PosterPDF Available

Analyzing microbiology In Vitro Diagnostics data using the OMOP CDM and SNOMED CT encoding Analyzing microbiology In Vitro Diagnostics data using the OMOP CDM and SNOMED CT encoding

Authors:
Poster

Analyzing microbiology In Vitro Diagnostics data using the OMOP CDM and SNOMED CT encoding Analyzing microbiology In Vitro Diagnostics data using the OMOP CDM and SNOMED CT encoding

Abstract

Antimicrobial resistance (AMR) is transforming the treatment of common infectious diseases, driving strong international call for actions which include improved data sharing to support secondary usage of medical data (Emberger, Tassone et al. 2018). The European VALUE-Dx project (https://value-dx.eu) that aims at combating antimicrobial resistance (AMR) and improve patient outcomes is one of those. The present study is part of the VALUE-Dx project. Initiatives in several countries demonstrate the added value of nomenclature standards to support national AMR surveillance (Gansel, Mary et al. 2019). When clinical microbiology laboratory data plays a key role in the fight against AMR (Fournier, Drancourt et al. 2013), we have shown that IVD systems tests and test results are very well described using LOINC® and SNOMED CT®, up to supporting SNOMED CT® mediated analytic (Le Gall, Vachon et al. 2019). Indeed, we are able to map 91% (1,349/1,482) of our taxa (VITEK® 2 and VITEK® MS IVD systems), 65% (13/20) of our ordinal test results, 89% (320/361) of our drugs and between 64% (7/11) and 98% (39/40) of our specimen breakdown to SNOMED CT®. With all this standardized information we were able to use SNOMED CT® encoded specimens to infer EUCAST alternate breakpoints following antibiotic susceptibility tests. In addition to the possibilities offered by SNOMED CT®, analysis of laboratory data may require to be independent from the institution data representation system and may call for a standardized data structure as the one offered by OHDSI (https://www.ohdsi.org). OHDSI and their OMOP Common Data Model (https://www.ohdsi.org/data-standardization/the-common-data-model), or OMOP CDM, are well-known to support decentralized data aggregation and analytics, while preserving data privacy (Hripcsak, Duke et al. 2015). Numerous observational studies using OHDSI exist up to the (theoretical) feasibility to support of clinical trials (Bartlett, Dhruva et al. 2019). To our knowledge, no observational study addresses microbiology IVD laboratory data. As part of the European VALUE-Dx project our use cases can be summarized as - Identify candidate laboratories for clinical studies based on AMR results - Support microbiology laboratory results demography observations o Describe tests implemented per laboratory or region in the network o Describe test results and observations (which may involve high level results interpretation such as Multi-Drug Resistant phenotype - Magiorakos et al. 2012 - where MRSA may be an archetype) obtained per laboratory or region in the network Consequently, our work aims at (i) demonstrate the capabilities and limitations of OMOP CDM to represents LOINC® and SNOMED CT® encoded microbiology IVD laboratory data; (ii) envisage options to solve those limitations aiming at prepare an OMOP mediated microbiology IVD laboratory data analytics in a case study.
Analyzing microbiology In Vitro Diagnostics data using the OMOP CDM
and SNOMED CT encoding
M. Le Gall, J-F. Gorse, X. Gansel | bioMérieux
Analyzing microbiology In Vitro Diagnostics data using the OMOP CDM
and SNOMED CT encoding
M. Le Gall, J-F. Gorse, X. Gansel | bioMérieux
INTRODUCTION METHODS RESULTS DISCUSSION
INTRODUCTION
Background Topic
OHDSI and its OMOP Common Data Model [2], is well-known to support decentralized data
aggregation and analytics, while preserving data privacy (see fig. 1).
In the context of VALUE-Dx, our use cases supporting usage of OMOP are
Identify candidate laboratories for clinical studies based on antimicrobial resistance tests
and test results
Support microbiology laboratory results demography observations per laboratory or region
oDescribe tests implemented
oDescribe test results and observations (including high level results interpretation
such as Multi-Drug Resistant phenotype)
To our knowledge, no observational study addresses microbiology laboratory data.
Objectives
Our work aims at
(i) Demonstrate the capabilities
and limitations of OMOP CDM
to represents LOINC® and
SNOMED CT® encoded
microbiology IVD laboratory
data
(ii) Envisage options to solve
those limitations aiming at
preparing future analysis
Analyzing specifically
laboratory microbiological
data implies capturing data at
a lower level than hospital or
regional EHR to gain a more
