Table 5.

Classification measures in each cohort using 3 different probability thresholds (0.25, 0.31, and 0.45). The output of the logistic regression model is a score between 0 and 1. To make a binary classification decision, a probability threshold is applied. If the calculated probability exceeds this threshold, it is categorized as having at least 1 clinically relevant discrepancy. The probability threshold is a crucial parameter that affects the trade-off between sensitivity and specificity.

Datasets and probability threshold			Specificity (%)	Sensitivity (%)	PPVa (%)		NPVb (%)	LR+c		LR–d		Alert rate (%)
Development dataset
	0.25	26		90	40	84		1.23	0.37		79
	0.31	54		74	46	78		1.61	0.48		56
	0.45	84		29	49	69		1.83	0.85		20
Temporal validation dataset
	0.25	26		89	42	80		1.20	0.42		80
	0.31	51		73	47	76		1.48	0.54		58
	0.45	88		25	55	66		2.04	0.85		17
Geographic validation dataset
	0.25	26		88	53	70		1.19	0.46		81
	0.31	49		79	60	71		1.56	0.42		65
	0.45	87		24	64	55		1.84	0.87		18

^aPPV: positive predictive value.

^bNPV: negative predictive value.

^cLR+: positive likelihood ratio.

^dLR–: negative likelihood ratio.