Background
Sepsis, caused by a dysregulated host response to infection [
1], is one of the most important conditions to identify within emergency care due to its high mortality and to a large extent treatable cause.
The mortality of severe sepsis (19–30%) [
2,
3] is more than three times higher than that of myocardial infarction (6–8%) [
4,
5], and rapid identification and therapy has traditionally been thought to be associated with improved outcome [
6‐
9]. However, the systematic review and meta-analysis by Sterling et al. [
10] questioned the benefit of early antibiotic treatment. Nevertheless, a recalculation by Yokee et al. questioned these conclusions [
11] and the recommendation of early antibiotic treatment remains a recommendation [
12].
Sepsis is a clinical diagnosis which can be defined as the presence of an infection in combination with two or more SIRS (Systemic Inflammatory Response Syndrome) criteria [
13,
14]. SIRS is in turn based mainly on vital parameters. However, 39% of the patients with serious infections lack abnormal vital parameters [
15] and 12% of the patients with severe sepsis do not fulfil the SIRS criteria [
16]. The inadequate sensitivity and specificity of the SIRS criteria has been a contributing factor to a recently suggested revision of the sepsis definition [
1]. Nevertheless, existing sepsis screening tools are still based mainly on SIRS criteria [
17,
18].
The diagnostic and prognostic significance of medical history is incompletely known regarding sepsis [
19]. Our hypothesis is that inclusion of variables related to septic patients’ reported symptom presentation may add value to a future screening tool.
The primary aim of the current study was to explore the presentations of adult septic patients in the prehospital setting as documented in EMS medical records and to identify and quantify keywords related to septic patients’ symptom presentation according to EMS documentation. The secondary aims were to compare keywords in relation to in-hospital mortality and the distribution of keywords in relation to age categories, survivors/ deceased and severe/ non-severe sepsis.
Discussion
The current study identified keywords related to septic patients’ presentation according to EMS documentation, using a mixed methods approach. The most frequently documented keywords related to patients’ symptom presentation were: abnormal, or suspected abnormal temperature, pain, acute altered mental status, weakness of the legs, breathing difficulties, loss of energy and gastrointestinal symptoms such as vomiting and diarrhoea.
Certain presentations were associated with increased in-hospital mortality and the distribution of keywords in relation to age categories, survivors/ deceased and severe/ non-severe sepsis varied.
Keywords related to symptom presentation are not included in the existing screening tools for sepsis identification within emergency care [
17,
18], which should be reconsidered. However, before this is done, prospective studies evaluating the sensitivity and specificity of these keywords needs to be evaluated.
Almost all patients that presented with the most common combined keyword; abnormal or suspected abnormal temperature had fever, while hypothermia was in general rare but more common among patients with severe sepsis (Table
2 and Additional file
7). Despite fever being frequently documented as a symptom in the EMS records; approximately one third of the patients lacked this finding. This observation is consistent with a previous study of bacteraemic ED patients by Lindvig et al. [
34], showing that 34.1% of bacteraemic patients had a normal temperature recorded at ED arrival.
Pain was frequently documented. The most common locations were the abdomen, extremity, back and urinary tract. The location often reflected the site of the original infection but general flu-like muscular pain was also common, in accordance to previous literature, describing diffuse pain as frequent [
19].
The combined keyword acute altered mental status, represents primary keywords ranging from altered behaviour to the deepest level of non-responsiveness and may reflect sepsis-associated encephalopathy (SAE) [
35,
36], known to affect up to 70% of patients with severe sepsis [
37]. It could be described by the patient in terms such as “feeling confused” or “feeling sleepy” or not remembering events in the last days, and by relatives as an observed disorientation, a lack of attention or an inability to verbally response [
19].
Weakness of the legs was another common symptom presentation. This has, to the best of our knowledge, not previously been described for septic patients in the prehospital setting. However, previous studies have indicated that sepsis induces a myopathy characterized by reduced muscle force-generating capacity, and loss of muscle mass [
38], and weakness of the legs is interpreted as an expression of this pathophysiology.
Breathing difficulties were frequently documented. Interestingly, only 39% of the patients with documented breathing difficulties had a pulmonary origin of the underlying infection, indicating that breathing difficulties are frequent in sepsis with a focus other than the lung. This may in turn suggest that the presentation of breathing difficulties is part of a systemic pathophysiological response to the underlying infection, which may include an anaerobic metabolism and metabolic acidosis.
In-hospital mortality varied in relation to the documented symptom presentation. The highest mortality rates were observed among patients with documentation of hypothermia, reduced urinary volumes and reduced intake of food or fluid. Interestingly, the mortality rate among patients presenting with these presentations exceeded that of patients presenting with keywords traditionally included in the definition of severe sepsis such as acute altered mental status. However, these findings need to be replicated in larger cohorts.
