A proposal for interoperable health information exchange with two Esperantos: ICF and LOINC. Presented at the 2010 NAAC ICF Conference: Enhancing our Understanding of the ICF.
4. Introduction
• Regenstrief’s 35-year history
• Indiana Network for Patient Care
– A working HIE for 15 years
– 200+ source systems
– 10.5 million patients, 3 billion results
– Regenstrief: 3rd party convener
• Regenstrief is the 1st WHO collaborating center in
medical informatics
• A fundamental challenge
– Local systems use idiosyncratic codes
• Vocabulary standards
– Provide the lingua franca of information exchange
6. LOINC Background
• Logical Observation Identifiers Names and Codes
• Organized by Regenstrief Institute in 1994
– Ongoing support from NLM and Regenstrief
• Covers domain of Clinical Observations
– Laboratory Observations (since 1995)
– Clinical Observations (since 1996)
• A universal code system that facilitates
exchange, pooling, and processing of results
7. LOINC’s General Role
• If an observation is a question, and the
observation value an answer:
– LOINC provides codes for the questions {OBR-4, OBX-3}
What is my patient’s hemoglobin level?
718-7:Hemoglobin:MCnc:Pt:Bld:Qn
How fast does my patient usually walk?
41959-8:Walking speed:Vel:1W^mean:^Patient:Qn:Calculated
8. Indiana Network for Patient Care
HL7 v.2.X Message
MSH|^~&|HOSPITAL_A|SAMPLE_HOSPITAL_A|||$YearMonthDay|||||||||||||||
PID|||$patientId$||$patientName$||||||||||||||||||||
PV1|||||||$attendingDoctor$||$consultingDoctor$||||||||
OBR|1|||44249-1^PHQ-9 Quick Depression Assessment Pnl^LN||$requestDate|||||||||
OBX|1|ST|44250-9^Little interest or pleasure in doing thing in last 2W^LN|1|3^More than
half the days^LN|||||||||||||
OBX|2|ST|44255-8^Feeling down, depressed, or hopeless in last 2W^LN|1|2^Several
days|||||||||||||
…
OBX|10|ST|44261-6^PHQ-9 Total Score^LN|1|11|||||||||||||
A
A
10. “Flat” Data Model
Patient_ID Provider_ID Date Height Weight Heart_Rate
1111 77777 2010 04 09 183 cm 90.7 kg 74 bpm
2222 77777 2010 04 09 152 cm 49.9 kg 65 bpm
One record per patient
11. “Stacked” Data Model
Patient_ID Provider_ID Date Observation_Code Observation_Name Value Units
1111 77777 2010 04 09 1234-5 Body Height 183 cm
1111 77777 2010 04 09 2345-6 Body Weight 90.7 kg
1111 77777 2010 04 09 3456-7 Heart Rate 74 bpm
2222 77777 2010 04 09 1234-5 Body Height 152 cm
2222 77777 2010 04 09 2345-6 Body Weight 49.9 kg
2222 77777 2010 04 09 3456-7 Heart Rate 65 bpm
One record per observation
15. A Highly ‘Open Source’ Model
• LOINC (the database) and RELMA (the
mapping program) are available freely
worldwide for nearly any purpose
• Much work is done by volunteers
• Content additions are end-user driven
20. Consolidated Health Informatics
• CHI Goal:
– Adopting interoperability standards for all US federal
health agencies
• Adopted LOINC as standard
– Laboratory result names (2003)
– Laboratory test order names (2006)
– Meds: structured product labeling sections (2006)
– Federally-required patient assessment instruments
with functioning and disability content (2007)
• Same process that adopted ICF as a standard for functioning
and disability domain
21. Other Key US Adoptions
• eLINCS
– Messaging standard for results delivery from LIS to an EHR
• NAACCR
– Volumes II (Data Standards/Dictionary) and V (Path Lab e-Reporting)
• CDISC
– Pharmaceutical research specs
• NCQA/HEDIS
– Used by 90% of US health plans to measure quality
• HITSP
– C80: vital signs, lab results, lab orders, genetic results, other results
– IS92: newborn screening
– C83: Patient assessment instruments (sections, questions, answers)
22. A Proposal for Effective
use of ICF and LOINC
Making complementary strengths productive
23. General Observations
• No computer-interpretable version of ICF
• Links with other vocabularies (UMLS, SNOMED)
don’t address qualified codes
• Several ICF item collections
– Full version, short version, ICF-CY, ICF core sets,
more…
• Challenge: ICF classification blends several
observation question/answer pairs into 1 code
– d410.1302 (changing basic body position) is really 4
“observations”
24. Goals
• Send a person (or population)’s ICF
classification using same machinery as other
health data
– To reach ICF’s goals, you need to share data
• Maximize strengths of each terminology
(minimize duplication of effort)
• Be informed by real world use
– Need some interested parties!
