Table 1 Test AUROCs for EHRFormer and XGBoost predicting one year mortality across key time periods in the UNOS dataset.

From: In silico perturbations provide multivariate interpretability in predicting post-lung transplant outcomes

Model

1987–2004

2005–2014

2015-now

All years

EHRFormer

0.61 [0.56,0.67] 95% CI

0.60 [0.56,0.65] 95% CI

0.56 [0.50,0.62] 95% CI

0.64 [0.61, 0.66] 95% CI

XGBoost

0.62 [0.55,0.72] 95% CI

0.63 [0.58,0.69] 95% CI

0.62 [0.58,0.66] 95% CI

0.63 [0.60, 0.67] 95% CI

Back to article page