Presentation at the 14th International Conference on e-Health Networking - Application and Services in 2012 .
See: http://www.mlhim.org http://gplus.to/MLHIM and http://gplus.to/MLHIMComm for more information about semantic interoperability in healthcare.
#mlhim #semantic_interoperability #health_informatics
Potential of AI (Generative AI) in Business: Learnings and Insights
Presentation HealthCom 2012
1. KNOWLEDGE ENGINEERING OF
HEALTHCARE APPLICATIONS
BASED ON MINIMALIST
MULTILEVEL MODELS
EXPANDING THE SCOPE OF EHEALTH:
FROM ELECTRONIC HEALTH RECORDS TO
BIOMEDICAL APPLICATIONS
Luciana Tricai Cavalini Timothy Wayne Cook
Department of Health Information Technology MLHIM Associated Laboratory
Medical Sciences College National Institute of Science and Technology –
Rio de Janeiro State University Medicine Assisted by Scientific Computing
2. DYNAMICS AND COMPLEXIT Y IN
HEALTHCARE
Time
Healthcare systems are
Space much more complex
than any other sector of
human
society, regarding 3
dimensions:
Ontology
3. WHY HEALTHCARE IS SO COMPLEX?
Healthcare is the only economic sector that deals with
biological production processes (which are created by nature)
All other economic sectors deal with industrial production
processes (which are created by the man)
Production processes that are created by the man are much
simpler than the biological processes, because:
Civilization starts just dozens of Evolution had millions of years to
thousands of years ago reach to that complexity
Biological systems are as complex as
Industrial systems are as simple as
necessary to guarantee the survival
possible to maximize profit
of the species
See Dawkins R. The greatest show on earth, pp. 204-5, and Marx K. Complete works.
4. THE ONTOLOGICAL COMPLEXIT Y
In practical terms;
building a
The greatest medical Thus, in
“megalithic system”
terminology medicine, there are
that all healthcare
(SNOMED-CT) has roughly 310,000
settings could use
more than 310,000 concepts, connected
would require a
terms, connected by to each other by
great amount of
more than millions of different
tables with 310,000
1,000,000 links ways
fields and millions of
relationships
Cavalini-Cook Conjecture: The probability of consensus between 2 or
more experts from the same field regarding which would be the
“maximum data model” for any given healthcare concept tends to zero
5. THE CONSEQUENCES OF HEALTHCARE
COMPLEXIT Y (1)
This complexity turns a computer science problem that does not exist (or at least it is not
critical) in any other sector of human society into a very important issue in healthcare.
This problem is:
6. Chest X-Ray:
- Nodule in
right apex
- Cough
- For 3 months - Cough
- Low fever - For 3 months
BAL: - Low fever
- TB Chest X-Ray:
- Nodule in
right apex
BAL:
- TB
Chest X-Ray:
- Nodule in
right apex
- Cough
- For 3 months
- Low fever
7. A UNDERESTIMATED PROBLEM
Semantic interoperability in healthcare is not perceived as a
problem by the vast majority of health informaticians because:
Apparently, it only
Academic projects
concerns national
Most software are usually focused
governments, and no
companies are on a very specific
country nowadays has
satisfied with their subject, and recording
the required
customer portfolio or their data in isolated
combination of
still dream the old silos is not seen as a
technical
monopolistic dream problem, because
capability, political
of taking over the they do not regard
will and transparency
whole global market their data as part of
to run a semantically
for themselves the patient’s Life
interoperable national
Health Record
ehealth project
8. THE CONSEQUENCES OF HEALTHCARE
COMPLEXIT Y (2)
Semantic interoperability is critical, but healthcare complexity brings
another intractable issue even for self-contained systems: maintenance
In healthcare, you define your data model today and it does not last 6 months, because
healthcare concepts evolve fast and new concepts come along every day
It is virtually impossible to make a customer satisfied with a default application; the
requisites are completely different, even for the simpler cases (e.g. two NHS GPs)
In real life, the average time for a medical software to be abandoned is
2 years and the abandon rate is 70% (source: CHAOS Report)
9. MULTILEVEL MODELING APPROACHES
Models openEHR MLHIM 13606
Approach Maximalist Minimalist Reductionist
RM residual
context Intense Minimal Intermediate
Data model
Maximum Any size Maximum
Possible Only message
EMR Any application
implementation exchange
10. KNOWLEDGE MODELING APPROACHES
Models openEHR MLHIM 13606
Concept Constraint
Structure Archetype Definition (CCD)
Archetype
Language ADL XML Schema ADL
# of
One Any number One
structures /
concept
Governance Top- Bottom-up, Top-
model down, consensus merit down, consensus
11. THE MLHIM SPECIFICATIONS
IMPLEMENTATION
The MLHIM Reference Model
XML Schema
Graphical representation
Examples of CCDs
ICD-10 4-digit codes for Respiratory Tuberculosis (A15. -)
Demography NCI Standard Template
The Data Model Converter to CCD
The CCD Repository Uploader
Code available at:
www.mlhim.org or https://launchpad.net/mlhim