Table 3. Performance of Full (21-Variable) Models and Reduced (5- or 3-Variable) Models in the Training Set.

Model	Accuracy	Sensitivity	Specificity	ROC AUC
Major response at week 12
Logistic regression
Full model	0.74	0.49	0.87	0.81
Reduced model	0.74	0.45	0.89	0.80
Random forest
Full model	0.74	0.42	0.89	0.80
Reduced model	0.72	0.46	0.85	0.78
No response at week 12
Logistic regression
Full model	0.75	0.47	0.90	0.77
Reduced model	0.75	0.44	0.90	0.77
Random forest
Full model	0.73	0.41	0.90	0.76
Reduced model	0.74	0.45	0.90	0.75

Abbreviation: ROC AUC, receiver operating characteristic area under curve.