Abstract
Objective: To identify risk factors for fever among patients treated in a neurologic intensive care unit (NICU).
Methods: The authors prospectively studied the frequency and causes of fever, defined as a patient’s first temperature ≥101 °F (38.3 °C), among 387 patients consecutively admitted to their NICU. After identifying risk factors for 1) any fever, 2) infectious fever, and 3) unexplained fever using logistic regression, they calculated disease-specific adjusted odds ratios for developing these types of fever among 12 diagnostic groups.
Results: Fever occurred in 23% (87/387) of patients. Fifty-two percent of fevers were explained by infection (predominantly pneumonia or bronchitis), and 28% were unexplained despite a complete diagnostic evaluation. NICU length of stay was a risk factor for all three types of fever (all p < 0.004); other risk factors included depressed level of consciousness for any fever (p = 0.005) and infectious fever (p = 0.048), endotracheal intubation for infectious fever (p = 0.01), and intraventricular catheterization for unexplained fever (p = 0.004). Subarachnoid hemorrhage increased the risk of both infectious and unexplained fever, even after adjusting for these risk factors (p = 0.006).
Conclusion: Fever occurs in nearly 25% of NICU patients, and is associated with increased length of stay and depressed level of consciousness. Endotracheal intubation is a risk factor for infectious fever, whereas intraventricular catheterization is a risk factor for unexplained fever, which suggests a role for ventricular hemorrhage in the pathogenesis of “central” fever. Subarachnoid hemorrhage increases the risk of developing fever of all types.
The adverse effect of temperature elevation on neurologic recovery in acutely brain-injured patients has become increasingly recognized in recent years. Even when delayed, 1 to 2 degrees of hyperthermia has deleterious effects on outcome in experimental models of focal and global ischemia and traumatic brain injury (TBI).1-3⇓⇓ The risk of poor functional outcome is increased with even mild temperature elevation (>37.5 °C) on admission after ischemic stroke4,5⇓ or intracerebral hemorrhage (ICH),6 and sustained high fevers are associated with poor outcome after aneurysmal subarachnoid hemorrhage (SAH).7 Brain temperature elevations have also been associated with elevated intracranial pressure after SAH and TBI.8
Given the potentially harmful effects of hyperthermia in neurocritical care patients, there is increasing interest in the application of devices and pharmacologic interventions that might prevent fever.9-13⇓⇓⇓⇓ Among patients with acute brain injury, it would be useful to identify those at high risk for fever, in order to target suitable candidates for prophylactic thermomodulatory therapy. One retrospective study of fever in neurologic intensive care unit (NICU) patients found an overall frequency of 47% and associations with increased ICU length of stay and cranial (as opposed to spinal) disease.14 We studied the frequency of fever in NICU patients and examined clinical and disease-specific risk factors for its occurrence.
Methods.
Patient population.
We prospectively screened all patients admitted to the Columbia-Presbyterian Medical Center NICU between February 11, 1999 and June 20, 1999 throughout their ICU stay for the presence of fever, defined as any temperature ≥38.3 °C (101 °F).15 As per existing NICU practice, temperature was recorded hourly in all patients using an infrared tympanic thermometer (Genius 3000A, Sherwood Medical, St. Louis, MO). Patients with fever were identified daily by reviewing each patient’s maximal temperature recorded during each 12-hour shift in the nursing charge report. The study protocol was approved by the Columbia-Presbyterian Institutional Review Board. Because of the observational nature of this study and lack of significant risk to patients, the need for written informed consent was waived.
Patient-level variables.
The following variables were recorded for all patients admitted to the NICU during this period at the time of ICU discharge: demographic variables—age, sex, race/ethnicity; clinical variables—fever (present/absent), ICU length of stay (LOS), craniotomy (yes/no), level of consciousness (alert/depressed), limb weakness (present/absent); procedures and instruments (all present/absent)—craniotomy, bladder catheter, arterial catheter, central venous catheter, pulmonary artery catheter, endotracheal tube, intraventricular catheter (IVC).
