Fever in Trauma Patients: Evaluation of Risk Factors, Including Traumatic Brain Injury

Trauma patients were identified by using the St Barnabas Hospital registry from 2008 to 2010. St Barnabas Hospital is a level 1 trauma center located in the Bronx, New York. Patients were included if they were age 15 years or older and admitted for at least 2 days to the ICU. Data were collected until discharge or up to 30 days.

Fever was defined as a body temperature of 100.4°F (38.0°C) or greater and was measured orally.⁸  Fever workup was based on clinical assessment⁹  and could include cultures of blood, sputum, urine, wound, or stool and a Clostridium difficile toxin assay. At the time of the study, patients with TBI were not undergoing a hypothermia protocol.

Patients were classified as having underlying medical conditions if those conditions predated their trauma admission. Patients were counted as intubated except if they were intubated for a surgical procedure and immediately extubated. Procedures included all surgical procedures except those related to long-term care such as tracheotomy and feeding tubes.

We examined fever at 2 different time points: in the first 48 hours to see if fevers correlated with mortality and during the entire ICU stay to investigate if fever reflected infection and mortality. This study was approved by the hospital’s institutional review board.

Groups were compared by using a χ² test. Means were compared by using an independent sample t test. Logistic regression was used for multivariate analysis, with 95% confidence intervals and Hosmer-Lemeshow goodness of fit. The threshold for significance is a P value of .05.

For the 162 trauma patients who met the criteria for inclusion in the study, 31% of admissions were due to falls, 16% to motor vehicle accidents, 14% to being hit by a car as a pedestrian, 14% to stab wounds, 12% to assault, and 11% to gunshot wounds. The study group was predominantly male (81%), with a mean (SD) age of 42 (20) years. The mean (SD) for APACHE II score was 13 (8), for score on the Glasgow Coma Scale was 11 (4), and for the Injury Severity Score (ISS) was 17 (6).

In order to determine if fever in the first 48 hours was a predictor of mortality, we divided all trauma patients into 2 groups: fever or no fever present in the first 48 hours after admission (Table 1).

Patients with fever had significantly lower scores on the Glasgow Coma Scale, higher APACHE II scores, and a higher ISS than did patients without fever. Patients with fever also differed in the number of underlying medical conditions, but no one specific underlying condition (diabetes, dialysis, or alcohol use) was identified as predisposing to fevers. Mortality was greater in febrile than in nonfebrile patients. More febrile patients required intubation and had a surgical procedure (correlated specifically for craniotomy and burr holes). No difference in highest WBC count, highest body temperature, or lowest platelet count was apparent between patients with and without fever.

Table 2 shows the result of the multivariate analysis. Fever correlated with intubation and underlying medical conditions. Fever was not a predictor of mortality.

A total of 82 patients (51%) had fever at some point in their ICU stay. Patients who had a source found for their fever stayed longer in the ICU (mean [SD], 19 [28] days vs 7 [5] days; P=.02) and in the hospital (28 [32] days vs 13 [10] days; P=.01) than did febrile patients without an infectious source found.

Despite fever workup, no source of infection was found in most patients (n=57, 70%). The ability to identify a source of infection did not differ among groups of patients with different risk factors, including TBI, intubation, or undergoing a surgical procedure (Table 3). Patients with an identifiable source for their fever, in contrast to patients where no source was found, had higher maximum body temperature, higher maximum WBC count, and lower platelet counts. Table 3 shows the multivariate analysis where peak WBC count remained significantly higher in febrile patients with an identifiable source for their infection.

The most common infectious source was pneumonia (34%), followed by urinary tract infection (20%), wound infections (14%), abdominal source (14%), and C difficile (12%). Two patients had a deep vein thrombosis. The timing for an infection developing is illustrated in the Figure. Three-quarters of patients with pneumonia were intubated.

The yield of blood cultures was low. Of 341 sets sent, 11 (3%) showed growth of microorganisms. Four grew gram-negative organisms, 6 grew coagulase-negative Staphylococcus, and 1 grew an anaerobe.

Patients with TBI compared with no TBI were older (mean [SD], 45 [21] years vs 38 [18] years; P=.02), had higher APACHE II scores (15 [7] vs 11 [8]; P=.03), and a higher ISS (18 [6] vs 15 [5]; P=.005). They were more likely to be intubated (P= .005), have a surgical procedure (P < .001), and die (P < .001). Age, blood transfusions, underlying medical disease, highest body temperature, highest WBC count, and lowest platelet count did not correlate with TBI. In a multivariate analysis, TBI correlated with mortality and undergoing a surgical procedure (Table 2).

In order to isolate the effect of TBI, patients with TBI were further divided into 2 groups as follows: those with TBI alone and those with TBI plus injuries at other sites. The groups were similar except that more patients in the group with additional injuries had higher ISS, higher APACHE II scores, and had undergone surgery. The multivariate analysis revealed no significant differences between these 2 groups. The mortality in the group with TBI alone was 17%, compared with 30% mortality in the group with TBI plus other injuries (P=.18).

Fever in the first 48 hours did not differ between patients with or without TBI or between patients with TBI alone or TBI plus other injuries: 41% in TBI only, 53% in TBI plus other injuries, 27% in patients without TBI (P = .35). The number of days febrile did not differ between patients with or without TBI.

The source of fever was identified in 30% of febrile patients with TBI and in 35% of febrile patients without TBI (P=.67). The most common source of fever was pneumonia. This occurred in 12% of febrile patients with TBI and 11% of febrile patients without TBI. The highest fever, the greatest WBC count, and the lowest platelet count did not differ between patients with and without TBI.

Fever in the first 48 hours after trauma was not a predictor of dying or of not dying. Fever was seen more often in patients who were intubated, even after the data were adjusted for severity of injury. These results suggest that the “stress” of trauma as manifested by fever did not protect the patient from death. One limitation is that the stress response can manifest in ways other than fevers. Further studies are needed to examine the roles of specific cytokine production after trauma and the roles that cytokines play in fever and survival.

The next question focused on the effect of TBI on fever and mortality. The patients with TBI were similar to the patients without TBI as far as developing fevers in the first 48 hours and the number of days febrile in the ICU. However, fever in TBI patients was not a predictor of mortality as Rincon et al⁷  had reported. Rincon et al, however, included patients with strokes and intracerebral hemorrage in addition to TBI. They reported that fever on admission to the ICU was common and was independently associated with in-hospital mortality.

Our results highlight the difficulty in finding a source of fever in trauma patients admitted to the ICU. Other parameters of infection such as maximum WBC count, highest body temperature, and lowest platelet counts were seen more often in febrile patients with a source of infection then in febrile patients without a source found. Very few patients had a bacteremia, but we did not control for patients who clinically appeared to have sepsis. This finding suggests that infections are not the main reason for fever in these trauma patients. Other possible causes of fever such as the severity of injuries or undergoing surgery were not significant factors on multivariate analysis.

Limitations of this study include the retrospective nature of the study, the number of patients recruited, the lack of measurement of cytokines, and the inability to compare febrile and afebrile patients with infections (as there were too few infected patients). Many questions remain about the etiology of fever and its benefits in trauma patients.

We thank AnnMarie McDonald and the St Barnabas Hospital surgery department for use of their trauma registry.