P0623 - Creation and Evaluation of the Gender, Sex, and Sexual Orientation (GSSO) Ontology

Nearly 17% of LGBTQIA+ (lesbian, gay, bisexual, transgender, queer/questioning, intersex, agender/asexual, and other umbrella gender/sexual identity minorities) persons report avoiding medical care due to discrimination, with 33% of transgender persons experiencing significant problems with medical providers. These problems are often connected to inappropriate or offensive language resulting in LGBTQIA+ persons feeling uneasy or unable to trust providers. Providers are often not trained in appropriate language use for LGBTQIA+ persons. In medicine, language is modeled using controlled vocabularies or ontologies. Ontologies are common, shared networks which explain information in a domain. Our objective was to create and evaluate an ontology system and resource guide which leveraged LGBTQIA+ terminology in use by medical professionals and the LGBTQIA+ community.

The Gender, Sex, and Sexual Orientation (GSSO) ontology includes over 10,000 terms, 14,000 external database mappings, 200 slang terms, 190 pronouns, and 200 nonbinary and culturally specific gender identities. It was built by iteratively scoping literature using seed terms across multiple databases and vocabulary sets and identifying terms common to those resources as candidates for inclusion.

We evaluated the computational efficiency, coverage, and usefulness of the GSSO on randomly selected journal articles, query-curated journal articles, and assembled gold-standard literature using a curated transgender bibliography. We designed a web interface (Fig. 1) for the GSSO with focus on its usage by clinical professionals and by LGBTQIA+ persons.

The GSSO is available via the NCBO BioPortal (https://bioportal.bioontology.org/ontologies/GSSO). All classes include human-readable definitions and were placed in a computationally accessible hierarchy.
We performed plaintext tagging using the GSSO on 14,019 MEDLINE titles, abstracts, and journal titles. The GSSO tagged 99.85% with terms. We also created a gold-standard transgender bibliography dataset for testing. In this dataset, the GSSO was able to tag all PubMed references and added more tags on average than MeSH. Results are summarized in Table 1.

The GSSO was more accurate at finding and identifying terms than MeSH tagging in all three of the literature elements searched. By identifying these terms in an automated fashion, we can identify and collate sparse literature in a way that can be used for research and medical education. The GSSO can assist providers in learning correct pronoun usages and understanding LGBTQIA+ medical terminology and slang. These processes can be automated to help in continuing education and assisting providers in knowing the best way to ensure their patients are comfortable. The GSSO can also be used in assisting in patient identification in electronic health records to find patients without additional interventions. It can further be used to help curate academic literature on LGBTQIA+ subjects for best practices and in creating culturally sensitive non-invasive training modules. The validated GSSO can be used to improve research, medical education, and patient care.