Early Discontinuation of Antibiotics in Patients Admitted With Clinically Suspected Serious Infection but Negative Cultures: Retrospective Cohort Study of Practice Patterns and Outcomes at 111 US Hospitals.

Background: The optimal duration for antibiotics in patients hospitalized with culture-negative serious infection (CNSI) is unknown. We compared outcomes in patients with CNSI treated with 3 or 4 vs ≥5 days of antibiotics. Methods: CNSI was identified among adults admitted to 111 US hospitals between 2009 and 2014 via electronic health record data, defined as suspected serious infection (blood cultures drawn and ≥3 days of antibiotics) and negative culture- and nonculture-based tests for infection. Patients treated with antibiotics on their last hospital day and patients with diagnosis codes for sepsis-mimicking conditions were excluded. Among patients without fevers/hypothermia or vasopressors by day 3, we calculated odds ratios for in-hospital mortality or discharge to hospice associated with 3 or 4 vs ≥5 days of antibiotics, adjusting for confounders. Results: Antibiotics were discontinued in 3 or 4 days in 1862 (9%) of 20 714 patients with CNSI. Early discontinuation was not associated with higher mortality odds overall (adjusted odds ratio [aOR], 1.27; 95% CI, .98-1.65), in patients presenting with (1.39; .88-2.22) and without sepsis (1.17; .81-1.69), and in those with pulmonary (1.23; .65-2.34) and nonpulmonary CNSI (1.30; .99-1.72). Early discontinuation appeared detrimental with propensity score weighting (aOR, 1.36; 95% CI, 1.03-1.80) and when retaining patients with sepsis mimics (1.38; 1.16-1.65), but it was protective (0.48; .37-.64]) when retaining patients who received antibiotics on their last hospital day. Conclusions: Early discontinuation of antibiotics in CNSI was not associated with significant harm in our primary analysis, but different conclusions based on alternative analytic decisions, as well as risk of residual confounding, indicate that randomized controlled trials are needed.