emphasis or approach • No one signed up for a needed task • Need for continuing work on maintaining a shared vision
Collaboratory issues • Ideal C’y needs shared vision and buy-in • strong business case • intellectual and practical appeal • Tight project management • funding dependent on performance • Can peer-based C’y be productive? • Can research-based C’y (vs. product-focused) be successful?