Extensions to decision curve analysis, a novel method for evaluating diagnostic tests, prediction models and molecular markers

Table 3 Simulation results for correction for over-fit: "best estimate" of net benefit minus net benefit after correction.

Threshold	Number of events	Correction method			Bootstrap – 10-fold
	(Total sample size)	Bootstrap	2-fold	10-fold
15%	100 (385)	0.0001	0.0021	0.0005	-0.0004
	50 (193)	0.0032	0.0048	0.0013	0.0019
	40 (154)	0.0050	0.0073	0.0025	0.0025
	30 (116)	0.0080	0.0119	0.0045	0.0035
	20 (77)	0.0106	0.0249	0.0081	0.0025
25%	100 (385)	0.0005	0.0007	0.0007	-0.0002
	50 (193)	0.0024	0.0038	0.0002	0.0022
	40 (154)	0.0036	0.0063	0.0012	0.0024
	30 (116)	0.0079	0.0109	0.0033	0.0046
	20 (77)	0.0098	0.0249	0.0069	0.0029
35%	100 (385)	0.0024	0.0047	0.0024	0.0000
	50 (193)	0.0063	0.0083	0.0039	0.0024
	40 (154)	0.0075	0.0111	0.0057	0.0018
	30 (116)	0.0111	0.0158	0.0084	0.0027
	20 (77)	0.0145	0.0304	0.0118	0.0027
60%	100 (385)	0.0049	0.0101	0.0052	-0.0003
	50 (193)	0.0114	0.0185	0.0106	0.0008
	40 (154)	0.0126	0.0231	0.0142	-0.0016
	30 (116)	0.0163	0.0292	0.0174	-0.0011
	20 (77)	0.0223	0.0465	0.0256	-0.0033
80%	100 (385)	0.0049	0.0085	0.0050	-0.0001
	50 (193)	0.0089	0.0181	0.0089	0.0000
	40 (154)	0.0101	0.0250	0.0121	-0.0020
	30 (116)	0.0098	0.0349	0.0167	-0.0069
	20 (77)	0.0179	0.0623	0.0295	-0.0116

ISSN: 1472-6947