Building a Cohort of Transgender and Nonbinary Patients from the Electronic Medical Record.

Abstract

Purpose: Sexual orientation, gender identity, and sex recorded at birth (SOGI) have been routinely excluded from demographic data collection tools, including in electronic medical record (EMR) systems. We assessed the ability of adding structured SOGI data capture to improve identification of transgender and nonbinary (TGNB) patients compared to using only International Classification of Diseases (ICD) codes and text mining and comment on the ethics of these cohort formation methods. Methods: We conducted a retrospective chart review to classify patient gender at a single institution using ICD-10 codes, structured SOGI data, and text mining for patients presenting for care between March 2019 and February 2021. We report each method's overall and segmental positive predictive value (PPV). Results: We queried 1,530,154 EMRs from our institution. Overall, 154,712 contained relevant ICD-10 diagnosis codes, SOGI data fields, or text mining terms; 2964 were manually reviewed. This multipronged approach identified a final 1685 TGNB patient cohort. The initial PPV was 56.8%, with ICD-10 codes, SOGI data, and text mining having PPV of 99.2%, 47.9%, and 62.2%, respectively. Conclusion: This is one of the first studies to use a combination of structured data capture with keyword terms and ICD codes to identify TGNB patients. Our approach revealed that although structured SOGI documentation was <10% in our health system, 1343/1685 (79.7%) of TGNB patients were identified using this method. We recommend that health systems promote patient EMR documentation of SOGI to improve health and wellness among TGNB populations, while centering patient privacy.