Fig. 4From: Susceptibility of AutoML mortality prediction algorithms to model drift caused by the COVID pandemicVenn diagrams after model modification. Diagrams show the 100 most important features of the best pre-pandemic model of each family after the methods “weight”, “6 months” and “scaled” were applied. The centre of each diagram shows the number of parameters that occur in each of the models, with “age” being the number one common feature in each of the methods. The text boxes on the left indicate which parameters are involved in all models sorted by their importance. RPC’s: red packed cells, INR: international normalized ratio, γGT: gamma-glutamyl transferase, ASA: American Society of Anaesthesiologists Physical Score [16], BUN: blood urea nitrogen, MCHC: mean corpuscular haemoglobinBack to article page