Gram-negative Bacilli Bloodstream Infections Research Articles

Novel scores relevant to antimicrobial resistance and hospital-acquired infections developed with data from a multi-hospital consortium in the Parisian region of France

Indicators for comparing and understanding differences in antimicrobial resistance (AMR) and healthcare-associated infections (HAIs) for benchmarking are essential to identify priorities for hospitals. This study measured the incidence of hospital-acquired or resistant Gram-negative bacilli bloodstream infections (GNB-BSIs) in a large public healthcare consortium in the Parisian region of France. Within each hospital, there was a strong positive correlation between the incidence of GNB-BSIs due to resistant GNB and the incidence of hospital-acquired GNB-BSIs. Two scores measuring AMR and HAI rates by combining different GNB-BSI incidence rates were developed as indicators. These scores were highly variable within the hospital consortium. On multi-variate analysis, AMR and HAI scores were significantly associated with the proportion of surgical beds, staff absenteeism and the consumption of alcohol-based hand rub, with the latter two characteristics being amenable to interventions. Carbapenem use was also linked to AMR, but this may be because carbapenems are the preferred drug for treating resistant infections. These results shed light on the incidence of HAIs and AMR in the study hospitals, and suggest possibilities for targeted interventions at healthcare facility level.

Mortality Attributable to Bloodstream Infections Caused by Different Carbapenem-Resistant Gram-Negative Bacilli: Results From a Nationwide Study in Italy (ALARICO Network).

Our aim was to analyze mortality attributable to carbapenem-resistant (CR) gram-negative bacilli (GNB) in patients with bloodstream infections (BSIs). Prospective multicentric study including patients with GNB-BSI from 19 Italian hospitals (June 2018-January 2020). Patients were followed-up to 30 days. Primary outcomes were 30-day mortality and attributable mortality. Attributable mortality was calculated in the following groups: Klebsiella pneumoniae carbapenemase (KPC)-producing Enterobacterales, metallo-β-lactamases (MBL)-producing Enterobacterales, CR-Pseudomonas aeruginosa (CRPA), CR-Acinetobacter baumannii (CRAB). A multivariable analysis with hospital fixed-effect was built to identify factors associated with 30-day mortality. Adjusted OR (aORs) were reported. Attributable mortality was calculated according to the DRIVE-AB Consortium. Overall, 1276 patients with monomicrobial GNB BSI were included: 723/1276 (56.7%) carbapenem-susceptible (CS)-GNB, 304/1276 (23.8%) KPC-, 77/1276 (6%) MBL-producing CRE, 61/1276 (4.8%) CRPA, and 111/1276 (8.7%) CRAB BSI. Thirty-day mortality in patients with CS-GNB BSI was 13.7% compared to 26.6%, 36.4%, 32.8% and 43.2% in patients with BSI by KPC-CRE, MBL-CRE, CRPA and CRAB, respectively (P < .001). On multivariable analysis, age, ward of hospitalization, SOFA score, and Charlson Index were factors associated with 30-day mortality, while urinary source of infection and early appropriate therapy resulted protective factors. Compared to CS-GNB, MBL-producing CRE (aOR 5.86, 95% CI 2.72-12.76), CRPA (aOR 1.99, 95% CI 1.48-5.95) and CRAB (aOR 2.65, 95% CI 1.52-4.61) were significantly associated with 30-day mortality. Attributable mortality rates were 5% for KPC-, 35% for MBL, 19% for CRPA, and 16% for CRAB. In patients with BSIs, carbapenem-resistance is associated with an excess of mortality, with MBL-producing CRE carrying the highest risk of death.

Clinical Impact of COVID-19 on Multi-Drug-Resistant Gram-Negative Bacilli Bloodstream Infections in an Intensive Care Unit Setting: Two Pandemics Compared.

