Search term selection

Selection of search terms drew upon professional subject expertise within the author team, personal lived experience of authors who are TGD themselves, and several existing resources. To create comprehensive search strategies, the terminology utilized in the search hedges intentionally transcend narrow definitions of TGD populations in medical contexts. The search hedges include culturally specific gender identity terms adapted from search strategies used in the transgender health guideline commissioned by the World Health Organization [19], the Trans Evidence Map [20], and a map of global gender-diverse identities published by Independent Lens [21]. Lastly, we consulted the initial searches used to populate Knowsy [22], a comprehensive online portal for 2SLGBTQIA+ evidence syntheses, which, while broader in scope, provided some additional terms. While no limits such as language, geography, and methodology were applied to our selection of search terms, inherent biases may be present in the terminology of these search hedges given their development by anglophones for use in English language databases. Culturally specific terms, such as “fa’afafine” or “muxe,” may be used out of context in ways that do not necessarily align with how they are used in cultural contexts, or in ways that subsume autochthonous understandings of gender under the heading of transgender or gender diverse.

While gender-affirming care is associated with TGD people, many gender-affirming interventions were developed for and are sought out by cisgender people, that is non-transgender individuals whose current gender identities correspond to the sexes they were assigned at birth. Therefore, the search hedges are intentionally selective in the inclusion of gender-affirming care terminology. The included terminology related to medical transition reflect common types of surgeries, procedures, and interventions primarily utilized for transgender and gender diverse populations — such as facial feminization, metoidioplasty, neovagina, and voice masculinization. Terminology like “hormone replacement therapy” and the names of individual hormones, medications, and surgeries that are used in both trans- and cisgender populations were excluded. Another example of medical transition terminology intentionally excluded from the search hedges are classes of drugs commonly used in gender-affirming care regimens, such as GnRH agonists. GnRH agonists can be prescribed as ‘puberty blockers’ for both transgender youth and for cisgender youth diagnosed with precocious puberty [23]. Specific medication names, such as cyproterone acetate (brand name Androcur), which is typically used for prostate cancer treatment as well as an anti-androgen in a gender-affirming hormone therapy regimen, were also excluded [24].

Data management process

To ensure transparency and reproducibility in the search hedge development process, all search terms were systematically collated and organized within a collaboratively maintained spreadsheet that was discussed between the two librarians at regular meetings. Organizing terms within the spreadsheet allowed for the tracking and testing of synonyms and word variants to maximize retrieval scope. Each librarian made notes on the spreadsheet to iteratively refine terminology to include in the hedges based on their performance in each database. Database-specific controlled vocabulary terms and free text keywords were separated within the spreadsheet to assess their relative performance, and the librarians ensured that any controlled vocabulary terms were reflected in corresponding free text terms to account for inconsistencies in database indexing. Search terms were also sorted into thematic clusters to ensure comprehensive representation of several different facets of TGD research. In addition to organizing terms, the spreadsheet facilitated conversations with the entire research team when it came to larger decisions on scope and sensitivity. When the team reached a consensus on whether to remove or retain individual terms in the search hedges, the spreadsheet assisted with documenting the decision-making process.

Search hedge structure

The search hedges were developed in the Ovid platform for two predominant academic databases, Ovid MEDLINE and Ovid APA PsycInfo. This decision reflects our determination that these are predominant databases in the health sciences and social sciences where there is a need for effective retrieval of TGD literature for research. Search hedges can be translated across different databases and between different interfaces and platforms that search the same database. Glanville et al. (2019) [25] provide a case study detailing how to translate a validated search hedge from one platform to another. For ease of translation across different databases, the controlled vocabulary terms for each search hedge have been grouped in the first line of each respective hedge. These database-specific controlled vocabulary terms are the MeSH terms for Ovid MEDLINE and the APA Thesaurus of Psychological Index terms for Ovid APA PsycInfo. The other database-specific elements that would need to be translated for other databases or platforms would be the field codes (such as “.ti,ab,kf,kw,ot,oa,cl.” or “.tw.”) and the adjacency operator (“adj5”) that we tried to minimize. The human limiter for Ovid MEDLINE (line 12) is specific to that database to mitigate indexing errors. We tried to limit the need to translate other database-specific syntax such as advanced wildcard operators to reduce potential errors in translation.