At the time of ICU discharge, patients were assigned by a single investigator (C.C.) to one of 12 principal diagnostic categories, based on a review of radiology reports and clinical data: ICH, SAH, cerebral infarction, TBI, CNS neoplasm, seizures, respiratory failure, unruptured aneurysm, carotid endarterectomy, interventional neuroradiology (INR) procedure, medical complication, and other neurosurgical procedure. Vital status at hospital discharge was classified as dead or alive.
Diagnostic evaluation of fever.
For each patient who developed a fever, all of the following diagnostic studies obtained within 48 hours of fever onset were recorded: chest radiography (CXR), blood cultures, sputum cultures, urine cultures, and CSF cultures. Blood cultures were obtained separately from a peripheral source and a central venous catheter, when present. Additionally, the results of all Clostridiumdifficile toxin assays and lower extremity Doppler studies obtained in the NICU were reviewed (all patients are routinely screened for deep vein thrombosis between 7 and 14 days in our NICU). These test results and the medical record were reviewed independently by two investigators (C.C. and N.S.), each of whom assigned a diagnostic category for fever according to the following criteria15: explained, infectious—pneumonia (purulent sputum with positive culture, with infiltrate on CXR), bronchitis (purulent sputum with positive culture, without infiltrate on CXR), sinusitis (positive nasal culture, with sinus opacification on CT), urinary tract infection (pyuria with positive urine culture of >100 K colonies), bloodstream infection (two positive blood cultures associated with local IV catheter site inflammation or the systemic inflammatory response syndrome [hyperthermia/hypothermia, tachycardia, tachypnea, and leukocytosis/leukopenia]16), cellulitis (warmth and blanching erythema), C. difficile enteritis (positive toxin assay), meningitis/ventriculitis (polymorphonuclear pleocytosis with positive CSF culture); explained, noninfectious—deep vein thrombosis (positive lower extremity Doppler), atelectasis (characteristic findings in two or more lung segments on CXR), drug fever (resolution of fever after drug discontinuation); unexplained (none of the above criteria identified); or incomplete diagnostic evaluation (lack of CXR or blood, urine, or sputum culture). When the two independent reviewers disagreed initially on the final diagnostic category, the data were re-examined until a consensus was reached.
Statistical analysis.
Differences in proportions were compared using Pearson χ2 test, and normally distributed continuous variables were compared using Student two-tailed t-test. After identifying demographic, clinical, and procedure/instrumentation variables associated with fever in a univariate analysis, significant independent predictors of fever were identified by entering all candidate variables identified in the univariate analysis (p < 0.05) into a backward stepwise logistic regression model, using p < 0.05 as the criterion for final retention in the model. To determine whether specific diagnoses predispose to fever, we calculated unadjusted and adjusted odds ratios (OR) for developing any fever, infectious fever, and unexplained fever within each diagnostic category. Adjusted OR were calculated by evaluating the partial coefficient for each diagnostic category after being added individually to the models of any, infectious, and unexplained fever that we developed previously. Assuming a two-sided χ2 test and a significance level of 0.05, a minimum of 25 patients within any diagnostic category would yield 80% power to detect an unadjusted OR for fever of 2.5 or greater.
Results.
Study population.
Over the 18-week study period, 387 patients were admitted to the NICU. Mean age was 54.1 ± 17.6 years (range 17 to 98). There were 208 men and 178 women studied. Mean NICU LOS was 3.3 ± 5.1 days (range 0 to 38). The number and proportion of patients assigned to each of the 12 diagnostic categories is shown in table 1. Of the 33 patients assigned to the other neurosurgical procedure group, the most common procedures included spinal laminectomy or fusion (n = 9), extracranial-intracranial bypass (n = 4), and encephaloduroangiosynostosis (n = 4). Of the 17 patients assigned to the medical complication group, the most common diagnoses were toxic metabolic encephalopathy (n = 5) and hypoxic-ischemic encephalopathy (n = 4).
