Skip to main content
. 2022 Aug 28;151:106024. doi: 10.1016/j.compbiomed.2022.106024

Table 3.

Performance of individual algorithms using the first (high sensitivity) and second (high specificity) operating points. The baseline models were the decision by the EMS clinicians whether to convey the patient to hospital or not and the recommendation arising from the use of the PRIEST clinical severity score [11].

Model Sensitivity Specificity npv ppv GM AUC
Baseline EMS clinician 0.84 0.39 0.92 0.23 0.57 NA
PRIEST score 0.97 0.41 0.98 0.26 0.63 0.83 ± 0.01

(a) High sensitivity LR 0.98 0.43 0.99 0.27 0.65 0.87 ± 0.01
XGB 0.98 0.39 0.99 0.26 0.62 0.86 ± 0.02
SVM 0.98 0.41 0.99 0.26 0.63 0.86 ± 0.01
ANN 0.96 0.72 0.99 0.43 0.80 0.90 ± 0.09

(b) High specificity LR 0.93 0.64 0.98 0.37 0.77 0.87 ± 0.01
XGB 0.95 0.56 0.98 0.31 0.72 0.86 ± 0.02
SVM 0.94 0.60 0.98 0.33 0.75 0.86 ± 0.01
ANN 0.95 0.77 0.98 0.48 0.86 0.90 ± 0.09