Table II.

Efficiency of AI in Dermatology Board Exam question generation

Subject	NOQ	R1 time (min)	Expected/question (min)	R2 time (min)	Expected/question (min)	Saved time (min) ± SD
Biopsy techniques	13	2.7	16.3	2.7	7.05	9.0 ± 4.6
CBCL	19	1.4	57.1	2.3	9.47	31.5 ± 24.3
CTCL	16	3	60	3.75	11.25	32.3 ± 24.8
HPV	13	3.3	10	2	12	8.4 ± 1.7
Alopecia	16	5.14	17.1	2.4	13.6	11.6 ± 0.4
Systemic disease	10	3.75	30	2	20	22.1 ± 4.1
Mycobacteria	8	2.4	12	7.5	22.5	12.3 ± 2.7
Darier disease	5	45	30	1.25	3	–6.6 ± 8.4
Ichthyoses	6	7.5	12	45	3	–18.8 ± 23.3
Acne	9	2.8	8.6	1	48	26.4 ± 20.6
Rosacea	3	2.2	20	100	20	–31.1 ± 48.9
Vasculitis	7	6	48	128	18	–34.0 ± 76.0

NOQ: Number of suitable questions: total suitable questions identified for each subject. R1 time (min): average time, in minutes, Reviewer 1 spent evaluating each AI-generated question. Expected time/Question (min): estimated average time, in minutes, that a reviewer would typically take to manually create a suitable question for a specific subject. R2 time (min): average time, in minutes, Reviewer 2 spent evaluating each AI-generated question. Saved time (min) ± SD: average time saved per question, in minutes, by using AI-generated questions compared with manually creating them, with standard deviation indicating the variability