Neonatal sepsis is a feared condition, at the same time as consensus for diagnostics criteria are missing, the potential negative effects of unnecessary antibiotic treatment plentiful, and AMR is an upcoming crisis [2, 8, 16]. A cornerstone in a factful approach to AMR is knowledge of current spectrum of bacterial pathogens, as well as their AMR patterns. In this study, we have described incidence, morbidity, mortality, AMR patterns, and biomarkers of GPB sepsis in neonates to increase the knowledgebase for clinicians to use in diagnostics and treatment of neonatal sepsis.

The incidence of GPB sepsis in this study was 1.47/1000 LB, which is more than 4 times the incidence for GNB sepsis (0.35/1000 LB) in our area, but with a lower SFR (2.8% versus 16.8%) and CFR (9.5% versus 28.0%) [14]. This result is in line with previous studies showing a higher mortality rate for GNB sepsis compared to GPB sepsis [10, 26, 27]. Few studies have reported on GPB-sepsis separately, but our GPB-SFR in LOS is like that of a recent Norwegian study [27]. In EOS, most affected infants were born full-term and GBS was the most common pathogen, concordant to previous international and Swedish studies [1, 28,29,30]. However, the GBS-EOS incidence (0.18/1000 LB) in our study was lower than both earlier Swedish and international reports [28,29,30], but higher than a recent report from Poland [31]. Interestingly, there were no fatal cases of GBS sepsis in this study, something that contradicts an often-described mortality in neonatal GBS sepsis of 7–9% [1, 28]. This is presumably a result of an effective infection control and the Swedish risk factor based intrapartum antibiotic prophylactic treatment regimen, implemented in 2008. A decrease of GBS-EOS has been reported during the same timeframe [32].

Noticeable was that all SFR cases in this study were LOS and that 9/10 of them were born extremely preterm, the latter in line with previous findings [26]. CoNS, where most isolates came from infants born extremely- or very preterm, was the most frequent causing pathogen in LOS. This is consistent with previous studies [10, 26, 27]. The CFR for CoNS seen in our study was in line with previous studies [10, 33]. CoNS are commonly presented as a group, despite the individual bacteria’s varying virulence [3]. A distinction of individual CoNS could be of interest for future studies, to see any potential difference in outcome in neonatal sepsis caused by different bacteria within this group. It is important to consider the vulnerability of this extremely preterm born patient group, many with more than one serious condition, which could explain that we found a SFR significantly lower than the CFR for CoNS (3.1% vs. 11.4%).

Further, it is thought-provoking that we found an increasing trend in S. aureus sepsis over the study period (Fig. 3), similar to findings from the USA (neonatal) [34] and Europe (adults and children) [35], but in contrast to finding in Australia (neonatal) [36], Korea (neonatal) [26] and a recent international systematic review in adult patients [37]. The stable CoNS incidence indicates that the increasing numbers of S. aureus infections is unlikely to be related to breaches in infection control measures.

None of the S. aureus isolates in our study were MRSA in contrast to results from studies in the USA (28%) [38], Australia (26%) [36] and a study covering several low-and middle income countries (61%) [39]. However, an increase of MRSA-positive isolates (all sample types) has been seen in our region over the study period, according to the Public Health Institute of Sweden [40], though mostly from asymptomatic colonization of the bacteria. Besides, a high percentage of CoNS in this study were AMR (95.3%), which highlights the groups tendency to acquire AMR genes. Therefore, the risk of both current AMR as well as the risk of causing an increase in the antibiotic resistant gene-load in infants’ intestinal microbiota, should be taken into consideration when treating CoNS. For instance, 14% of the CoNS isolates were resistant to amikacin, a first line antibiotic commonly used in combination with cephalosporins or cloxacillin in LOS of unknown cause in our setting. Still, none of the CoNS isolates showed resistance to vancomycin, something that has been described from several countries [41, 42], which establishes vancomycin as first line treatment for culture-verified CoNS sepsis in neonates in our area. However, it is important to emphasize the limitation of vancomycin use to confirmed CoNS sepsis, to avoid contributing to the further development of AMR. Further, a recent Norwegian study found a wide variation in amounts of prescribed vancomycin in between sites, with no clear relation to the sepsis related mortality rates [43], which emphasis the need to use vancomycin restrictively. In addition, the WHO has classified vancomycin as “watch” in the AwaRe category. This classification indicates that vancomycin belongs to a group of antibiotics that should be used only for specific infections [16].

As noted, diagnostics of neonatal sepsis is challenging. Biomarkers are helpful, but our study showed that 9.8% of all GPB sepsis had no elevated CRP concentration, specifically apparent for infants in the lower GWs. This, together with a well described delayed increase in CRP-levels at the occurrence of an infection, makes CRP as a biomarker unreliable for early detection of neonatal sepsis [44]. It has been suggested that a combination of CRP and procalcitonin (PCT) testing could improve the diagnostic sensitivity as PCT-levels increase after just 6-hours [44, 45]. However, neither an increased CRP- nor an elevated PCT-level are specific for infections, as they can both be elevated by non-infectious factors as well [45]. Most infants with GPB born before GW 32 had a lowered PLT count (< 100). The variation in virulence levels among GPB, including CoNS, presents a complex scenario. Despite being GPB, CoNS exhibit properties that results in comparatively less inflammatory response than other GPB (Supplementary Figure S1 and S2). This distinction can be challenging to attribute solely to host factors, such as the immaturity of the host´s immune system in premature infants.

The retrospective design of this study is a limitation, due to the lack of possibility to control missing data or to re-test cultures for accuracy. Further, since we have not accounted for infants seeking care after being discharged to home, our incidence calculations could be falsely slightly underestimated, especially regarding GBS sepsis. Moreover, the vulnerability and co-morbidity of the study population cause for several possible confounders when it comes to interpreting symptoms of sepsis. Lastly, the high SFR and CFR that we found for mixed infections are difficult to interpret due to the low numbers, the fact that one of the bacteria could be a possible contaminant, and the unusual occurrence of several bloodstream bacteria at once. The population-based design and the availability to extract data from several different systems (patient records, the Swedish Neonatal Quality Register (SNQ), the microbiological laboratory at Karolinska University Hospital and from the Public Health Agency of Sweden) is a strength as it is less likely that we have missed culture positive GPB sepsis cases.