Evaluation of bias and gender/racial concordance based on sentiment analysis of narrative evaluations of clinical clerkships using natural language processing

Table 2 Sentiment scores based on evaluator and student demographics

Evaluator	Student	Mean (SD) sentiment score	p-value
Gender^a
Female	Female	0.29(0.22)	0.88
Female	Male	0.29(0.21)
Male	Female	0.32(0.25)
Male	Male	0.33(0.24)
Race^b
Asian	Asian	0.31(0.22)	0.64
	African American	0.33(0.23)
	White	0.33(0.24)
	Other	0.30(0.19)
African American	Asian	0.35(0.10)
	African American	0.19(0.09)
	White	0.27(0.15)
	Other	0.25(0.01)
White	Asian	0.31(0.24)
	African American	0.24(0.27)
	White	0.29(0.23)
	Other	0.33(0.23)
Other	Asian	0.29(0.22)
	African American	0.19(0.22)
	White	0.26(0.20)
	Other	0.24(0.16)
Ethnicity^c
Hispanic or Latino	Hispanic or Latino	0.28(0.20)	0.06
Hispanic or Latino	Not Hispanic or Latino	0.31(0.23)
Not Hispanic or Latino	Hispanic or Latino	0.30(0.21)
Not Hispanic or Latino	Not Hispanic or Latino	0.30(0.23)

^a Gender_Concordance (i.e., mean sentiment score among faculty/student pairs with concordant gender vs. mean sentiment score among faculty/student pairs without concordant gender)
^b Race_Concordance (i.e., mean sentiment score among faculty/student pairs with concordant race vs. mean sentiment score among faculty/student pairs without concordant race)
^c Ethnicity_Concordance (i.e., mean sentiment score among faculty/student pairs with concordant ethnicity vs. mean sentiment score among faculty/student pairs without concordant ethnicity)

ISSN: 1472-6920