interoperability in health informatics: saving data from obsolescence, and putting the patient first...
TRANSCRIPT
Interoperability in Health Informatics:saving data from obsolescence,
and putting the patient first
Thomas BealeOcean Informatics, Australia
Sponsored by: www.OceanInformatics.biz
© 2005 Ocean Informatics
* * * Programme * * *
• The problems of health IT
• The problem of Interoperability
• Ontological basis for solutions
• Standards
• Strategy
© 2005 Ocean Informatics
The ProblemClinical Reality ICT Requirement
Life-time Healthcare EHR valid for 120 years
Complexity & change in medicine
Future-proof HIS
Multi-contact health Interoperability of sites
Preventative medicine,
epidemiology, population studies
Real decision support; queryable (=standardised) longitudinal, patient-centric EHR
Mobile patient – leisure, military, professional, regugee
National & global interoperability
© 2005 Ocean Informatics
The human view
© 2005 Ocean Informatics
The engineering view
© 2005 Ocean Informatics
* * * Programme * * *
• The problems of health
• The problem of Interoperability
• 3 principles for solutions
• Standards
• Strategy
© 2005 Ocean Informatics
The Clinical Information Trail
referral
orderresult
discharge
referral
order result
referral
orderresult
Chest infection GP reviewGP visit Back to foot clinic
Main GP
Foot ulcer foot clinic (hospital)
hospital
Diabetol.
See specialist
DI & path
ImagingRenal function test
Stroke – hospital
hospital2
GP2
See other GP on holiday CT scan
Soc. worker
rehabilitation
Non-interoperable data
discharge
referral
workflow?interoperability
Repeated tests
Clinical errors
System-wide planningLimited Dec support
CostsPatient-centred view?
© 2005 Ocean Informatics
First Improvement - Messages
referral
Chest infection GP reviewGP visit Back to foot clinic
Main GP
Foot ulcer foot clinic (hospital)
hospital
Diabetol.
See specialist
DI & path
ImagingRenal function test
Stroke – hospital
hospital2
GP2
See other GP on holiday CT scan
Soc. worker
rehabilitation
Still non-interoperable data!
discharge
referral
order
order result
discharge referral
orderresult
referral
result
Which items to copy?Fixed Structures
Interop = stovepipesNo EHR…
Who is managing care?Relevant clinical info?
Medico-legal support?
© 2005 Ocean Informatics
Introducing the EHR
referral
hospital
Diabetol.
Main GP
DI & path
hospital2
GP2
Soc. worker
discharge
referral
order
order result
discharge referral
orderresult
referral
result
StandardisedShared EHR
EHR VISIBILITY
S h a r e d C a r e, L o n g i t u d i n a l, p a t i e n t – c e n t r e dS h a r e d C a r e, L o n g i t u d i n a l, p a t i e n t – c e n t r e d
The Patient
Copying controllableFixed Structures
Interop = std busPatient-centred view
Carers can communicateRelevant clinical info
Medico-legal support
© 2005 Ocean Informatics
Difficult Questions revisited (1)
• Which items are copied to shared EHR?– Rule-based – e.g. “current meds”, “therapeutic
precautions” + major event categories + ad hoc requests ...these rules must be able to evolve...security/consent...
• Where is the master copy of each item?– Usually in the shared EHR, to enable standardised
reads...but if EHR technology used at care sites, could be there as well – “system of EHR systems”
• Where & how to achieve interoperability?– Standards-based middleware + web services– highly enabled if EPRs are converted to EHR
technology
© 2005 Ocean Informatics
Difficult Questions revisited (2)
• Where is the “EHR”?– The patient-centred, longitudinal, shared care EHR is in
its own servers – where? Depends...– Governance by e.g. GP organisations, local health
authorities or other trusted bodies
• Who is in charge of the EHR? – The patient can be in charge of consent– GP or other carer can be clinical “co-driver”
• How does decision support work?– It now has a standardised & longitudinal database to
work with
© 2005 Ocean Informatics
Difficult Questions revisited (3)• Who is managing the patient’s care?