Two mutually related pandemics are ongoing worldwide: the COVID-19 and antimicrobial resistance pandemics. This study aims to evaluate the impact of COVID-19 on multi-drug-resistant Gram-negative bacteria (MDR-GN) bloodstream infections (BSIs) in a single intensive care unit (ICU). We conducted a retrospective study including patients admitted to the ICU, reorganized for COVID-19 patients’ healthcare, with at least one confirmed MDR-GN BSI during 2019–2020. We compared clinical and microbiological features, incidence density, antibiotic therapy and mortality rate in pre- and during-COVID-19 pandemic periods. We estimated the impact of COVID-19 on mortality by means of univariate Cox regression analyses. A total of 46 patients were included in the study (28 non-COVID-19/18 COVID-19). Overall, 63 BSI episodes occurred (44/19), and non-COVID-19 patients had a higher incidence of MDR-GN BSIs and were more likely to present K. pneumoniae BSIs, while the COVID-19 group showed more A. baumannii BSIs with higher per pathogen incidence. COVID-19 patients presented more critical conditions at the BSI onset, a shorter hospitalization time from BSI to death and higher 30-day mortality rate from BSI onset. COVID-19 and septic shock were associated with 30-day mortality from MDR-GN BSIs, while early active therapy was a protective factor. In conclusion, COVID-19 showed a negative impact on patients with MDR-GN BSIs admitted to the ICU.

89. Follow-Up Blood Cultures (FUBC) in the Management of Gram-Negative Bacilli (GNB) Bloodstream Infections (BSIs): Frequently Obtained and Rarely Helpful

BackgroundWhile GNB BSIs remain a major cause of morbidity and mortality, no clear guidelines exist on the utilization of FUBCs to guide management. Despite the recognition of persistent bacteremia as a risk factor for increased mortality, early studies suggested FUBCs were low yield in this setting, and thus had low utility. More recently, some controversy has arisen with multivariate analyses suggesting FUBC acquisition may be associated with lower mortality. We sought to characterize the utilization and yield of FUBCs for GNB BSIs at our institution.MethodsWe performed a retrospective review of 514 episodes of consecutive blood cultures from unique adult inpatients with GNB BSI between July 2017-July 2019. Exclusion criteria included prior positive culture, polymicrobial Gram stain, or discharge, death, or comfort measures only within 24 hours of Gram stain. FUBCs were defined as blood cultures collected between 24 hours to 7 days after the index blood culture. Baseline clinical and microbiologic characteristics were compared between groups, as well as clinical outcomes.ResultsOf 514 episodes, 338 (66%) had FUBCs performed, with a median of 2 FUBCs/episode. The majority of FUBCs (322/338; 95%) were negative, with 9 (3%) yielding the same organism and 9 (3%) yielding a different organism. Most initial FUBCs were obtained prior to index antimicrobial susceptibility results (227/338; 67%). Patients with FUBCs performed had a higher median Pitt bacteremia score (2 vs 1; p = 0.015) and were more likely to have hospital onset (36% vs 22%; p = 0.002), severe neutropenia (16% vs 4%; p < 0.001) and a catheter-associated source (13% vs 4%; p = 0.001). 30-day mortality did not differ between patients with or without FUBCs (10% vs 11%; p = 0.84).ConclusionFUBCs were frequently obtained, but were of low yield even in comparison to recent similar studies. Though FUBCs were performed in more severe cases, a difference in mortality was not observed. Delaying the decision of whether to obtain FUBCs until after index antimicrobial susceptibility results are available would reduce unnecessary testing in most cases. Further study could better define where FUBCs after antimicrobial susceptibility testing would be most helpful.Disclosures All Authors: No reported disclosures

Implementation of a Rapid Phenotypic Susceptibility Platform for Gram-Negative Bloodstream Infections With Paired Antimicrobial Stewardship Intervention: Is the Juice Worth the Squeeze?

