Analysis of narrative assessments of internal medicine resident performance: are there differences associated with gender or race and ethnicity?

Background: Equitable assessment is critical in competency-based medical education. This study explores differences in key characteristics of qualitative assessments (i.e., narrative comments or assessment feedback) of internal medicine postgraduate resident performance associated with gender and race and ethnicity.

Methods: Analysis of narrative comments included in faculty assessments of resident performance from six internal medicine residency programs was conducted. Content analysis was used to assess two key characteristics of comments- valence (overall positive or negative orientation) and specificity (detailed nature and actionability of comment) - via a blinded, multi-analyst approach. Differences in comment valence and specificity with gender and race and ethnicity were assessed using multilevel regression, controlling for multiple covariates including quantitative competency ratings.

Results: Data included 3,383 evaluations with narrative comments by 597 faculty of 698 residents, including 45% of comments about women residents and 13.2% about residents who identified with race and ethnicities underrepresented in medicine. Most comments were moderately specific and positive. Comments about women residents were more positive (estimate 0.06, p 0.045) but less specific (estimate - 0.07, p 0.002) compared to men. Women residents were more likely to receive non-specific, weakly specific or no comments (adjusted OR 1.29, p 0.012) and less likely to receive highly specific comments (adjusted OR 0.71, p 0.003) or comments with specific examples of things done well or areas for growth (adjusted OR 0.74, p 0.003) than men. Gendered differences in comment specificity and valence were most notable early in training. Comment specificity and valence did not differ with resident race and ethnicity (specificity: estimate 0.03, p 0.32; valence: estimate - 0.05, p 0.26) or faculty gender (specificity: estimate 0.06, p 0.15; valence: estimate 0.02 p 0.54).

Conclusion: There were significant differences in the specificity and valence of qualitative assessments associated with resident gender with women receiving more praising but less specific and actionable comments. This suggests a lost opportunity for well-rounded assessment feedback to the disadvantage of women.