Secondary Logo

Journal Logo

Body Temperature at the Emergency Department as a Predictor of Mortality in Patients With Bacterial Infection

Yamamoto, Shungo MD, DTM&H, DrPH; Yamazaki, Shin PhD; Shimizu, Tsunehiro MD, PhD; Takeshima, Taro MD, PhD; Fukuma, Shingo MD, PhD; Yamamoto, Yosuke MD, PhD; Tochitani, Kentaro MD; Tsuchido, Yasuhiro MD; Shinohara, Koh MD; Fukuhara, Shunichi MD, DMSc

Section Editor(s): Asensi., Victor

doi: 10.1097/MD.0000000000003628
Research Article: Observational Study

Hypothermia is a risk factor for death in intensive care unit (ICU) patients with severe sepsis and septic shock. In the present study, we investigated the association between body temperature (BT) on arrival at the emergency department (ED) and mortality in patients with bacterial infection.

We conducted a retrospective cohort study in consecutive ED patients over 15 years of age with bacterial infection who were admitted to an urban teaching hospital in Japan between 2010 and 2012. The main outcome measure was 30-day in-hospital mortality. Each patient was assigned to 1 of 6 categories based on BT at ED admission. We conducted multivariable logistic regression analysis to adjust for predictors of death.

A total of 913 patients were enrolled in the study. The BT categories were <36, 36 to 36.9, 37 to 37.9, 38 to 38.9, 39 to 39.9, and ≥40 °C, with respective mortalities of 32.5%, 14.1%, 8.7%, 8.2%, 5.7%, and 5.3%. Multivariable analysis showed that the risk of death was significantly low in patients with BT 37 to 37.9 °C (adjusted odds ratio [AOR]: 0.2; 95% confidence interval [CI] 0.1–0.6, P = 0.003), 38–38.9 °C (AOR: 0.2; 95% CI 0.1–0.6, P = 0.002), 39–39.9 °C (AOR: 0.2; 95% CI 0.1–0.5, P = 0.001), and ≥40 °C (AOR: 0.1; 95% CI 0.02–0.4, P = 0.001), compared with hypothermic patients (BT <36 °C).

The higher BT on arrival at ED, the better the outcomes observed in patients with bacterial infection were.

From the Department of Healthcare Epidemiology (SY, SY, TT, SF, YY, SF), School of Public Health in the Graduate School of Medicine, Kyoto University, Kyoto; Division of Infectious Diseases (SY), Kobe University Hospital, Hyogo; Current affiliation: Center for Environmental Health Sciences (SY), National Institute for Environmental Studies, Ibaraki; Department of Infectious Diseases (TS, KT, YT, KS), Kyoto City Hospital, Kyoto; Division of Community and Family Medicine (TT), Center for Community Medicine, Jichi Medical University, Tochigi; Center for Innovative Research for Communities and Clinical Excellence (CIRC2LE) (SF, SF), Fukushima Medical University, Fukushima; Department of Clinical Laboratory Medicine (current affiliation for YT), Graduate School of Medicine, Kyoto University, Kyoto; and Disease Control and Prevenion Center (current affiliation for KS), National Center for Global Health and Medicine, Tokyo, Japan.

Correspondence: Shunichi Fukuhara, Department of Healthcare Epidemiology, School of Public Health in the Graduate School of Medicine, Kyoto University, Yoshida Konoemachi, Sakyo-ku, Kyoto 606 8501, Japan (e-mail:

Abbreviations: AOR = adjusted odds ratio, BT = body temperature, CI = confidence interval, ED = emergency department, RR = respiratory rate.

Presented in Part: 25th European Congress of Clinical Microbiology and Infectious Diseases. April 25 to 28, 2015, Copenhagen, Denmark.

SY had full access to all data in the study, takes responsibility for the integrity of the data and accuracy of the data analysis, and wrote the 1st draft; TS, KT, YT, and KS collected and interpreted the data and drafted the paper; and SY, TT, SF, YY, and SF supervised the research, interpreted the data, and helped draft the manuscript.

The authors have no funding and conflicts of interests to disclose.

