Keywords reflecting sepsis presentation based on mode of emergency department arrival: a retrospective cross-sectional study

This was a retrospective cross-sectional study of 479 adult patients admitted to the ED of Södersjukhuset between January 1, and December 31, 2013, and discharged with an International Classification of Disease, tenth revision (ICD-10) code compatible with sepsis.

The ED of Södersjukhuset is an urban 603-bed teaching hospital in Stockholm, Sweden, with more than 129,000 adult ED visits annually [9].

Sepsis was defined as discharge from in-hospital care with an ICD-10 code compatible with sepsis (see the “Inclusion criteria” section above). The study material was collected during the period when the Sepsis-2 criteria [11] were in use.

Severe sepsis was defined, in accordance to a previously developed definition adapted to emergency care [12], see Additional file 1.

The EMS group was defined as patients arriving to the ED by EMS, i.e., ambulance or ambulance helicopter. The non-EMS group was defined as patients arriving to the ED by any means of transportation other than EMS, i.e., walking, by private car/taxi, or police transport.

Primary keywords were defined as symptoms or factors that describe the septic patient’s clinical presentation in the ED setting, e.g., “vomiting,” “fever,” or “hypotension.” The combined keywords consist of several primary and/or combined keywords. The primary and combined keywords were those identified in a previous study among Swedish EMS patients [5] with the addition of any new keyword describing the patient’s clinical presentation in the ED, i.e., not previously identified in the prior prehospital study [3].

Keywords reflecting clinical presentation were quantified, both as primary keywords and as combined keywords within two subgroups of patients based on mode of arrival: by EMS or by means of transportation other than EMS.

ED and in-hospital records were reviewed (TakeCare® v. 18.3.10 CompuGroup Medical Stockholm Sweden) and analyzed with respect to prevalence of keywords mode of arrival and demographics (age gender and sepsis severity)

The chief complaint, current history, and preliminary assessment sections of the ED charts were analyzed for the quantification of keywords. The current history section focuses on the acute situation and not a comprehensive medical history. Only symptoms with new onset, defined as onset within 3 weeks prior to ED arrival, were considered relevant for the current study. Each patient could fulfill multiple primary and several different combined keywords.

P values are presented without adjustment for multiple comparisons and the Bonferroni-adjusted significance level is described in the footnote of each table by dividing the significance level 0.05 with the number of performed comparisons. Only two tailed P values that remained significant according to the Bonferroni-adjusted significance level are considered significant in the current study.

Stockholm Regional Ethical Review Board approval was obtained for this study (reference number 2012/1288-31/3).

See Fig. 1 for flow chart for inclusion and exclusion of patients.

A total of 479 patients were included in current study: 357 EMS patients and 122 non-EMS patients.

Characteristics of the patients are presented in Table 1. The median age in the entire sample of septic ED patients was 75 years (Interquartile range, IQR 61-85), 255/479 (53.2%) were men and 248/459 (54.0%) with sufficient documentation to determine severity grade had severe sepsis. The overall in-hospital mortality was 99/479 (20.7%).

The median age in the EMS group was 15.5 years older than in the non-EMS group (78.0 years; IQR 68-86 vs 62.5 years; IQR 39-73, p value < 0.001). Patients in the EMS group had a significantly higher prevalence of severe sepsis than patients in the non-EMS group (218/346; 63.0% vs 30/113; 26.5%, p value < 0.001) and a higher in-hospital mortality rate (94/357; 26.3% vs 5/122; 4.1%, p value < 0.001). There was no significant difference with respect to gender between the patients arriving by EMS vs non-EMS, see Table 1.

Ninety primary keywords and 14 combined keywords describing the presentation of septic patients were extracted from ED charts, see Additional files 2 and 3.

The most common combined keyword among both patients arriving by EMS and by non-EMS was “abnormal, or suspected abnormal temperature.” Both fever and hypothermia have previously been described as symptoms related to the septic patient [3, 13]. However, despite “abnormal, or suspected abnormal temperature” being the most prevalent keyword in the current study, approximately one-third of the included patients did not present with fever. This finding is consistent with previous studies [3, 14].

“Temporal deterioration” represents an acute change in the patient’s habitual state, but it does not describe the details of this change. The prevalence of temporal deterioration was high both among EMS and non-EMS patients. These findings are consistent with those demonstrated by Bohm et al. in emergency calls involving septic patients [15].

Vomiting and diarrhea were common among both patients arriving by EMS and by non-EMS. The high frequency of “gastrointestinal symptoms” in the current study is supported by previous studies [3, 13].

