<!DOCTYPE article PUBLIC "-//NLM//DTD JATS (Z39.96) Journal Archiving and Interchange DTD v1.0 20120330//EN" "JATS-archivearticle1.dtd">
<article xmlns:xlink="http://www.w3.org/1999/xlink">
  <front>
    <journal-meta />
    <article-meta>
      <title-group>
        <article-title>The New SNOMED CT International Medicinal Product Model</article-title>
      </title-group>
      <contrib-group>
        <contrib contrib-type="author">
          <string-name>Olivier Bodenreider</string-name>
          <email>olivier.bodenreider@nih.gov</email>
          <xref ref-type="aff" rid="aff0">0</xref>
          <xref ref-type="aff" rid="aff1">1</xref>
        </contrib>
        <aff id="aff0">
          <label>0</label>
          <institution>Blue Wave Informatics LLP Exeter</institution>
          ,
          <country country="UK">United Kingdom</country>
        </aff>
        <aff id="aff1">
          <label>1</label>
          <institution>U.S. National Library of Medicine National Institutes of Health Bethesda</institution>
          ,
          <addr-line>Maryland</addr-line>
          ,
          <country country="US">USA</country>
        </aff>
      </contrib-group>
      <pub-date>
        <year>2018</year>
      </pub-date>
      <fpage>7</fpage>
      <lpage>10</lpage>
      <abstract>
        <p>- Objectives. To present the new SNOMED CT international medicinal product model. Methods. We present the main elements of the model, with focus on types of entities and their interrelations, definitional attributes for clinical drugs, and categories of groupers. Results. We present the status of implementation as of July 2018 and illustrate differences between the original and new models through an example. Conclusions. Benefits of the new medicinal product model include comprehensive representation of clinical drugs, logical definitions with necessary and sufficient conditions for all medicinal product entities, better high-level organization through distinct categories of groupers, and compliance with international standards.</p>
      </abstract>
    </article-meta>
  </front>
  <body>
    <sec id="sec-1">
      <title>-</title>
      <p>Keywords—SNOMED CT; Medicinal products; Knowledge
representation.</p>
    </sec>
    <sec id="sec-2">
      <title>I. INTRODUCTION</title>
      <p>
        The representation of medications in SNOMED CT [
        <xref ref-type="bibr" rid="ref1">1</xref>
        ] has
long been suboptimal for reasons that are briefly analyzed
below. Over the past ten years, several working groups of
SNOMED International have worked at addressing various
issues related to the representation of medicinal products and
substances, including:
      </p>
    </sec>
    <sec id="sec-3">
      <title>Which levels of granularity should be represented in the international release of SNOMED CT? (e.g., do branded drugs belong in the international release?)</title>
    </sec>
    <sec id="sec-4">
      <title>How to optimally represent drug classes? (i.e., at the</title>
      <p>substance level vs. product level?)</p>
      <p>These questions have been debated in the community, but
only recently has SNOMED CT published a new model for the
representation of medicinal products in the international release
and engaged in retrofitting its content to this model.</p>
      <p>The main objective of this work is to present the new
SNOMED CT international medicinal product model available
in the July 2018 version of SNOMED CT – a major revision of
the original model, with focus on the types of entities defined
and their interrelations. In addition, we discuss the benefits of
the new medicinal product model for the SNOMED CT user
community.</p>
    </sec>
    <sec id="sec-5">
      <title>II. BACKGROUND To set the stage for the new SNOMED CT international medicinal product model, we briefly discuss the limitations of</title>
      <p>stating exactly which active ingredients a clinical drug contains
vs. at least which ingredients it contains).</p>
      <sec id="sec-5-1">
        <title>B. Use cases</title>
        <p>
          Unlike many aspects of clinical medicine, description of
medications tends to be country-specific. For example, brand
names, formulations and strengths of active ingredient
substances may all differ across countries. Moreover, national
regulatory agencies may have different criteria for authorizing
medicinal products, and all medicinal products can only be made
available for sale or supply when they have an authorization of
some kind. For these reasons, many countries have developed
their own medicinal product terminologies (e.g., RxNorm [6] in
the U.S, the NHS Dictionary of medicines and devices (dm+d)
[7] in the U.K, the Australian Medicines Terminology (AMT)
[
          <xref ref-type="bibr" rid="ref8">8</xref>
          ] in Australia, and the G-Standaard [
          <xref ref-type="bibr" rid="ref9">9</xref>
          ] in the Netherlands).