This is an open access article distributed under the Creative Commons Attribution-NonCommercial-NoDerivatives License 4.0, where it is permissible to download, share and reproduce the work in any medium, provided it is properly cited. The work cannot be changed in any way or used commercially.

Received December 7, 2015

Received in revised form March 29, 2016

Accepted April 13, 2016

Back to Top | Article Outline


Fever is a common feature of infection, and the febrile response exerts a number of beneficial effects by fighting sepsis and enhancing chemotaxis, neutrophil migration, phagocytosis, antibody production, and T-cell proliferation.1–4 A recent observational study reported that elevated peak temperature during the 1st 24 hours in an ICU is associated with decreased mortality in critically ill patients with infection.5 Several reports have also shown an association between hypothermia and increased mortality in ICU patients with infection.6–10 However, fever can also harm the host, as an increase of 1 °F (0.56 °C) leads to an increase in pulse rate of 10 beats/minute, and the increased metabolic rates caused by fever may exacerbate cardiopulmonary function.1 Additionally, extreme hyperthermia can damage the central nervous system.1 Fever in patients with noninfectious diseases is associated with worse outcomes.5,11

Body temperature (BT) at emergency department (ED) admission is more relevant to clinicians in the ED than is BT at admission to an ICU. To our knowledge, no studies have investigated the relationship between BT at presentation to the ED and prognosis in patients with bacterial infection.

In this study, to facilitate the assessment of a patient's severity of condition on admission to the ED, we investigated the association between BT on arrival at the ED and mortality in patients with bacterial infection.

Box 1

Box 1

Back to Top | Article Outline


Study Design, Setting, and Population

This retrospective cohort study used data from patients suspected to have sepsis in our previous study.12 The cohort was designed to evaluate the clinical usefulness of serum C-reactive protein in patients with suspected sepsis. First, SY extracted the consecutive ED patients over 15 years of age admitted to the Kyoto City Hospital, an urban Japanese teaching hospital with 548 beds, after having a blood culture drawn in the ED between January, 2010 and December, 2012. Then, SY, TS, KT, YT, and KS extracted the following data anonymously from electronic medical records for each patient with suspected sepsis: age, gender, underlying disease, diagnosis for admission, vital signs, laboratory findings, and outcome in our previous study. That cohort included 1310 patients.12 Of these patients, 926 were ultimately diagnosed with bacterial infection and deemed eligible for the present study. Classification of bacterial infection was determined based on the agreement between the diagnosis of the treating physician at the time of discharge from the hospital and the investigators’ assessment. To promote data independence, only the index admission was included for patients with multiple admissions during the study period. Patients transferred from another hospital or who had cardiopulmonary arrest on arrival at the hospital were excluded.

Back to Top | Article Outline

Study Protocol

The following data were extracted from electronic medical records for each patient: age, gender (male/female), use of corticosteroids (yes/no), malignancy (present/absent), bacteremia (present/absent), vital sign values (blood pressure, respiratory rate [RR], mental confusion, and BT), blood urea nitrogen value, and outcome. The following predictors were defined based on previous studies: mental confusion (present/absent), blood urea nitrogen >7 mmol/L (20 mg/dL), RR ≥30/minute, and systolic blood pressure <90 mm Hg or diastolic blood pressure ≤60 mm Hg (either alone or in combination).12–16 Mental confusion was defined as disorientation in person, place, or time or the presence of a stupor or coma, in accordance with a previous study.13 For vital signs and laboratory data, initial values for the hospital visit were recorded. Each patient was assigned to 1 of 6 categories based on BT at admission to the ED: <36, 36 to 36.9, 37 to 37.9, 38 to 38.9, 39 to 39.9, and ≥40 °C. BT was mainly measured via electronic thermometer at the axilla, except in patients with extremely low BT (<35 °C), in which case the core BT was measured using a bladder or rectal probe.

