Sepsis leads to high morbidity and mortality (1). Although numerous studies have explored the relationship between various patient characteristics and mortality, presenting symptoms outside of fever have received comparatively little consideration. Presenting symptoms may be important to consider for several reasons: 1) they may influence initial sepsis recognition and timing of antibiotic administration and other important therapies; 2) they may represent underlying variation in disease pathophysiology; and 3) they may have associations with outcome as a result of one or both of these reasons.
To address the significance of presenting symptoms in sepsis, we studied two cohorts of patients presenting to our emergency department (ED) with septic shock: those with “explicit” infectious symptoms and those with “vague” symptoms. We hypothesized that patients with vague symptoms of infection would have higher in-hospital mortality. We postulated a mechanism whereby vague symptoms lead to delayed recognition, delayed antibiotic administration, and therefore increased mortality. We also explored the prevalence of individual symptoms and their relationship to mortality.
We conducted a retrospective cohort study of all adult (age, ≥ 18 yr old) ED patients at a large, urban academic hospital over a 2-year period from April 1, 2014, to March 31, 2016. The study protocol was approved by our Institutional Review Board with a waiver of informed consent.
Inclusion and Exclusion Criteria
We included patients for analysis who met the following definition for septic shock, adapted from the current Centers for Medicare and Medicaid Services (CMS) Severe Sepsis/Septic Shock Early Management Bundle (SEP-1) definition (2): 1) a hospital International Classification of Diseases, 9th Edition, diagnosis code for sepsis, and confirmed source of infection or high suspicion for infection upon hospital admission based on chart review of ED and hospital admission notes; 2) presence of two or more systemic inflammatory response syndrome criteria while in the ED; and 3) persistent hypotension in the ED (systolic blood pressure [SBP], < 90 mm Hg on at least two measurements), elevated lactate greater than 4.0 mmol/dL, or initiation of vasopressor infusion while in the ED. We excluded patients transferred from an outside facility after having received fluid or vasopressor resuscitation or broad-spectrum antibiotics, and those deemed not eligible for aggressive care.
Data Collection and Definition of Covariates
Data were abstracted from the electronic medical record system via both electronic query and chart review. Time-to-appropriate antibiotic was defined as time elapsed from hypoperfusion, defined as first hypotension recorded (SBP, < 90 mm Hg) or high lactate resulted (> 2.0 mmol/dL) to administration of broad-spectrum antibiotics (2). Broad-spectrum antibiotics were defined as per CMS guidelines (i.e., a single broad-spectrum agent, or both antibiotics from an approved combination therapy regimen) (2). Additionally, a “nonapproved” antibiotic was considered appropriate if it was indicated for a specific known organism at the time of administration. We also reported time-to-antibiotics from triage.
The initial Sequential Organ Failure Assessment (SOFA) score (3) was calculated using component inputs documented during the patient’s ED stay. We also calculated the weighted Charlson Comorbidity Score (4). Triage location was dichotomized into acute versus nonacute areas, where the acute area of the ED has the highest level of resources. The “Sepsis Flag” is part of a protocol introduced before the period of study whereby ED providers trigger a color-coded warning on the electronic track board indicating suspicion for possible sepsis. The protocol included trigger criteria (possible infection, risk factors, and SBP < 100 or shock index ≥ 1) and was linked to prioritizing care, obtaining diagnostics, and considering antibiotics and fluid bolus within 1 hour (5).
Explicit and Vague Presenting Symptoms
Presenting symptoms were abstracted from nursing triage, ED physician, resident, and/or physician assistant’s notes. We developed an a priori definition of explicit presenting symptoms, as those we thought would immediately lead the clinician to consider infection, for example, apparent enough to trigger the sepsis alert (5). Symptoms were considered explicit if they included fever, chills, or rigors, cough with productive sputum, dysuria, reported skin redness or concern for soft-tissue infection, or referral for specific infection diagnosis. Additionally, measured temperature greater than or equal to 100.4°F at triage was included as explicit given its likely influence on the treating clinician to immediately consider infection. Presenting symptoms were defined as vague if they did not include any of the explicit symptoms listed above, thus making infection less apparent. For example, presenting symptoms of fatigue, weakness, and abdominal pain without fever were considered vague. Vague presenting symptom complex was the primary predictor of interest.