detailed level of information
laboratory information
system
middleware or even in vitro
diagnostics (IVD) devices
OMOP CDM makes extensive use of standardized vocabularies. Notably LOINC®and SNOMED
CT®are used to encode the OMOP «Standardized Clinical Data Tables» (see fig. 2) SPECIMEN,
MEASUREMENT and OBSERVATION.
We reused our previous work [3] showing that in vitro diagnostics (IVD) systems tests and
test results are very well described using LOINC®and SNOMED CT®,up to supporting
SNOMED CT®mediated analytic. Indeed, we were able to map 91% (1,349/1,482)of our taxa
(VITEK®2 and VITEK®MS IVD systems), 65% (13/20)of our ordinal test results, 89%
(320/361) of our drugs and between 64% (7/11) and 98% (39/40)of our specimen
breakdown to SNOMED CT®.
Figure 1- How OHDSI works (from G. Hripcsak - MedInfo conference, 2019)
Figure 2- OMOP CDM v6 tables in Yellow are candidate to host laboratory data. In Green table targeted to
store LOINC®and SNOMED®CT encoded element
This work was undertaken as part of the European VALUE-Dx project [1],
aiming to combat antimicrobial resistance (AMR) and improve patient
outcomes.
Analyzing microbiology In Vitro Diagnostics data using the OMOP CDM
and SNOMED CT encoding
M. Le Gall, J-F. Gorse, X. Gansel | bioMérieux
Analyzing microbiology In Vitro Diagnostics data using the OMOP CDM
and SNOMED CT encoding
M. Le Gall, J-F. Gorse, X. Gansel | bioMérieux
INTRODUCTION METHODS RESULTS DISCUSSION
Model analysis & data mappingProblem statement
METHODS
Aligning data logical view from a laboratory perspective with the OMOP CDM v6 is not
straightforward as shown in fig. 3.
The concepts of “isolate”, hierarchy of tests and of test results are absent from OMOP CDM
v6. Note that Some active discussions exist on the OMOP forum
Data model analysis was performed using all OMOP CDM v6 available documentation and tools
from the OHDSI sites & forum. It also reuses laboratory workflow analysis as in [4].
Terminology mapping uses our previous work [3]. Mapping strategy is described in fig. 4.
Figure 3-
This figure
shows a logical
representation
of laboratory
microbiology
data (in green)
compared to
the OMOP CDM
(in dark blue).
Figure 4- This figure presents the SNOM²ED CT®concepts mapped to data present in OMOP CDM table PERSON,
VISIT_OCCURRENCE, CARE_SITE, SPECIMEN, MEASUREMENT, OBSERVATION
Analyzing microbiology In Vitro Diagnostics data using the OMOP CDM
and SNOMED CT encoding
M. Le Gall, J-F. Gorse, X. Gansel | bioMérieux
Analyzing microbiology In Vitro Diagnostics data using the OMOP CDM
and SNOMED CT encoding
M. Le Gall, J-F. Gorse, X. Gansel | bioMérieux
INTRODUCTION METHODS RESULTS DISCUSSION
Results
RESULTS
We designed two options to represent laboratory microbiology data in OMOP v6. Both
involve a root MEASUREMENT with the intention to (i) mimic the isolates ; (ii) anchor the
specimen and all subsequent tests. Links are implemented through FACT_RELATIONSHIP.
The first one (fig. 5) is only base on MEASUREMENTs and stores susceptibility results in one
single record. Limitations are that organism as SNOMED CT concept is not a permitted value
to MEASUREMENT in CDM v6. The model only permits LOINC answers, that are not a
sustainable option to describe identification results. Storing quantitative value (i.e. MIC)
and qualitative interpretation (i.e. S/I/R) for drug susceptibility results in a single
MEASUREMENT is not clearly allowed / prohibited in CDM v6.
Figure 6- Model is
based on both
MEASUREMENTs and
OBSERVATIONs. One
MEASUREMENT as a
root anchoring both
(i) OBSERVATION
carrying the organism
concept and (ii)
MEASUREMENTs for
drug susceptibility
tests.
Drug susceptibility
tests are stored as 2
individual records for
the quantitative MIC
and its interpretation
(S/I/R)
Figure 5- Model is
based only on
MEASUREMENTs. The
root MEASUREMENT
may be any lab test. In
case of identification
test, organism as
SNOMED CT concept is
not a permitted value
to MEASUREMENT .
Child MEASUREMENTs
of identification are
drug susceptibility
tests.
The second model was implemented in a data end-point, populated with data extracted
from a middleware and an IVD device. Ongoing analysis show that OHDSI Athena tool does
not support FACT_RELATIONSHIP.
The second model (fig. 6) is based on both MEASUREMENTs and OBSERVATIONs. It allows to
store the chain of tests along a lab process and susceptibility results as two separate
MEASUREMENTs, one for the MIC and second one for the corresponding category (using
dedicated LOINC codes). The root MEASUREMENT host root identification test and
corresponding identified organism is hosted in the anchored OBSERVATION.
Analyzing microbiology In Vitro Diagnostics data using the OMOP CDM
and SNOMED CT encoding
