Table 3.

Diagnostic accuracy measures of the developed model (n = 1032) for predicting anticipated benefit from an EPCC at different cut-off values

Cut-off value	0.1	0.2	0.3a	0.4a	0.5	0.6	0.7	0.8	0.9
NPV, %	96.3	87.7	83.1	79.7	77.2	73.1	68.7	67.3	66.2
False negative in the high anticipated benefit score group, %	0.0	7.1	17.9	25.9	33.9	54.5	75.0	87.5	97.3
Youden index	0.07	0.35	0.43	0.41	0.38	0.27	0.11	0.06	0.02
Patients predicted by the model as having anticipated benefit (moderate and high), % of total	94.8	64.4	44.4	32.7	24.4	16.8	8.1	4.3	0.7
Patients scored by the GP as having anticipated benefit (moderate and high), % of total	34.1	34.1	34.1	34.1	34.1	34.1	34.1	34.1	34.1
Sensitivity, %	99.4	87.2	72.4	59.9	49.9	34.4	15.6	8.2	1.7
Specificity, %	7.6	47.4	70.1	81.5	88.5	92.4	95.7	97.8	99.9
PPV, %	35.8	46.2	55.7	62.6	69.0	69.9	65.5	65.9	85.7

^a The cut-off values 0.3 and 0.4 were considered, with the cut-off value 0.3 chosen as optimal.

EPCC = extended person-centred consultation. NPV = negative predictive value. PPV = positive predictive value.