Since access to specific databases and database platforms may vary depending on institutional affiliation and subscriptions, database operators such as advanced wildcards and adjacency were intentionally avoided or minimized to facilitate translation to other database platforms. Because of the prominence and accessibility of PubMed in health sciences research, the MEDLINE hedge was developed in the Ovid MEDLINE® ALL < 1946-Present> database multi-segment, which is identical in coverage to PubMed [26]. The Ovid MEDLINE search hedge utilizes the field codes “.ti,ab,kf,kw,ot,oa,cl,” which searches the title, abstract, author keywords, other titles, other abstracts, and collection titles of a record [27]. These field codes were intentionally selected to match PubMed’s [tiab] field code. For the Ovid APA PsycInfo hedge, the “.tw” field code was selected due to its close alignment with [tiab] in PubMed [28,29]. We chose field codes that align with [tiab] in PubMed in order to capture search terms in the title, abstract, or keywords fields of search results while still reducing irrelevant results from additional metadata fields. Neither Ovid MEDLINE nor Ovid APA PsycInfo allow searching ful text and other metadata fields (e.g., journal title and cited references) typically introduce unwanted or irrelevant results. Searching title, abstract, controlled vocabulary terms, and author-provided keywords is standard practice in evidence synthesis. Despite efforts to avoid using database syntax operators that may be difficult to translate to other databases, line 10 in each search hedge was developed in February 2024 after pilot validation to further reduce noise related to the keyword phrase “gender neutral.” This line was kept separate in each hedge to ease translation to databases that do not have adjacency operators.

The search hedges were designed in structured blocks to enhance adaptability for specific research questions. This approach allows for more modularity, making it easier to refine and adapt individual components of the hedges without disrupting the entire search strategy. For example, when MeSH headings relating to gender-affirming procedures were updated in 2025, it was easy to update the validated hedge [30]. By isolating different elements of transgender and gender diverse terminology in modular blocks, users can make more intentional choices to tailor searches for different goals, research questions, contexts, or databases. This method aligns well with evidence synthesis principles, where structured reproducible searches that are well-balanced between sensitivity and specificity are essential. After iterative testing, these search hedges were developed to have nine independent blocks (Table 1). Some of the blocks have a functional component for ease of translation across multiple databases, such as the separation of controlled vocabulary terms from other free-text keywords. However, most of the blocks are organized around a semantic component that researchers may want to narrow down a search on, such as transmasculine or transfeminine identities, culturally specific gender identity terms, or keywords related to gender-affirming care and medical transition. Terms are alphabetized within each block to facilitate ease of navigation and review for errors in translation. Table 1 illustrates the block structure utilized across both hedges, using the Ovid MEDLINE hedge as validated as an example.

Download:

• PNG
  larger image
• TIFF
  original image

Table 1. Modular block structure of the search hedges.

https://doi.org/10.1371/journal.pone.0351303.t001

Initial testing of the Ovid MEDLINE search hedge uncovered problematic indexing inconsistencies in the search results regarding the medical subject heading (MeSH) for “Humans.” Filtering search results utilizing the “Humans” MeSH term was important for the functionality of the Ovid MEDLINE search hedge due to the common phenomenon of sex change present across other species. In the MeSH thesaurus, “Humans” is a more specific subject heading under the broader “Animals” MeSH term, and there are studies within MEDLINE indexed with both the “Animals” and “Humans” subject headings to indicate mixed human and non-human studies. The recommended approach from the Cochrane Handbook for Systematic Reviews of Interventions to limit MEDLINE search results to human studies is to append the search string “NOT (animals NOT humans)” to identify and remove animal-only studies and exclude them from the search results using a double negative. Compared to human filters that may be built into databases, this strategy is a more sensitive way to limit searches to human studies only [31]. However, the utilization of even this sensitive approach recommended by Cochrane unintentionally excluded a significant number of relevant studies where human-only studies were incorrectly indexed as “Animals” instead of “Humans.” Thus, modifications in line 12 of the Ovid MEDLINE search hedge were made to identify and remove results indexed under specific portions of the “Animals” MeSH tree structure to avoid erroneously excluding human studies that were incorrectly indexed with only the “Animals” MeSH term.