– Could be anyone, but now there is a mechanism to communicate who it is
• How does each clinician determine where the relevant information for next decision is?– By using shared EHR: problem- and issue-threading,
other derived information structures; relies on fidelity of EPR->EHR synchronisation
• How are medico-legal problems addressed?– All decisions in the EHR are linked back to their input
items – via causal links, issues, problem classifiers etc– Only one place – shared EHR – needs to have reliable
technology for non-repudiation, notary, etc
© 2005 Ocean Informatics
Strategic Issues for the interoperable EHR• Technical
– Consolidated versus “pure federated” EHR– Where are EHRs, what governance?– Human users and other systems need access to EHR– Centralised and distributed deployment possible– Problem of language and vocabulary
• Socio-political– Doctors’ Fear of making their information so available– Confidentiality needs of patients– Fear of doctors losing control (?!) over patient care– Differing national legislation on privacy & consent– Clinician fear of more data entry– Cost and consequences of user training
© 2005 Ocean Informatics
Practice MgtEPREPR EPR
A multi-layer EHR strategy
Care
Delivery
Patient-centred EHR
Shared-carePatient-centred
Longitudinal
secure
EHRShared-care
Patient-centred
Longitudinal
secure
Wide-areaaccess Meta-data items (EHR Index)
additional indexing?
EHR Extracts
© 2005 Ocean Informatics
* * * Programme * * *
• The problems of health IT
• The problem of Interoperability
• 3 principles for solutions
• Standards
• Strategy
© 2005 Ocean Informatics
Principles for “good” models
• 1 - Limited scope – targeted at one area such as demographics, workflow, ehr– Why? Same principle as low-coupled software
• 2 - Separation of viewpoints - RM/ODP EV, IV, CV– Why? Separates information (fine-grained) and service (coarse-
grained) semantics; don’t hardwire policies & bus process into the software
• 3 - Ontologically layered– Why? Separates progressively more specific & changeable
concepts into modular layers, allows division of what is hard-wired into software and what is knowledge available at runtime
• A good standard will usually have components reflecting these separations…
© 2005 Ocean Informatics
Is “ontology” just academic preciousness?
• Ontologies exist wherever there are definitions of the meaning of symbols, such as words
• Definitions can be in the form of a UML model (implicit), a functional specification, a clinical terminology (explicit), etc
• So if we care about any kind of model, we are using ontology
• But…usually we are not conscious of it, and mix ontological levels up…
• Leading to brittle, unmaintainable software
© 2005 Ocean Informatics
Ontologies and modelling
• ISO OSI is a well-known attempt to consciously design layers of models based on ontological thinking, I.e. layers of meaning, and is still influential
• TCP/IP is a post-hoc description of layers which emerged from organic development
• The big question is always: how much of any domain’s ontology will you hard-wire into the software & db schemas?
© 2005 Ocean Informatics
Ontological levels in health
Level 1 – data-sharing(persistence/exchange)
Level 2 – invariantdomain concepts
Level 3 – variantre-usable domain concepts
Level 4 – variantlocal & use-specific
base ontological commitments of domain, e.g.
