Fig. 1
From: Susceptibility of AutoML mortality prediction algorithms to model drift caused by the COVID pandemic
Strobe Diagram. Applying exclusion criteria as shown, 109,719 anaesthetic cases remained in the whole dataset. Training and validation collective for the “native” as well as the “scaled” and “weighted” models consisted of 102,666 cases. The “6 months” method dataset comprised 11,161 cases. CT computed tomography, PET positron emission tomography, MRT magnetic resonance tomography, ICU intensive care unit