Back to Top | Article Outline


The main outcome measure was 30-day in-hospital mortality. Patients who were discharged or transferred from the hospital within 30 days of admission or who remained in the hospital for more than 30 days were considered alive in this analysis.17

Back to Top | Article Outline

Data Analysis

We conducted multivariable logistic regression analysis to adjust for the predictors of death by introducing prespecified variables (age, gender, condition severity, steroid use, malignancy, and bacteremia) based on the findings of previous studies and clinical relevance.12,14–16,18–21 We used prespecified variables to adjust for the predictors because the present study was not an exploratory analysis to select statistically significant variables among many other candidate variables. To adjust for condition severity, we used the CURB-65 score, which was originally developed as a severity score for community-acquired pneumonia and later validated in patients with suspected sepsis, regardless of source, and patients admitted for nonsurgical illness.12,14–16 To evaluate which of the parameters were more closely associated with mortality, we treated each component of the CURB-65 score, as a separate explanatory variable. As normal BT varies with age, the age component was used as a continuous variable.22 We calculated both unadjusted and adjusted odds ratios (aORs) and 95% confidence intervals (CIs) and considered a 2-sided P value <0.05 statistically significant. We assessed the calibration of the model using the Hosmer–Lemeshow goodness-of-fit test. A P value <0.05 indicates a lack of good fit for the model. Regarding the model discrimination, we also computed the area under the receiver operating characteristic curve with a 95% CI using 500 bootstrap resampling.23 We did not conduct formal sample size calculations, and all available data were used to maximize the power. Previous studies have suggested that at least 8 to 10 events per variable are needed for reliable multiple logistic regression analysis.24,25 As for missing values, we planned to conduct a complete case analysis if the missing values were below 5%, as such an analysis might have been feasible in that case.26 If the missing values were at or above 5%, we planned to perform the appropriate imputation. All data were analyzed using Stata software, version 13 (StataCorp, College Station, TX).

Back to Top | Article Outline

Ethical Approval

The Ethics Committees of the Kyoto University Graduate School and Faculty of Medicine approved this protocol. As the study was observational and data were collected anonymously, the institutional review board waived the need for patient consent. Instead, we gave the participants the opportunity to disclaim their participation.

Back to Top | Article Outline


Of the 926 participants with bacterial infection, only RR data were missing for 13 patients (1.4%). We therefore conducted a complete case analysis, leaving 913 patients for further evaluation. The 30-day in-hospital mortality was 9.6% (88 deaths). Demographics, underlying illnesses, vital signs, laboratory findings, and diagnoses are presented in Table 1. The most common diagnosis was pneumonia, followed by urinary tract infection, skin and soft tissue infection, acute cholangitis, and acute cholecystitis. Mortality ranged from 32.5% among patients with BT <36 °C to 5.3% among patients with BT ≥40 °C (Table 1), and mortality decreased as BT increased.



The unadjusted ORs for the mortality of each BT category relative to the reference range of <36 °C are presented in Table 2. Multivariable analysis showed that the risk of death was significantly low in patients with BT 37 to 37.9 °C (AOR: 0.2; 95% CI 0.1–0.6, P = 0.003), 38 to 38.9 °C (AOR: 0.2; 95% CI 0.1–0.6, P = 0.002), 39 to 39.9 °C (AOR: 0.2; 95% CI 0.1–0.5, P = 0.001), and ≥40 °C (AOR: 0.1; 95% CI 0.02–0.4, P = 0.001), compared with hypothermic patients (BT <36 °C) (Table 2). The multivariable model showed good calibration for mortality, with a Hosmer–Lemeshow test of 7.05 (df = 8, P = 0.53), indicating good fit. The area under the receiver operating characteristic curve of the model was 0.84 (95% CI: 0.80–0.87).



Back to Top | Article Outline


In the present study, we found that the risk of mortality decreased as the BT on arrival at the ED increased in patients with bacterial infection. Our findings are consistent with those of previous observational studies, which reported that a low BT in critically ill patients with infection in ICUs is associated with increased mortality.5–10 Moreover, the results are consistent with those of studies reported that hypothermia in patients with bacteremia, including noncritically ill patients, was associated with a worse survival outcome than in patients with a normal or high BT.27–30 In our study, BT at admission to an ED was associated with mortality in patients with infection, including mildly to moderately ill nonbacteremic patients. These findings should help raise awareness among ED doctors that a patient without fever may have a severe bacterial infection.