“observation”, “subject-of-care”, “protocol”…
minimal ontological commitments – sufficient for “recording” and “sharing”, e.g.“composition”, “committer”, “attestation”…
atomic domain concepts, e.g. “lab result”, “patient”,
“apgar score”, “BP measurement”, …
use context-specific concepts, e.g. “asthma note”, “ante-natal exam”
Level 0 – foundationalObject meta-model (objects, attributes etc)
built-in data types,
© 2005 Ocean Informatics
…and corresponding models
Domain Base IM
Persistence IM
archetypes
Semantic templates
Terminology/ontology
100% bidir. mapping
Base ontology for
Re-usable pieces
Bind/mediate
Exchange IM
100% bidir mapping
Level 1 – data-sharing(persistence/exchange)
Level 2 –invariantdomain concepts
Level 3 – variantre-usable domain concepts
Level 4 – domain-variantlocal & use-specific
Level 0 – foundational
© 2005 Ocean Informatics
* * * Programme * * *
• The problems of health IT
• The problem of Interoperability
• 3 principles for solutions
• Standards
• Strategy
© 2005 Ocean Informatics
What is a “good” standard?• Based on the 3 principles• Componentised• Validated by implementation feedback
© 2005 Ocean Informatics
Are good standards being built?• Some are based on the 3 principles, and will lead to
modular software and interoperable platforms (CEN 13606, HISA, openEHR)
• Some are very mixed, e.g. HL7v3 RIM – contains Act, Observation, but also Medication, Invoice, RMIMs more focussed…
• HL7/OMG services project – updating Corbamed
• In Australia, archetypes being developed for standardisation
• But … still no good standard for clinical data types - Quantity, Coded_term, Ordinal, Text, etc
• Emergence is ad hoc, uncontrolled, and sometimes competitive….
© 2005 Ocean Informatics
Standards today
Level 1 – data-sharing(persistence/exchange)
Level 2 – invariantDomain concepts
Level 3 – domain-variantre-usable concepts
Level 4 – domain-variantlocal & use-specific
ISO 11404 – general purpose data typesLevel 0 – foundational
ISO RM/ODPOMG MOF/UML
CEN 13606-1 – EHR communicationISO ???? – clinical data types
Corbamed – PIDS
Content-dependent Content-independent
HL7v2EDIFACT
HL7v3 RIMopenEHR – EHR IMUN/CEFACT – ebXML-8
W3C XML-schema
openEHR – common, DT, DS IMs
emerging – national archetypes
CEN 13606-2 – categorial structuresHL7v3 CDA
WHO ???? – basic archetypes
Denmark – G-EPJ
openEHR – ADL
CEN EN13940 – continuity of care
ISO 8601 – date/time
Corbamed – TQS
© 2005 Ocean Informatics
CEN 13606
Proposal: a new CEN standard (WG2-led)
Archetype Base IM
Exchange IM
archetypes
Semantic templates
Part 1
Part 2 – archetype
interoperability
CEN nnnnn
Archetype ontology
Base ontology IM
Information/ terminology
brokering
IM/IM mappingguidance
Archetype repository
service model+ examplearchetypes
Reduceddata types
Part 3 – ?????
© 2005 Ocean Informatics
* * * Programme * * *
• The problems of health IT
• The problem of Interoperability
• 3 principles for solutions
• Standards
• Strategy
© 2005 Ocean Informatics
What should government do?• The market will not solve the problem - it is a
‘commons’ problem• Define & design national EHR infrastructure
based on standardised information, services and knowledge (like defining the internet)
• Centrally manage knowledge resources - data sets, archetypes, terminologies - but allow distributed development!
• Use economic encouragements and legislative pressure to get compliance
• Engage clinicians and consumers all the way
© 2005 Ocean Informatics
What should Standards Orgs do?
• Follow 3 principles of modularity
• Analyse todays standards and improve
• Improve development MO to more disciplined engineering, less ad hoc discussion
• Use constant implementation to provide validation – work with implementation orgs
• (these are the principles of openEHR)
© 2005 Ocean Informatics
What should vendors do?
• Get involved in standards development
• Choose standards judiciously
• Plan for the health computing platform
• Modularise your software using 3 principles
• Expect service-oriented computing environment, e.g. PIDS, CTS etc
• Be ready to use archetypes, templates, terminology
© 2005 Ocean Informatics
Final thoughts
© 2005 Ocean Informatics
Where we need to go
© 2005 Ocean Informatics
Where we can get in 5 years
© 2005 Ocean Informatics
Where most of us are today