Skip to main content
[Preprint]. 2023 Nov 22:rs.3.rs-3593490. [Version 1] doi: 10.21203/rs.3.rs-3593490/v1

Table 4.

Average performances across different feature sets based on the held-out testing sets in the initial cohort after label adjustment.

Feature Sets Feature Dimension Average AUROC Average AUPRC
Phecodes [structured] + Demographics 1228 0.831 ± 0.040 0.917 ± 0.022
HPO + Demographics 26 0.737 ± 0.051 0.852 ± 0.042
Phecodes [unstructured] + Demographics + note counts 422 0.803 ± 0.049 0.90 ± 0.025
Phecodes [structure] + Demographics + note counts 1229 0.822 ± 0.045 0.918 ± 0.023
HPO + Demographics + note counts 27 0.736 ± 0.054 0.864 ± 0.035
Phecodes [structure & unstructured] + Demographics + note counts 1297 0.838 ± 0.052 0.917 ± 0.028