Table 5. Classifier model prediction performance on CPOnly (controls and culture positive cases) for fixed sensitivity ratio of 0.8.

The probability of sepsis threshold was adjusted individually for each model in each cross validation run to achieve 0.8 sensitivity. Each metric value is computed as the mean over 10 iterations of cross-validation. Values in brackets indicate performance range over the 10 iterations. Bold text indicates highest performance in each column.

Model	Specificity	PPV	NPV
AdaBoost	0.71 [0.64, 0.81]	0.23 [0.18, 0.30]	0.97 [0.97, 0.98]
Gradient boosting	0.69 [0.58, 0.91]	0.23 [0.16, 0.47]	0.98 [0.97, 0.99]
Gaussian process	0.53 [0.32, 0.80]	0.16 [0.11, 0.29]	0.97 [0.95, 0.98]
K-nearest neighbors	0.20 [0, 0.77]	0.13 [0.09, 0.26]	0.29 [0, 1]
Logistic regression	0.71 [0.61, 0.82]	0.23 [0.17, 0.31]	0.97 [0.97, 0.98]
Naïve Bayes	0.69 [0.49, 0.77]	0.22 [0.14, 0.28]	0.98 [0.96, 0.99]
Random forest	0.71 [0.62, 0.83]	0.23 [0.19, 0.32]	0.98 [0.97, 0.99]
Support vector machine*	0.72 [0.63, 0.83]	0.23 [0.18, 0.32]	0.98 [0.97, 0.98]

PPV: positive predictive value; NPV: negative predictive value

*The radial basis function kernel was used for the support vector machine