=Paper=
{{Paper
|id=Vol-2807/shortO
|storemode=property
|title=An Ontological Representation of Sex and Gender Information (short paper)
|pdfUrl=https://ceur-ws.org/Vol-2807/shortO.pdf
|volume=Vol-2807
|authors=Paul Fabry,Adrien Barton,Jean-François Ethier
|dblpUrl=https://dblp.org/rec/conf/icbo/FabryBE20a
}}
==An Ontological Representation of Sex and Gender Information (short paper)==
https://ceur-ws.org/Vol-2807/shortO.pdf
An Ontological Representation of Sex and
Gender Information
Paul FABRYa,1, Adrien BARTONa,b,1 and Jean-François ETHIER a,1
a
Groupe de Recherche Interdisciplinaire en Informatique de la Santé (GRIIS),
Université de Sherbrooke, Quebec, Canada
b
Institut de Recherche en Informatique de Toulouse (IRIT), CNRS, France
Abstract. Sex and gender are important health determinants. It is therefore valuable
to represent them in medical records. However, these entities are complex to define.
In this paper we review some existing representations of sex and gender, with a
special focus on the informational entities representing them. We detail our proposal
for formalizing sex and gender informational entities according to the OBO Foundry
methodology. In particular, we introduce classes enabling to represent information
that may be either well-defined or ambiguous relatively to whether they represent
sex or gender.
Keywords. Sex, Gender, Information content entity
1. Introduction
The terms “sex” and “gender”, despite having clearly different meanings, are still
sometimes used interchangeably in some contexts to characterize a human individual.
While “sex” refers to characteristics that are biologically determined such as
chromosomes distribution and reproductive/sexual anatomy and physiology, “gender”
refers to socially constructed characteristics [1].
Sex is commonly categorized as female or male, especially in administrative
documents, but there is a great variety of biological categories beyond those two. There
is also a considerable diversity in gender differentiation, depending on how people
perceive themselves (gender identity) and how they express their gender (gender
expression).
Evidence of sex and gender differences have been reported in chronic disease,
physiological processes and the impact of lifestyle on health [2]. Both sex and gender
are important health determinants, therefore information about them is valuable to
support care providers. This being the case, it must also be included as variables in
clinical research in order to fully evaluate the impact of studied interventions.
Learning Health Systems (LHS) are conceptual frameworks that enable the tight
coupling of care delivery, research and knowledge transfer. Starting from data generated
1
Corresponding Authors. GRIIS, Université de Sherbrooke, 2500, boul. de l’Université, Sherbrooke
(Québec), J1K 2R, Canada. IRIT, Université Toulouse III, q118, route de Narbonne, 31062 Toulouse Cedex 9,
France. E-mails: paul.fabry@usherbrooke.ca, adrien.barton@irit.fr; jf.ethier@usherbrooke.ca.
Copyright © 2020 for this paper by its authors. Use permitted under Creative Commons License Attribution 4.0 International (CC BY 4.0).
through care delivery, research results are more likely to be relevant and the findings are
re-injected in care through knowledge transfer processes like audit/feedback tools and
decision support tools. This then triggers a new cycle [3]. LHSs rely on a common,
source-independent representation of clinical information in order to support
interoperability between various data sources and across the different activities. The goal
is to build a better understanding of a human individual through their data across multiple
sources. Applied ontologies can provide such a model.
As part of the LHS PARS3 (“Plateforme apprenante en recherche en santé et en
services sociaux” - https://griis.ca/en/solutions/pars3), we have developed several
ontologies for various domains, such as the Prescription of Drugs Ontology PDRO [4,5]
These ontologies can support the creation of a relational schema that is then mapped to
databases from various healthcare institutions, in order to support a system of data
mediation [6]. As a result, these ontologies are focused on the representation of
informational entities (IAO:information content entity – ICE) pertaining to health
information. The ontological representation of informational entities referring to sex and
gender is therefore of primary importance to us.
Many electronic medical record systems integrate sex and gender information into
patient demographic information. However, most medical records still only specify the
sex of an individual, often labelled “sex assigned at birth”, without providing the whole
spectrum of possible biological sexes, or the possibility to specify a gender instead of a
sex. Sometimes a field asking to choose “M” or “F” will not even specify explicitly
whether it refers to sex or gender. Given the ambiguity surrounding both data capture
(e.g. unclear question) or data storage (e.g. a database field labelled “M-F” without other
information to clearly understand the nature of the information stored in the field), our
ontology needs to be flexible enough to allow annotation of ambiguous data elements
while providing the means to annotate with a much higher degree of precision. In this
paper, we review existing terminological and ontological representations of sex and
gender, detail our proposal for formalizing sex and gender informational entities in our
demographic data ontology DEMO, and discuss some of its implications.
2. Representation of Sex and Gender
2.1. Representation of Sex and Gender in Health Data Standards
Distinction between sex and gender is acknowledged in the major international health
standards such as HL7’s Fast Healthcare Interoperability Resource (FHIR) or SNOMED
CT [7], as well as in laboratory and radiology standards such as LOINC [8] or DICOM
[9]. A comprehensive review is made available by Canada Health Infoway’s Sex and
Gender working group [10]. We will highlight here the most important aspects as they
pertain to our goals.