The association between high fever and low risk of mortality in patients with infection may have several explanations. First, patients with hypothermia may simply have been dying of severe bacterial infection and may have been beyond the point of treatment.

Second, patients with poor febrile response may have been vulnerable to infectious disease because fever per se is an important host defensive mechanism against infection, with a number of beneficial effects. Fever enhances the phagocytosis of extracellular organisms, intracellular killing of ingested intracellular bacteria, chemotaxis, neutrophil migration, production and activity of antibodies, T-cell proliferation, and complement activation.1–4 Additionally, fever also directly exerts adverse effects on many pathogens, because elevated BT suppresses the replication of a number of pathogenic organisms. Levels of serum iron, an important virulence factor in pathogenic bacteria, are decreased under hyperthermic conditions, resulting in decreased microbial virulence. Fever also induces lysis of many bacteria if BT is sufficiently elevated.1,3,31,32

Third, the association between BT and mortality risk may be less direct, as low BT/absence of fever may merely delay the diagnosis of a bacterial infection. Given that fever is a common feature of infection,33 if a patient presents a high fever, a physician typically considers the possibility of infection. However, if a patient does not have a fever, a physician may not consider infection, thereby delaying the accurate diagnosis of patients who actually have a bacterial infection, thereby leading to increased mortality.34 As such, physicians should be alert to the possibility of bacterial infection even in patients without a fever. Considering the possibility of bacterial infection then allows for the application of clinical prediction rules, such as CURB-65, which may help evaluate condition severity in patients suspected of any source of sepsis.12,14–16

Back to Top | Article Outline


Several limitations to the present study warrant mention. First, the method of measuring BT was not standardized, and BT was measured mainly at the axilla using a digital thermometer. The axilla is not an optimal site for measuring BT, because axillary temperature is susceptible to error and can be misleading.35 A recent systematic review reported that peripheral (tympanic membrane, temporal artery, axillary, or oral) thermometers did not have clinically acceptable accuracy.36 However, measuring BT with central (pulmonary artery catheter, urinary bladder, esophageal, or rectal) thermometers in all ambulatory patients in EDs would not be realistic. Previous studies in Japanese populations have shown that axillary temperature is only 0.1 to 0.2 °C lower than oral temperature.37,38 In critically ill Japanese patients, mean difference for axillary temperature compared with bladder temperature was −0.33 ± 0.55 °C.39 Therefore, the axilla is considered an acceptable site for measuring BT among Japanese.22 Second, we were unable to determine the difference between BT at admission to the ED and each patient's normal temperature, preventing us from identifying which was more relevant to the patient's outcome: the absolute value of BT or the change in BT from normal. Third, we were unable to determine the usage of antipyretics before patients came to the ED because the precise data as to whether or not patients took antipyretics, including over-the-counter drugs, were unavailable from the medical record. The benefits of antipyretics in treating fever have not been confirmed; in fact, some antipyretic agents have been shown to cause coronary vasoconstriction in patients with coronary artery disease.33,40 Additionally, the administration of nonsteroidal antiinflammatory drugs or acetaminophen has been associated with increased mortality in septic patients.41,42 Taking antipyretic drugs before visiting the ED might have thus contributed to a poor prognosis in patients without a fever. In contrast, a recent randomized controlled trial showed that the early administration of acetaminophen for fever in critically ill patients with probable infection did not affect the number of ICU-free days and mortality.43 Given that antipyretics do not affect mortality in infection, a misclassification of patients taking antipyretics prior to coming the ED into a lower BT group than their true BT would attenuate the relationship between BT on admission at the ED and mortality. Fourth, bacterial infections may cause abrupt changes in BT, especially in cases of bacteremia. Thus, a patient who arrived at the ED afebrile or with mild fever would have been classified as such, even though 1 hour before or after that recording, this patient could have had a high fever. Therefore, these recordings may have been influenced by chance and may not be representative of the true BT. Even so, the dose–response relationship between BT on admission at the ED and mortality would be helpful in screening high risk patients in the ED. Finally, this study was conducted retrospectively, and the data were collected from electric medical records, which may be incomplete and inaccurate. However, the number of missing values was relatively small (less than 5%), and we used mortality as an outcome due to its robust nature. This was a single center study, and further prospective studies on different patients are warranted before conclusions can be drawn on the true impact of BT on patient outcomes.