The primary outcome was in-hospital mortality.
Summary statistics, including in-hospital mortality, were calculated for all subjects during the 2-year study period and reported by vague versus explicit symptom cohorts.
Multivariate logistic regression was used to determine the adjusted association between vague symptoms and in-hospital mortality. With in-hospital mortality as the dependent variable, we included the primary exposure and all potential confounders, based on a priori knowledge, into a forward-selection logistic regression model (for all candidate covariates included, see the online supplement, Supplemental Digital Content 1, http://links.lww.com/CCM/D751). We set the covariate stay criteria to p less than or equal to 0.10 for significance. We forced the term time-to-appropriate antibiotic from first documented hypoperfusion into the model, given our particular interest in its confounding effect on the relationship between vague symptoms and mortality. We examined the univariate effect of time-to-antibiotics on mortality and its effect in multivariate modeling with and without vague symptoms. Lactate value was dichotomized using a cut-off of 4.0 mmol/dL. We calculated the C-statistic to evaluate model fit.
Next, we constructed a forward-selection multivariate model using delayed first appropriate antibiotic as the dependent variable, and vague symptom complex as the primary predictor of interest. We chose a threshold for delayed antibiotic administration as greater than 1 hour after first SBP less than 90 mm Hg or lactate greater than 2.0 mmol/dL, according to Surviving Sepsis Campaign treatment recommendations (6). Candidate covariates were the same as in the first model.
Finally, we constructed a model using in-hospital mortality as the dependent variable, including the individual symptom covariates that had univariate association with mortality. These symptom variables were forced into the model, given that they were of primary interest as exposure variables. We also forced SaO2/FIO2 (S/F) ratio and Glasgow Coma Scale (GCS) less than 15 because of their potential confounding effects on the presenting symptoms shortness of breath and altered mental status.
We used SAS version 9.4 (SAS Institute, Cary, NC) and R.3.4.4 (R Foundation, Vienna, Austria; https://R-project.org) for all analyses. A two-tailed p value of less than or equal to 0.05 was used as the cut-off for all tests of statistical significance. Odds ratios (ORs) are reported with 95% CIs and p values.
During the 2-year study period, 654 patients met criteria for septic shock, 245 (37%) of whom presented with vague symptoms of infection (Table 1). The in-hospital mortality rates of those with vague versus explicit symptoms of infection were 34% and 16% (p < 0.01), respectively. On multivariate analysis, the adjusted OR of vague presenting symptoms for mortality was 2.12 (95% CI, 1.32–3.40) (Table 2). Of note, only eight patients (2%) included in the explicit cohort met the explicit criteria based on measured triage fever alone.
Patients who presented with vague symptoms were older (median age, 68 vs 65 yr; p < 0.01) and sicker compared with those presenting with explicit symptoms (Table 1), as evidenced by hypotension (SBP, < 90 mm Hg) at triage (27% vs 20%; p = 0.06), mean SOFA score (7.5 vs 6.6; p < 0.01), initial serum lactate greater than 4.0 mmol/L (58% vs 39%; p < 0.01), and intubation in the ED (26% vs 12%; p < 0.01). Patients with vague symptoms were more frequently triaged to acute (83% vs 70%; p < 0.01). Despite triage recognition of acuity, acknowledgement of possible sepsis was lower for those with vague symptoms (55% vs 74% having the Sepsis Flag set in ED by provider; p < 0.01). Patients with vague symptoms also received antibiotics later after onset of hypoperfusion (hypotension or elevated lactate) (1.6 vs 0.8 hr; p < 0.01).
In multivariate analysis, vague symptoms were independently associated with delays in antibiotic administration by greater than 1 hour after documented hypoperfusion (adjusted OR, 2.03; 95% CI, 1.41–2.93). Despite a univariate association with mortality (OR, 1.06; 95% CI, 1.00–1.12; p = 0.04), time-to-appropriate antibiotic was not significantly associated with mortality in multivariate analysis that included vague symptoms (adjusted OR, 1.01; 95% CI, 0.94–1.08; p = 0.78). Removing vague symptoms from the mortality model, the magnitude of the effect size of time-to-antibiotics increased (adjusted OR, 1.04; 95% CI, 0.97–1.11; p = 0.27) but remained insignificant. The rate of positive blood cultures was greater in patients with explicit symptoms compared with vague symptoms (Table 1); the isolated pathogens are listed in the Appendix (Table A1). Presence of positive blood culture was not significantly associated with mortality in multivariate analysis. Table 2 reports all covariates that were independently associated with mortality.
