Table 3.

Classification accuracy using different probability cutoffs for logistic models

Age phase	Model 1						Models 2/3
	Probability level	Classification correct, %	Sensitivity, %	Specificity, %	True-Positive cases	False-positive cases	Probability level	Classification correct, %	Sensitivity, %	Specificity, %	True-Positive cases	False-positive cases
3.5	0.5	80.4	2.5	100	1	0	0.5	79.2	2.4	100	1	0
	0.25	68.3	57.5	71.1	23	46	0.25	66.7	61	68.2	25	48
	0.103	42.2	90	30.2	36	111	0.15	53.1	90.2	43	37	86
4.5	0.5	81.9	0	100	0	0	0.5	87.5	36.4	96.7	8	4
	0.25	74.4	47.2	80.4	17	32	0.25	86.1	68.2	89.3	15	13
	0.088	36.7	91.7	24.5	33	123	0.12	76.4	90.9	73.8	20	32
5.5	0.5	76.7	6.7	95.29	3	8	0.5	84.5	44.7	95.3	21	8
	0.25	74.9	62.22	78.2	28	37	0.25	79.5	63.8	83.7	30	28
	0.131	53.5	91.11	43.5	41	96	0.13	68.5	89.4	62.8	42	64
6–7	0.5	80.7	23.91	95.9	11	7	0.5	87.8	63.8	94	30	11
	0.25	77.1	69.6	79.1	32	36	0.25	84.3	76.6	86.3	36	25
	0.094	49.1	89.1	38.4	41	106	0.13	79	89.4	76.4	42	43

Probability cutoffs .25 and .5 are given as they are common in the literature. However, the best diagnostic models between T1 and T4 range from .08 to .15 cutoffs, giving 90% sensitivity in the sample.