Table 3.

Results of multivariable analysis using the LASSO method for the outcome of ‘Completely Stone Free’ showing chosen predictor variables and their corresponding coefficients.

Variable	LASSO co-efficient
(Intercept)	4.5
Sex = Male	−0.582
Age	−0.006
Number of stones being treated	−0.817
Size of the major axis (of the largest stone)	−0.176
Size of vertical axis (of the largest stone)	0.002
Skin-to-stone distance	0.004
Stone location not in the VUJ	−0.751
ROC analysis of model:	AUC = 0.66
At a predicted probability cut-off of 0.51 (for a ‘Completely Stone Free’ successful outcome of SWL treatment):
Sensitivity = 59.8%
Specificity = 56.3%

Variables used as input into this model were those who had a significance of p > 0.1 on univariate analysis. The chosen predictor variables shown in this table with a negative coefficient indicate that higher values are linked to greater probability of unsuccessful SWL with outcome of ‘Not Completely Stone Free’. The LASSO co-efficient is not standardized and relates to the values of the variables and not the relative weight of influence of the variables. For example, for every one-year increase in age, there is a 0.006 percentage points decrease in the predicted probability of SWL success. The dataset was split into 2/3 training data to develop the predictive model whose performance is evaluated on the remaining 1/3 test data for ROC analysis. SWL, extracorporeal shockwave lithotripsy; LASSO, least absolute shrinkage and selection operator; ROC, receiver operator curve; VUJ, vesicoureteric junction.