Our study identified a total of 9076 patients with Gram-negative BSI over a five-year period (2016–2020) in six German university hospitals, with E. coli (59.6%) being the most prevalent pathogen, followed by Klebsiella spp. (23.7%), reflecting trends observed both nationally and across Europe [8]. A. baumannii BSI incidence was very low in the 5-year study period (31 cases). E. coli and Klebsiella spp. as very relevant pathogens are long known for their clinical relevance and have been recently highlighted as two of the most relevant pathogens responsible for the "burden of infection" [9], which is supported by other studies [10].

The increasing trend of AMR in our study echoes the broader situation in Germany and across Europe, where AMR remains a major public health challenge [8, 11]. Our detailed analysis contributes specific insights into the German context, offering valuable data for shaping national and regional antimicrobial stewardship and infection control policies.

Our data show rather stable 3GCR rates for E. coli and Klebsiella spp., contrasting with global concerns about the spread of AMR [12]. The incidence of carbapenem resistance remains relatively low in our study with 0.4% highest for E. coli and 0.3% for Klebsiella spp. (Table 3) being different from findings from other European studies. Nevertheless the presence of 3GCREB raises alarms, echoing the concerns highlighted by both German and European health authorities about the escalating challenge of AMR [8]. For our E. coli and Klebsiella spp. it can be assumed that 3GCREB is mostly due to ESBLs and less frequently due to AmpCs [13, 14], though specific tests for ESBL and AmpC were not part of the study design. A systematic review and meta-analysis reported a steady increase in the prevalence of ESBL-producing Enterobacterales in European countries, with significant variability between countries highlighting the emergence of community-acquired BSI caused by ESBL-producing Enterobacterales which expands the threat of these organisms beyond hospital settings, necessitating broader surveillance and containment strategies [15].

The relationship between antimicrobial resistance and clinical outcomes is a key concern in managing BSI. In our analysis the 3GCR phenotype in E. coli or Klebsiella spp. BSI was not a risk factor for adverse outcomes. These statements are central to our work and are based on the subanalysis carried out without taking into account the empirical and specific antibiotic therapy used and also without 6-month follow-up. While several studies have indicated that AMR is associated with higher mortality rates, longer hospital stays, and increased healthcare costs, mostly without taking into account specific regional differences or the impact of antimicrobial stewardship programmes. [16], other studies did not find an increased short- or long-term mortality in E. coli BSI either, particularly for BSI with predominance of community-acquisition like in our cohort [17]. Although less frequent, BSI due to ESBL-positive K. pneumoniae carried a worse prognosis in another German study including a higher in-hospital mortality [18]. A meta-analysis showed that bacteremia with ESBL-producing Enterobacterales was associated with a higher mortality compared with bacteremia due to non-ESBL producers [19]. In a BSI study due to Enterobacterales at ten European hospitals the 3GCR phenotype increased the hazard of death compared with third-generation cephalosporin susceptibility (adjusted hazard ratio 1.63; 95% CI 1.13–2.35) [20]. An analysis that included non-BSI infections, the estimated number of attributable deaths due to infections with 3GCREB E. coli and Klebsiella spp. increased more than four-fold [21]. However, in patients with strict community-onset bacteremia, this association is less well established, with studies producing conflicting results [22,23,24]. Differences in mortality might, however, partly be explained by heterogeneous studies, including/excluding polymicrobial bacteremia, different patient features, different circulating regional strains and treatment strategies. According to some work, the increased risk of death has been related to delays in the initiation of effective antimicrobial therapy [25]. Nevertheless, the impact of an adequate empirical antibiotic treatment is of course high and has to be considered [26]. For example, others found a higher crude 14 day mortality for bacteraemic patients due to ESBL producers, but the association disappeared when adjusting for inappropriate antibiotic therapy [24]. Data on empirical and targeted antimicrobial treatment were not available for the present analysis and has to be investigated by a different study design (case–control study), so that the interpretation has to be drawn with caution.

In other works, further risk factors of ESBL BSI were identified like obstructive urinary tract disease, previous surgical history and the prior use of a cephalosporin antibiotic within 3 months [27]. Our findings further demonstrated that patients with a higher PBS faced worse outcomes, a relationship that has been well-documented. Specifically, we observed that a score ≥ 4 was associated with significantly increased odds of death (OR = 8.86, 95% CI: 2.92–26.87) but not discharge with sequelae, underscoring the utility of this score in predicting BSI mortality and morbidity.

The high incidence of BSI observed in haematology/oncology departments underscores the unique vulnerability of this patient population, primarily due to immunosuppression caused by both the malignancies themselves and the intensive chemotherapy regimens required for treatment. Interestingly resistance rates to piperacillin also slightly decreased over time (Table 3) without notable changes in therapeutic regimens to our knowledge. Patients in haematology/oncology were in general more susceptible to invasive infections, including those caused by multi-drug resistant organisms, due to their compromised immune systems and frequent use of invasive devices, such as central venous catheters [28,29,30].

Strengths of the current study include the prospective study design, the stratification by department and the high number of cases which allows to highlight distinct epidemiological aspects of Gram-negative BSI in ICU and non-ICU settings over the study period. The multicentre analysis allows finding general trends over several years. Our study is not without limitations. The study did not start at the beginning of 2016, but Q4 2016 and ended in Q2 2020, i.e. the analysis does not include complete years which may lead to uncertainties of measurements. A challenge in evaluating the datasets was merging the data from two different databases (R-NET and BLOOMY cohort studies). Additional clinical data, e.g., information on empiric and definitive as well as appropriate antibiotic therapy, is missing. This is an important factor, especially in infections with multidrug-resistant organisms, where delayed or inappropriate therapy can lead to worse outcomes. Due to the study design this data is not available and can therefore not be included in the analysis. Our data exclusively reflect the epidemiological scenario in tertiary care hospitals which may overestimate the incidence of AMR compared to hospitals of primary or secondary care level and limits generalizability of our data.