• Facilitate addressing challenges in ICF use
– Relationship to standardized assessments and clinical
measures
25. Original Option 1
• Simplest Approach: One LOINC code
– NNNN-N:Functioning Classification:Imp:^Patient:Pt:Ord:ICF
– Expected “answer” in OBX-5 would be a ICF classification
• Problems with Simplest Approach
– Still have blending of question/answer in OBX-5
– No indications of sets
26. Original Option 2
• Full LOINC Modeling including panels for
ICF Sets
• Example: d420 – Transferring oneself
– N-N:Transferring oneself.Performance:Imp:^Patient:Pt:Ord:ICF
– N-N:Transferring oneself.Capacity:Imp:^Patient:Pt:Ord:ICF
– Expected “answers” in OBX-5 would be the ICF qualifiers
0
–
No
setup
or
physical
help
from
staff
1
–
Setup
help
only
2
–
One
person
physical
assist
3
–
Two+
person
physical
assist
8
–
ADL
acBvity
itself
did
not
occur
during
enBre
7
days
27. Original Option 2
• Problems with this approach
– Labor intensive
• Each ICF component + qualifier combination
would be a different LOINC code (assessing
different attributes)
• Keeping up with sets would be very difficult
– Some modeling challenges (e.g. anatomy)
– Negotiating IP issues
31. HL7 CDA Framework for
Questionnaire Assessments
• Specifies a document package representing the
full assessment “form”
• For each observation/answer, enables
concurrent transmission of:
– Model of Use (LOINC)
• Exact measurement, as on the assessment
– Model of Meaning (SNOMED, ICF) [optional]
• Representation of the conceptual assertion in another
(standard) terminology/classification
– Supporting Clinical Observations (LOINC, SNOMED)
[optional]
• Data from the EHR that supports the assessment decision
32. Proposed ICF Result Package in LOINC
ICF
classificaBon
panel
ICF
collecBon,
populaBon
descriptor,
observaBon
Bme
period,
other
descriptors
of
the
observaBon
period
1 to many
ICF
classificaBon
results
panel
ICF
component,
any
applicable
qualifiers,
fully-‐qualified
ICF
item
0 to many ICF
supporBng
clinical
observaBons
panel
Any
supporBng
clinical
measurements
for
that
ICF
classificaBon
(direct
measures,
assessment
scores,
etc)
33. Example ICF Result Package in LOINC
R/O/C
Example
Answers
NN-‐N
ICF
classifica9on
panel
NN-‐N
ICF
classificaBon
collecBon
R Full
NN-‐N
PopulaBon
descripBon
O Clinic population >65 years
NN-‐N
DuraBon
of
observaBon
period
O Point in time
R
1 to N
NN-‐N
ICF
classifica9on
results
panel
R d450
NN-‐N
ICF
code
stem
O d450.12
NN-‐N
ICF
funcBoning
classificaBon
C 1 – MILD difficulty
NN-‐N
AcBviBes
and
parBcipaBon
performance
qualifier
C 2 – MODERATE difficulty
NN-‐N
AcBviBes
and
parBcipaBon
capacity
without
assistance
qualifier
O
0 to N NN-‐N
ICF
suppor9ng
clinical
observa9ons
panel
59460-‐6
Morse
Fall
Risk
Total
55
0.9 m/sec
4195703
Mean
walking
speed
24H
34. Benefits of Nested Model
• Uses HL7-LOINC messaging framework while minimizing
redundant modeling
• Accommodates ‘meta-data’ about the result package
• Flexes to accommodate large or small sets of ICF codes
• Enables explicit connection between ICF classification
and supporting clinical data
• Accommodates sending alternate identifiers (e.g. UMLS
or SNOMED) for ICF components
• Could also use the ICF classification result panel in
another context
– nested under a regular clinical observation to convey the
higher level interpretation of that result
35. Next Steps
• Looking for collaborators with live
systems that have a need to exchange ICF
classifications electronically
– And want to used established messaging
standards
• Present to Clinical LOINC Committee
7/16/2010
• To infinity and beyond…