Sepsis is recognized as a dysregulation of the inflammatory response in patients with infection. Progression to severe forms with organ dysfunction develops in one out of three patients, commonly resulting in long-term hospitalization and death [1, 2]. Algorithms for identification and management of severe infection hence emphasize recordings of vital signs and laboratory data, aiming at discovering circulatory failure and other organ dysfunctions at an early stage [3, 4]. In recent years, one of the main focuses in sepsis related research has been on candidate biomarkers in the host and their possible role in targeted therapy. However, definite novel therapeutic approaches have not been established and optimized anti-infective therapy is still the mainstay of treatment in severe sepsis. Delayed administration of the initial dose of antimicrobial agents, the adequacy of antimicrobial therapy and, where applicable, delayed surgical source control, are all independent prognostic factors [5–8]. Care bundles with guidance on how to diagnose and handle affected patients, including up-to-date recommendations on empirical treatment, therefore emerges as one of the most important measures to improve outcomes in sepsis. In 2007, Llewelyn and Cohen addressed the relevance of monitoring the microbial epidemiology of sepsis on a local basis [9]. Based on such knowledge, empirical recommendations may be customized. Accordingly, we established the present study of adult community acquired severe sepsis to describe the occurrence, characteristics and handling of affected patients admitted to our hospital. Secondary objectives were to evaluate our physicians’ compliance with local guidelines and to identify potential predictors of in-hospital mortality in severe sepsis.

We estimated the annual incidence of community acquired severe sepsis in our hospital to be 2.2/1000 admissions and 0.5/1000 inhabitants. In-hospital mortality was 25% and the 28-day all-cause mortality rate was 13/100 000 inhabitants per year.

Previous data on the occurrence of sepsis in Norway is limited to a retrospective study using data from the Norwegian Patient Registry, in which the total incidence of severe sepsis was calculated to 3.0/1000 admissions and 0.47/1000 inhabitants [10]. In contrast to this study, we did not include nosocomial sepsis which constitutes approximately 50% of all cases [5, 6, 11–14]. During three months prior to the present study we performed a prospective pilot survey of case-defined sepsis in our ED. Subsequent comparison with cases identified retrospectively by discharge ICD-10 codes revealed that 50% of all sepsis-cases were missed when using the retrospective method alone (Nygård, unpublished results). Thus, the calculation in our current prospective study probably represents a more accurate estimation of the Norwegian incidence of community acquired severe sepsis than previous results. Previous prospective studies with systematic inclusion from both ICUs and non-ICUs, reporting details on the occurrence of severe sepsis, are to our knowledge limited to a single survey from Spain [15]. They calculated an annual incidence of 1.0/1000 inhabitants, including nosocomial cases. Other studies of community acquired severe sepsis have used different methods for patient inclusion, and many do not offer data on incidence or are not applicable for such purpose [16–22].

The respiratory tract was the most frequent origin of infection in our patients. This is consistent with results from other studies [5, 11–14, 23–25]. We diagnosed abdominal infections less frequently, while GUI and STI were found more often than in many previous reports [6, 11–13, 25–27]. However, studies with inclusion of patients treated outside ICUs have found a distribution of diagnoses more similar to ours [15, 16, 23, 28, 29]. Patients with abdominal infections have been shown to have a high demand for intensive care treatment [11]. There are in addition more abdominal infections in nosocomial compared to community acquired sepsis [15, 30]. In our study most patients with this diagnosis were treated in an ICU, whereas patients with GUI on the contrary mainly were treated outside ICUs. Moreover, patients with abdominal infection had a prolonged length of stay (data not shown). In many studies on sepsis epidemiology, frequencies are estimated on the basis of prevalence data from ICUs only. Together, these observations suggest that distribution of various infections is influenced by study design, e.g. our low occurrence of abdominal infections is likely conditional to inclusion of patients from outside ICUs.

A likely pathogen has been found in 60-75% of eligible subjects in other studies on sepsis [5, 6, 13, 19, 24, 31], comprising positive blood cultures within the range of 22-37% [5, 6, 12, 19, 30–33]. This is comparable to our data. In line with related observations, we identified more Gram-positive than Gram-negative bacteria [1, 12, 24]. Antigen- and toxin tests were included in our analysis of microbiological aetiology. A positive urinary antigen assay was the only laboratory documentation in 48% of subjects with S. pneumoniae. This is in accordance with a prospective report on community acquired pneumonia, where 44% of 171 cases with S. pneumoniae were diagnosed by urinary antigen detection alone [34]. In Norway, guidelines for pneumococcal vaccination of adults ≥ 65 years and pre-defined younger persons at risk have been established since 1996, but vaccine coverage has not been satisfactory. However, a pneumococcal conjugated vaccine has been a part of the routine vaccination program for children with nearly completely coverage since 2006. From this point there has been a marked decline in the number of cases with invasive S. pneumoniae, also among adults. Thus, the proportion of patients with severe sepsis originating from this microbe is probably decreasing.