BackgroundImplementation of the Accelerate PhenoTM Gram-negative platform (RDT) paired with antimicrobial stewardship program (ASP) intervention projects to improve time to institutional-preferred antimicrobial therapy (IPT) for Gram-negative bacilli (GNB) bloodstream infections (BSIs). However, few data describe the impact of discrepant RDT results from standard of care (SOC) methods on antimicrobial prescribing.MethodsA single-center, pre-/post-intervention study of consecutive, nonduplicate blood cultures for adult inpatients with GNB BSI following combined RDT + ASP intervention was performed. The primary outcome was time to IPT. An a priori definition of IPT was utilized to limit bias and to allow for an assessment of the impact of discrepant RDT results with the SOC reference standard.ResultsFive hundred fourteen patients (PRE 264; POST 250) were included. Median time to antimicrobial susceptibility testing (AST) results decreased 29.4 hours (P < .001) post-intervention, and median time to IPT was reduced by 21.2 hours (P < .001). Utilization (days of therapy [DOTs]/1000 days present) of broad-spectrum agents decreased (PRE 655.2 vs POST 585.8; P = .043) and narrow-spectrum beta-lactams increased (69.1 vs 141.7; P < .001). Discrepant results occurred in 69/250 (28%) post-intervention episodes, resulting in incorrect ASP recommendations in 10/69 (14%). No differences in clinical outcomes were observed.ConclusionsWhile implementation of a phenotypic RDT + ASP can improve time to IPT, close coordination with Clinical Microbiology and continued ASP follow up are needed to optimize therapy. Although uncommon, the potential for erroneous ASP recommendations to de-escalate to inactive therapy following RDT results warrants further investigation.

90. Impact of Discrepant Rapid Diagnostic Test (RDT) Results on Antimicrobial Stewardship Program (ASP) Interventions in Patients with Bloodstream Infections (BSI) due to Gram-Negative Bacilli (GNB)

BackgroundImplementation of the Accelerate PhenoTM Gram-negative platform (AXDX) paired with ASP intervention projects to improve time to definitive institutional-preferred antimicrobial therapy (IPT). However, few data describe the impact of discrepant RDT results from standard of care (SOC) methods on antimicrobial prescribing. Here we evaluate the prescribing outcomes for discrepant results following the first year of AXDX + ASP implementation.MethodsConsecutive, non-duplicate blood cultures for adult inpatients with GNB BSI following combined RDT + ASP intervention were included (July 2018 – July 2019). AXDX results were emailed to the ASP in real time then released into the EMR upon ASP review and communication with the treating team. SOC identification (ID; Vitek® MS/Vitek® 2) and antimicrobial susceptibility testing (AST; Trek SensititreTM) followed RDT as the reference standard. IPT was defined as the narrowest susceptible beta-lactam, and a discrepancy was characterized when there was categorical disagreement between RDT and SOC methods. When IPT by AXDX was found to be non-susceptible on SOC, this was characterized as “false susceptible“. Conversely, “false resistance” was assessed when a narrower-spectrum agent was susceptible by SOC. Results were also deemed discrepant when the AXDX provided no/incorrect ID for on-panel organisms, no AST, or a polymicrobial specimen was missed.ResultsSixty-nine of 250 patients (28%) had a discrepancy in organism ID or AST: false resistance (9%), false susceptible (5%), no AST (5%), no ID (4%), incorrect ID (2%), and missed polymicrobial (2%). A prescribing impact occurred in 55% of cases (Table 1), where unnecessarily broad therapy was continued most often. Erroneous escalation (7%) and de-escalation to inactive therapy (7%) occurred less frequently. In-hospital mortality occurred in 4 cases, none of which followed an inappropriate transition to inactive therapy. ConclusionThough the AXDX platform provides rapid ID and AST results, close coordination with Clinical Microbiology and continued ASP follow up are needed to optimize therapy. Although uncommon, the potential for erroneous ASP recommendations to de-escalate to inactive therapy following AXDX results warrants further investigation.DisclosuresAmy J. Mathers, MD, D(ABMM), Accelerate Diagnostics (Consultant)

Randomized Trial Evaluating Clinical Impact of RAPid IDentification and Susceptibility Testing for Gram-negative Bacteremia: RAPIDS-GN.

