Table 2.

Comparison of diagnostic and triage accuracy of GPT-4 with racial and ethnic conditions. All the CIs were calculated using the Clopper-Pearson method and are reported in percentages.

Accuracy		Correct answers without race and ethnic conditions, n (%; 95% CI)	Correct answers with racial and ethnic conditions, n (%; 95% CI)
			Black	White	Asian	Hispanic
Diagnosis
	Overall (n=45)	44 (97.8; 88.2-99.9)	45 (100; 92.1-100)a	45 (100; 92.1-100)a	45 (100; 92.1-100)a	45 (100; 92.1-100)a
	Emergent care (n=15)	15 (100; 78.2-100)	15 (100; 78.2-100)a	15 (100; 78.2-100)a	15 (100; 78.2-100)a	15 (100; 78.2-100)a
	Nonemergent care (n=15)	15 (100; 78.2-100)	15 (100; 78.2-100)a	15 (100; 78.2-100)a	15 (100; 78.2-100)a	15 (100; 78.2-100)a
	Self-care (n=15)	14 (93.3; 68.1-99.8)	15 (100; 78.2-100)a	15 (100; 78.2-100)a	15 (100; 78.2-100)a	15 (100; 78.2-100)a
Triage
	Overall (n=45)	30 (66.7: 51.0-80.0)	28 (62.2; 46.5-76.2)b	30 (66.7; 51.0-80.0)a	30 (66.7; 51.0-80.0)a	28 (62.2; 46.5-76.2)c
	Emergent care (n=15)	13 (86.7; 59.5-98.3)	11 (73.3; 44.9-92.2)b	12 (80.0; 51.9-95.7)a	15 (100; 78.2-100)b	15 (100; 78.2-100)b
	Nonemergent care (n=15)	15 (100; 78.2-100)	15 (100; 78.2-100)a	14 (93.3; 68.1-99.8)a	14 (93.3; 68.1-99.8)a	12 (80.0; 51.9-95.7)d
	Self-care (n=15)	2 (13.3; 1.7-40.5)	2 (13.3; 1.7-40.5)a	4 (26.7; 7.8-55.1)b	1 (6.7; 0.2-31.9)a	1 (6.7; 0.2-31.9)a

^aP value=.99.

^bP value=.5.

^cP value=.69.

^dP value=.25.