Table 1 Risk of developing fever (T ≥38.3 °C) by diagnostic category
Fever of any cause.
Eighty-seven patients (23%) developed fever, defined as at least one recorded temperature ≥38.3 °C. Pneumonia or bronchitis was by far the most common cause of infectious fever, accounting for 42% of cases overall (table 2). The second most common category was unexplained fever (28%). Nineteen percent of patients had no source of fever identified, but did not have a complete diagnostic evaluation.
Table 2 Causes of fever
Seven variables were associated with fever in the univariate analysis (all p < 0.0001): longer ICU LOS; non-alert level of consciousness; limb weakness; placement of a central venous, arterial, or ventricular catheter; and endotracheal intubation (table 3). After entering these seven candidate variables into a stepwise backward logistic regression model, ICU LOS (OR 1.30 per day in ICU, 95% CI 1.19 to 1.41, p < 0.0001) and depressed level of consciousness (OR 2.36, 95% CI 1.30 to 4.31, p = 0.005) were identified as significant independent risk factors for fever.
Table 3 Univariate analysis of risk factors for fever among neuro-ICU patients
Table 1 shows the proportion of patients within each category with fever, and for each diagnostic group, the adjusted OR for developing fever after correcting for ICU LOS and level of consciousness. SAH was the only diagnostic group associated with any fever after adjusting for level of consciousness and ICU LOS; 65% of patients with SAH were febrile overall, with an uncorrected OR of 8.1 (p < 0.001) and a corrected OR of 3.36 (p = 0.006). Patients with TBI also were at increased risk for developing fever, with an overall frequency of 40% and an unadjusted OR of 2.46 (p = 0.03), but the adjusted OR for this association did not reach significance. With the exception of patients who underwent craniotomy for clipping of an unruptured aneurysm or resection of neoplasm, patients admitted to the ICU for monitoring after neurosurgical or INR procedures were at low risk for developing fever.
Infectious fever.
Forty-five patients (12%) were classified as having infectious fever. In a univariate analysis comparing patients with and without infectious fever, the same seven variables that were associated with fever of any cause (see table 3) were also associated with fever of infectious cause (all p < 0.0004, data not shown). In addition, craniotomy was associated with an increased risk of infectious fever (p = 0.01, χ2 test). After entering these eight candidate variables into a backward stepwise logistic regression model, ICU length of stay (OR 1.14, 95% CI 1.07 to 1.21 per day in ICU, p < 0.0001), endotracheal intubation (OR 3.33, 95% CI 1.33 to 8.33, p = 0.01), and depressed level of consciousness (OR 2.40, 95% CI 1.01 to 5.71, p = 0.048) were identified as independent risk factors for infectious fever. SAH was the only diagnostic category independently associated with infectious fever after accounting for these risk factors (adjusted OR 2.96, p = 0.02, see table 1).
Unexplained fever.
Twenty-six patients (7%) were classified as having unexplained (or possibly central) fever. In a univariate analysis comparing patients with and without unexplained fever, the same seven variables that were associated with fever of any cause were also associated with unexplained fever (see table 3, all p < 0.01, data not shown). After entering these seven candidate variables into a backward stepwise logistic regression model, ICU length of stay (OR 1.7, 95% CI 1.02 to 1.13 per day in ICU, p < 0.004) and the presence of an IVC (OR 4.63, 95% CI 1.62 to 13.21, p = 0.004) were identified as independent risk factors for unexplained fever. Of the 27 IVCs placed, 88% were placed in SAH (n = 10) or ICH patients (n = 14) with intaventricular hemorrhage (IVH), whereas 12% (n = 3) were placed in patients with CNS neoplasms. SAH was the only diagnostic category independently associated with unexplained fever after accounting for other risk factors (adjusted OR 2.76, p = 0.046, see table 1).
Mortality.
Hospital mortality was 9.2% (8/87) among patients with fever, compared to 4.1% (12/295) among those without (p = 0.09).