Although SNOMED CT is the largest clinical terminology in the
world, it would be difficult for its developers to maintain a
medication formulary for the entire world and support a large
variety of use cases, from e-prescribing to clinical analytics.
        </p>
        <p>For this reason, SNOMED International has decided to focus
on four major use cases:
•
•
•
•</p>
      </sec>
    </sec>
    <sec id="sec-6">
      <title>To facilitate international interoperability of medication concepts (e.g., for use in patient summaries and for crossborder care)</title>
    </sec>
    <sec id="sec-7">
      <title>To provide a strong foundation for member countries to develop their national medicinal product terminology (e.g., by adding package and branded product information)</title>
    </sec>
    <sec id="sec-8">
      <title>To support medication analytics for research purposes</title>
    </sec>
    <sec id="sec-9">
      <title>To support the development of international medication</title>
      <p>decision support, such as allergy checking and duplicate
therapy checking</p>
      <p>Additionally, since SNOMED CT is based on description
logic (DL), emphasis was put on defining medication concepts
with necessary and sufficient conditions, so that a DL classifier
could be used for checking consistency and automatically
inferring a subclass hierarchy among products.</p>
      <sec id="sec-9-1">
        <title>C. International standards</title>
        <p>
          In 2012, the health informatics technical committee in the
International Standards Organization (ISO), under the auspices
of Working Group 6, Pharmacy and Medications, published ISO
11615: Health informatics -- Identification of medicinal
products -- Data elements and structures for the unique
identification and exchange of regulated medicinal product
information. This was updated last year (2017) with minimal
changes to its core structures and definitions. This is the core of
a suite of standards known as IDMP: the Identification of
Medicinal Products [
          <xref ref-type="bibr" rid="ref10">10</xref>
          ], with supporting standards in the suite
describing supporting concepts such as substances, dose forms,
routes of administration, units of presentation and units of
measure. ISO 11615 provides a conceptual model and definition
of the key data items that are needed to describe a medicinal
product as it is authorized by a regulatory agency, in terms of its
pharmaceutical composition, its main clinical characteristics, its
supporting authorization and key information from its
manufacturing processes. ISO:11615 therefore describes three
types of things that are given global identifiers in the regulatory
domain (the medicinal product, the packaged medicinal product
and the pharmaceutical product).
        </p>
        <p>Since all medicinal product terminology must be sourced
and abstracted from what exists, and the authorized medicinal
product information from a regulatory agency within a
jurisdiction provides this, it is important that any clinical
medicinal product terminology fully understands ISO:11615
and manages its own information structures in harmony with it.
In addition, having the structure and population of clinical
medicinal product terminology in harmony with that used in the
regulatory environment facilitates the flow of information
between the two domains of use. This is particularly vital for
pharmacovigilance and patient safety, but also for analysis and
ongoing clinical research and pharmacoepidemiology.</p>
      </sec>
    </sec>
    <sec id="sec-10">
      <title>III. METHODS In this section, we present the main elements of the SNOMED CT international medicinal product model and discuss its implementation.</title>
      <sec id="sec-10-1">
        <title>A. Defining the new medicinal product model</title>
      </sec>
      <sec id="sec-10-2">
        <title>1) Types and interrelations</title>
        <p>The new SNOMED CT international medicinal product
model keeps the distinction between substances (chemical or
biological entities) and products (manufactured objects). The
two are intimately related in that substances are the ingredients
of products (e.g., the medicinal product “Product containing
atorvastatin” has the substance “Atorvastatin” as its active
ingredient).</p>
        <p>However, instead of a single type of entity for medicinal
products regardless of the level of granularity (“product”),
SNOMED CT now has multiple types and distinguishes whether
the product contains at least (existential restriction) or only
(universal restriction) a given active ingredient substance.