While the need for a more complex human limiter was discovered before pilot screening and validation of the search hedges began, some additional changes to the search hedges were made based on irrelevant articles that appeared frequently in our pilot validation process. The term “eunuch,” which has historically been used as a gender identity term [32], only brought in irrelevant results, so the decision was made to remove the term entirely. The acronym TGD (transgender and gender diverse) also pulled in irrelevant results discussing the Three Gorges Dam of the Yangtze River Estuary, so the term was adapted to exclude “Yangtze” from those results (line 5 of the Ovid MEDLINE hedge, line 6 of the Ovid APA PsycInfo hedge). The final adaptation made to reduce significant sources of noise identified in pilot validation was the introduction of a proximity operator (ADJ on the Ovid platform) to the term “gender neutral” to only include results where “gender neutral” was used within five words of keywords related to pronouns, language, or bathrooms. While the hedges were developed with the intention to avoid operators that are not fully available in other frequently used databases like PubMed, this decision was made to reduce the number of irrelevant articles that were found describing concepts such as gender-neutral HPV vaccinations.

Validation

As discussed in the study protocol [18], the validation process for the search hedges involved randomly sampling from the search results our hedges produced in Ovid MEDLINE and Ovid APA PsycInfo. The results from each database were uploaded to the web-based review management software EPPI-Reviewer [33] for screening. De-duplication between databases was not performed to preserve the distinctive sample for each database. These randomized samples from each database were screened to validate the sensitivity and precision of the search hedges.

The precision validation process utilized the two-stage screening model of title and abstract review followed by full text review. During the title and abstract screening phase, we categorized articles using the following five categories: 1) Exclude, 2) No_Abstract, 3) Mixed_Topics, 4) LGB_maybe_T, and 5) Include. Articles were excluded if they exclusively used binary gender language without clarification of the inclusion of transgender people (e.g., women and men, female and male), if they focused exclusively on intersex people and people with differences in sexual development, or if they were focused on the gender identities of cisgender people. Articles were included if they discussed TGD people or TGD-related topics. More specifically, records were included if they had disaggregated data reflecting inclusion of TGD participants in the case of empirical studies, and non-empirical records were included if they distinguished a substantive discussion of TGD people in specific ways rather than strictly as part of a larger umbrella category (such as LGBT). Articles that were categorized as exclude or include at title and abstract screening did not undergo full text screening because they had clearly demonstrated their relevance or lack of relevance to TGD people and topics. There were three additional categories used in title and abstract screening that determined which articles would be considered for the full text screening phase. The No_Abstract category was used for articles without abstracts that needed to be reviewed in further depth to address how substantive their engagement with TGD populations was in the full text. The category Mixed_Topics was applied to articles that were unclear whether or not TGD people would be mentioned. The LGB_maybe_T category was used for articles that used some variation of an umbrella 2SLGBTQIA+ acronym for their study population, yet it was unclear if there was a significant indication of transgender study subjects or discussions applicable to TGD individuals.

During the full text screening phase, the only two categories available for screeners to choose from were Include or Exclude. Articles were excluded if they did not substantively engage with TGD individuals or related topics. This means that if the word “transgender” was in the abstract of an article through authors spelling out an acronym like 2SLGBTQIA+ without significant discussion of transgender or gender diverse individuals specifically, those articles were excluded. On the other hand, articles were included if they substantially engaged with TGD populations, were directly relevant to TGD topics, or if participant data was reported separately for TGD study subjects.

While the validation process has been described in detail in the published protocol [18], it is worth noting that the screening process involved a core team of four reviewers who independently reviewed randomized sets from within the random samples for each database. For both title and abstract and full text screening phases, articles were independently screened by at least two trained reviewers following the established inclusion and exclusion criteria for each category. Those four reviewers were overseen by two other team members responsible for the study design. In the case of disagreements across reviewers on individual records, consensus was reached among the study team to reconcile disagreements and determine which of the five categories a title and abstract record belonged to, or, in the final phase, whether to include or exclude a record based on a full text review. We discuss the consensus process more fully in the published protocol [18], and in the present publication we describe the results of that process, including agreement rates across the review team.