Separating explicit and vague symptom complexes into their component elements, the symptom fever/chills/rigors was a qualifying symptom in 326 patients (80%) of the explicit cohort. The following individual symptoms all had univariate associations with mortality: fever/chills/rigors, altered mental status, shortness of breath, and headache (Table 3). Fever/chills/rigors was the only symptom that had a univariate association with antibiotic delay greater than 1 hour (OR, 0.41; 95% CI, 0.29–0.57; p < 0.01). Only absence of fever/chills/rigors (adjusted OR, 2.70; 95% CI, 1.63–4.47) and presence of shortness of breath (adjusted OR, 1.97; 95% CI, 1.23–3.15) were independently associated with mortality (Table 4). The effect of fever/chills/rigors symptom on mortality was independent of measured fever in the ED, and of larger magnitude.
Consistent with our hypothesis, we found differences in mortality, 34% versus 16%, for septic patients who presented to the ED with vague versus explicit symptoms of infection, respectively. The presence of vague symptoms upon hospital presentation was independently associated with higher mortality when adjusted for commonly reported confounders. Although vague presenting symptoms were associated with both delayed antibiotic administration and in-hospital mortality, our data did not support the corollary to our hypothesis that antibiotic delay was a primary driver of mortality. Furthermore, we found that absence of fever, chills, or rigors was the primary driver of the mortality association. Additionally, shortness of breath was independently associated with mortality even after adjusting for measured S/F ratio in the ED and other important confounders.
The observation that patients with vague symptoms had longer time-to-antibiotics warrants consideration. Patients with vague symptoms were generally identified as being seriously ill at triage, as evidenced by the higher rate of triage to the acute area of the ED. Yet the treating team did not recognize infection as often, based on the electronic Sepsis Flag being set less frequently. This highlights an interesting disconnect between recognition of “acuity” versus recognition of “infection.” Both cohorts had similar rates of meeting quick SOFA (qSOFA) criteria within the ED, with equally low sensitivities at triage. Assuming early recognition is important, these findings suggest that better education or clinical decision rules are needed to help clinicians identify sepsis on presentation, and that vague symptoms are a risk factor for diagnostic delay.
However, the substantial difference in mortality between the two cohorts appears to be due to factors beyond just sepsis recognition and antibiotic delay. Time-to-antibiotics was not a significant predictor of mortality in multivariate analysis, although this study was not powered to show effect sizes associated with hourly delays that have been previously published (7–12). Septic patients with explicit symptoms may tend to receive earlier antibiotics and also have higher survival, but not necessarily due to earlier antibiotic administration. The strong mortality association with vague symptoms persisted even after adjusting for common predictors of mortality in sepsis. This suggests that there could be value of including presenting symptoms into sepsis risk prediction rules. Other clinical decision tools (e.g., thrombosis in myocardial infarction risk score or pulmonary embolism rule out criteria rule) have incorporated presenting symptoms (13 , 14); however, this has not been explored in depth in sepsis.
Our findings that septic patients with vague symptoms have an elevated risk of mortality upon presentation, and are also prone to treatment delays, suggest that the presenting symptom complex may be an unmeasured confounder in epidemiologic studies that report associations between early sepsis treatments and mortality. The largest study to date by Seymour et al (9) reported an OR for death of 1.04 for each hour of antibiotic delay, but did not adjust for presenting symptoms. In univariate analysis, we found an OR for death of 1.06 for each hour of antibiotic delay, which was statistically significant. In multivariate analysis, including symptom complex, the OR was 1.01 and no longer significant (removing symptom complex from the model, the OR for each hour of antibiotic delay was 1.04, also insignificant albeit in a much smaller study population than that of Seymour et al ). Our findings suggest that adjusting for presenting symptoms may attenuate the effect of antibiotic delays that have been previously reported.
We found that the presenting symptom of fever/chills/rigors was the primary driver of the observed mortality effect in our a priori definition of explicit symptoms. This finding is consistent with those of Henning et al (15), who identified the association of fever and mortality in sepsis in a smaller patient sample (14). However, Henning et al (15) did not distinguish between symptom of fever and measured fever in the ED. Our findings suggest that each of these sources of information about fever is independently associated with mortality. Other studies have reported the association between measured fever and mortality (15–21).