Discussion.
We studied the frequency, causes, and predictors of fever in 387 consecutively admitted NICU patients. Fever defined as any temperature ≥38.3 °C occurred in 23% of subjects. In slightly over half the fever was associated with an infection, and in approximately one quarter no cause could be identified, suggesting central fever. ICU LOS and depressed level of consciousness increased the risk of fever in general, endotracheal intubation predicted infectious fever (which was most often related to lung infection), and IVC placement was associated with unexplained fever. SAH increased the risk of both infectious and unexplained fever, even after adjusting for these risk factors, which may reflect hypothalamic dysfunction and abnormal thermo-regulation in these patients.
The frequency of fever in our neurocritical care population (23%) is similar to that among hospitalized patients in general (29%),17 but lower than has been reported in neurosurgical ICU patients (47%),14 which may reflect differences in patient acuity: we treated substantially fewer TBI patients than the aforementioned study, and had more patients who underwent uncomplicated neurosurgical or endovascular procedures. According to admission diagnosis, the highest rate of fever in our study occurred in patients with SAH (65%), followed by TBI (40%) and ICH (31%). A higher frequency of fever (approximately 70%) was reported within 3 days of admission for SAH and TBI in a small retrospective study of randomly selected ICU patients,18 and among general medical-surgical ICU patients.19
We identified non-alert level of consciousness and increased ICU LOS as independent risk factors for developing fever from any cause. Depressed level of consciousness may predispose to fever because of impaired thermoregulation related to brain injury, or from complications resulting from prolonged immobilization, such as atelectasis, pneumonia, and urinary tract infection. Increased ICU LOS has previously been associated with fever in neurosurgical and general medical-surgical ICU patients14,19⇓; however, it remains unclear whether fever per se is a cause or effect of increased ICU LOS. We did not calculate admission Acute Physiology and Chronic Health Evaluation II scores, a global measure of disease severity and mortality risk for ICU patients, which are predictive of fever in medical-surgical ICU patients.19
Fever was associated with infection in 50% of our subjects (see table 2), which is nearly identical to the proportion of fevers attributed to infection in medical-surgical ICU patients (53%).19 Pulmonary infections (bronchitis and pneumonia) accounted for 82% of infections in our study, and are the predominant cause of infectious fever in patients with stroke,20,21⇓ whereas a more diverse spectrum of infections occur in hospitalized or critically ill medical-surgical patients with fever.17,19⇓ Coma is a well-established risk factor for nosocomial pneumonia in neurosurgical ICU patients.22,23⇓ In addition to depressed level of consciousness and ICU LOS, we also identified endotracheal intubation as an independent risk factor for infectious fever, which probably reflects the additional risk of ventilator-associated pneumonia in these patients.24 Noninfectious processes such as deep vein thrombosis and atelectasis were infrequently associated with fever in our population, in accordance with studies challenging the notion that these conditions often generate a febrile response.25,26⇓
No cause of fever was identified despite a complete diagnostic evaluation in 28% of our patients, suggesting fever of central origin. Although regularly encountered and routinely diagnosed by exclusion in clinical practice, very little empiric research regarding central fever has been performed to date, which may relate in part to difficulty in establishing the diagnosis with certainty. Refractory high fever (>42 °C) in the immediate aftermath of massive supratentorial or brainstem ICH is well described, and is the basic observation supporting central fever as a clinical entity.27-29⇓⇓ In 330 hospitalized patients with stroke, unexplained fever occurred in 15%, and was associated with earlier onset than infection-related fever; no other predictors were identified.20 In our NICU patient population, increased ICU LOS and IVC placement (which was usually performed for IVH) were associated with unexplained fever. Although fever in general has been associated with IVH in ICH patients6 and IVC placement after SAH,7 to our knowledge the association between IVC placement and unexplained fever that we found is the first clinical observation to corroborate animal experiments demonstrating hyperthermia after intraventricular blood or albumin injection,30,31⇓ and to support the notion that IVH can cause central fever in humans. The mechanism by which IVH may alter hypothalamic function and cause central fever is speculative: direct hemotoxic damage to thermoregulatory centers in the preoptic region, interference with tonic inhibitory inputs from the lower midbrain that ordinarily suppress thermogenesis, and stimulation of prostaglandin production leading to temperature set-point elevation have all been invoked.32