The keywords “abnormal skin,” “abnormal urination,” and “loss of energy” were also present among patients arriving by both EMS and by non-EMS to a similar extent.

“Abnormal circulation,” “abnormal breathing,” “acute altered mental status,” and “decreased mobility” were all significantly more common among patients arriving by EMS. The first three are directly connected to the former criteria for severe sepsis [11] and the findings most likely reflect that EMS patients are more severely ill than non-EMS patients, which is also supported by previous studies [8].

“Pain” is an unspecific symptom and the single most common chief complaint among ED patients in general [16]. That pain was more common among non-EMS patients could potentially be a consequence of the increased prevalence of acute altered mental status among patients arriving by EMS, in turn impairing the patient’s capability to express pain. Furthermore, keywords reflecting abnormal vital signs were more common among patients arriving by EMS patients and abnormal vital signs may render more attention among emergency care providers as compared to symptoms, e.g., pain.

The combined keyword “risk factors for sepsis” includes several primary keywords such as ongoing or recent infection/invasive procedures/immunosuppressive treatment but did not include traditional risk factors for sepsis such as age and pre-existing comorbidity. However, traditional risk factors should be considered in a future screening tool. The current study included components registered in the chief complaint, current history, and preliminary assessment sections of the ED charts, where the current history section focuses on the acute situation and not a comprehensive medical history, e.g., pre-existing comorbidity. Patients with “risk factors for sepsis,” as defined in the current study, are often informed to seek medical attention if they deteriorate and may therefore seek medical attention at an earlier stage and hence need an ambulance to a lesser extent. In addition, patients with “risk factors for sepsis” have previously been shown to be younger [3] which may affect the capacity to use means of transportation other than the ambulance, but these speculations remain to be investigated.

Several factors are thought to contribute to the observed differences in keyword prevalence based on mode of arrival. We do not believe that the arrival mode per se exhibits causality on the presentation but rather that the observed differences reflect that patients arriving by EMS are older and have a higher prevalence of severe sepsis as compared with patients arriving by other means. This is supported by the stratification analyses showing that “abnormal circulation,” “acute altered mental status,” and “decreased mobility” did not remain significantly more frequent among EMS patients and “risk factors for sepsis” did not remain significantly more frequent among non-EMS patients when stratified for sepsis severity, indicating that differences in sepsis severity accounts for the observed differences for these keywords between EMS vs non-EMS patients.

The observed differences between septic EMS vs non-EMS patients support the principle of adapting sepsis screening tools to the population where they are planned to be implemented; in this case, the EMS vs the ED, and can be exemplified as a screening tool applied in the EMS may include variables, e.g., “abnormal circulation,” “abnormal breathing,” “acute altered mental status,” and “decreased mobility,” while a screening tool proposed to be used for “walk-in” ED patients may include other variables such as “pain” and “risk factors for sepsis.”

Nevertheless, despite these observed differences, we would like to emphasize that most keywords demonstrated a similar distribution regardless of mode of arrival and the most prevalent keywords related to sepsis presentation in the current study confirm prior results from the prehospital [3, 15] and ED settings [13].

There are several limitations to the current study.

The definition of sepsis based on ICD-code can be questioned, as it is well known that identification of septic patients based on ICD codes leads to an underestimation of the true number of septic patients [17] and hence, there is a risk that the study sample may not be representative of all ED patients with sepsis. Inclusion based on ICD-code sepsis may involve a selection of the most severely ill septic patients. However, the method has been used in large epidemiological studies of sepsis [18] and is the only realistic method for larger registry studies.

Furthermore, the current study was a retrospective study with the inherent limitations of missing data. The prevalence of keywords was based on documented observations in ED records. Documentation may depend on various factors such as inter-individual variation among ED personnel and ED workload. Many of the keywords represent symptoms, and the identification of symptoms requires a thorough history taking by the ED doctor in addition to a communicable patient.

Multiple comparisons were performed, with the inherent risk of inferring type I errors. Therefore, the level of significance was adjusted by applying a Bonferroni correction. However, the Bonferroni correction is excessively strict with the inherent risk of inferring type II errors [19]. Hence, although erring on the side of caution, this may have resulted in true differences being regarded as non-significant.

Furthermore, when comparing the prevalence of especially primary keywords, the number of patients in the compared groups was few and hence the results must be interpreted with caution and need to be confirmed in larger samples.

Finally, the current study is a single center study which may limit the generalizability of the results. However, patients were included over a period of 1 year which enables the seasonal variation of sepsis to be accounted for. In addition, the study setting was the largest ED in Scandinavia at the time [9].