Rapid blood culture diagnostics are of unclear benefit for patients with gram-negative bacilli (GNB) bloodstream infections (BSIs). We conducted a multicenter, randomized, controlled trial comparing outcomes of patients with GNB BSIs who had blood culture testing with standard-of-care (SOC) culture and antimicrobial susceptibility testing (AST) vs rapid organism identification (ID) and phenotypic AST using the Accelerate Pheno System (RAPID). Patients with positive blood cultures with Gram stains showing GNB were randomized to SOC testing with antimicrobial stewardship (AS) review or RAPID with AS. The primary outcome was time to first antibiotic modification within 72 hours of randomization. Of 500 randomized patients, 448 were included (226 SOC, 222 RAPID). Mean (standard deviation) time to results was faster for RAPID than SOC for organism ID (2.7 [1.2] vs 11.7 [10.5] hours; P < .001) and AST (13.5 [56] vs 44.9 [12.1] hours; P < .001). Median (interquartile range [IQR]) time to first antibiotic modification was faster in the RAPID arm vs the SOC arm for overall antibiotics (8.6 [2.6-27.6] vs 14.9 [3.3-41.1] hours; P = .02) and gram-negative antibiotics (17.3 [4.9-72] vs 42.1 [10.1-72] hours; P < .001). Median (IQR) time to antibiotic escalation was faster in the RAPID arm vs the SOC arm for antimicrobial-resistant BSIs (18.4 [5.8-72] vs 61.7 [30.4-72] hours; P = .01). There were no differences between the arms in patient outcomes. Rapid organism ID and phenotypic AST led to faster changes in antibiotic therapy for gram-negative BSIs. NCT03218397.

640. Randomized Clinical Trial Evaluating Clinical Impact of RAPid IDentification and Antimicrobial Susceptibility Testing for Gram-Negative Bacteremia (RAPIDS-GN)

BackgroundRapid blood culture diagnostics increase cost and have unclear benefit for patients with Gram-negative bacilli (GNB) bloodstream infections (BSIs). We conducted a multicenter, prospective randomized controlled trial (RAPIDS-GN), comparing outcomes of patients with GNB BSI who had blood culture testing with standard of care (SOC) culture and antibiotic susceptibility testing (AST) vs. rapid organism identification (ID) and phenotypic AST using the Accelerate Pheno System (AXDX).MethodsSubjects with blood culture Gram stain showing GNB were randomized to receive SOC testing with antimicrobial stewardship review (AS) or AXDX plus SOC testing with AS, at two academic medical centers between October 2017 and October 2018. SOC testing included rapid MALDI-TOF mass spectrometry ID and agar dilution or broth microdilution AST. In a modified intention to treat analysis, subjects were excluded if: Gram stain was erroneous, culture was positive during off-hours, blood culture in the prior week had GNB, they were deceased/on comfort care, or admitted to a nonparticipating hospital. The primary outcome was time to first antibiotic modification within 72 hours after randomization. Subjects without antibiotic modifications were assigned a time of 72 hours. No censoring was observed. T-tests and Wilcoxon rank-sum tests were used for statistical analyses.ResultsOf 500 randomized subjects, 448 were included (226 SOC, 222 AXDX). Groups did not differ in baseline characteristics (Table 1). Median (IQR) hours to first antibiotic modification was faster in the AXDX vs. SOC group [8.6 (2.6, 27.6) vs. 14.9 (3.3, 41.1)], P = 0.02 (Figure 1). Median (IQR) hours to first Gram-negative antibiotic modification (including escalation and de-escalation) was faster in the AXDX than SOC group [17.4 (4.9, 72) vs. 42.1 (10.1, 72)], P < 0.001 (Figure 2). Groups did not differ in clinical outcomes (Table 2). Mean (S.D.) time to results was faster for AXDX than SOC for organism ID [2.7 (1.2) h vs. 15.6 (20.3) h, P < 0.001] and AST [13 (55.7) h vs. 54.6 (45.5) h, P < 0.001].ConclusionIn the largest trial to evaluate the clinical impact of a blood culture diagnostic for GNB BSI, we found that rapid organism ID and phenotypic AST led to faster changes in antibiotic therapy for Gram-negative bacteremia. DisclosuresRitu Banerjee, MD, PhD, Accelerate Diagnostics: Grant/Research Support; BioFire: Research Grant; Biomerieux: Research Grant; Roche: Research Grant Robin Patel, MD, ASM and IDSA: Other Financial or Material Support, Travel reimbursement, editor’s stipends; CD Diagnostics, Merck, Hutchison Biofilm Medical Solutions, Accelerate Diagnostics, ContraFect, TenNor Therapeutics Limited, Shionogi: Grant/Research Support; Curetis, Specific Technologies, NextGen Diagnostics, PathoQuest, Qvella: Consultant; NBME, Up-to-Date, the Infectious Diseases Board Review Course: Honorarium recipient, Other Financial or Material Support; Patent on Bordetella pertussis/parapertussis PCR issued, a patent on a device/method for sonication with royalties paid by Samsung to Mayo Clinic, and a patent on an anti-biofilm substance issued: Other Financial or Material Support, Patents.