We found that patients with SAH have enhanced susceptibility to both infectious and unexplained fever, even after controlling for the risk factors that we identified (see table 1). Of note, patients with TBI appeared to have increased susceptibility to infectious fever, but this association failed to reach significance (see table 1). The association of fever with SAH is well described, and its development 4 to 6 days after onset has been correlated with the development of cerebral vasospasm.33 Our data suggest that patients with SAH may have a generalized disturbance of thermoregulation, which might result from either hypothalamic dysfunction and set-point elevation related to the frequent presence of thick clot in the suprasellar cistern or impaired heat-dissipating mechanisms resulting from intense activation of the sympathetic nervous system and generalized peripheral vasoconstriction.34 In this view, whether the initial pyrogenic stimulus is infectious or central in origin, body temperature may tend to run higher in SAH than in other neurocritical care patients.
Several weaknesses of our study deserve mention. For the sake of simplicity, we only analyzed first fever in the NICU, so many patients initially classified as having infectious fever may have had unexplained fevers later on, and vice versa. We used tympanic thermometers, which may be unreliable in critically ill patients35,36⇓; although this does not limit the internal validity of our study, it may limit its applicability to patients with core temperature measurements. We did not record many clinical details, such as the day of onset and duration of fever, maximum body temperature, or response to treatment. We also did not systematically evaluate brain CT scans in order to evaluate radiographic predictors of fever. Some of the diagnostic categories we analyzed were represented in small numbers, resulting in inadequate statistical power to detect associations between these conditions and fever. Finally, because this study was strictly observational, meaning that the evaluation of fever was ultimately left to the discretion of the attending physician, 19% of our febrile patients lacked a complete diagnostic evaluation as we defined it. Hence, we have probably underestimated the proportion of febrile NICU patients who have infectious or unexplained fever.
Our findings may have implications for clinical practice. Mortality was more than twice as high in febrile than in nonfebrile patients (9.2 vs 4.1%). Although not significant, this trend supports previous observations that fever is associated with poor outcome after acute brain injury.4-7⇓⇓⇓ Our results suggest that the risk of fever in mechanically ventilated and comatose acute stroke and neurotrauma patients is sufficiently high to warrant early active intervention to maintain normothermia and prevent fever. A similar management strategy might also be reasonable for noncomatose SAH and IVC-treated ICH patients. Clinical trials are needed to determine whether interventions of this type are feasible, safe, and effective in terms of improving outcome.
Acknowledgments
Acknowledgment
The authors thank Y. Evelyn Du, PhD, for statistical advice, and Mary Prescutti, RN, for coordinating the nursing protocols involved in this study.
- Received May 17, 2002.
- Accepted October 28, 2002.
References
Letters: Rapid online correspondence
REQUIREMENTS
You must ensure that your Disclosures have been updated within the previous six months. Please go to our Submission Site to add or update your Disclosure information.
Your co-authors must send a completed Publishing Agreement Form to Neurology Staff (not necessary for the lead/corresponding author as the form below will suffice) before you upload your comment.
If you are responding to a comment that was written about an article you originally authored:
You (and co-authors) do not need to fill out forms or check disclosures as author forms are still valid
and apply to letter.
Submission specifications:
- Submissions must be < 200 words with < 5 references. Reference 1 must be the article on which you are commenting.
- Submissions should not have more than 5 authors. (Exception: original author replies can include all original authors of the article)
- Submit only on articles published within 6 months of issue date.
- Do not be redundant. Read any comments already posted on the article prior to submission.
- Submitted comments are subject to editing and editor review prior to posting.
