188. Which patients with gram-negative bacteremia need follow-up blood cultures?

Universal follow-up blood culture (FUBC) in gram-negative bacteremia (GNB) is not recommended, but it has been routinely conducted in many acute-care hospitals. In contrast with Staphylococcus aureus bacteremia, risk factors for positive FUBC in GNB have not been well investigated. Therefore, we tried to identify the risk factors for and develop predictive scores of positive FUBC.

All adults (≥18 years-old) with GNB were identified in a tertiary-care hospital during the 2-year period, retrospectively. Death within 2 days of GNB and polymicrobial infection with gram-positive bacteria or fungus were excluded. GNB were classified into eradicable and non-eradicable source of infection groups, according to the possibility of source removal. We performed multivariate analyses for identifying risk factors for positive FUBC and built prediction scores using the coefficients of the multivariate logistic regression models.

Of total 1,473 GNB, FUBC was drawn in 1,268 (86%) patients and 122 (9.6%) had positive results. In patients with eradicable source of infection, ESBL-producing microorganism, catheter-related bloodstream infection, unfavorable treatment response, and quick sequential organ failure assessment (qSOFA) score (≥2) were associated. On the other hand, administration of effective antibiotics and adequate source control were negatively associated with positive FUBC. In non-eradicable source of infection, end-stage renal disease on hemodialysis, and unfavorable treatment response were related to positive FUBC and administration of effective antibiotics was negatively associated (Table 1). When we built prediction scores according to the coefficients, the areas under the curves were 0.864 (95% confidence interval [CI95] 0.816–0.912) and 0.792 (CI95, 0.721–0.861), respectively. When we applied a cutoff of 0, specificities/negative predictive values in eradicable and non-eradicable source of infection groups were 84.7%/95.6% and 95.5%/95.0%, respectively (Table 2).

Our prediction scores based on adequate source control and use of effective antibiotics showed high specificities and negative predictive values. Therefore, we could expect these score systems to contribute to reducing unnecessary FUBC in GNB.

All authors: No reported disclosures.