Skip to main content
. 2021 Mar 1;4(1):ooab015. doi: 10.1093/jamiaopen/ooab015

Table 2.

Concordance of model predictions across test dataset

EN (n, %)
XGB (n, %)
CNN (n, %)
Model correct 1435 81.6 1452 82.6 1528 86.9
All models correct 1357 77.2 1357 77.2 1357 77.2
Model correct, 1 or both other models wrong 78 4.4 95 5.4 171 9.7
 Positive case 10 0.6 5 0.3 4 0.2
 Negative case 68 3.9 90 5.1 167 9.5
Note:

Total number in holdout test dataset = 1758.