Race and Gender Bias in Narrative Letters of Recommendation for Plastic Surgery Residency Applicants

Victoria N Yi; J Eleanor Seo; Colleen McDowell; Brett T Phillips; Kristen Rezak; Alexander C Allori; Ash Patel

doi:10.1016/j.jss.2024.12.017

Race and Gender Bias in Narrative Letters of Recommendation for Plastic Surgery Residency Applicants

J Surg Res. 2025 Jan 7:306:152-164. doi: 10.1016/j.jss.2024.12.017. Online ahead of print.

Authors

Victoria N Yi¹, J Eleanor Seo², Colleen McDowell³, Brett T Phillips³, Kristen Rezak³, Alexander C Allori³, Ash Patel³

Affiliations

¹ Duke University School of Medicine, Durham, North Carolina. Electronic address: victoria.yi@duke.edu.
² Duke University School of Medicine, Durham, North Carolina.
³ Department of Surgery, Division of Plastic, Maxillofacial, and Oral Surgery, Duke University Medical Center, Durham, North Carolina.

PMID: 39778234
DOI: 10.1016/j.jss.2024.12.017

Abstract

Introduction: It has been previously shown that gender bias exists in standardized letters of recommendation for plastic surgery residency. However, similar analysis has not been conducted for narrative letters of recommendation (NLORs). Therefore, this study aims to determine if there exists linguistic bias in NLORs for plastic surgery residency based on applicants' and writers' gender and race.

Methods: The sample included applicants to a 6-y integrated plastic surgery residency program within application cycles from 2021 to 2023. The exclusion criteria included reapplicants and applicants without NLORs. Applicant demographics were self-identified. Faculty gender was identified through public online platforms and faculty race through Namsor, validated artificial intelligence software for name classification. The Linguistic Inquiry and Word Count program was used to calculate the amount of words in each NLOR that fell within predetermined linguistic categories, such as power and social behavior. Descriptive statistics and Wilcoxon rank-sum tests were used in the analysis were appropriate.

Results: Six hundred twenty-six unique applicants had at least one NLOR. Female writers used more descriptive evaluations of applicants compared to male writers with greater emphasis on applicants' social and emotional qualities. White writers used more words conveying negative tone to describe non-White applicants. Non-White writers used more word describing accomplishment, drive, and social qualities of non-White applicants compared to White Applicants.

Conclusions: Analysis of NLORs revealed differences related to gender and race. Female writers produced more favorable recommendations than male writers. Racial discordance between writer and applicant resulted in differences in the letter quality. Faculty education to improve the objectivity of NLORs is necessary.

Keywords: Gender bias; Letter of recommendation; Race bias; Residency evaluation; Residency selection.