While many health care standards provide distinct terms or placeholders for sex and
gender, the definitions chosen for these elements make them more challenging to
integrate into an ontology. For example, in the version 4.0.1 of FHIR, at least three
elements are available for gender: administrative gender, person gender and patient
gender. administrative gender is defined as: “The gender of a person used for
administrative purposes.” An extension adds patient gender identity defined as the
gender the patient identifies with. The value set for patient gender identity contains more
choices than for patient gender above. It is important to note that while these four
elements contain the term “gender” and refer to it in their definitions, the term “gender”
itself is never defined, leaving much uncertainty around the terms. Moreover, it is not
clear whether the term refers to gender itself or to an informational entity that is about a
gender. Lastly, the use of field value “Other” adds an epistemic dimension incompatible
with a stable ontological representation (as the extension of the class “Other” depends
on the extension of its sibling classes, which might change if new sibling classes are
introduced [11]). So while multiple standards acknowledge the sex and gender terms,
more work is required to integrate them in a realist ontology.
2.2. Representation of Sex and Gender in Ontologies
2.2.1. Sex and Gender Entities
The gender, sex and sexual orientation ontology (GSSO) identifies, categorizes and
associates to other terminologies thousands of terms related to gender, sex, and sexual
orientation [12]. GSSO is available through NCBO BioPortal. However, GSSO is only
very partially aligned with the Basic Formal Ontology and articulated with the OBO
Foundry principles [13] and it does not integrate the ICEs that are relevant to us.
Biological sex is formalized in the Phenotype And Trait Ontology (PATO) [14] as
an organismal quality that determines the bearer’s ability to undergo sexual reproduction.
PATO:biological sex has the subclasses PATO:genotypic sex and PATO:phenotypic sex
depending on whether the biological sex quality inheres in the bearer’s composition of
sex chromosomes or the physical expression of sexual characteristics.
Sex and gender classes in these ontologies are quality or social roles. However,
related information content entities are necessary to support our LHS as explained above.
2.2.2. Informational Entities about Sex and Gender
Informational entities about sex are represented in the Vaccine Ontology (VO) [15].
VO:biological sex datum* is categorized as a child of IAO:measurement datum and
defined as: “A measurement datum that represents the biological sex of an animal.”
Subclasses have been created for representing ICEs about multiple sex kinds. For
example: VO:female biological sex datum=def. “A biological sex datum that represents
the biological sex of an animal (including human) as being female.”
Terms referring to informational entities related to gender can be found in the
Ontology of Medically Related Social Entities (OMRSE) [16] as subclasses of
OMRSE:social identity information content entity. OMRSE:gender identity information
content entity* is defined as: “A social identity information content entity that is about
whether some person identifies as some gender.” This class has also several subclasses
representing ICEs about multiple gender kinds. For example: OMRSE:female gender
identity information content entity=def. “A gender identity information content entity that
is about some person’s identifying as female in gender.”
3. Sex and Gender in DEMO Ontology
As part of our ontology suite to represent clinical informational entities, we are
developing an ontology, named “DEMO” (stands for DEMographics Ontology) that
focuses on demographics data, including sex and gender. Our goal concerning those
latter data is threefold:
• To formalize different informational entities about sex and gender;
• To not be limited to a male/female or man/woman binary representation, but be able
to represent the other possibilities within sex or gender;
• To be able to annotate electronic health records containing representations that are
ambiguous with respect to whether they refer to sex or gender (such as a field with
two values ‘M’ and ‘F’, without more specifications about what those refer to).
In accordance to the OBO Foundry methodology, we imported the pre-existing
classes mentioned above (*) in our ontology. However, those classes were limited to
binary alternatives (man/woman for gender, male/female for sex) and we proposed to
VO and OMRSE representatives to add classes about biological intersex and non-binary
gender. As a result, the following classes were created and imported by DEMO:
• VO:intersex biological sex datum=def. “A biological sex datum that represents the
biological sex of an animal (including human) as being intersex.”
• OMRSE:non-binary identity ICE=def. “A gender identity ICE that is about some
person’s identifying as non-binary in gender.”
To take into account ambiguous representations of sex and gender, we created the
class DEMO:biological sex or gender identity ICE=def. “An ICE that is intended to denote
a biological sex or a gender identity.” This class is defined as a class equivalent to:
VO:biological sex datum OR OMRSE:gender identity ICE. Three subclasses were added
to account for multiple possibilities: DEMO:female biological sex or woman gender
entity ICE, DEMO:male biological sex or man gender entity ICE, and DEMO:intersex
biological sex or non-binary identity ICE. These subclasses are defined as “A biological
sex or gender identity ICE that is intended to denote a female [resp. male, intersex]
biological sex or a woman [resp. man, non-binary] gender identity”.
The addition of informational entities about intersex and non-binary gender enables
us to capture more diverse information about patients’ sex and gender. Alongside “male”,
“female”, “man” and “woman”, they allow us to represent field values representing
intersex, hermaphrodite or gender-fluid identity for example.