487. Prevalence of Antimicrobial Resistance in Gram-Negative Bacilli Bloodstream Infections at a Tertiary Teaching Hospital in the Dominican Republic

BackgroundBloodstream infections (BSI) with gram-negative bacilli (GNB) are a major cause of morbidity and mortality worldwide. Sepsis due to BSI can carry a mortality rate as high as 40%, with higher mortality in developing nations. Early and appropriate empiric therapeutic selection plays an important role in survival. The rising incidence of antimicrobial resistance (AMR) limits empiric treatment options. Local susceptibility patterns can vary per region, institution or setting. Understanding local AMR may help guide empiric treatment choices. We seek to describe resistance rates for GNB BSI in the Dominican Republic (DR).MethodsThis is a retrospective review of antimicrobial susceptibility patterns from bloodstream infections in a tertiary hospital in the DR. Susceptibility data from all adult inpatient blood cultures were collected from January 1 to December 31, 2017.ResultsA total of 124 blood cultures were reported. The most common organisms were Escherichia coli (43%) and Klebsiella pneumoniae (23%). Fluoroquinolone resistance was present in 70% of E. coli. Phenotypic susceptibility patterns consistent with extended-spectrum β-lactamase (ESBL) producing GNB were present in 46% of isolates. Carbapenem resistance was found in 4 samples and was most common in P. aeruginosa. Susceptibility profile is described on Table 1.ConclusionAMR was high in GNB BSIs in the DR. High rates of ESBL render common cephalosporins sub-optimal for empiric treatment. PTZ retains in vitro susceptibilities despite cefepime resistance but clinical efficacy is controversial. CTX-M ESBLs may cause these resistance pattern in vitro. Further studies are needed to determine genetic mechanisms of resistance. Establishing antimicrobial stewardship programs with rapid diagnostic testing that identify mechanisms of resistance may promote judicious use of carbapenems and reduce further the risk of further development of AMR. Disclosures All authors: No reported disclosures.