The observed association between fever/chills/rigors and mortality supports the concept of differing underlying patient phenotypes at play in sepsis populations (22 , 23). Prior investigators have explored immunologic differences between septic patients with and without fever. Marik and Zaloga (16) found that a febrile response was not associated with higher levels of circulating proinflammatory cytokines and suggested that “the hypothermia of sepsis may be due to hypothalamic dysfunction with alternation in the thermal set-point...” It is also possible that virulence factors of the inciting pathogen may contribute to the presence or absence of febrile symptoms: we found that afebrile patients were significantly less likely to have positive blood cultures. However, positive blood cultures were not independently associated with mortality, consistent with Kethireddy et al (8) who report that culture-positive and culture-negative septic patients have similar outcomes. We examined blood culture pathogen counts and found no one species was disproportionate between groups.
Shortness of breath emerged as an independent predictor of mortality in our cohort. The Sepsis-3 consensus definition and qSOFA derivation address the importance of respiratory abnormality in identifying high-mortality septic patients (24 , 25). However, S/F ratio measured in the ED and respiratory rate measured at triage had lower predictive value than the shortness of breath symptom. This has clinical implications, as reliable measurement of respiratory rate can be difficult. Although respiratory infections can contribute to patients developing the sensation of shortness of breath, this symptom might indicate decompensation from an underlying sepsis-related etiology. Therefore, shortness of breath warrants consideration in risk stratification upon presentation and in adjusted analyses of mortality.
The effect of the symptom altered mental status, which had a significant univariate association with mortality, was attenuated when analyzed in a model with documented GCS in the ED. This argues that the reported symptoms altered mental status and GCS measurement in the ED are similar, and distinguishing the two has no additive value. This does not, however, diminish the importance of mental status assessment upon hospital presentation.
This analysis has several limitations. First, this was a single-center study in a large, urban academic medical center; therefore, local patient mix, practice patterns, and mortality variation may exist that are not representative of the general population or patient cohorts at other hospitals. Second, this was a retrospective study that relied on chart review to obtain patient history components. Third, a clinician may not obtain (or document) as detailed a history on a patient who is extremely ill or in extremis, leading to ascertainment bias for those classified with vague symptoms. Fourth, patients with vague symptoms were older and sicker, and this may have had an influence on mortality above what we were able to adjust for. Finally, as with any observational study, there may still exist unmeasured bias that we did not account for.
More than one third of patients admitted with septic shock present to the ED with vague symptoms that are not specific to infection. These patients have significantly higher mortality than those with explicit infectious symptoms that is independent of demographics, illness acuity, and time-to-antibiotics. These findings suggest that the nature of presenting symptoms may play an important role in sepsis clinical phenotyping and may be an important confounder in sepsis epidemiologic studies.
1. Rhee C, Dantes R, Epstein L, et al. Incidence and trends of sepsis in US hospitals using clinical vs claims data, 2009–2014. JAMA 2017; 318:1241–1299
2. (CMS) CfMMS: Sepsis Bundle Project (SEP), National Hospital Inpatient Quality Measures. 2015. Available at: https://www.qualitynet.org
. Accessed February 11, 2018
3. Jones AE, Trzeciak S, Kline JA. The Sequential Organ Failure Assessment score for predicting outcome in patients with severe sepsis and evidence of hypoperfusion at the time of emergency department
presentation. Crit Care Med 2009; 37:1649–1654
4. Charlson ME, Pompei P, Ales KL, et al. A new method of classifying prognostic comorbidity in longitudinal studies: Development and validation. J Chronic Dis 1987; 40:373–383