Moreover, when we are confronted with records where these fields are insufficiently
defined, the biological sex or gender identity ICE class works as a catch-all solution. For
example, if we have a data source with an undefined field labelled “Sex/Gender” and its
possible value are “M” or “F”, we can state that this field is about a biological sex or a
gender. If this field has the value “F” for a given patient, we can classify it as an
information content entity that denotes a female biological sex or a woman gender. While
such data constructs are undesirable because of their ambiguity, they are certainly in
existence; therefore, not allowing data access to such fields would not be acceptable to
users.
4. Discussion and Conclusion
To represent ambiguous information, we could have instead introduced a class
ambiguous biological sex or gender identity information content entity, whose instances
are ICEs that denote ambiguously to either a sex or a gender. However, the ambiguity
depends on the context and the receiver of the information: a piece of information can be
ambiguous for a user (say, someone who is retrieving the information from an institution
to which she does not belong), and non-ambiguous for another user (say, someone who
knows how the database has been built). Therefore, the representation of an epistemically
charged notion like an “ambiguous” representation is much more complex, and it is not
clear that such complexity would bring commensurable gains. This is why we introduced
the catch-all class biological sex or gender identity ICE.
Axioms using the IAO:is about relation should also be added in the future [17]. Note
also that although they were created initially for medical records, these classes should be
useful whenever sex and gender information are used. Finally, it is important to keep in
mind that sex and gender data are highly sensitive data from an ethical point of view. For
example, while analyzing gender information may allow us to discover specific health
problems for more fragile populations, it entails a risk of categorizing people, possibly
against their will or knowledge, in a way that might be detrimental to them.
References
[1] WHO | Gender and Genetics [Internet]. WHO. World Health Organization; [cited 2020 Aug 25].
Available from: https://www.who.int/genomics/gender/en/index1.html.
[2] Burgess C, Kauth MR, Klemt C, et al. Evolving Sex and Gender in Electronic Health Records. Fed Pract.
2019;36:271–277.
[3] Kaggal VC, Elayavilli RK, Mehrabi S, et al. Toward a Learning Health-care System – Knowledge
Delivery at the Point of Care Empowered by Big Data and NLP. Biomed Inform Insights. 2016;8:13–22.
[4] Ethier J-F, Barton A, Taseen R. An ontological analysis of drug prescriptions. Appl Ontol. 2018;13:273–
294.
[5] Barton A, Fabry P, Ethier J-F. A classification of instructions in drug prescriptions and pharmacist
documents. Proceedings of the 10th International Conference on Biomedical Ontology (ICBO 2019).
Buffalo, New York, USA; p. 1–7.
[6] Ethier J-F, Curcin V, Barton A, et al. Clinical data integration model. Core interoperability ontology for
research using primary care data. Methods Inf Med. 2015;54:16–23.
[7] IHTSDO. SNOMED CT [Internet]. Leading healthcare terminology, worldwide. 2015 [cited 2015 Jan
12]. Available from: http://www.ihtsdo.org/snomed-ct/.
[8] Regenstrief. LOINC [Internet]. Logical Observation Identifiers Names and Codes (LOINC®). 2015
[cited 2015 Jan 12]. Available from: http://loinc.org/.
[9] Gibaud B. The DICOM Standard: A Brief Overview. In: Lemoigne Y, Caner A, editors. Molecular
Imaging: Computer Reconstruction and Practice. Dordrecht: Springer Netherlands; 2008. p. 229–238.
[10] Sex and Gender [Internet]. [cited 2020 Jul 2]. Available from: https://infocentral.infoway-
inforoute.ca/en/collaboration/wg/sex-gender.
[11] Bodenreider O, Smith B, Burgun A. The Ontology-Epistemology Divide: A Case Study in Medical
Terminology. Form Ontol Inf Syst. 2004;2004:185–195.
[12] Kronk CA, Dexheimer JW. Development of the Gender, Sex, and Sexual Orientation ontology:
Evaluation and workflow. J Am Med Inform Assoc. 2020;27:1110–1115.
[13] Smith B, Ashburner M, Rosse C, et al. The OBO Foundry: coordinated evolution of ontologies to support
biomedical data integration. Nature Biotechnology. 2007;25:1251–1255.
[14] The Phenotype And Trait Ontology [Internet]. 2020 [cited 2020 Jul 2]. Available from:
https://github.com/pato-ontology/pato.
[15] He Y, Cowell L, Diehl A, et al. VO: Vaccine Ontology. Nat Prec. 2009;1–1.
[16] Hicks A, Hanna J, Welch D, et al. The ontology of medically related social entities: recent developments.
J Biomed Semantics. 2016;7:1–4.
[17] Schulz, Stefan, Martínez-Costa, Catalina, Karlsson, Daniel, et al. An Ontological Analysis of Reference
in Health Record Statements. Frontiers in Artificial Intelligence and Applications [Internet]. IOS Press;
2014 [cited 2020 Sep 3].