CDC/FDA/NLM/ONC/CMS

Public Workshop on Promoting Semantic Interoperability of Laboratory Data

Addressing LOINC coding challenges

Daniel J. Vreeman, PT, DPT, MSc

Regenstrief-McDonald Scholar in Data Standards, Indiana University School of Medicine
Director, LOINC and Health Data Standards, Regenstrief Center for Biomedical Informatics

@djvreeman

2016

Key Issues
1. Expertise and resources
2. Genetic testing
3. Sources of mapping variability
4. Maintenance
photo via Vernio77 | cc-by-sa

Mapping might not be

rocket science, but
It does require specialized expertise and
dedicated resources

Recall that mapping can occur at many points in

the lab data exchange flow
Causes of common mapping errors I see
Not recognizing key differences in aspects of the LOINC name
Not having all the information needed at hand
Life is hard(more on that later)

Education, training, other support can help

Free LOINC workshops and tutorials
CDC-funded Lab Interoperability Cooperative
LOINC Essentials book*
LOINC Mapping Validity Check as part of LOINC Premium Membership
*Disclosure of financial conflict of interest: Im the author of that book.

Large labs are making progress

Smaller labs struggle more

All would benefit from a process

simplified by having a preselected list from the IVD vendor

Not all juice is worth

the squeeze.
Pareto principle > Long Tail in standards.
Ignore challenge tests for now.
photo via elwillo

A few tests account for most

result volume
70% of the tests take 30% of your
mapping time (and vice versa)
Not every variable will have a
LOINC code

Cumulative Laboratory Observation Volume (%)

100%
90%
80%
70%
60%
50%
40%
30%
20%
10%
0%

500

1000

1500

2000

2500

3000

3500

4000

Number of Laboratory Observation Codes (N)

Vreeman DJ, Finnell JT, Overhage JM. A rationale for parsimonious laboratory term mapping by frequency. AMIA Annu Symp Proc. 2007 Oct 11:771-5. PMID: 18693941.

Cumulative Laboratory Observation Volume (%)

100%
90%
80%
70%
60%
50%
40%
30%
20%
10%
0%

500

1000

1500

2000

2500

3000

3500

4000

Number of Laboratory Observation Codes (N)

In INPC, 80 codes made 80% of volume

Vreeman DJ, Finnell JT, Overhage JM. A rationale for parsimonious laboratory term mapping by frequency. AMIA Annu Symp Proc. 2007 Oct 11:771-5. PMID: 18693941.

Some fruit is lower

than others.
Dont need everything all at once.
photo via iancarroll

Genetic Testing
Some docket comments suggested unfamiliarity
with LOINCs approach

photo via blprnt | cc-by

An evolving field
Changing reporting styles
LOINC terms reflect how data is
being reported
About 2000 terms in current
release

Deckard J, McDonald CJ, Vreeman DJ. Supporting interoperability of genetic data with LOINC. J Am Med Inform Assoc. 2015 Feb 5. PMID: 25656513.

4 Main Categories
Full mutation analysis
82534-9CASR gene full mutation analysis in Blood or Tissue by Sequencing

Targeted mutation analysis

81830-2 AGL gene targeted mutation analysis in Blood or Tissue by Molecular genetics
method

Mutation analysis limited to known familial mutations

81867-4 NOTCH3 gene mutation analysis limited to known familial mutations in Blood
or Tissue by Molecular genetics method

Specific genetic variations/other

Trinucleotide repeats
Large deletion/duplications
Chromosomal alterations
Aneuploidy/Uniparental disomy (UPD)
Karyotyping
Clonality in leukemia and lymphomas

Deckard J, McDonald CJ, Vreeman DJ. Supporting interoperability of genetic data with LOINC. J Am Med Inform Assoc. 2015 Feb 5. PMID: 25656513.