5. Reisner AT, Roberta C, Thorsen JE, et al. Development and validation of the SPoT Sepsis (Shock Prevention on Triage) clinical decision rule. Abstr. Acad Emerg Med 2015; 22:S43
6. Rhodes A, Evans LE, Alhazzani M, et al. Surviving Sepsis Campaign: International Guidelines for Management of Sepsis and Septic Shock
: 2016. Crit Care Med 2017; 45:1–67
7. Pruinelli L, Westra BL, Yadav P, et al. Delay within the 3-hour Surviving Sepsis Campaign guideline on mortality for patients with severe sepsis and septic shock
. Crit Care Med 2018; 46:500–505
8. Kethireddy S, Bilgili B, Sees A, et al; Cooperative Antimicrobial Therapy of Septic Shock
(CATSS) Database Research Group: Culture-negative septic shock
compared with culture-positive septic shock
: A retrospective cohort study. Crit Care Med 2018; 46:506–512
9. Seymour CW, Gesten F, Prescott HC, et al. Time to treatment and mortality during mandated emergency care for sepsis. N Engl J Med 2017; 376:2235–2244
10. Liu VX, Fielding-Singh V, Greene JD, et al. The timing of early antibiotics and hospital mortality in sepsis. Am J Respir Crit Care Med 2017; 196:856–863
11. Ferrer R, Martin-Loeches I, Phillips G, et al. Empiric antibiotic treatment reduces mortality in severe sepsis and septic shock
from the first hour: Results from a guideline-based performance improvement program. Crit Care Med 2014; 42:1749–1755
12. Sterling SA, Miller WR, Pryor J, et al. The impact of timing of antibiotics on outcomes in severe sepsis and septic shock
: A systematic review and meta-analysis. Crit Care Med 2015; 43:1907–1915
13. Antman EM, Cohen M, Bernink PJ, et al. The TIMI risk score for unstable angina/non-ST elevation MI: A method for prognostication and therapeutic decision making. JAMA 2000; 284:835–842
14. Kline JA, Mitchell AM, Kabrhel C, et al. Clinical criteria to prevent unnecessary diagnostic testing in emergency department
patients with suspected pulmonary embolism. J Thromb Haemost 2004; 2:1247–1255
15. Henning DJ, Carey JR, Oedorf K, et al. The absence of fever is associated with higher mortality and decreased antibiotic and IV fluid administration in emergency department
patients with suspected septic shock
. Crit Care Med 2017; 45:e575–e582
16. Marik PE, Zaloga GP. Hypothermia and cytokines in septic shock
. Norasept II Study Investigators. North American study of the safety and efficacy of murine monoclonal antibody to tumor necrosis factor for the treatment of septic shock
. Intensive Care Med 2000; 26:716–721
17. Peres Bota D, Lopes Ferreira F, Mélot C, et al. Body temperature alterations in the critically ill. Intensive Care Med 2004; 30:811–816
18. Sundén-Cullberg J, Rylance R, Svefors J, et al. Fever in the emergency department
predicts survival of patients with severe sepsis and septic shock
admitted to the ICU. Crit Care Med 2017; 45:591–599
19. Laupland KB, Shahpori R, Kirkpatrick AW, et al. Occurrence and outcome of fever in critically ill adults. Crit Care Med 2008; 36:1531–1535
20. Kushimoto S, Gando S, Saitoh D, et al; JAAM Sepsis Registry Study Group: The impact of body temperature abnormalities on the disease severity and outcome in patients with severe sepsis: An analysis from a multicenter, prospective survey of severe sepsis. Crit Care 2013; 17:R271
21. Young PJ, Saxena M, Beasley R, et al. Early peak temperature and mortality in critically ill patients with or without infection. Intensive Care Med 201238:437–444
22. Seymour CW, Rosengart MR. Septic shock
: Advances in diagnosis
and treatment. JAMA 2015; 314:708–717
23. Rimmelé T, Payen D, Cantaluppi V, et al; ADQI XIV Workgroup: Immune cell phenotype and function in sepsis. Shock 2016; 45:282–291
24. Singer M, Deutschman CS, Seymour CW, et al. The Third International Consensus Definitions for Sepsis and Septic Shock
(Sepsis-3). JAMA 2016; 315:801–810
25. Seymour CW, Liu VX, Iwashyna TJ, et al. Assessment of clinical criteria for sepsis: For the Third International Consensus Definitions for Sepsis and Septic Shock
(Sepsis-3). JAMA 2016; 315:762–774
diagnosis; emergency department; presenting symptoms; septic shock; time-to-antibiotics; unmeasured confounding
Supplemental Digital Content
Copyright © by 2018 by the Society of Critical Care Medicine and Wolters Kluwer Health, Inc. All Rights Reserved.