Background: Little evidence exists for de-escalation of nosocomial pneumonia therapy without positive cultures. Objective: The purpose of this study was to identify potential predictors of treatment failure following de-escalation to a fluoroquinolone in culture-negative nosocomial pneumonia. Methods: The study involved a single-center, retrospective cohort of patients admitted with diagnosis of nosocomial pneumonia and positive chest radiography who received at least 24 hours of fluoroquinolone monotherapy following at least 24 hours of appropriate empirical antibiotics. Treatment failure was defined using a composite of all-cause death within 30 days of discharge, treatment re-escalation, or readmission for pneumonia within 30 days of discharge. The Cox proportional hazards model was used to analyze predictors of treatment failure. Duration of empirical antibiotics and significant univariable exploratory predictors were included in multivariable analysis. Results: Of 164 patients, 23 (14%) failed de-escalation. Duration of empirical antibiotics (68.5 ± 32.1 vs 65.8 ± 35 hours) was not associated with treatment failure in univariable (Hazard Ratio [HR] = 1.002 [95% CI = 0.991-1.013]) or multivariable analyses (HR = 1.003 [95% CI = 0.991-1.015]). Significant exploratory predictors on univariable analysis included active cancer, intensive care unit (ICU) admission at empirical initiation, APACHE II score, and steroid use ≥20-mg prednisone equivalent. ICU admission at empirical initiation (HR = 2.439 [95% CI = 1.048-5.676]) and steroid use ≥20-mg prednisone equivalent (HR = 2.946 [95% CI = 1.281-6.772]) were associated with treatment failure on multivariable analysis. Conclusion and Relevance: Duration of empirical antibiotics does not appear to influence failure of de-escalation to fluoroquinolone monotherapy in culture-negative nosocomial pneumonia.