Table 4.

Comparison of the performance of predictive models for the outcome of ‘Completely Stone Free’ or ‘Stone Free with CIRFs’ using three methods of multivariable analyses; first without the inclusion of CTTA variables and then with the addition of CTTA variables to the models.

		Outcome
		Completely Stone Free		Stone Free with CIRFs
		Patient and stone related variables	With the addition of CTTA variables	Patient and stone related variables	With the addition of CTTA variables
Multivariable models evaluated	Statistic of model performance
LASSO	Area under the curve on ROC analysis for discrimination performance	0.66	0.64	0.67	0.67
	Hosmer-Lemeshow p-value	p < 0.001	p < 0.001	p < 0.001	p < 0.001
Partial Least Squares	Q2 (quality assessment) statistic	0.117	0.086	—	—
Random Forests	Area under the curve on ROC analysis	0.67	0.65	—	—

A significant Hosmer–Lemeshow p-value, as shown in this table, indicates the model is poorly calibrated. Partial Least Squares and Random Forests methods were not performed for the outcome of ‘Stone Free with CIRFs’. CIRFs, clinically insignificant residual fragments; CTTA, computed tomography texture analysis; LASSO, least absolute shrinkage and selection operator; ROC, receiver operator curve.