Abstract
Background:
Fever is a common response to both infectious and non-infectious physiologic insults in the critically ill, and in certain populations it appears to be protective. Fever is particularly common in trauma patients, and even more so in those with infections. The relationship between fever, trauma status, and mortality in patients with an infection is unclear.
Patients and Methods:
A review of a prospectively maintained institutional database over a 17-year period was performed. Surgical and trauma intensive care unit (ICU) patients with a nosocomial infection were extracted to compare in-hospital mortality among trauma and non-trauma patients with and without fever. Univariable analyses compared patient and infection characteristics between trauma and non-trauma patients. A multivariable logistic regression model was created to identify predictors of in-hospital mortality, with a focus on fever and trauma status.
Results:
Nine hundred forty-one trauma patients and 1,449 non-trauma patients with ICU-acquired infections were identified. Trauma patients were younger (48 vs. 59, p < 0.001), more likely to be male (73% vs. 56%, p < 0.001), more likely to require blood transfusion (74% vs. 47%, p < 0.001), had lower Acute Physiology and Chronic Health Evaluation (APACHE) II scores (18 vs. 19, p = 0.02), and had lower rates of comorbidities. Trauma patients were more likely to develop a fever (72% vs. 43%, p < 0.001) and had lower in-hospital mortality (9.6% vs. 22.6%, p < 0.001). In multivariable analysis, non-trauma patients with fever had a lower odds of mortality compared with non-trauma patients without fever (odds ratio [OR] 0.63, p = 0.004). Trauma patients with fever had the lowest odds ratio for mortality when compared to non-trauma patients without fever (OR 0.25, p < 0.001).
Conclusions:
In this large cohort of trauma and surgical ICU patients with ICU-acquired infections, fever was associated with a lower odds of mortality in both trauma and non-trauma patients. Further investigation is needed to determine the mechanisms behind the interplay between trauma status, fever, and mortality.
Fever is common in critically ill patients and can result from both infectious and non-infectious physiologic insults [1,2]. Fever is particularly common after traumatic injury with an incidence of up to 50% [3]. Although fever is associated with worse outcomes in neurologic injury [4-6], there is evidence that it can be protective in other settings including non-neurologic trauma and infection [7–9]. Overall, febrile critically ill patients without an infection appear to have higher mortality [10], whereas fever in the presence of infection seems to be protective [9,11].
There are conflicting theories regarding the role and importance of fever in infection. Some suggest that fever itself is protective, whereas others argue that fever merely denotes the presence of an adequate immune response. At the molecular level, the development of fever is mediated by pyrogenic cytokines (e.g., interlekin-1 and tumor necrosis factor-α) to stimulate acute phase reactants and carry out numerous immune and end-organ system interactions [12]. This thermoregulation—both hyperthermia and hypothermia—has been hypothesized to be a tolerogenic physiologic defense strategy against both infection and traumatic injury [13].
Given the complex biologic interplay between inflammation, infection, and injury, there are many possible confounding factors. The relationship between fever and trauma in infection, and their combined effect on clinical outcomes, remains unclear. We hypothesized that fever would be associated with a lower odds of mortality in both trauma and non-trauma patients with an intensive care unit (ICU)-acquired infection.
Patients and Methods
Patient population
A prospectively maintained database containing information on all patients admitted to the combined surgical and trauma ICU at a single academic medical center from November 1996 to July 2014 was searched for patients with the presence of nosocomial infection. A cohort was constructed retrospectively to compare the primary outcome of in-hospital mortality among trauma and non-trauma patients with and without fever. Only the index infection was included, and patients were not duplicated in the analysis unless a separate hospital admission was present. The treatment of fever, including within the traumatic brain injury (TBI) population, was not standardized and was at the clinician's discretion; a protocol to prevent fever in these patients was not utilized. Data collection was conducted every other day during each included patient's admission by ICU staff via review of the electronic medical record, patient interview, and patient examination. All data were collected until hospital discharge or patient death.