Back to Top | Article Outline


In conclusion, the higher the BT on arrival at the ED, the better the outcomes observed in patients with bacterial infection were. An ED clinician should be aware that patients without fever may be more worrisome than those with fever.

Back to Top | Article Outline


1. Cunha BA. Vincent JL, Carlet J, Opal SM. Shoulder fever be treated in sepsis? The Sepsis Text. London: Kluwer Academic Publisher; 2002. 705–717.
2. Roberts NJ, Sandberg K. Hyperthermia and human leukocyte function. II. Enhanced production of and response to leukocyte migration inhibition factor (LIF). J Immunol 1979; 122:1990–1993.
3. Roberts NJ. Temperature and host defense. Microbiol Rev 1979; 43:241–259.
4. Jampel HD, Duff GW, Gershon RK, et al. Fever and immunoregulation. III. Hyperthermia augments the primary in vitro humoral immune response. J Exp Med 1983; 157:1229–1238.
5. Young PJ, Saxena M, Beasley R, et al. Early peak temperature and mortality in critically ill patients with or without infection. Intensive Care Med 2012; 38:437–444.
6. Clemmer TP, Fisher CJ, Bone RC, et al. Hypothermia in the sepsis syndrome and clinical outcome. The Methylprednisolone Severe Sepsis Study Group. Crit Care Med 1992; 20:1395–1401.
7. Peres Bota D, Lopes-Ferreira F, Mélot C, et al. Body temperature alterations in the critically ill. Intensive Care Med 2004; 30:811–816.
8. Laupland KB, Gregson DB, Zygun DA, et al. Severe bloodstream infections: a population-based assessment. Crit Care Med 2004; 32:992–997.
9. Tiruvoipati R, Ong K, Gangopadhyay H, et al. Hypothermia predicts mortality in critically ill elderly patients with sepsis. BMC Geriatr 2010; 10:70.
10. Kushimoto S, Gando S, Saitoh D, et al. 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 (Lond, Engl) 2013; 17:R271.
11. Greer DM, Funk SE, Reaven NL, et al. Impact of fever on outcome in patients with stroke and neurologic injury: a comprehensive meta-analysis. Stroke 2008; 39:3029–3035.
12. Yamamoto S, Yamazaki S, Shimizu T, et al. Prognostic utility of serum CRP levels in combination with CURB-65 in patients with clinically suspected sepsis: a decision curve analysis. BMJ Open 2015; 5:e007049–e17049.
13. Lim WS, van der Eerden MM, Laing R, et al. Defining community acquired pneumonia severity on presentation to hospital: an international derivation and validation study. Thorax 2003; 58:377–382.
14. Marwick CA, Guthrie B, Pringle JE, et al. Identifying which septic patients have increased mortality risk using severity scores: a cohort study. BMC Anesthesiol 2014; 14:1.
15. Armiñanzas C, Velasco L, Calvo N, et al. CURB-65 as an initial prognostic score in internal medicine patients. Eur J Intern Med 2013; 24:416–419.
16. Howell MD, Donnino MW, Talmor D, et al. Performance of severity of illness scoring systems in emergency department patients with infection. Acad Emerg Med 2007; 14:709–714.
17. Graham PL, Cook DA. Prediction of risk of death using 30-day outcome: a practical end point for quality auditing in intensive care. Chest 2004; 125:1458–1466.
18. Shapiro NI, Wolfe RE, Moore RB, et al. Mortality in Emergency Department Sepsis (MEDS) score: a prospectively derived and validated clinical prediction rule. Crit Care Med 2003; 31:670–675.
19. Laupland KB, Davies HD, Church DL, et al. Bloodstream infection-associated sepsis and septic shock in critically ill adults: a population-based study. Infection 2004; 32:59–64.
20. Gleckman R, Hibert D. Afebrile bacteremia. A phenomenon in geriatric patients. JAMA 1982; 248:1478–1481.
21. Laupland KB, Zahar J-R, Adrie C, et al. Severe hypothermia increases the risk for intensive care unit-acquired infection. Clin Infect Dis 2012; 54:1064–1070.
