In Reply We appreciate the Letters to the Editor by Hsu and Lai, Darlow et al, and Chao in response to our retrospective cohort study.¹ Hsu and Lai inquired regarding the large number of patients with an unclear source of infection. By design, we emulated the conditions of a randomized clinical trial and thus considered only data available in the first 24 hours of presentation (when decisions regarding empirical antibiotics are made). After reviewing data available at the time of discharge, only 862 of 4523 patients treated with piperacillin-tazobactam (19.1%) and 606 of 3046 patients treated with cefepime (19.9%) had an unclear infectious source. This rate is comparable to that of the recent ACORN (Antibiotic Choice on Renal Outcomes) trial.² A post hoc sensitivity analysis including infectious diagnosis at discharge did not impact our point estimate (4.8% increase in 90-day mortality [95% CI, 1.7%-8.0%]). Given that the Surviving Sepsis international guidelines recommend empirical antibiotics with 1 to 3 hours of presentation,³ antibiotic selection commonly occurs before the pathogen or anatomic site of infection is known.