Patient demographics (age, gender, race, trauma/non-trauma status, transplant status), comorbidities (hypertension, diabetes mellitus, hyperlipidemia, coronary artery disease, peripheral vascular disease, chronic kidney disease, pulmonary disease, liver disease, malignancy, steroid use, and human immunodeficiency virus status), and Acute Physiology and Chronic Health Evaluation (APACHE) II score at time of ICU admission were recorded. Primary ICU admitting diagnoses were captured to include the categories cardiovascular/thoracic, gastrointestinal, hepatopancreatobiliary/spleen, head and neck, musculoskeletal, urologic, neurologic (including spinal pathology), skin/soft tissue, and other. Primary surgical procedures were also captured for each patient including laparotomy, laparoscopy, cardiothoracic, vascular, transplant, head and neck, debridement/burn surgery, orthopedic, and neurologic (including spine surgery). Rates of repeat surgical procedures were also captured as was length of ICU stay. In addition, variables describing the infectious episode—including fever (defined as temperature ≥38.5°C), white blood cell count, bacterial or fungal organism, and primary site of infection—were also documented. We hypothesized that those patients with an ICU-acquired infection who developed fever after trauma would have improved survival compared with trauma patients who did not develop a fever, non-trauma patients who did develop a fever, and non-trauma patients who did not develop a fever.
Data analysis
Univariable analyses were conducted to compare patient and infection characteristics between trauma and non-trauma patients. Characteristics were also compared between trauma and non-trauma patients with and without fever. Differences were compared using the χ2 or Fisher exact test for categorical variables and the Wilcoxon rank-sum test for continuous variables. A multivariable logistic regression model was created to adjust for clinically relevant concurrent effects and to identify predictors of in-hospital mortality, particularly focusing on the interplay between fever and trauma status. A p value of <0.05 was used for statistical significance, and all statistical analyses were performed using SAS 9.4 (SAS Institute, Cary, NC).
Results
A total of 2,390 patients with ICU-acquired infections were identified: 941 trauma patients and 1,449 non-trauma surgical patients. There were large differences in the demographics of these two groups, with trauma patients being younger, more likely to be male, and with fewer comorbidities (Table 1). Trauma patients were more likely to receive a blood transfusion and be mechanically ventilated, however, were less likely to require re-operation. Neurologic admitting diagnoses dominated in the trauma population, whereas gastrointestinal diagnoses were most common among non-trauma patients. Laparotomy was performed commonly in both groups, however, trauma patients underwent slightly more orthopedic procedures.
Patient Characteristics, Hospitalization and Infection Details, and Outcomes for Trauma and Non-Trauma Patients
Categorical variables listed as N (%) and continuous variables listed as median (interquartile range)
BMI = body mass index; HIV = human immunodeficiency virus; ICU = intensive care unit; APACHE II = Acute Physiology and Chronic Health Evaluation II; WBC = white blood cell count.
There were more frequently documented fevers (72.3% vs. 43.3%, p < 0.001) among trauma patients, and this group also exhibited a higher maximum temperature. The site of infection was substantially different between the two groups, with trauma patients most likely to have a pulmonary source of infection and non-trauma patients an abdominal source of infection. Additionally, fungal infections were more common among non-trauma patients. Overall, trauma patients had a lower mortality rate compared with non-trauma patients (9.6% vs. 22.6%; p < 0.001) (Table 1).
Trauma patients
Trauma patients with and without fever were then compared with several demographic and comorbidity differences noted as reported in Table 2. Trauma patients who developed a fever were more likely to be mechanically ventilated, receive a blood transfusion, and have pulmonary and blood stream sources of infection. Neurologic primary diagnoses were most common for both febrile and afebrile patients, and orthopedic procedures were also most prevalent, although narrowly. Re-operation rates were comparable at 27.1% for febrile and 25.3% for afebrile trauma patients (p = 0.582). Length of ICU stay was similar with a median of 9.0 days for both groups (p = 0.828). Notably, febrile trauma patients had a substantially lower in-hospital mortality compared with afebrile trauma patients (7.7% vs. 14.6%, p = 0.001) (Table 2).