The proportion of patients with detection of a plausible pathogen differed among the various origins of infection in our study. The detection rate was low in RTI and abdominal infection in particular. This is also previously documented [20, 31]. Patients with negative microbiological samples had a significant greater risk of death than patients with positive samples. Although not significant in all categories, this result is consistent when stratification of infection categories is performed, supporting that the increased risk of death in cases with negative samples was retained throughout our multivariate model. Detection of a plausible pathogen gives the opportunity to administer validated adequate antimicrobial treatment. In subjects with negative microbiological samples, correct antimicrobial therapy can be delayed or not provided. An increased risk of death has been demonstrated in patients with severe infection receiving inadequate antimicrobial therapy [6, 7, 35–37]. In many of these studies multiresistant isolates were often encountered. We identified no isolates of MRSA, MRSE or VRE, and only two cases of ESBL-producing Gram-negative bacilli (of which one was even suspected, and treated accordingly, at admission). On the contrary, S. pneumoniae and E. coli were frequently identified. The CFRs in these two categories were low; 7 and 18.5% respectively (data not shown). Hence, a possible contributor to the different risk of death among patients with positive versus negative microbiological samples might be our high frequency of microbes with unproblematic resistance properties. Inability to tailor treatment with effective antimicrobial agents might also have influenced outcomes of the patients which received pre-hospital antimicrobial agents. In a study by Garnacho-Montero et al., previous antibiotic therapy within the last month was independently related to administration of inadequate antimicrobial therapy [6].

We were able to demonstrate a correlation between correct use of empirical antimicrobial agents and susceptibility in detected pathogens. It is therefore of concern that one in five patients did not receive the recommended regimen. One possible explanation for this is that sepsis is diagnosed on the basis of unspecific criteria, independent of primary focus, microbiological aetiology and host factors. Current diagnostic algorithms focus on identifying complications of an infection rather than its aetiology. The level of precision in establishing the focus of infection was variable and often low in our study. Only 17 of 38 patients with confirmed abdominal infection or endocarditis were assigned with the correct focus, and half of them received initial treatment as recommended. The longest delays before initiation of antimicrobial therapy were found in these two groups, and their hospital mortality was high. An evaluation of the quality of clinical diagnostic practice in severe sepsis, comparing the suspected focus at admission with a confirmed diagnosis at discharge, has to our knowledge not been published previously. We identified severe sepsis in nine subjects not suspected with infection at admission. Opposite, a study of the aetiology of illness in suspected severe sepsis found that 18% of the patients had noninfectious diagnoses mimicking sepsis [20]. Seen together these observations suggest a two-sided limitation in commonly used sepsis algorithms.

We found that a six hour delay or more in administration of the initial dose of antimicrobial treatment was associated with an increased risk of death, but could not demonstrate independent impact on mortality during the preceding hours, as reported by Kumar et al. in a study of patients with septic shock [5]. Other studies investigating the impact of early administration of antimicrobial therapy have also failed to demonstrate an hourly decrease in survival [36, 38]. Some have demonstrated beneficial effects on survival with administration of antimicrobial therapy within the first hour [36, 39]. Kumar et al. limited their inclusion to cases given effective antimicrobial therapy. Likewise, the beneficial effect found by Gaieski et al. was significant only when antimicrobial therapy was considered appropriate [36]. Analyses of our data with the same limitations as Kumar and Gaieski resulted in significant impact of hourly increasing time to administration of antimicrobial agents on mortality in multivariate analysis. However, the hourly effect was low. Since we have included a broad selection of the population with severe sepsis, ranging from septic shock to cases with other and less severe organ dysfunctions, our results concerning timing is inevitably influenced by the different levels of severity in our population. We were not able to severity stratify our patients and cannot investigate this matter any further.

In patients ≥ 75 years, antimicrobial therapy was less in compliance with current recommendations and more delayed. An age dependent risk of in-hospital mortality has been demonstrated in severe infection [40]. Following the results in our study, we question whether this is solely caused by host factors. Increasing age did not emerge as an independent risk factor. Subjects ≥ 75 years had on average one additional comorbidity, a significant higher presence of cognitive impairment and a significant higher creatinine level at admission than younger patients (data not shown). These data indicates that there were more potentially complicating factors among elderly patients. However, our Hospital guidelines are clear in terms of instructions on adjusting the doses of antimicrobial agents when needed, as well as recommending alternative treatment if the primary choice of drug is contraindicated. This was taken into account when we evaluated the level of compliance. Hence, we consider that there is room for improvements in the handling of our elderly patients, especially given the small difference in mortality after hospital discharge between patients aged 50–75 versus ≥75 years during the long-term follow-up (Figure 2).

We have found a high incidence of community acquired severe sepsis in a Norwegian university hospital. Initial clinical aetiology was often revised and the diagnosis sometimes overlooked in the emergency department. Adequate antimicrobial therapy improved outcome, while undefined microbiological aetiology, endocarditis, abdominal infections and delayed administration of antimicrobial agents increased the risk of death. A need for improved handling of elderly patients was identified. Our results emphasizes that quick identification of correct source of infection, proper sampling for microbiological analyses, and fast administration of adequate antimicrobial agents are crucial points in the management of severe sepsis.