GALEN Where we started Building Classifications with GALEN

GALEN Where we started Building Classifications with GALEN A look into the future…? Key Questions 1 GALEN

GALEN-IN-USE Partners University of Manchester, VAMP Health, UK University of Nijmegen, Holland EFCC and its affiliates: l l WCC, NOMESCO, CNR Rome, U St Etienne IASIST, IDGmb. H, CMITD, STAKE, U Louvain University Hospital of Geneva, LNAT, Switzerland OLE, RAMIT, Datasoft, Belgium GSF Medis, U Hildesheim, Germany CNR, Rome, Italy SPRI, U Linköping, Sweden VTT, Oulu Univ Hospital, Medici Data, Finland 2 GALEN

GALEN Where we started 3 GALEN

4 Clinical Terminology Data Entry Clinical Record Decision Support Best Practice GALEN

Choose terms from a coding scheme enter search: cystitis Cholecystitis, Cystitis, Iatrogenic NOS cystitis NOS Acute Chronic Chemical cystitis cholecystitis Subacute Acute Postoperative cholecystitis, cystitis NOS Follicular Bacterial Drug induced cystitis Cholecystitis Idiopathic Bacterial Radiation cystitis etc 5 next page Too Big. . . picking lists too long Too Small. . . not enough clinical detail GALEN

A new clinician-friendly option: structured data entry severity mild cystitis moderate severe onset gradual sudden chronicity acute sub-acute chronic Assemble a phrase. . . …from a set of useful pieces. . . …displayed on the screen. “moderately severe, acute cystitis of sudden onset” 6 GALEN

Only one problem: l How can the computer know in advance what are the sensible pieces to display? l Traditional coding schemes do not have the information required cystitis 7 Read ? ED M SNO OPC ICD S GALEN

A new resource What is cystitis combined with ‘acute and mild? how would you like the answer ? . . . a resource that can cope with arbitrarily complex clinical concepts. . . 8 GALEN

A new resource as an English phrase and as an ICD Code <“mild, acute cystitis”, D 595. 0 >. . . and cope with different natural languages and coding schemes. 9 GALEN

Computer knows ‘rules’ for: l how do you classify things? l l when are two things the same ? l l ‘fractured eyebrow causing donkey’ are any combinations redundant? l 10 ‘Inflammation of Liver’ vs ‘Hepatitis’ are there any illegal combinations? l l Fractures of femur, and of femoral neck ‘finger which is part of hand’ GALEN

GALEN Building Classifications with GALEN 11 GALEN

Manual Classifications l Most medical ‘classifications’ are really ‘thesauri’ l Thesauri have mixed hierarchies – kind, part, cause, modified, . . . heart valve aortic valve cusp of aortic valve l For interpretation by people l l 12 Only people can reorganise (and that isn’t easy or quick) l Must read rubrics Model of use implicit and fixed GALEN

‘Computed’ Classifications l Strict hierarchies using ‘kind of’ ‘Disorder of heart’ ‘Disorder of valve in left ventricle in heart’ ‘Disorder of cusp in valve in left ventricle in heart’ l Computable l l People can understand (but don’t have to) Computers can reorganise – because all meaning is explicit in formal expression 13 GALEN

Bridges from patient data to classifications Coding & Classification Decision support UMLS Clinical Applications Me. SH SNOMED Axes GALEN Medical Records MEGATERM ICPC ACRNEMA Data entry READ OPCS ICD-9 ICD-10 GALEN

Added value of Computed Classifications l Hide the complexity l l 15 Let the computer do the work Small on the outside, huge on the inside l Reorganise classifications when needed l More complete, consistent and automatic indexing l Automatic cross mapping/merging l Quicker to update - worry about what, not where l ‘Coherence without Uniformity’ GALEN

Issues to consider Can not prove result is complete or correct (but can’t do that with manual result either) Initially expensive to build GALEN

GALEN A look into the future…? 17 GALEN

What will the classifications of today look like in the future? They won’t exist Each will be replaced by several different views on something else, each serving individually the specific (conflicting) functions which were formerly tortured out of the one classification NOT a book NOT a picking list 18 GALEN

What will these ‘views’ be like ? Each view is like a classification, but. . . Multiaxial (very) One, single clear purpose Humans say what terms are included and what they mean Computers decide how to arrange them Not fixed in stone - malleable, customisable, localisable but always coherent 19 GALEN

‘Views’ of what ? ‘Clinical Terminology’ Something clinicians use to record information about real, individual patients Somebody else’s problem. . as long as the bridge to classification views exists 20 GALEN

GALEN Key Questions 21 GALEN

Key Questions (for users of terminology systems) l What do you want it to do? l Clinical systems are the end l terminology the means – Don’t get the cart before the horse l Can it do it? l l l 22 Can anything do it? A universal coding scheme for everything is implausible Coherence without Uniformity GALEN

Key Questions (for builders of terminology systems) l Can it be built? implemented? maintained? l l l Will it scale? l l Organisationally? Technically? Can applications use it? l 23 Quickly enough? Are there short-term milestones and a migration pathway? Can we maintain them? Can vendors sell them GALEN

Warnings l Be cautious of hype l l l 24 It’s not as easy as it looks Beware scaling It’s not how big it is that counts - it’s what you can do with it l Don’t reinvent the wheel l Simple to understand doesn’t mean simple to use l Coding systems are now software GALEN

GALEN-IN-USE Advanced terminological services to support integrated clinical information systems galen@cs. man. ac. uk http: //www. cs. man. ac. uk/mig/galen/ 25 GALEN