Table 2.

Machine learning prediction accuracies

Variable	Predicting non-hazardous consumption (n = 290, 55.4%)	Predicting lowered AUDIT severity group (n = 269, 51.4%)
NCV average accuracy	0.64	0.48
NCV average ROC	0.69	0.52
RCV sensitivity	0.79	0.63
RCV specificity	0.53	0.44

NCV Nested cross-validation models. RCV repeated cross-validation. ROC Receiver operating characteristics