4. Panels (Batteries) in LOINC
• Panel term linked to enumerated child elements
– Child elements can be panels themselves (nesting)
• Panel term names (under discussion)
– Component often have “panel”, include authoritative source
– Property typically “-” because child elements will vary
– Scale typically “-” because child elements will vary
– Class PANEL.*
• Child elements linked and identified as:
– Required (R) Element always reported with panel
– Optional (O) Element may not be reported depending on
institutional policies or capabilities
– Conditional (C) Element is a key finding and thus should be
assumed to be negative, absent, or not present if
panel result does not include data for this
element.
10. Introduction
• Patient assessments are widely used to measure a broad
range of health attributes
– Functional status, depression, health-related quality of life, etc
– Survey instruments, questionnaires, assessment forms, etc
• Observations from patient assessments (whether
clinician-observed or self-report) are in many respects
very similar to other kinds of clinical observations
• Survey instruments have psychometric properties
• Question meaning tightly coupled with answers
General Aim: LOINC could serve as a “master question file”
and provide a uniform representation
11. Approach
• Iterative refinement of the base panel model as
we added new content
– Kept uncovering new wrinkles
• Collaborated with many people
– Tom White, CHI Functioning and Disability
workgroup, APSE, AHIMA, CMS, RTI, HL7, HITSP, and
others
• Represent the full assessment content with
attributes at three levels
– Individual item, answer list, panel-specific item
instance
12. Attributes of Assessment Items
• Question (item) name/text
– Exact question text, form-specific display name
• Data type
• Definition/description
• For numeric values: units of measure, range checks
• For categorical results: answers in an answer list
• Copyright and terms-of-use notices
• HL7 field sub-id
• HL7 data types (v2 and v3)
13. Structured Answer Lists
• Many items have highly specialized, fixed answer lists
– Often the answer lists define the meaning of the question
– Few are represented by existing codes in reference terminologies
• LOINC has created answer codes where needed
– Have “LA” prefix and a mod-10 check digit
– Are unique by lexical string (ignoring capitalization)
– Intentionally do NOT distinguish based on context-specific meaning
• In some cases, the answer list is identified with a
Regenstrief-assigned OID (for HL7 CDA use)
– Identify lists as “normative” vs “example”
• Answer list shows sequence, but not bound by it
• Store local codes for items and have place to store
universal code (e.g. SNOMED) if we’re able
14. Attributes of Items in a Panel Instance
• Some non-defining attributes of an item vary by panel
– Vary across instruments or different forms of the same
assessment
• Represented at the level of the item instance in a panel
– Display name override (e.g. “BMI” vs “Body Mass Index”)
– Cardinality
– Observation ID in form (local code)
– Skip logic
– Data type in form
– Answer sequence override
– Consistency/validation checks
– Relevance equation
– Coding instructions
15.
16.
17.
18.
19. Advantages of the Master Catalog
• Single database (LOINC) contains the details
about individual observations and sets
– In the database, all forms (sets) look the same
– Automatic standardization
• Separates the form structure, question details,
the rendered version (paper or screen), and the
program that manages it
• Can easily reuse observations (and attributes)
in different forms/sets
21. Rules for Display of Items
• SURVEY_QUEST_TEXT (if populated). Used when item is asked as
a question. Sometimes the item has a label and a question, so we
store both as [label].[question text]
Pain Presence. Ask resident: “Have you had pain or hurting at
any time in the last 7 days?”
• DISPLAY_NAME_FOR_FORM (if populated). Provides an override
display linked to the instance of the LOINC in a particular form.
Allows for presentation variation that doesn’t affect meaning and
for where the LOINC naming conventions require some difference
b/w the item and the LOINC Component.
Item label = “Body Mass Index (BMI)”
LOINC Component = “Body mass index”
• COMPONENT. This is the default display
23. Currently in LOINC
• US Government Forms
– CARE, MDSv2, MDSv3, OASIS B1, OASIS C RFC
– US Surgeon General’s Family Health Portrait
• Brief Interview for Mental Status (BIMS)
• Confusion Assessment Method (CAM)
• Geriatric Depression Scale (GDS)
• HIV Signs and Symptoms Checklist
• Home Health Care Classification
• howRU
• Living with HIV (LIV-HIV)
• Morse Fall Scale
• OMAHA
• PHQ (9 and 2)
• Quality Audit Marker (QAM)
25. ASPE as Key Supporter
• ASPE (Jennie Harvell) has championed use of
HIT standards for assessment instruments in
many venues
• Initial Reports
– Making the "Minimum Data Set" Compliant with
Health Information Technology Standards
– Standardizing the MDS with LOINC® and Vocabulary
Matches
26. 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)
27.
28.
29. Many Other Opportunities
• PhenX Measures
• PROMIS
• Neuropsychological testing instruments (APA)
• Lots of other commonly-used instruments
(SF-36, etc)
• CDC case report and other forms
• National physical therapy outcomes database
measures
32. Variation Abounds
• Despite many instruments now in LOINC,
reuse of items has been minimal
– E.g. extremely few of same items b/w MDSv2
and MDSv3
– MDSv3 has greater similarity to CARE, but the
lookback period is different (7D vs 2D)
• We noticed differences that might have
been avoided
• Urge developers to weigh the cost of
losing comparability before inventing
36. A Uniform Data Model Would Help
• We usually started from paper forms, though some instruments had
their own software and data structures
• Forced to reconcile many potential discrepancies
– “Unknown”/“unable to determine” as answer choices vs flavors of null
– How do you store “Other specified ______”
– Units of measure implied?
– Which text is the item and which is “help”
• Incongruence with the HL7+LOINC model
– Items for things that could go in PID, etc
– Flat data model vs stacked
– Every ‘Check all that apply’ stored as separate yes/no item
• Interoperable data exchange standards haven’t been in the minds
of survey developers
• Starting with the LOINC model may elucidate hidden challenges
39. Intellectual Property Issues
• Must negotiate separate agreements with each
copyright/IP holder for inclusion in LOINC
• Many instruments have difficult restrictions
– Protection against change and attribution are understandable
– Some want royalties
– Commercial use in LOINC’s context is tricky
• Even more complicated when several instruments
included in larger CMS ones (MDS, CARE, etc)
• Funders should require developers to avoid such
restrictive licenses
41. Always New Challenges
• Answer list sequences
– Same answers across instruments but different order
• Skip logic shown at level of answer
– Current strategy is to aggregate up to question level
• Items with pictures
• Computer-adaptive testing coefficients and
attributes