Moreover, although SNOMED CT did not introduce new types
for substances, it modified the relationship between base
ingredients (e.g., atorvastatin) and their salt or ester
modifications (e.g., atorvastatin calcium and atorvastatin
calcium trihydrate). In the past, base substances subsumed
substance modifications, which is incorrect and can lead to
wrong inferences (e.g., clinical drugs containing an ingredient
modification are distinct from products containing the
corresponding base ingredient and should not be subsumed by
them). To this end, SNOMED CT introduced a new relationship
“is modification of” used in replacement of “isa” between a
substance and its modification (e.g., atorvastatin calcium is
modification of atorvastatin, instead of atorvastatin calcium isa
atorvastatin).</p>
        <p>The main type for medications in SNOMED CT is “clinical
drug” (combining precise ingredient substance, strength and
dose form). In addition to “clinical drug”, SNOMED CT also
provides several types for aggregation and navigation purposes,
including “medicinal product form” (combining ingredient
substance and dose form or dose form group) and “medicinal
product” (only specifying the ingredient substance). These two
types exist in two flavors, with existential restrictions
(subsuming any medication containing a given ingredient
substance or its modification, which is convenient for clinical
analytics purposes) or universal restriction (subsuming all
medications that only contain this ingredient substance or its
modification, which is necessary for decision support and for
interoperability of clinical care). Optionally, SNOMED CT also
supports the representation of medicinal products defined in
reference to ingredient substance modifications (in the universal
restriction flavor), to support use cases in which the substance
modification is clinically significant (e.g., for many of the
glucocorticosteroid preparations).</p>
      </sec>
      <sec id="sec-10-3">
        <title>2) Definitional attributes for clinical drugs</title>
        <p>Presentation strength. One feature of the new SNOMED
CT international medicinal product model is to provide a
complete and detailed representation of medications. In the past,
clinical drug entities were all primitive classes. While active
ingredient substance and dose form were explicitly represented
through roles, strength was not. In contrast, the new model
provides necessary and sufficient conditions for all clinical
drugs. Moreover, to support computation, strength is
represented not as a string (e.g., “10 mg”), but as a series of
discrete elements (numerator value, numerator unit,
denominator value, denominator unit) and presentation strength
is represented instead of concentration strength for the majority
of products. The practical importance of the presentation
strength is to support the distinction among iso-concentration
products. For example, the distinction between the following
two clinical drugs would be lost if only the concentration
strength were recorded: “Product containing precisely
norepinephrine (as norepinephrine bitartrate) 2 milligram/2
milliliter conventional release solution for infusion ampule” and
“Product containing precisely norepinephrine (as
norepinephrine bitartrate) 20 milligram/20 milliliter
conventional release solution for infusion ampule”.</p>
        <p>Basis of strength substance (BoSS). The BoSS is the
substance in reference to which strength is expressed in a clinical
drug. The BoSS typically corresponds to the active ingredient
substance or the active moiety, but can also be an arbitrary
substance selected as the reference. In the new SNOMED CT
international medicinal product model, both the active
ingredient and the BoSS are explicitly recorded. For example, in
“Product containing precisely norepinephrine (as
norepinephrine bitartrate) 2 milligram/2 milliliter conventional
release solution for infusion ampule”, the active ingredient is
“norepinephrine bitartrate”, but the BoSS is “norepinephrine”.</p>
        <p>Dose form and unit of presentation. In the past, SNOMED
CT only provided a Has dose form attribute. In addition to
harmonizing dose forms with the international standard from
IDMP (ISO:11239) implemented by the European Directorate
for Quality in Medicines (EDQM), the new SNOMED CT
international medicinal product model distinguishes between
manufactured dose form (e.g., conventional release oral tablet)
and unit of presentation (oral solid dose forms, such as tablet;
containers, such as vial; and actuations, when the quantity of
product released is determined by a metering valve).</p>
      </sec>
      <sec id="sec-10-4">
        <title>3) Categories of groupers</title>
        <p>Past versions of SNOMED CT have integrated groupers
corresponding to different types of medication classes, including
classes reflecting a mechanism of action or physiologic effect
(e.g., “Hydroxymethylglutaryl-coenzyme A reductase
inhibitor”, “Antiplatelet agent”), a chemical structure (e.g.,
“Aminoglycosides”), an intended site of administration (e.g.,
“Otic dosage form product”), and a therapeutic role (e.g., “Anti
glaucoma preparation”, “Antimalarial agent”, “Alopecia
preparation”). These entities were just primitive classes under
which the appropriate drug products were arranged manually.
