Table 2

Prognostic performance of the 5 min Apgar score-based model for identifying children with developmental vulnerability on each or any of the five Early Developmental Instrument (EDI) domains

	Calibration ability*		Stratification capacity*			Classification accuracy*
EDI domains	Probability of vulnerability if in high-risk group, %	Probability of vulnerability if in low-risk group, %	Low-risk category†, %	Mid-risk category†, %	High-risk category†, %	Detection rate/sensitivity, % (95% CI)	False-positive rate, % (95% CI)
Physical	25	7.6	39	52	9.4	19 (18 to 20)	8 (7 to 8)
Social	24	6.0	43	48	9.7	20 (19 to 22)	8 (8 to 9)
Emotional	23	6.0	48	44	7.8	16 (14 to 17)	7 (6 to 7)
Language and cognitive	26	6.8	49	39	12	26 (24 to 27)	10 (9 to 10)
Communication	24	7.3	54	41	5.2	11 (10 to 12)	4 (4 to 5)
Any EDI domain*	51	21	60	31	9.3	17 (16 to 18)	6 (6 to 7)

*Calibration ability identifies whether the risk of vulnerability calculated from the model reflects the actual fraction of children in the population with developmental vulnerability (ie, positive predictive value, whether the high-risk group had a high rate of developmental vulnerability). Risk stratification capacity shows whether the proportion in which children are assigned to low-risk and high-risk categories is reasonable from an early intervention standpoint. Classification accuracy describes the extent to which the model assigns children with vulnerability to the high-risk category (sensitivity) and children with no vulnerability to the low-risk category (specificity).
†Risk categories for vulnerability on each EDI domain were based on model-predicted probabilities of developmental vulnerability: low risk <10%, mid risk 10%–19%, high risk ≥20%). Risk categories for vulnerability on any EDI domain were based on model-predicted probabilities of developmental vulnerability: low risk <30%, mid risk 30%–44% and high risk ≥45%).