The impact of follow-up blood cultures on mortality and management in patients with gram-negative bloodstream infections: a validation cohort study

BACKGROUND: The practice of obtaining follow-up blood cultures (FUBCs) in patients with gram-negative bloodstream infections (GN-BSI) remains controversial. The current study had two goals: (1) to internally validate our earlier findings of decreased mortality in patients with FUBCs and increased mortality in those with persistent GN-BSI, and (2) to identify mechanisms by which FUBCs might impact clinical care and outcome. METHODS: Adults with GN-BSI at Duke University were enrolled from 2015 to 2021. Cox proportional hazards models with propensity score-weighting were used to examine associations of FUBCs and positive FUBCs with attributable and all-cause in-hospital mortality. Immortal time bias was addressed by treating FUBCs as a time-dependent variable and through sensitivity analyses. To understand the potential impact of FUBCs on clinical care, matched cohorts were generated (n = 41 in each group): (1) no FUBCs, (2) FUBCs negative for growth, and (3) FUBCs positive for growth (i.e., persistent GN-BSI). Infection-related consultations, imaging studies, and procedures were determined for each matched cohort. RESULTS: Of the 1291 patients in this study, FUBCs were obtained in 78% (1012 patients) and positive in 20% (199 patients). After adjusting for immortal time bias, obtaining FUBCs remained associated with decreased attributable mortality (Hazard ratio [HR] 0.39; 95% confidence interval [CI] 0.32-0.49; p < 0.0001). Persistent GN-BSI was associated with increased attributable mortality (HR 1.77; 95% CI, 1.24-2.52; p = 0.002). Obtaining FUBCs was associated with subsequent infectious diseases (ID) consultation (No FUBCs: 2/41 [5%]; FUBCs: 21/82 [26%]; p < 0.0001). This finding was primarily due to patients with persistent GN-BSI (FUBCs negative: 6/41 [15%]; FUBCs positive: 15/41 [37%]; p = 0.04). Patients with persistent GN-BSI had a higher rate of source control procedures (FUBCs positive: 12/41 [29%]; All others: 11/82 [13%]); p = 0.05). CONCLUSIONS: Obtaining FUBCs was associated with decreased mortality. This mortality effect could stem from increased ID consults after FUBCs and increased source control procedures following positive FUBCs.