This ontology maintenance approach was both labor-intensive
and error-prone, and did not benefit from the support of DL
classifiers.</p>
        <p>Recently, the new SNOMED CT international medicinal
product model reorganized these groupers, such that all
medicinal products, including the groupers, could be defined
with necessary and sufficient conditions. To this end, groupers
were clustered into four main categories:</p>
        <p>Based on disposition (e.g., “Product containing
3-hydroxy3-methylglutaryl-coenzyme A reductase inhibitor” [product
grouper], defined as having active ingredient “Substance with
3hydroxy-3-methylglutaryl-coenzyme A reductase inhibitor
mechanism of action” [substance grouper], itself defined as
having disposition “3-hydroxy-3-methylglutaryl-coenzyme A
reductase inhibitor” [disposition]. Of note, individual substances
exhibiting this disposition are (manually) subsumed by the
substance grouper, while products having these individual
substances as their active ingredient are automatically classified
under the product grouper.</p>
        <p>Based on chemical structure (e.g., “Product containing
aminoglycoside” [product grouper], defined as having as its
active ingredient the substance “Aminoglycosides” [substance
grouper]. Of note, individual substances exhibiting this chemical
structure are (manually) subsumed by the substance grouper,
while products having these individual substances as their active
ingredient are automatically classified under the product
grouper.</p>
        <p>Based on intended site of administration (e.g., “Product
manufactured as parenteral dosage form” [product grouper],
defined as having manufactured dose form “Parenteral dose
form” [dose form grouper], itself defined as having intended site
“Parenteral” [intended site]. Of note, individual parenteral dose
forms are (manually) subsumed by the dose form grouper, while
products having these individual manufactured dose forms are
automatically classified under the product grouper.</p>
        <p>Based on therapeutic role. In some cases, therapeutic role
and mechanism of action are intimately related and both are
definitional properties for the drug (e.g., “Product containing
antimalarial”). In this case, the therapeutic role is simply treated
as a disposition, i.e., a definitonal characteristic. In other cases,
the therapeutic role reflects the culture and practice of healthcare
and may not be universally acknowledged by all regulatory
agencies. Here, the therapeutic role is a non-definitional
characteristic. The organization of medications under
therapeutic roles in SNOMED CT is still under development.</p>
      </sec>
      <sec id="sec-10-5">
        <title>B. Implementation</title>
        <p>New definitions. As mentioned earlier, one major difference
from the original medicinal product model is that all medicinal
products are now defined with necessary and sufficient
conditions in the new SNOMED CT international medicinal
product model, which enables the SNOMED CT developers to
detect inconsistencies and infer hierarchies automatically.</p>
        <p>New types. The implementation of the medicinal product
model has required the creation of several new types of
medications, initially all represented as “product”. These types
include Clinical drug, Medicinal product form, and Medicinal
product. Similarly, types had to be created for the new
definitional characteristics (e.g., Disposition, Dose form,
Intended site). Finally, the Substance hierarchy was significantly
revised to accommodate these changes.</p>
        <p>New attributes. Similarly, new attributes had to be created
to support the definition of medications in relation to these new
types. These attributes include Has disposition, Has
manufactured dose form, Has active ingredient, Has precise
active ingredient, Has basis of strength substance, and Has
presentation strength (for unit and value, and for numerator and
denominator).</p>
        <p>The vast majority of the changes described here will be
available as part of the July 2018 release of SNOMED CT
(international release).</p>
      </sec>
    </sec>
    <sec id="sec-11">
      <title>IV. RESULTS</title>
      <sec id="sec-11-1">
        <title>A. Quantitative results</title>
        <p>Table 1 shows the approximate number of concepts for the
various types of entities discussed above projected to be
available in the July 2018 international release of SNOMED CT,
along with examples for each type.</p>
        <p>In this example, we contrast the representation of the clinical
drug “Amlodipine 10 mg and atorvastatin 10 mg oral tablet” in
the original SNOMED CT medicinal product model (circa 2015)
and in the new SNOMED CT international medicinal product
model (preview of the July 2018 release).</p>
      </sec>
      <sec id="sec-11-2">
        <title>1) Axioms</title>
        <p>As shown in Fig. 1, the original model represented clinical
drugs as primitive concepts (i.e., only with necessary conditions
and SubClassOf axioms), while all clinical drugs (and most
medication entities) are fully defined (with necessary and
sufficient conditions and EquivalentTo axioms) in the new
model. In other words, the original model only provided a partial
representation of the clinical drugs (with no representation of the
strength, for example), while the new model now represents
