Identification of transgender people with cancer in electronic health records (EHR): Recommendations based on CancerLinQ observations.

Abstract

e19046 Background: Data regarding people who are gender minorities are not well-captured in oncology practices or large cohorts. Given this, collection of cancer prevalence and outcomes data, which are necessary to understand disparities in this population, are significantly hampered. Real-world data may be the most readily available source to explore outcomes in transgender populations. A database of EHR data on people with cancer, CancerLinQ is housed by the American Society of Clinical Oncology and collected from nation-wide practices and multiple EHRs. Methods: In order to identify people with cancer who may be transgender within CancerLinQ, we used three criteria: 1. ICD 9/10 diagnosis (Dx) code indicating likely transgender or non-binary gender conforming status 2. Male gender and Dx of cervical, endometrial, ovarian, or Fallopian tube cancers 3. Female gender and Dx of prostate, testicular, or penile cancers . We reviewed medical records for gender, cancer diagnosis, Dx indicating transgender identity and surgical history of transition-related surgeries, in other words surgeries that align bodies with identities and constructed a de-identified dataset of people who met one of the three criteria listed above. People without evidence of transgender identity were assigned as: (1) likely Dx error, because the person had either (a) no cancer, (b) an alternate cancer Dx consistent with sex assigned at birth, or (c) a cancer not listed in the above criteria (e.g. lung); or (2) likely error in gender data, if the cancer Dx was confirmed, but gender data was not; or (3) unknown. Results: Of ~1.3 million records in CancerLinQ at time of case selection, 557 matched inclusion criteria and 242 were abstracted. 76% of patients with ICD9/10 gender related Dx codes had evidence confirming transgender identity. By contrast, only 2% and 3% of the people identified by criteria 2 and 3 had evidence of transgender identity, respectively. Conclusions: Given the need for data regarding transgender people with cancer and the deficiencies of current data resources, a national concerted effort is needed to broaden terminology in EHRs to include whether people are transgender or not as routine and required data elements, provided by patients at their discretion. [Table: see text]