The documented presentations varied between age categories which may reflect a variation in the physiological response to an infection relating to age. However, it may also reflect that health care personnel direct their questions differently when encountering elderly patients, focusing on more basic functions e.g., food/fluid intake and whether they can stand and walk.
Finally, presentations differed between survivors and deceased. Known or suspected fever and shivering were more frequently documented among survivors which may indicate that these patterns reflect an appropriate immunological response or possibly a protective effect per se. This is consistent with previous studies demonstrating a decreased mortality in septic patients with moderate fever [
39].
Limitations
The analysis of sepsis presentation was based on EMS documentation which is associated with inherent restrictions. Documentation can be affected by many factors e.g., what EMS ask the patient, the patient’s ability to explain his/her experience and the presence of relatives who may or may not be able to describe the situation at hand. It is, as described above, not always possible to discern the origin of the documented information. The EMS records present the symptoms as documented by EMS personnel. To perform open interviews with septic patients would be an alternative approach to explore sepsis symptom presentation. However, interviews in the ambulance would be difficult to perform for logistical reasons. In addition there would be a bias towards less sick patients due to the most sick septic patients being unable to participate in an interview. Moreover, EMS personnel have been shown to have difficulties identifying septic patients [
25], which would lead to an inclusion bias. Furthermore, a third of the septic patients present with altered mental status, which would impair their ability to participate in interviews in the acute setting and affect their recall if the interview would be performed in retrospect.
Since the EMS records are brief and often lack detail, there is a risk that not all possible keywords are documented. However, even if the true prevalence of various keywords is expected to be higher than that documented, the relative proportions between the keywords are assumed to be similar.
Inclusion based on ICD codes has been used in several previous studies [
27,
40] and is the only reasonable way for database searches, but can be questioned as it is well known that diagnostic coding is a problem [
41], and consistently underestimates the incidence [
42]. Hence, assumedly more patients with sepsis were admitted by EMS and cared for in-hospital during the study period but discharged with ICD codes other than those compatible with sepsis, e.g., those indicating the focus of infection i.e., pneumonia or meningitis instead of sepsis. Inclusion by the means of ICD code could potentially entail a selection of more sick patients, i.e., a higher proportion of severe sepsis as well as patients with symptoms more typical of the common picture of sepsis, e.g., fever and hemodynamic instability since these patients may be more readily identified in the clinical setting. Hence, the inclusion based on ICD codes may limit the generalizability or transferability [
22] (the corresponding term within qualitative research) of our results to all possible septic patients.
Furthermore, the creation of keywords may have been influenced by the preconceptions of the authors, which is inherent in all qualitative analyzes, and the creation of exclusive subcategories was sometimes difficult as many of the complaints resembled each other. The authors have different backgrounds and met regularly to ensure trustworthiness and a consistent approach to analysis of the data.
The mixed methods approach [
20,
21], starting with an inductive content analysis [
43,
44], is used to explore previously unstudied areas. Hence, the current study should be viewed as the necessary first step in upcoming studies aiming to identify parameters with a high predictive value with respect to sepsis identification. As a first step it was necessary to identify keywords which could be analyzed in prospective studies and compared between septic and non-septic patients. The keywords in the current study are most likely not pathognomonic for sepsis. Moreover, it is unlikely that there are unique keywords pathognomonic for sepsis as the presentation is so diverse, but we do believe in the predictive value of combinations of keywords related to presentation and possibly together with other parameters measurable in the ambulance.
The frequency of keywords was sometimes associated with broad 95% Confidence Intervals (CIs). Especially in the subgroup analyses presented in Additional files
5,
6 and
7 the CIs indicate that larger study samples would be required for an increased precision.
Finally, the identified keywords were those documented by EMS within a cohort of septic patients admitted to the ED of Södersjukhuset by EMS and discharged with ICD code sepsis. It is possible that septic patients discharged with more organ specific ICD codes / arriving by other means than EMS, as well as septic patients in other settings could present with other symptoms. Hence the results of the current study may not be generalizable / transferable to other settings. Prospective studies are needed to analyze whether the identified keywords are representative for septic patients in general and to understand their predictive value. We suggest that it is necessary to include keywords in sepsis screening tools, however, which specific keywords or combinations thereof remain to be studied.
Acknowledgements
Many thanks to Stefan Solbrand for help with the data acquisition and to Ella Danielson and Kenneth Asplund for valuable discussions regarding the content analysis of the current study.