(210) Gender Differences in Otolaryngology Residency Recommendation Letters and Personal Statements

Introduction: The personal statement (PS) and letters of recommendation (LORs) are essential components of the residency application process. However, their subjective nature may put them at risk for unconscious bias during their interpretation in the application process. This study evaluated general and linguistic differences in PSs and LORs between the applicant genders in a single otolaryngology residency program using deep learning techniques.

Methods: PSs and LORs for a subset of 2021-2022 and 2022-2023 applicants to a single otolaryngology residency program were reviewed. A natural language processing (NLP) package generated sentiment scores for each PS and LOR. A peer-reviewed tool, Empath, categorized words and phrases through a deep learning method of neural embedding over 1.8 billion words of modern fiction. T-tests compared between all groups. This data represents preliminary analyses of an ongoing study.

Results: Among 617 LORs, 317 were for male applicants, and 300 were for female applicants. No significant difference in word count was identified between LORs for male and female applicants (p=0.21). On deep learning analysis, LORs for female applicants had more feminine words (p=0.01) and words of trust (p=0.01). By contrast, male applicant LORs contained more words of optimism (p=0.01) and cheerfulness (p=0.02). For PSs, female applicants wrote statements with more positive emotions (p=0.02), while male applicants wrote more content about leadership (p=0.03).

Conclusions: Though otolaryngology has historically been a more predominantly male specialty, this study had a similar number of female applicants to male applicants, reflecting a recent trend toward gender equality. In LORs, females were described with more words of trust, while males were described more optimistically and enthusiastically. In PSs, male applicant discussed their leadership skills and experiences more often than female applicants. NLP analyses of LORs and PSs can be implemented into a machine learning prediction model for receiving interview invites, which may identify patterns but also potentially uncover implicit biases.