2334. Risk Factors of Multidrug-Resistant Gram-Negative Bacterial Bloodstream Infections in Children’s Hospitals in Japan

BackgroundAlthough multidrug-resistant (MDR) Gram-negative bacilli (GNB) are a serious and growing concern worldwide, the epidemiological data on children are still limited. Our aim was to evaluate the risk factors for MDR GNB bloodstream infections (BSI) in children.MethodsPatients with GNB BSI were enrolled between April 2010 and March 2017 at eight children’s hospitals in Japan. Clinical and microbiological data were collected retrospectively. The 2012 criteria of the Centers for Disease Control and Prevention were used to define MDR. MDR and non-MDR GNB BSI were then compared in terms of risk factors.ResultsIn total, 629 GNB BSI cases were identified. The median age and proportion of males was 2 years (IQR 0.3–8.7) and 50.7%, respectively. Underlying diseases were found in 94% of the patients. The proportion of GNB BSI cases developing after >48 hours from admission was 76.2%. The most common GNB was Escherichia coli (29.3%, 184/629), followed by Klebsiella pneumoniae (19.7%, 124/629) and Pseudomonas aeruginosa (16.4%, 103/629). MDR comprised 24.5% (154/629) of cases. The MDR rate for E. coli, K. pneumoniae, and P. aeruginosa was 44.0% (81/184), 23.4% (29/124), and 16.5% (17/103), respectively. The coverage rate of the initial empiric therapy for the MDR and non-MDR GNB BSI cases was 60.4% and 83.4%, respectively (P < 0.001). The all-cause mortality rate at 28 days of GNB BSI was 10.7% (67/629), 13.6% (21/154), and 9.7% (46/475) for MDR- and non-MDR GNB BSI, respectively (P = 0.167). The all-cause mortality rate at 28 days was 10.4% (14/135) and 7.7% (27/351) for MDR and non-MDR Enterobacteriaceae BSI (P = 0.341) and 41.2% (7/17) and 18.6% (16/86) for MDR- and non-MDR P. aeruginosa BSI, respectively (P = 0.056). Multivariate logistic regression analysis showed that MDR GNB BSI was independently associated with anticarcinogenic drug use within 30 days (OR: 43.90; 95% CI: 4.69–411.08), older age (OR: 1.05; 95% CI: 1.02–1.09), and admission to the neonatology ward (OR 0.019; 95% CI: 0.005–0.076).ConclusionOne-fourth of GNB BSI cases were MDR. Anticarcinogenic drug use and older age were risk factors for MDR GNB BSI in children’s hospitals. MDR P. aeruginosa infections were associated with higher all-cause mortality.Disclosures All authors: No reported disclosures.

Pre-engraftment bloodstream infections in acute leukemia patients undergoing unrelated cord blood transplantation following intensified myeloablative conditioning without ATG.

The aim of this study is to investigate the impact of pre-engraftment bloodstream infections (BSIs) on the outcomes in acute leukemia patients undergoing myeloablative cord blood transplantation (CBT). A total of 226 acute leukemia patients who received unrelated CBT were enrolled in this study, and all these patients received an intensified myeloablative conditioning without ATG. Pre-engraftment BSIs occurred in 72 patients (31.9%), and the median time of onset was 4.5days after cord blood infusion, BSIs of gram-negative bacilli, and gram-positive cocci comprised of 63.8 and 36.2%, respectively. The cumulative incidences of neutrophil and platelet engraftment, acute or chronic graft versus host disease (GVHD) were comparable among the non-BSI, gram-negative bacilli BSI, and gram-positive cocci BSI groups. The cumulative incidence of transplant-related mortality (TRM), relapse, overall survival (OS), and disease-free survival (DFS) was similar between the non-BSI and the BSI groups. For subgroups analysis, TRM was lower in gram-positive cocci BSI patients compared with that of gram-negative bacilli BSI patients (8.3 vs 39.3%) (p = 0.01) (HR = 0.39, p = 0.034), and the 5-year OS was higher in gram-positive cocci BSI cohort (79.1 vs 44.2%) (p = 0.01) (HR = 0.36, p = 0.046). Our study demonstrated that, for acute leukemia patients who received CBT after myeloablative conditioning that omitted ATG, pre-engraftment BSI had no impact on engraftment, GVHD, TRM, relapse, and long-term survival. Due to the fact that gram-negative bacilli BSI was associated with poor outcomes compared with gram-positive cocci BSI, appropriate early empirical antimicrobial management strategies and better supportive care are required to decrease the gram-negative bacilli BSI-related mortality.

