Fig. 8. Additional analyses on a HiRID time-series validation dataset.

a Comparison of external validation performance on static vs temporal HiRID validation datasets. b Correctly classified ‘discharged alive’ and ‘discharged dead’ labels made on static and temporal HiRID external validation datasets. c Odds ratio distributions of the Logistic Regression model trained on HiRID predicted labels 1–5 h before discharge/death.