strength. In practice, it also means that the product hierarchy is
entirely inferred by the DL classifier in the new model, whereas
it was created manually in the original model.</p>
        <p>Moreover, in addition to strength, the new model offers a
more complete representation of clinical drugs, including unit of
presentation and basis of strength substance.</p>
        <p>Of note, the classic approach to restricting a multi-ingredient
drug to specific ingredients is through closure axioms, for
example, the last axiom (underlined) in: has_ingredient SOME
amlodipine and has_ingredient SOME atorvastatin and
has_ingredient ONLY (amlodipine or atorvastatin). However,
the flavor of DL used in SNOMED CT, ℇℒ++, does not support
universal restrictions (needed to express has_ingredient ONLY
…). The workaround implemented in SNOMED CT consists in
adding an axiom for the count of active ingredients instead, to
distinguish this clinical drug from other clinical drugs
containing additional ingredients. While somewhat unorthodox,
this workaround is effective for the representation of medicinal
products.</p>
      </sec>
      <sec id="sec-11-3">
        <title>2) Subclass hierarchy</title>
        <p>As mentioned earlier, the subclass hierarchy consisted of
stated SubClassOf relations in the original model, whereas it is
entirely inferred in the new model. Moreover, as shown in Fig.
2A, different types of groupers were entangled in the original
model (e.g., for dispositions and therapeutic roles). While
dispositions are definitional properties (e.g., atorvastatin has
always the disposition of being an inhibitor of a specific
enzyme), therapeutic roles are not (i.e., atorvastatin may not be
recognized as an antilipemic agent by all regulatory agencies).
In the new model (Fig. 2B), therapeutic roles are no longer part
of the subclass hierarchy. In the past, having therapeutic roles in
the subclass hierarchy could lead to incorrect inferences (e.g.,
timolol is an antiglaucoma drug only when administered in eye
drops), and ad hoc solutions were required to avoid such wrong
inferences (e.g., link the therapeutic role to the medicinal
product form rather than the medicinal product).</p>
      </sec>
    </sec>
    <sec id="sec-12">
      <title>V. DISCUSSION</title>
      <sec id="sec-12-1">
        <title>A. Benefits of the new medicinal product model</title>
        <p>The new SNOMED CT international medicinal product
model provides a more comprehensive representation of clinical
drugs (e.g., explicit representation of strength, unit of
presentation, basis of strength substance). Providing necessary
and sufficient conditions for each clinical drug not only supports
automatic inference of the subclass hierarchy, but also supports
the identification of equivalent classes. The latter will be
required for interoperability between the international release of
SNOMED CT and national medicinal product extensions, in
which clinical drugs will be defined, along with country-specific
medication entities, such as packages and branded products. In
other words, the new model provides a strong foundation for
national medication extensions. Member countries without a
medicinal product extension can enrich the international model
with the information required for their use cases (e.g., packaging
information and brand names for e-prescribing purposes).
Member countries that already have medicinal product
terminologies (e.g., RxNorm, NHS dm+d, AMT) should be able
to align the clinical drugs in these terminologies to SNOMED
CT’s and will benefit from the high-level organization in
SNOMED CT to organize their content. Once national
extensions are aligned with SNOMED CT, medication entities
in various extensions are de facto interoperable through their
shared relations to SNOMED CT.</p>
        <p>The distinction introduced in the new model among the
various types of groupers (related to dispositions, to chemical
structure, to intended site, and to therapeutic roles) not only
facilitates the organization and navigation of drug entities, but
also avoids wrong inferences without having to use ad hoc
solutions (such as linking the medicinal product form rather than
the medicinal product to a therapeutic role). In other words, the
new model offers better distinction between definitional and
assertional knowledge.</p>
        <p>Finally, the new model is compliant with international
standards, such as IDMP, which is required to support
pharmacovigilance beyond national boundaries. Features of the
new model required for compliance with IDMP include
providing an exact list of ingredients for clinical drugs (i.e.,
“Product containing ONLY…”), providing the manufactured
dose form (not only the administrable dose form) and
presentation strength (not only concentration strength). Related
to the representation of strength, the substance in reference to
which strength is expressed (i.e., the basis of strength substance)
must also be explicitly represented. Similarly for the unit of
presentation (in reference to the EDQM standard). In addition,
the SNOMED CT model extends and complements the IDMP
model in a compatible and harmonious way, providing
international identifiers for types of representation of medicines
for clinical use not required in the regulatory domain: in
particular the Clinical Drug, in its representation and
identification of the product using its manufactured dose form.
