Fig. 1: Study cohort selection.
From: Personalized antibiograms for machine learning driven antibiotic selection

a 119,840 hospital admissions corresponding to 69,069 unique adult patients admitted from Stanford emergency rooms between 2009 and 2019 were initially examined for inclusion. b 42,448 admissions had a microbial culture and intravenous or intramuscular empiric antibiotic order placed within the first 24 h of the encounter. c Admissions were excluded if microbial cultures had been ordered in the 2 weeks leading up to the encounter. d Admissions resulting in negative microbial cultures were excluded in the primary analysis, leaving 8342 infections from 6920 unique patients.