Skip to main content
. 2019 Aug 7;2(8):e198719. doi: 10.1001/jamanetworkopen.2019.8719

Figure 2. Performance in Setting 1 and 2.

Figure 2.

A and B, Area under the receiver operating characteristic (AUROC) performance of ElasticNet model (A) and pulmonary embolism (PE) Neural model (B) on holdout internal Stanford hospital and clinics (SHC) and external Duke validation. C and D, Performance in setting 3: comparison with 3 clinical scoring systems on 100 outpatient samples (medical record manually reviewed by experts) seen at SHC (C) and 101 from Duke (D): PERC (Pulmonary Embolism Rule-out Criteria), Wells, rGeneva against our proposed ElasticNet and PE neural model.