Table 2

Discrimination of three models between cases and controls

Classifier	Codes only				Codes and keywords
Classifier	AUC	Sensitivity	Specificity	PPV at 7.1% prevalence	AUC	Sensitivity	Specificity	PPV at 7.1% prevalence
Random forest	0.90	0.77	0.92	0.46	0.94	0.85	0.92	0.46
Logistic regression	0.90	0.76	0.93	0.49	0.94	0.84	0.93	0.51
Naïve Bayes	0.87	0.78	0.91	0.45	0.90	0.80	0.91	0.43

AUC, area under the receiver operating characteristic curve; PPV, positive predictive value.