Search term insights

The development and validation process uncovered search terms with unclear relevance to TGD topics, prompting a more detailed examination of the context in which these terms appeared and whether these unclear terms appeared alongside other TGD-related terms. These terms with unclear relevance included: “gender inclusive” or “gender inclusivity,” “gender diverse” or “gender diversity,” “gender-nonconformity” or “gender-nonconforming,” and “transperson.” The precision of each of these terms was calculated (Table 3). Records that included these search terms with a lower precision rate as compared to their respective databases were manually searched for other TGD-related terms that were also included in the search hedges in their titles, abstracts, and MeSH terms or keywords (Table 4). The aim was to see how much these individual search terms and phrases were contributing to less precise results, and how many of these irrelevant results would have been captured by each hedge if these terms were removed. The articles were further divided based on their include or exclude status at each screening phase.

Download:

• PNG
  larger image
• TIFF
  original image

Table 3. Breakdown of polysemic term prevalence in included and excluded records from MEDLINE and APA PsycInfo.

https://doi.org/10.1371/journal.pone.0351303.t003

Download:

• PNG
  larger image
• TIFF
  original image

Table 4. Co-occurrence of polysemic terms with other search terms in included records across MEDLINE and APA PsycInfo.

https://doi.org/10.1371/journal.pone.0351303.t004

Mixed_Topics and LGB_maybe_T categories

The majority of articles from Ovid MEDLINE classified as Mixed_Topics mentioned gender identity as part of social determinants of health, an equity-deserving group, or a possible focused topic for cultural competence training in education, health care, or the workplace. The term “gender non-conformity” was used in this category to describe externally observable behavior in children or adolescence without reporting on the participants’ gender self-identification. Similarly, another five articles discussed sex assignment surgery in Intersex people without consideration for the self-identification of the patients. Articles from APA PsycInfo in this category focused on theoretical discussion of gender identity, as well as the gender development of youth, intersex patients, and gay men. Less than half of the articles in this category were coded as TGD-relevant during full-text review (Table 2).

Download:

• PNG
  larger image
• TIFF
  original image

Table 2. Records included and excluded after full text review in each database stratified by ‘maybe’ classification type at title and abstract stage.

https://doi.org/10.1371/journal.pone.0351303.t002

Overall, the majority of articles classified as LGB_maybe_T substantially discussed TGD people and/or topics. Articles from Ovid MEDLINE in this category discussed gender minority people, along with sexual minority people, in the context of equity consideration in education, healthcare, and training for service providers. While we have defaulted to TGD as an overarching category, gender minority and related phrases are part of the search hedges and are used widely in the extant literature. In addition to these topics, articles from APA PsycInfo discussed broader topics, including substance use and abuse, marriage equality, queer historical perspectives, and religion and spirituality. For both databases, the majority of the articles in this category were coded as TGD-relevant upon examination of the full-text (Table 2).

Polysemic search terms

Specific search terms that had lower precision rates than their respective database precision included: “gender inclusiv*” in MEDLINE and APA PsycInfo (48% and 46%, respectively), “gender diversity” in MEDLINE and APA PsycInfo (37% and 44%, respectively), “gender diverse” in APA PsycInfo (30%), and “transperson*” in MEDLINE and APA PsycInfo (6% and 0%, respectively) (Table 3).

Most included articles also contained other search terms that allowed these articles to be picked up by the search hedge without the polysemic terms. “Gender inclusive,” “gender inclusivity,” and “gender diversity” were used in articles in MEDLINE pertaining to advocating for inclusion of TGD patients in medical and health settings. Similarly, in APA PsycInfo, these terms were used to discuss the experience of TGD people with discrimination in educational and professional settings. The only instance of “transperson*” that met inclusion criteria was used in the context of male-to-female transpersons. However, this article also contained the search term “transgender” in its title and abstract, as well as the MeSH heading term “transgender person” (Table 4).