Patient Characteristics, Hospitalization and Infection Details, and Outcomes among Trauma Patients with and without Fever
Categorical variables listed as n (%) and continuous variables listed as median (interquartile range)
BMI = body mass index; ICU = intensive care unit; APACHE II = Acute Physiology and Chronic Health Evaluation II; WBC = white blood cell count
Non-trauma patients
Non-trauma patients with and without fever were also compared as shown in Table 3. Similar to trauma patients, non-trauma patients who developed a fever were younger, had fewer comorbidities, and had higher rates of blood transfusion and mechanical ventilation. Pulmonary and blood stream sources of infection were also similarly higher among non-trauma patients with fever. Gastrointestinal primary diagnoses were most common for both febrile and afebrile patients, and laparotomy was the most prevalent operation for both groups. Re-operation rates were comparable at 45.0% for febrile and 40.5% for afebrile trauma patients (p = 0.088). Length of ICU stay was longer for non-trauma patients with fever with median 10.0 days compared with 8.0 days in patients without fever (p < 0.001). However, no mortality difference was observed between febrile and afebrile non-trauma patients (21.3% vs. 23.5%; p = 0.342) (Table 3).
Patient Characteristics, Hospitalization and Infection Details, and Outcomes among Non-Trauma Patients with and without Fever
Categorical variables listed as N (%) and continuous variables listed as median (interquartile range)
BMI = body mass index; HIV = human immunodeficiency virus; ICU = intensive care unit; APACHE II = Acute Physiology and Chronic Health Evaluation II; WBC = white blood cell count.
Multivariable analysis
In the multivariable model, several factors were associated with mortality among critically ill patients with nosocomial infections, including age, coronary artery disease, hepatic insufficiency, malignancy, blood transfusion, APACHE II score, and fungal infection. A decreased odds of mortality was also noted for intra-abdominal and urinary sources of infection. Most importantly, the lowest odds of mortality was observed in trauma patients with fever, followed by trauma patients without fever, and finally non-trauma patients with fever (Table 4).
Predictors of In-Hospital Mortality among Patients with ICU-Acquired Infections
ICU = intensive care unit; OR = odds ratio; CI = confidence interval; BMI = body mass index; HIV = human immunodeficiency virus; APACHE II = Acute Physiology and Chronic Health Evaluation II; WBC = white blood cell count.
Discussion
This single institution review of a prospectively maintained dataset compared febrile and afebrile critically ill trauma and non-trauma surgical patients with an ICU-acquired infection and identified independent risk factors for in-hospital mortality. Analysis was chiefly designed to evaluate the interaction between trauma status, presence of fever, and in-hospital mortality. In both trauma and non-trauma surgical patients with ICU-acquired infections, the presence of fever was associated with lower in-hospital mortality. Trauma status was also associated with lower in-hospital mortality.
As expected, trauma patients were younger and healthier than non-trauma patients, but they were also much more likely to require mechanical ventilation and receive blood transfusions. Although this increased incidence of pulmonary and hemodynamic support was seen in trauma patients, the APACHE II score was slightly higher in non-trauma patients suggesting greater physiologic stress in this group [14]. Overall, the greater physiologic reserve of younger, healthier trauma patients likely played a key role in their overall lower in-hospital mortality as advanced age has long been recognized as a strong independent risk factor for mortality [15,16].
The primary source of infection was substantially different between groups, with nearly three-quarters of trauma patients developing pneumonia compared with one-quarter of non-trauma patients having a pulmonary source of infection. This finding is not unexpected, given that trauma patients were more likely to require mechanical ventilation. Additional work has also shown that trauma patients are at higher risk of pneumonia independent of mechanical ventilation, with intubated trauma patients nearly four times as likely to develop pneumonia compared with intubated non-trauma patients [17]. This is consistent with the increased rate of pulmonary infections identified among trauma patients in the current study. A more even distribution of infectious sources was seen in non-trauma patients, with intra-abdominal infections being the most common. Many of these surgical patients underwent abdominal operations, which would likely explain the comparatively high rate of abdominal sources of infection in this population.
