The present prospective study showed that most patients who triggered the sepsis alert in our emergency department fulfilled both Sepsis-3 and Sepsis-2 criteria, and the rate of blood culture positivity was similar but poor in both Sepsis-3 and Sepsis-2.
In the present study, episodes fulfilling Sepsis-3 criteria had a bacteremia rate of 36.2% and episodes fulfilling Sepsis-2 criteria had similar bacteremia rate of 35.0%. One previous study reports 22% bacteremia in patients with Sepsis-3 , and in another study, 48.4% patients presenting with Sepsis-3 had positive blood cultures, however it was not stated if they consisted of only clinically relevant growth . At least in the case of the higher reported rate of 48.2% only patients admitted to the ICU were studied, and probably does not reflect the frequency of bacteremia in the vast majority of patients with sepsis in the ED. In addition, previous data on patients with Sepsis-2 has shown bacteremia rates of around 30% [7-9, 17].
In the present study, the bacteremia rate was generally higher than previously reported. The reason for this discrepancy is unknown, however it might be related to differences in study design. In the present study, patients were included prospectively in the ED and there were no patients that received intravenous antibiotic treatment immediately prior to blood culture sampling. In addition, all included episodes had six blood culture bottles compared to four bottles that were used in the previous studies. As it is well known that blood culture yield correlates with sampled volume [6, 18, 19], this may have increased sensitivity by increasing the blood culture sample volume. We also employed a previously published method [14, 15] to define clinically relevant growth and contaminant growth. Notably, previous studies in both Sepsis-3 and Sepsis-2 patients included mostly retrospectively identified patients with sepsis [7, 20] or did not specify whether contaminants were excluded in the blood culture positivity analysis [7, 9, 20, 21].
The microbiological findings in Sepsis-3 and Sepsis-2 positive episodes were similar regarding both clinically relevant pathogens, contaminants and microbiological spectrum. It is reasonable to suggest that the underlying reason for this might be the large overlap between the two groups in the present study. However, in the 88 episodes that had discrepant sepsis categorizations (Table 2), there were more episodes with clinically relevant growth and fewer contaminations in the Sepsis-3 positive group compared to the Sepsis-2 positive group. As the sample size of episodes with discrepant sepsis categorization was small, it was not possible to exclude a difference in blood culture results in this group.
The microorganisms implied in community acquired sepsis has previously been well studied, and the most common pathogens consist of E. coli, other Enterobacterales and S. aureus . In the present study, the rank order of microorganisms in Sepsis-3 positive episodes is consistent with the findings in previous studies. However, in episodes that did not fulfill Sepsis-3 criteria, Enterobacterales other than E. coli as well as S. aureus was not as commonly isolated as in the Sepsis-3 positive episodes. It is possible that this difference reflects the importance of certain microbial virulence factors and interplay with immunological mechanisms as a major cause of organ dysfunction in sepsis .
There is a significant overlap between patients fulfilling Sepsis-3 and Sepsis-2 criteria. In the present results, 80% of patients fulfilling criteria for either sepsis criteria also fulfilled the other criteria. This contrasts to a recently published large study by Engoren et al  which included 18,183 patients who had either Sepsis-3 or Sepsis-2, where only 6,841 (37.6%) fulfilled criteria for both definitions. Additionally, our results contrast with another study by Todorovic et al. in which 24% of patients fulfilled both definitions of sepsis . All patients included in the present study had a clinical suspicion of sepsis and were admitted to the ED at the time of inclusion. The discrepancy between the present results and previously published data is significant and may therefore be a result of patient selection criteria. In the previous study by Engoren et al, patients were screened retrospectively for sepsis, whereas in the present study, patients with suspected sepsis were included prospectively using a sepsis alert system. As trigger parameters for the sepsis alert system included components also present in SOFA (e.g. altered mental status, oxygen saturation) it is probable that the use of a sepsis alert system increased the likelihood that patients who fulfilled Sepsis-3 also fulfilled Sepsis-2 in our study.
The main strength of this study is its prospective design, using a clinical screening tool to select patients with suspected sepsis at presentation to the ED. The prospective design allowed for inclusion of patients with a “a priori” high likelihood of sepsis, regardless of the final diagnosis. The present results reflect the real-world scenario, where the final diagnosis is not available to the health care provider in the emergency department. Also, due to the prospective design, the blood culture sampling process was standardized, requiring 6 blood culture bottles to be sampled before initiation of antibiotic therapy in all patients included in the study.
The present study has several limitations. First, our study included only patients with a suspected infection upon presentation to the ED. This might introduce a bias in patient selection. However, the patients were carefully evaluated and represent a well-defined cohort for sepsis. Second, most patients included in the study had community acquired sepsis. However, the clinical criteria for Sepsis-3 and Sepsis-2 is general and applies for both community acquired sepsis and hospital acquired sepsis.