M. Le Gall, J-F. Gorse, X. Gansel | bioMérieux
Analyzing microbiology In Vitro Diagnostics data using the OMOP CDM
and SNOMED CT encoding
M. Le Gall, J-F. Gorse, X. Gansel | bioMérieux
INTRODUCTION METHODS RESULTS DISCUSSION
Conclusions
Future Directions
1. https://value-dx.eu/
2. Hripcsak, G. et al. (2015). "Observational Health Data Sciences and Informatics (OHDSI):
Opportunities for Observational Researchers." Stud Health Technol Inform. 2015; 216:
574578
3. Le Gall, M. et al (2019). "SNOMED CT Coding and Analytics of in vitro Diagnostics
Observations." Stud Health Technol Inform 264: 1460-1461
4. Fournier, P.-E. et al. (2013). "Modern clinical microbiology: new challenges and solutions."
Nature Reviews Microbiology 11: 574-585
Discussion
DISCUSSION
This project has received funding from the Innovative Medicines Initiative 2
Joint Undertaking under grant agreement No 820755. This Joint
Undertaking receives support from the European Union’s Horizon 2020
research and innovation programme and EFPIA and bioMérieux SA, Janssen
Pharmaceutica NV, Accelerate Diagnostics S.L., Abbott, Bio-Rad
Laboratories, BD Switzerland Sàrl, and The Wellcome Trust Limited.
www.imi.europa.eu // www.value-dx.eu
Representing laboratory data into OMOP is challenging
MEASUREMENT do not allow using SNOMED CT concepts to represent microbial or viral
identification results. The model allows using LOINC answers that is not a sustainable
solution for identification results.
Combination of measurements and observations allow to represent lab tests / results at
the cost of clarity and heavy usage of FACT_RELATIONSHIP.
A shared usage of FACT_RELATIONSHIP across all OMOP end-points is a blocking issue. If
this is achievable in a given project it is very challenging across independent projects
thus causing interoperability issues.
Under the OMOP CDM v6, a merge between the two models may give good results provided
the implementation is shared across all end-points. Ideally an evolution of OMOP CDM is
needed to accurately represent laboratory data and prevent usage of FACT_RELATIONSHIP.
OMOP CDM supports the representation of lab. microbiology data. with limitations. Loaded
into a series of nodes within an OHDSI-ARACHNE federated architecture, it also proved to be
usable with some limits that we are investigating.
Usage of FACT_RELATIONSHIP, may jeopardize implementations across projects and the
wanted interoperability.
In line with some active discussions on the OHDSI forum, laboratory data need
Clarity on interpretation of the so call “convention” notably using SNOMED CT concepts
to describe identification results.
Evolution(s) of OMOP CDM to better represent lab data, get rid as much as possible of
FACT_RELATIONSHIP.
Finally we foresee the lack of concept model in the SNOMED CT®Organism hierarchy as a
future limitation if we are to use of the ontological nature of SNOMED®CT to support data
analytics.
In the close future, we will deepen our ongoing analysis of model implemented (fig. 6),
pursuing definition of a better representation of lab. data under OMOP CDM v6 & upper, and
implement a live Proof of Concept.
ResearchGate has not been able to resolve any citations for this publication.
Article
Full-text available
The vision of creating accessible, reliable clinical evidence by accessing the clincial experience of hundreds of millions of patients across the globe is a reality. Observational Health Data Sciences and Informatics (OHDSI) has built on learnings from the Observational Medical Outcomes Partnership to turn methods research and insights into a suite of applications and exploration tools that move the field closer to the ultimate goal of generating evidence about all aspects of healthcare to serve the needs of patients, clinicians and all other decision-makers around the world.
Article
In the twenty-first century, the clinical microbiology laboratory plays a central part in optimizing the management of infectious diseases and surveying local and global epidemiology. This pivotal role is made possible by the adoption of rational sampling, point-of-care tests, extended automation and new technologies, including mass spectrometry for colony identification, real-time genomics for isolate characterization, and versatile and permissive culture systems. When balanced with cost, these developments can improve the workflow and output of clinical microbiology laboratories and, by identifying and characterizing microbial pathogens, provide significant input to scientific discovery.
SNOMED CT Coding and Analytics of in vitro Diagnostics Observations
  • Le Gall
Le Gall, M. et al (2019). "SNOMED CT Coding and Analytics of in vitro Diagnostics Observations." Stud Health Technol Inform 264: 1460-1461