The association between fever and mortality in non-trauma patients has been described previously and is supported by the current study. The presence of fever in infected surgical patients has been shown to be associated independently with decreased mortality [8]. Well-conducted prior works have also demonstrated no mortality benefit with the use of acetaminophen for treatment of fever in critically ill patients with sepsis, suggesting that fever on its own is not harmful in sepsis [18,19]. In fact, other studies have actually found an increase in mortality with antipyretics [10,20]. Because of these findings, the decision to treat fever in critically ill surgical patients with sepsis remains unclear and merits further study.
Unlike the non-trauma population, the relationship between fever and mortality in trauma patients with infection has not been well described prior to the current analysis. In a retrospective review of critically ill trauma patients, there was no relationship between presence of fever in the first 48 hours after trauma and mortality [3]. A randomized trial of aggressive treatment of fever with antipyretics and cooling blankets in non-TBI trauma ICU patients found a trend toward increased mortality, however, results did not reach statistical significance [7]. In contrast, the current study did identify a survival benefit for trauma ICU patients with fever. To our knowledge, this work is one of the first to suggest that fever may be a positive prognostic sign in trauma patients with an infection.
Trauma patients with fever in our cohort of ICU patients with nosocomial infections had the lowest odds of mortality. Trauma patients without fever experienced the second lowest odds of mortality, even lower than non-trauma patients with fever. One explanation for this apparent relative survival benefit of trauma may be the theory of pre-conditioning. First reported in myocardial infarction [21], ischemic pre-conditioning is the finding that brief periods of ischemia protect myocardium against future, longer periods of ischemia. This protective effect of a stressful insult has since been expanded to other fields including trauma, dubbed traumatic pre-conditioning [22–25]. Our findings support this hypothesis of traumatic pre-conditioning, suggesting that trauma itself may induce a stress response that minimizes further physiologic insults. However, prior work has suggested the exaggerated non-specific immune response to trauma may increase susceptibility to infection [26]. The role of fever, as shown in our study, might indicate an appropriate and specific response to infection within the trauma patient population. The mechanisms behind these findings represent an exciting opportunity for future research.
Our study has several limitations. Its retrospective design contains the inherent possibility of selection bias. Although confounders were controlled for in multivariable regression, critically ill trauma and non-trauma surgical patients do have fundamental differences that are difficult to exclude completely with statistical methods, however, together these patients comprise a common population cared for in surgical ICUs and are thus often considered together despite their differences. Other potentially important data points, such as presence of multiple infections, degree of hypodynamic compromise and shock, injury severity score for trauma patients, as well as quantitative TBI and spinal cord injury data were not reliably available given the study's retrospective nature. Additionally, the single-institution design may limit generalizability, especially given the variability in nosocomial pathogens between medical centers and regions. Despite this, we present a large review of a prospectively collected cohort of critically ill trauma and non-trauma surgical patients with infection and identified multiple independent predictors of in-hospital mortality.
In summary, the presence of fever was found to be associated with lower mortality in both critically ill trauma and non-trauma patients with ICU-acquired infections. The presence of fever in trauma patients was associated with the lowest odds of in-hospital mortality, a finding not described previously. Further research is needed to elucidate the physiologic mechanisns behind and clinical implications of these findings.
Footnotes
Authors' Contributions
Concept and design: N.R.E., T.E.H., T.L.H., R.G.S. Data acquisition: T.E.H., E.D.K., K.A.P., R.G.S. Analysis and interpretation: W.J.K., T.E.H., N.R.E., E.D.K., T.L.H., R.G.S. Manuscript drafting and editing: All authors.
Funding Information
This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors.
Author Disclosure Statement
No competing financial interests exist.