Consequences of avoiding β-lactams in patients with β-lactam allergies

The choice of empiric antibiotics for the treatment of gram-negative bacilli (GNB) bloodstream infections (BSIs) in patients presenting with a β-lactam (BL) allergy is often a difficult decision given that these agents are first-line treatment in many guidelines. We sought to compare rates of clinical failure between patients with a history of BL allergy who received either a BL or a non-β-lactam (NBL). Adult patients with a past medical history of BL allergy and receipt of antibiotics for treatment of a GNB BSI were included from 3 academic medical centers. Treatment groups were classified as BL or NBL groups based on the empiric antibiotics received. Clinical failure was assessed 72 to 96 hours after initiation of empiric antibiotics. Hypersensitivity reactions during receipt of antibiotic therapy for the index BSI were recorded. A total of 552 patients were included for analysis: 433 patients in the BL group and 119 patients in the NBL group. Clinical failure was higher in the NBL group compared with the BL group (38.7% vs 27.4%, P = .030). The most common cause of clinical failure was a temperature of greater than 38.0°C 72 to 96 hours after receipt of empiric antibiotics (NBL group: 22.7% vs BL group: 13.9%, P = .016). Hypersensitivity occurred in 16 (2.9%) of 552 patients. Thirteen (2.5%) of 552 patients experiencing hypersensitivity reactions were exposed to a BL during treatment for GNB BSI. Among patients with a BL allergy, use of BL antibiotics is associated with a lower rate of clinical failure. The low rate of hypersensitivity provides further evidence about the risk of cross-reactivity between BL classes. These results support the practice of using a BL from an alternative class for patients in need of gram-negative antibiotic coverage.

Concordance of Gastrointestinal Tract Colonization and Subsequent Bloodstream Infections With Gram-negative Bacilli in Very Low Birth Weight Infants in the Neonatal Intensive Care Unit

Gram-negative bacilli (GNB) cause as many as 20% of episodes of late-onset sepsis among very low birth weight (VLBW, birth weight < or =1500 g) infants in the neonatal intensive care unit. As the gastrointestinal (GI) tract can serve as a reservoir for GNB, we hypothesized that VLBW infants with prior GI tract colonization with gentamicin-susceptible GNB who developed bloodstream infections (BSI) would do so with gentamicin-susceptible GNB. A prospective cohort study of VLBW infants was performed in 2 level III neonatal intensive care units from September 2004 to October 2007. GI tract surveillance cultures were obtained weekly. Risk factors for GNB BSI and for GI tract colonization with GNB were assessed. Fifty-one (7.3%) of 698 subjects experienced 59 GNB BSIs of which 34 occurred by 6 weeks of life and 625 (90%) of 698 subjects were colonized with GNB. Overall, 25% of BSI and 16% of GI tract isolates were nonsusceptible to gentamicin and colonization with the same species and same gentamicin susceptibility profile preceded 98% of GNB BSIs. Vaginal delivery, birth weight < or =750 g, GI tract pathology, increased use of central venous catheters, use of vancomycin, mechanical ventilation, and H2 blockers/proton pump inhibitors were associated with GNB BSI. Vaginal delivery, birth weight >1000 g, and treatment with carbapenem agents were associated with GNB colonization. These data support the use of empiric gentamicin to treat late-onset sepsis in infants colonized with gentamicin-susceptible GNB. Targeted GI tract surveillance cultures of infants with specific risk factors during weeks 2 to 6 of life could be used to guide empiric therapy for late-onset sepsis.