This is of significant value for international interoperability of
medication information for individual patients.</p>
      </sec>
      <sec id="sec-12-2">
        <title>B. Limitations and future work</title>
        <p>As mentioned earlier, the absence of universal restrictions in
ℇℒ++, the flavor of description logic used by SNOMED CT,
requires a workaround to ensure that a clinical drug does not
have any other active ingredients than the active ingredients
listed as part of the definition. This is achieved through the
addition of an axiom for the count of active ingredients in the
clinical drug. For example, as shown in Fig. 1, “Amlodipine 10
mg and atorvastatin 10 mg oral tablet” is made distinct from
(hypothetical) clinical drugs containing amlodipine, atorvastatin
and another ingredient, because it is also specified that the count
of active ingredients of this clinical drug is 2 – the count would
be 3 for the other drug. When SNOMED CT transitions to a
more expressive DL, the use of universal restrictions will make
it possible to express proper closure axioms, rendering this
ingredient counting axiom unnecessary.</p>
        <p>While a majority of oral solid dose form drugs will be
available if the new model in the upcoming international release
(July 2018), the conversion effort with continue over the next
few releases for liquids (oral solutions and parenteral drugs) and
topical drugs.</p>
        <p>
          While SNOMED CT is expected to provide a strong
foundation for the development or anchoring of national drug
extensions, this remains to be evaluated in practice. No national
drug extensions have been developed based on the international
model yet. Following earlier attempts (e.g., [
          <xref ref-type="bibr" rid="ref11">11</xref>
          ]), efforts are
underway to harmonize existing drug terminologies (RxNorm,
dm+d, AMT) with SNOMED CT, but no metrics of success
have been reported yet. Assuming these efforts success, the
national terminologies harmonized with SNOMED CT could
provide a mechanism for the long-term maintenance of the
SNOMED CT drug content, i.e., new clinical drugs in national
extensions could be automatically promoted to the international
release of SNOMED CT, and ancillary entities (medicinal
product and medicinal product form) could be automatically
derived from the definition of clinical drugs.
        </p>
      </sec>
    </sec>
    <sec id="sec-13">
      <title>VI. CONCLUSIONS</title>
      <p>The new SNOMED CT international medicinal product
model represents a major revision of the original model in use
over the past ten years. Benefits of the new model include
comprehensive representation of clinical drugs, logical
definitions with necessary and sufficient conditions for all
medicinal product entities, better high-level organization
through distinct categories of groupers, and compliance with
international standards.</p>
    </sec>
    <sec id="sec-14">
      <title>ACKNOWLEDGMENT</title>
      <p>This work was supported in part by the Intramural Research
Program of the NIH, National Library of Medicine. The authors
thank the contributors to the SNOMED International Drug
Model Working Group for useful discussion, particularly Jim
Case, Yongsheng Gao, Emma Melhuish, Toni Morrison,
Guillermo Reynoso, and Phuong Skovgaard. Special thanks to
Yongsheng Gao and Kai Kewley for sharing the OWL preview
of SNOMED CT used in the Protégé screenshots.</p>
    </sec>
  </body>
  <back>
    <ref-list>
      <ref id="ref1">
        <mixed-citation>
          [1]
          <string-name>
            <given-names>SNOMED</given-names>
            <surname>International</surname>
          </string-name>
          , “SNOMED CT,”
          <year>2018</year>
          .
        </mixed-citation>
      </ref>
      <ref id="ref2">
        <mixed-citation>
          [2]
          <string-name>
            <given-names>A.L.</given-names>
            <surname>Rector</surname>
          </string-name>
          , et al.,
          <article-title>“Getting the foot out of the pelvis: modeling problems affecting use of SNOMED CT hierarchies in practical applications</article-title>
          ,
          <source>” J Am Med Inform Assoc</source>
          , vol.
