The impact of different imputation methods on estimates and model performance: an example using a risk prediction model for premature mortality

Table 2 Performance measures

	Complete	Mode	Single	Multiple
	Case	Imputation	Imputation	Imputation
FEMALE
Nagelkerke R2¹	0.1084	0.1116	0.1120	0.1111–0.1120
Integrated Brier Score²	0.0055	0.0063	0.0063	0.0063–0.0063
C-index³	0.8666	0.8719	0.8719	0.8711–0.8719
Discrimination Slope⁴	0.0892	0.0930	0.0936	0.0926–0.0938
Calibration in the large⁵	-0.0016	-0.0017	-0.0017	-0.0017, -0.0017
Calibration Slope⁶	1.0000	1.0000	1.0000	1.0000–1.0000
MALE
Nagelkerke R2¹	0.1050	0.1078	0.1078	0.1078–0.1081
Integrated Brier Score²	0.0086	0.0093	0.0093	0.0093–0.0093
C-index³	0.8549	0.8548	0.8550	0.8550–0.8553
Discrimination Slope⁴	0.0904	0.0988	0.0990	0.0990–0.0993
Calibration in the large⁵	-0.0020	-0.0022	-0.0022	-0.0022,-0.0022
Calibration Slope⁶	1.0000	1.0000	1.0000	1.0000–1.0000

¹The Nagelkerke R2 measures the percent of variance explained by the model with a target value of one
²The Integrated Brier Score measures the average squared difference between the outcome and the predicted risk (while taking censoring into account) with a target value of zero
³Harrell’s concordance index (c-index) which is the fraction of the number of concordant pairs over the number of concordant pairs and discordant pairs
⁴Time-specific discrimination slope is the difference in the average predicted risk of those who had an event and those who did not have an outcome
⁵Calibration in the large is the difference between the average observed risk (normally calculated using Kaplan-Meier curves) and the average predicted risk
⁶The calibration slope assesses if the betas are well-calibrated for the model (with a slope of one indicating perfect calibration)

ISSN: 1478-7954