In addition to the narrative text they

report now, we recommend that
laboratories always report as discrete
variables chromosome analysis results,
genetic variation(s) found, and genetic
variation(s) tested forAll genetic studies
should also include an observation that
reports the reference
sequence / reference assembly build.
Deckard J, McDonald CJ, Vreeman DJ. Supporting interoperability of genetic data with LOINC. J Am Med Inform Assoc. 2015 Feb 5. PMID: 25656513.

Not all data producers

have the same worldview

photo via
[credit]
photo via ericmay
| cc-by-nc

Which leads to
Variations!

photo via vilseskogen | cc-by-nc

Different labs have different reporting styles for

the same test:
Document-level codes for representing the report as a whole
Nominal codes to report the variant found
Ordinal codes to report the presence or absence of a particular variant
Narrative impression codes for reporting the overall impression
Ordinal codes to report the overall impression as Positive or Negative

Further variation in the style of reporting

variations at different alleles
Same test could potentially be represented in
different ways in LOINC depending on how the
reporting was structured

The Solution

photo via Dawn Huczek | cc-by

Evolving towards a more

graceful reporting style
Top-level LOINC document codes for
different genes and types of studies
Structure other data using flexible, generic
information structure (HL7v2 or FHIR):
Still using LOINC codes for the observation IDs

Possible Sections of Genetics Report

1.
2.
3.
4.
5.

Overall study variables

Simple variants
Structural variants
Pharmacogenomics
Complex variants

Other evolving
conversations
Which distinctions in specimen are
clinically relevant
At what level should we make
distinctions among PCR-based methods

Other Sources
of Variation
Method vs methodless terms
Mass vs substance reporting
Similar specimens

Recently received funding for a new project to create

a roll-up mechanism of Equivalence Classes
Continue to promote the approach of mapping to the
level of specificity you have: i.e. dont throw away
information by mapping to a more generic term
We will be creating a flexible, extensible, and
computable mechanism for rolling up LOINC codes
into clinically relevant equivalence groups
Could help in constructing flowsheets, research
queries, quality measurement tracking, etc

Potassium [Moles/volume] (Blood)

6298-4 Potassium [Moles/volume] in Blood
32713-0 Potassium [Moles/volume] in Arterial blood
39790-1 Potassium [Moles/volume] in Capillary blood
41656-0 Potassium [Moles/volume] in Mixed venous blood
39789-3 Potassium [Moles/volume] in Venous blood
Several authors have studied agreement between arterial and venous blood measurements of potassium with blood
gas analyzers and found good intra-patient agreement.[1416] However, it is important to remember that the
concentration of key electrolytes is influenced by the presence of hemolysis. In vitro hemolysis of red cells in a blood
sample will cause release of K+ from RBCs and may produce an artifactual hyperkalemia. Laboratory assays of
venous samples are screened for hemolysis, but blood gas electrolyte measurements may not be.[17,18]

Stay up to date

photo via Arul Irudayam | by-nd

LOINC recommends
that users update to
the current version of
LOINC within 90 days
of its publication.

Why?
The most recent release is always our best
release ever.
The world moves fast.
Maximize the potential benefit of
standardized data.

Tools Like RELMA Can Help

Find Local Terms Mapped to Deprecated LOINCs

Tools Like RELMA Can Help

Find Local Terms Mapped to Deprecated LOINCs

Review commentary

Tools Like RELMA Can Help

Find Local Terms Mapped to Deprecated LOINCs

Review commentary
Check the alternatives

Tools Like RELMA Can Help

Find Local Terms Mapped to Deprecated LOINCs

Map when ready

Review commentary
Check the alternatives

Other Updates to Your Terms

Your testing may have changed
Your codes may have changed
Watch changes to codes, names, or units

Across many INPC

institutions, in 2 years after
go live we saw half as many
new local terms appear as
what we started with.
Vreeman DJ, Stark M, Tomashefski GL, Phillips DR, Dexter PR. Embracing change in a health information exchange. AMIA Annu Symp Proc. 2008 Nov 6:768-72.