Additional file 1: of Multicenter derivation and validation of an early warning score for acute respiratory failure or death in the hospital Mikhail Dziadzko Paul Novotny Jeff Sloan Ognjen Gajic Vitaly Herasevich Parsa Mirhaji Yiyuan Wu Michelle Gong 10.6084/m9.figshare.7270214.v1 https://springernature.figshare.com/articles/journal_contribution/Additional_file_1_of_Multicenter_derivation_and_validation_of_an_early_warning_score_for_acute_respiratory_failure_or_death_in_the_hospital/7270214 Table S1. Variables selected for model. Table S2. APPROVE performance in derivation cohort to predict for event within 48 h of score. Table S3. Median (interquartile range) time to event from when APPROVE first exceeds threshold during hospitalization to primary event of death or intubation that leads to MV > 48 h, or to individual events of hospital death, or intubation leading to MV > 48 h in 2013 and 2017 validation cohorts. For composite qualifying event, earliest event (death or intubation leading to MV> 48 h) was considered the time of the event. Figure S1. Area under the curve (AUROC) for APPROVE, MEWS, and NEWS to predict for hospital mortality (a), mechanical ventilation (MV) > 48 h (b) and MV of any duration (c) in retrospective 2013 validation hospital and (d, e, f, respectively) prospective 2017 validation hospital. APPROVE, MEWS and NEWS calculated at multiple random time points for each patient and evaluated for a qualifying event after score calculation. (DOCX 141 kb) 2018-10-30 05:00:00 Acute respiratory failure Prediction Electronic health records Early warning scores Random forest