22. Iriki M, Kosaka M, Murakami N, et al. Studies on the axilary temperature of the Japanses aged. Nippon Ronen Igakkai Zasshi 1975; 12:172–177.Article in Japanese.
23. Steyerberg EW, Bleeker SE, Moll HA, et al. Internal and external validation of predictive models: a simulation study of bias and precision in small samples. J Clin Epidemiol 2003; 56:441–447.
24. Peduzzi P, Concato J, Kemper E, et al. A simulation study of the number of events per variable in logistic regression analysis. J Clin Epidemiol 1996; 49:1373–1379.
25. Cepeda MS, Boston R, Farrar JT, et al. Comparison of logistic regression versus propensity score when the number of events is low and there are multiple confounders. Am J Epidemiol 2003; 158:280–287.
26. Royston P, Moons KGM, Altman DG, et al. Prognosis and prognostic research: developing a prognostic model. BMJ 2009; 338:b604.
27. Bryant RE, Hood AF, Hood CE, et al. Factors affecting mortality of gram-negative rod bacteremia. Arch Intern Med 1971; 127:120–128.
28. Diekema DJ, Beekmann SE, Chapin KC, et al. Epidemiology and outcome of nosocomial and community-onset bloodstream infection. J Clin Microbiol 2003; 41:3655–3660.
29. Rebelo M, Pereira B, Lima J, et al. Predictors of in-hospital mortality in elderly patients with bacteraemia admitted to an Internal Medicine ward. Int Arch Med 2011; 4:33.
30. Lillie PJ, Allen J, Hall C, et al. Long-term mortality following bloodstream infection. Clin Microbiol Infect 2013; 19:955–960.
31. Kluger MJ. Is fever beneficial? Yale J Biol Med 1986; 59:89–95.
32. Duff GW. Is fever beneficial to the host: a clinical perspective. Yale J Biol Med 1986; 59:125–130.
33. Plaisance KI, Mackowiak PA. Antipyretic therapy: physiologic rationale, diagnostic implications, and clinical consequences. Arch Intern Med 2000; 160:449–456.
34. Norman DC. Fever in the elderly. Clin Infect Dis 2000; 31:148–151.
35. Mackowiak PA. Mackowiak PA. Clinical thermometric measurements. Fever Basic Mechanisms and Management 2nd ed.Philadelphia, New York: Lippincott-Raven; 1997. 27–34.
36. Niven DJ, Gaudet F E, Laupland K B, et al. Accuracy of peripheral thermometers for estimating temperature: a systematic review and meta-analysis. Ann Intern Med 2015; 163:768–777.
37. Tasaka S, Yoshitoshi Y. A Report on the axillar temperature of the healthy Japanese people. Nisshinigaku 1957; 44:633–638.Article in Japanese.
38. Iriki M, Tsuchiya K, Kinno I, et al. The oral temperature of healthy Japanese. Jpn J Biometeorol 1988; 25:163–171.Article in Japanese.
39. Nonose Y, Sato Y, Kabayama H, et al. Accuracy of recorded body temperature of critically ill patients related to measurement site: a prospective observational study. Anaesth Intens Care 2012; 40:820–824.
40. Friedman PL, Brown EJ, Gunther S, et al. Coronary vasoconstrictor effect of indomethacin in patients with coronary-artery disease. N Engl J Med 1981; 305:1171–1175.
41. Arons MM, Wheeler AP, Bernard GR, et al. Effects of ibuprofen on the physiology and survival of hypothermic sepsis. Ibuprofen in Sepsis Study Group. Crit Care Med 1999; 27:699–707.
42. Lee BH, Inui D, Suh GY, et al. Association of body temperature and antipyretic treatments with mortality of critically ill patients with and without sepsis: multi-centered prospective observational study. Crit Care (Lond, Engl) 2012; 16:R33.
43. Young P, Saxena M, Bellomo R, et al. Acetaminophen for fever in critically Ill patients with suspected infection. N Engl J Med 2015; 373:2215–2224.
Copyright © 2016 The Authors. Published by Wolters Kluwer Health, Inc. All rights reserved.