          <volume>18</volume>
          , no.
          <issue>4</issue>
          ,
          <issue>2011</issue>
          , pp.
          <fpage>432</fpage>
          -
          <lpage>440</lpage>
          .
        </mixed-citation>
      </ref>
      <ref id="ref3">
        <mixed-citation>
          [3]
          <string-name>
            <given-names>A.</given-names>
            <surname>Zakharchenko</surname>
          </string-name>
          and
          <string-name>
            <given-names>J.</given-names>
            <surname>Geller</surname>
          </string-name>
          , “
          <article-title>Auditing of SNOMED CT's Hierarchical Structure using the National Drug File - Reference Terminology,” Stud Health Technol Inform</article-title>
          , vol.
          <volume>210</volume>
          ,
          <year>2015</year>
          , pp.
          <fpage>130</fpage>
          -
          <lpage>134</lpage>
          .
        </mixed-citation>
      </ref>
      <ref id="ref4">
        <mixed-citation>
          <string-name>
            <given-names>K.</given-names>
            <surname>Greibe</surname>
          </string-name>
          , “
          <article-title>Development of a SNOMED CT based national medication decision support system,” Stud Health Technol Inform</article-title>
          , vol.
          <volume>192</volume>
          ,
          <year>2013</year>
          , pp.
          <fpage>1147</fpage>
          .
        </mixed-citation>
      </ref>
      <ref id="ref5">
        <mixed-citation>
          <string-name>
            <given-names>F.R.</given-names>
            <surname>Goss</surname>
          </string-name>
          , et al.,
          <article-title>“Evaluating standard terminologies for encoding allergy information</article-title>
          ,
          <source>” J Am Med Inform Assoc</source>
          , vol.
          <volume>20</volume>
          , no.
          <issue>5</issue>
          ,
          <issue>2013</issue>
          , pp.
        </mixed-citation>
      </ref>
      <ref id="ref6">
        <mixed-citation>
          <string-name>
            <surname>U.S. National Library</surname>
          </string-name>
          of Medicine, “RxNorm,”
          <year>2018</year>
          .
        </mixed-citation>
      </ref>
      <ref id="ref7">
        <mixed-citation>
          <string-name>
            <surname>NHS Business Service</surname>
            <given-names>Authority</given-names>
          </string-name>
          , “
          <article-title>Dictionary of medicines and devices (dm+d</article-title>
          ),”
          <year>2018</year>
          .
        </mixed-citation>
      </ref>
      <ref id="ref8">
        <mixed-citation>
          [8]
          <string-name>
            <given-names>Australian</given-names>
            <surname>Digital Health Agency</surname>
          </string-name>
          , “Australian Medicines Terminology,”
          <year>2018</year>
          .
        </mixed-citation>
      </ref>
      <ref id="ref9">
        <mixed-citation>
          [9]
          <string-name>
            <given-names>Royal</given-names>
            <surname>Dutch Pharmacists</surname>
          </string-name>
          <article-title>Association (KNMP), “The G-Standaard: the medicines standard in healthcare</article-title>
          ,”
          <year>2018</year>
          .
        </mixed-citation>
      </ref>
      <ref id="ref10">
        <mixed-citation>
          [10]
          <string-name>
            <given-names>European</given-names>
            <surname>Medicines</surname>
          </string-name>
          <string-name>
            <surname>Agency</surname>
          </string-name>
          , “Introduction to ISO Identification of Medicinal Products,”
          <year>2018</year>
          .
        </mixed-citation>
      </ref>
      <ref id="ref11">
        <mixed-citation>
          [11]
          <string-name>
            <given-names>S.J.</given-names>
            <surname>McBride</surname>
          </string-name>
          , et al.,
          <article-title>“Using Australian medicines terminology (AMT) and SNOMED CT-AU to better support clinical research,” Stud Health Technol Inform</article-title>
          , vol.
          <volume>178</volume>
          ,
          <year>2012</year>
          , pp.
          <fpage>144</fpage>
          -
          <lpage>149</lpage>
          .
        </mixed-citation>
      </ref>
    </ref-list>
  </back>
</article>