Table 1.
Design of selected risk scores.
Table 2.
Variables included in each risk score.
Table 3.
Models and corresponding tuning parameters.
Fig 1.
A. Distribution of the sampling variables. Age: age at surgery; eject_fracti: ejection fraction; atrium_size: left atrial size; hypertensio: pulmonary hypertension; reoperation: number of previous cardiac surgeries; emergency: emergency surgery; cardiac_sh: cardiogenic shock; aortic valve: aortic valve surgery B. Distribution of sampling variables. Valve_revasc: heart valve surgery and CABG; tricuspid: tricuspid valve surgery; pacemaker: pacemaker dependency; ami48h: acute myocardial infarction 48h after cardiac surgery; dialysis: renal replacement therapy after cardiac surgery; renal_failure: acute kidney injury after cardiac surgery; high_creatinine: creatinine levels higher than 2mg/dl.
Table 4.
Characteristics of sample.
Table 5.
MINE results.
Fig 2.
Receiver operating curves compared across models.
Table 6.
Performance for all 13 models.
Fig 3.
Variable importance grid for the top models.
Table 7.
Area under the curve for the RheSCORE, 2000 Bernstein-Parsonnet, EuroSCORE II, InsCor, Ambler score, Guaragna score and the New York score.