Table 2.

Random forests classification metrics in cross-validation

Classification task*	Validation method	Accuracy	Specificity	Sensitivity	PPV	NPV
Active GCA versus controls	10-fold CV	95.0%	96.7%	93.3%	96.7%	93.5%
Inactive GCA versus controls	10-fold CV	98.3%	100%	96.7%	100%	96.8%
Active GCA versus inactive GCA	Leave-one-patient-out	51.7%	46.7%	56.7%	51.5%	51.9%

^*Classification tasks were performed using the default configurations of the random forest classifier from the sci-kit learn library (V.1.3.2) in Python. Classification metrics were calculated as the ratio of total correct predictions to total predictions, as defined for each metric, across each fold of cross-validation method.

GCA, giant cell arteritis; NPV, negative predictive value; PPV, positive predictive value.