Development of a Keyword-Based Search Strategy to Identify Gender Diverse Youth in Electronic Medical Records

Abstract

A retrospective cross-sectional study of nearly 3 years of emergency department (ED) encounters was performed with the goal of developing an electronic medical record (EMR) search strategy to identify transgender and gender diverse (TGD) youth. A search of nine keywords to identify TGD youth was conducted in the EMR, and text from clinician notes was extracted and reviewed. Among 15,413 ED encounters, 1126 keywords were identified, and 91.2% were classified as TGD. The final dataset classified 605 ED encounters (428 unique youth) as TGD. The study offers a feasible technique to identify and prioritize TGD youth in health research using EMR data.

Introduction

The term transgender and gender diverse (TGD) refers broadly to the population of individuals whose gender identities differ from their sex assigned at birth. ¹ In health care settings, most research involving TGD youth (defined for the purposes of this study as youth 25 years of age or younger ² ) has been limited to specialized clinic populations, which has limited generalizability to other settings. ³ In addition, the use of gender-specific diagnostic codes to identify TGD youth in nonspecialized health care settings likely underestimates prevalence, because youth may not be asked about their gender identity or it may not be documented as a diagnostic code. Accurate identification of TGD youth in the electronic medical record (EMR) is a necessary prerequisite for studying health service use and health outcomes among this population.

Recent studies have utilized keyword search strategies to improve identification of TGD individuals in EMRs. In multiple studies conducted with adults, the use of keywords to identify transgender status in EMRs increases sensitivity relative to using diagnostic codes alone, but youth may require a different algorithm than adults to identify transgender status.^4,5 To our knowledge, only one study has used keywords to identify TGD youth, but this study identified transgender status only (i.e., transmasculine or transfeminine) and did not identify other youth on the gender diversity spectrum (e.g., nonbinary, agender). ⁶

As many as 9.2% of high school youth identify as gender diverse, nearly one-third of whom identify as nonbinary, ⁷ highlighting the need to improve identification strategies for TGD youth. While this population would likely be most accurately identified if all youth were directly asked their gender identity during health care encounters, that is not standard practice in most pediatric emergency departments (EDs). The current study aimed to develop a strategy for identifying TGD youth in the EMR using both a keyword search strategy and the demographics section to identify TGD youth presenting to a pediatric ED.

Methods

Study design

Data were collected as part of a larger retrospective, cross-sectional study examining suicide screening rates in the ED of an urban tertiary children’s hospital from September 2019 to August 2022. The children’s hospital has approximately 55,000 ED visits annually. Encounters were included if youth received Ask Suicide-Screening Questions (ASQ), which all youth ages 10 years or older in the ED are eligible to receive under universal suicide screening protocols. Approximately 60% of eligible youth received the ASQ during the study period. This study was deemed exempt by the hospital’s institutional review board and consent and assent were waived.

Procedures

Encounters by TGD youth were identified based on documentation of TGD status in the demographics section of the EMR or via a keyword-based search of ED clinician notes. Nine keywords were selected based on previous literature^4–6,8 and consensus from study team members, as follows: (1) transgender; (2) they/them; (3) preferred name; (4) male-to-female; (5) female-to-male; (6) nonbinary; (7) gender dysphoria; (8) pronouns; (9) agender. Many keywords were considered but these were selected in combination with the literature and team consensus, as they were suspected to be highest yield in identifying youth based on knowledge of typical documentation practices at the study site. If a keyword was present in the free text of any ED clinician note during the encounter, the surrounding text was extracted. Extracted text was independently reviewed by two members of the research team and categorized as: (1) keyword indicates TGD status, (2) keyword does not indicate TGD status, or (3) unclear based on the text provided. Coders ensured that the text clearly indicated the keyword was referring to the TGD status of the patient.

Interrater reliability was measured using Cohen’s Kappa (κ). Disagreements among coders were resolved through discussion among the research team until consensus was achieved. If TGD status was unclear based on the text snippet, a manual review of the full ED note text was conducted to determine TGD status. If TGD status was identified based on a keyword in one ED encounter, all other ED encounters by the same youth were classified as having TGD status. If an encounter did not have any keywords coded to identify TGD status, the gender in the EMR was used for that youth.

Data analysis

Among ED encounters by TGD youth, we described the number of distinct TGD-related keywords within each encounter. We calculated means and percentages of characteristics of ED encounters by age, race and ethnicity, insurance payor, and metro-normed Child Opportunity Index 2.0 (COI). The COI 2.0 is a composite metric of components of a child’s neighborhood that promote healthy development including education, social, and economic factors. ⁹ Analyses were conducted in the open-source R software environment. ¹⁰

Results

There were 15,413 ED encounters with ASQ screening performed during the study period (encounters were by youth ages 8–25, median = 14). Of these, 1126 keywords were identified and surrounding text snippets were manually reviewed. Of those, 91.2% (n = 1027) keywords were classified as indicating TGD status (Table 1). Agreement among coders ranged from 94.7 − 100% across keywords with an average agreement of 98.5% (κ = 0.875). Text did not clearly indicate TGD status for 3.0% of encounters (n = 31), prompting review of the complete note which resolved the status in all cases.

