BACKGROUND: The causes and clinical characteristics of recurrent gram-negative bacterial bloodstream infections (GNB-BSI) are poorly understood. METHODS: We used a cohort of patients with GNB-BSI to identify clinical characteristics, microbiology, and risk factors associated with recurrent GNB-BSI. Bacterial genotyping (pulsed-field gel electrophoresis [PFGE] and whole-genome sequencing [WGS]) was used to determine whether episodes were due to relapse or reinfection. Multivariable logistic regression was used to identify risk factors for recurrence. RESULTS: Of the 1423 patients with GNB-BSI in this study, 60 (4%) had recurrent GNB-BSI. Non-White race (odds ratio [OR], 2.35; 95% confidence interval [CI], 1.38-4.01; P = .002), admission to a surgical service (OR, 2.18; 95% CI, 1.26-3.75; P = .005), and indwelling cardiac device (OR, 2.73; 95% CI, 1.21-5.58; P = .009) were associated with increased risk for recurrent GNB-BSI. Among the 48 patients with recurrent GNB-BSI whose paired bloodstream isolates underwent genotyping, 63% were due to relapse (30 of 48) and 38% were due to reinfection (18 of 48) based on WGS. Compared with WGS, PFGE correctly differentiated relapse and reinfection in 98% (47 of 48) of cases. Median time to relapse and reinfection was similar (113 days; interquartile range [IQR], 35-222 vs 174 days; IQR, 69-599; P = .13). Presence of a cardiac device was associated with relapse (relapse: 7 of 27, 26%; nonrelapse: 65 of 988, 7%; P = .002). CONCLUSIONS: In this study, recurrent GNB-BSI was most commonly due to relapse. PFGE accurately differentiated relapse from reinfection when compared with WGS. Cardiac device was a risk factor for relapse.