The majority of excluded articles did not contain other search terms in the search hedges (Table 5). Excluded articles in both databases, use “gender inclusive” or “gender inclusivity” in a gender binary context, with frequent co-occurrence of the terms “male,” “female,” “women,” and “men.” These articles often focused on promoting greater presence of women in traditionally male-dominated fields (e.g., surgery, business) or advocating for better consideration of the experience of men in women-focused topics, such as domestic abuse and unintended pregnancy. For excluded articles in APA PsycInfo, “gender diversity*” and “gender diverse” have high co-occurrence rates. Similarly to their MEDLINE counterpart, articles with these terms often focused on promoting women and girls in sports, business, and education. In the sample used for validation, the term “transperson*” occurred exclusively in the context of transpersonal psychology in APA PsycInfo with no relevance to TGD topics, despite there being some relevant uses of the term “transperson*” in the full APA PsycInfo database. The hedge in the supplement has been modified to narrow matching terms to transperson or transpersons.

Download:

• PNG
  larger image
• TIFF
  original image

Table 5. Co-occurrence of polysemic terms and other search terms in excluded records across both databases.

https://doi.org/10.1371/journal.pone.0351303.t005

Finally, the review process, which was conducted concurrently across both Ovid MEDLINE and Ovid APA PsycInfo, resulted in 642 disagreements in the title and abstract screening out of 4,623 records screened, meaning the rate of agreement across reviewers in aggregate was 13.9%. In the full text screening, 128 disagreements arose from the review of 1,013 full texts, resulting in a 12.6% rate of agreement. This indicates an improvement in agreement across reviewers in the later phase, despite the significant increase in material necessary to review in a full text when compared to a title and abstract record.

Limitations

Several important considerations should be noted when applying this hedge. First, as a product of its development by English language speakers and within North American academic contexts, the hedges reflect both linguistic and geographic bias. While it includes terms such as “Two-Spirit,” which hold cultural specificity, these terms do not necessarily align with Western constructions of transgender identity. They should not be assumed to do so. This underscores the importance of critical engagement with terminology, particularly when applying the hedge to research on culturally distinct gender identities. A strength of this study is that it is grounded in the community and that the majority of co-authors identify as TGD.

Second, these hedges were designed specifically for use in Ovid MEDLINE and Ovid APA PsycInfo. While they may be adapted for use in other databases, such as CINAHL, or other platforms, such as PubMed or EBSCOhost, care should be taken when translating syntax, field codes, and indexing conventions. It is also important to note that the syntax and field codes databases use are constantly evolving. For example, during the validation of these search hedges, the National Library of Medicine updated its thesaurus of medical subject headings (MeSH), updating the terminology used for two MeSH terms in our Ovid MEDLINE hedge. As of 2025, “Sex Reassignment Procedures” changed to “Gender-Affirming Procedures” and “Sex Reassignment Surgery” changed to “Gender-Affirming Surgery” in the National Library of Medicine thesaurus [30]. At the same time, new features may become available, such as APA PsycInfo adding a new population check tag during this study — “nonbinary.po” as a method of filtering search results to only include articles indexed as including non-binary participants [29]. While updated subject headings and new features to identify transgender or non-binary populations within databases are helpful, it is important to be aware of the limitations of relying on these strategies without a comprehensive search that is able to catch articles that are not indexed or indexed incorrectly.

Third, we validate these hedges against a gold standard data set which a co-author [REDACTED] created. As we have acknowledged elsewhere [18], this is a potential conflict of interest. In our view, it is not one because the librarian co-authors [REDACTED] and [REDACTED] discovered Knowsy independently when beginning to develop an early version of the hedges featured in this article. Collaboration with [REDACTED] was warranted given their expertise in the field and topic. That data set contained articles up to 2019; however, our searches were finalized in 2024. This could be interpreted as introducing a temporal bias because we cannot know how Knowsy would perform on the literature published between its end date in 2019 and our study’s end date in 2024. However, this is true of any gold standard data set.

Finally, users should note that terminology related to TGD identities continue to evolve. Language shifts — such as the introduction of “non-binary” or the declining use of terms like “transsexual” — require that the hedge remain dynamic. Accordingly, this is intended to be a living hedge. Librarian co-authors Gagen and Marsalis will maintain and periodically update the hedge to reflect language, indexing practices, and community usage developments as part of the dissemination plan via platforms frequently consulted by expert researchers in the health and social sciences.