Optimal risk groups

	Group 1 (Error Rate)^g	Group 2 (Error Rate)	Group 3 (Error Rate)	Group 4 (Error Rate)	Odds Ratio 1 (P Value)	Odds Ratio 2 (P Value)	Odds Ratio 3 (P Value)	LRT P Value^h
1 cut point^a	26 and below (1.59)	27+ (2.58)	–	–	1.63 (<.001)^d	–	–	<.001
2 cut points^b	19 and below (1.34)	20–28 (1.88)	29+ (2.6)	–	1.40 (.10)^e	1.94 (<.001)^e	–	<.001
3 cut points^c	24 and below (1.59)	25–66 (2.44)	67–90 (3.03)	91+ (2.07)	1.54 (<.001)^f	1.91 (<.001)^f	1.30 (.25)^f,ⁱ	<.001

Note:—LRT indicates likelihood ratio test.
↵a Search done for each shift volume level.
↵b Search done by fours.
↵c Search done by tens.
↵d Interpretation: The odds of containing an error are 63% higher for studies in group 2, compared with studies in group 1.
↵e Interpretation: The odds of containing an error are 40% higher for studies in group 2, compared with studies in group 1; the odds of containing an error are 94% higher for studies in group 3, compared with studies in group 1.
↵f Interpretation: The odds of containing an error are 54% higher for studies in group 2, compared studies in group 1; the odds of containing an error are 91% higher for studies in group 3, compared with studies in group 1; the odds of containing an error are 30% higher for studies in group 4, compared with studies in group 1.
↵g Error rate per 1000 exams.
↵h Optimal cutoff points are defined as the ones with the most significant (likelihood ratio test) split.
↵i Odds ratio 3 decreases, seemingly suggesting that group 4 is safer (than 2); however, this is due to sample size limitations in group 4.