Table 1.

Text Snippets Identified by Keyword Search

Keyword	Total text snippets identified with keyword	Text snippets classified as gender diverse, n (%)
Pronouns	306	262 (85.6)
Agender	14	14 (100)
Transgender	207	172 (83.1)
Male-to-female or female-to-male	71	69 (97.2)
They/them	197	195 (99.0)
Nonbinary	126	121 (96.0)
Gender dysphoria	147	142 (96.6)
Preferred name	58	52 (89.7)
Total	1126	1027 (91.2)

The 1027 keywords determined to reflect TGD status were present in 462 encounters by 346 unique youth. Thirty-one percent (n = 143) of encounters were identified using just one keyword, 32% (n = 149) contained two keywords, 23.8% (n = 110) three keywords, and 13% (n = 60) contained four or more keywords. For encounters identified using one keyword, that keyword was most often pronouns (41%), followed by they/them (21%), gender dysphoria (14%), transgender (12.6%), and nonbinary (8.4%).

Following the keyword search, the demographics section of the EMR was used to identify an additional 143 encounters with TGD status, representing 82 unique patients. This resulted in a total of 605 encounters by 428 unique youth (3.9% of overall sample). See Table 2 for descriptive information of the sample.

Table 2.

Characteristics of Emergency Department Visits by Gender Diverse Status

ED visit characteristic	Overall, n (%)	ED visits by gender diverse youth, n (%)
Age (year)
<10	1373 (9)	8 (1)
10–12	3643 (24)	128 (21)
13–15	5337 (35)	299 (49)
16–17	3469 (23)	134 (22)
18+	1591 (10)	36 (6)
Race and ethnicity
White, non-Hispanic	4006 (26)	268 (44)
Black, non-Hispanic	2876 (19)	61 (10)
Hispanic/Latinx	7263 (47)	210 (35)
Asian, non-Hispanic	495 (3)	21 (3)
Other	773 (5)	45 (7)
Insurance
Private/commercial	5838 (38)	345 (57)
Government/Medicaid	9363 (61)	254 (42)
Self-Pay/other/unknown	212 (1)	6 (1)
Child Opportunity Index (metro-normed)
Very low	5478 (36)	165 (28)
Low	4166 (28)	135 (23)
Moderate	2238 (15)	122 (21)
High	1299 (9)	52 (9)
Very high	1853 (12)	118 (20)

ED, emergency department.

Discussion

Identification of TGD youth was feasible using the gender demographics section of the EMR and a keyword search strategy involving review of clinician note text. One-third of TGD youth were identified using only one keyword, highlighting the importance of using a comprehensive list of keywords to identify TGD youth. The population of TGD youth identified in the current study was 3.9%, which is slightly lower than recent population-based estimates. ⁷ This is likely in part due to a lack of standardized data collection of TGD status in the ED. Further, TGD status may be underreported in medical settings, given structural discrimination and inequities in health care for TGD individuals and for racially minoritized youth.^11–13 Notably, nonbinary individuals are even less likely to disclose their gender identity to health care providers than transgender individuals. ¹⁴ There may also be bias on the part of health care providers that influence the documentation of gender identity in the EMR.

Limitations and strengths

Limitations of the study include that the sample consisted of youth who had received an ASQ as it was part of a broader study on suicide risk, so the generalizability of the search strategy for all ED encounters remains unclear. Youth presenting for a mental health chief complaint are more likely to receive the ASQ even under universal screening protocols and may also be more likely to receive more detailed social history screens compared with youth presenting with a primarily medical chief complaint. In addition, only ED encounters were searched for keywords; it is possible that relevant keywords existed in other sections of the EMR and this should be a focus of future research. Keyword searches of free-text are also limited by misspellings and abbreviations that may not be captured. Finally, given the ever-expanding nomenclature among TGD youth, other keywords that were not included in the study could have identified other youth.

Conclusion

Overall, this study demonstrates that keyword searching is a promising technique to identify and prioritize TGD youth in health services research. This study also highlights the importance of using a variety of keywords to best capture youth on the gender diverse spectrum. Future research should continue to explore the generalizability of this strategy.

Authors’ Contributions

A.N.B.: Conceptualization, investigation, formal analysis, writing—original draft. J.A.H.: Conceptualization, methodology, writing—original draft. D.L.: Formal analysis, writing—review and editing. J.W.: Investigation, writing—review and editing. J.B.: Methodology, writing—review and editing. A.J.: Conceptualization, methodology, writing—review and editing.

Abbreviations

ASQ

Ask Suicide-Screening Questions

COI

Child Opportunity Index

ED

emergency department

EMR

electronic medical record

TGD

transgender and gender diverse

 

Cite this article as: Burnside AN, Hoffmann JA, Lorenz D, Walter J, Berona J, Janssen A (2026) Development of a keyword-based search strategy to identify gender diverse youth in electronic medical records, Transgender Health 11:3, 220–223, DOI: 10.1089/trgh.2023.0182.