Linguistic Analysis of Race and Gender Bias in Letters of Recommendation for Neurology Residency Applicants
Vivian Chen1, Andrea Linares Lopez2, Michelle Maher2, Nhu-y Phan2, Cassidy Allison2, Alexis Lorio2, Christopher Eckstein2, Suma Shah2
1Duke University School of Medicine, 2Neurology, Duke University Medical Center
Objective:

To evaluate the association of gender, race, and language use in letters of recommendation (LORs) for neurology residency applicants

Background:

Gender bias in language in LORs for residency programs has been documented across specialties, such as plastic surgery, ENT, OB/GYN. Differences in language used for underrepresented minorities in medicine have also been reported. To our knowledge, this topic has yet to be studied in neurology residency applications, necessitating further exploration as LORs are becoming increasingly important in residency selection.

Design/Methods:

This cross-sectional study analyzed narrative LORs submitted to a single neurology residency program during the 2023-2024 cycle. Applicants without LORs were excluded, and demographics were self-identified. The Linguistic Inquiry and Word Count (LIWC) software was used to assess linguistic categories such as Analytical thinking, Achievement, and Social Behavior. Median group differences were analyzed using Mann-Whitney U tests for gender and a Kruskal-Wallis test for race, followed by post-hoc Wilcoxon tests.

Results:

Among 683 applicants, 45.7% were male, 54.2% female, and 0.1% identified with another gender. Racial demographics included 38.7% White, 34.1% Asian, 6.3% Black or African American, 7.0% Hispanic or Latino, 0.9% Middle Eastern or North African, 6.3% multiracial, 4.1% other, and 2.6% preferred not to answer. 2490 LORs were analyzed. Letters for male applicants contained more authentic language (9.05 vs 8.27, p=0.0039), while female applicants’ letters were more positive (1.15 vs 1.10, p=0.0393) and included more social (16.5 vs 15.9, p<0.0001), and leadership language (81.5 vs 79.6, p<0.0001). Letters for Asian applicants compared to White applicant letters contained more long words (33.8 vs 33.0, p=0.0006), while other group differences were not detected.

Conclusions:

Language differences in LORs exist along gender and racial lines in neurology residency applications. There is a need to further investigate these differences and address potential implicit biases to enhance objectivity in LOR writing and promote equity in residency selection.

10.1212/WNL.0000000000208469
Disclaimer: Abstracts were not reviewed by Neurology® and do not reflect the views of Neurology® editors or staff.