The traditional normal temperature of 98.6 (37C) is derived from axillary measurements made in 1871. A more recent survey of oral temperatures revealed an average of 98.2 (36.8C) – by this survey, fever was defined as 99 (37.2C) in the early AM and 100 (37.8C) in the evening (Tmin is at 4 – 8:00 AM; Tpeak is 4 – 6:00 PM) [JAMA 268: 1578, 1992]. Also note that elderly have a normal temperature 0.9 (0.5C) less than younger patients. Still, the operational definition of fever has been any temperature greater than 100.4 (38C) [Chest 101: 1644, 1992]. The Society of Critical Care Medicine practice guideline recommends that 101 (38.3C) be used as the cutoff for further evaluation [Crit Care Med 26: 392, 1998]

Fever is a sign of inflammation, not necessarily infection. In fact, ~ 50% of fevers in the ICU are not due directly to infection [JAMA 273: 117, 1995; Intensive Care Med 25: 668, 1999; Neurology 60: 837, 2003]. Note also that the severity of the febrile response does not correlate to the likelihood of infection [Chest 117: 855, 2000]

Post-Operative Fever (surgical trauma, infection, VTE, malignant hyperthermia)

Infection may be absent in as many as 2/3 of fevers in the immediate post-operative period [Surgery 94: 358, 1983]. Most of these fevers will occur as a single episode (if they are recurrent, beware) and infection can generally be ruled out via physical exam. Fever may be a natural response to the “controlled trauma” of surgery.

Tissue Injury

Universal in post-op patients. Can trigger SIRS.


There is no correlation between atelectasis and fever [Heart Lung 17: 166, 1988]. Often, post-operative “atelectasis” is the result of supine (as opposed to upright) films. These fevers are probably the result of surgical trauma. A study of 100 consecutive open heart surgery patients showed that 87% of febrile patients had atelectasis on POD#1, however of the atelectatic patients, only 25% had fever [Chest 107: 81, 1995]


Most post-operative wounds are uncomplicated (skin + subcutaneous tissue) and can be treated with debridement. Antimicrobial therapy should be reserved for evidence of deep tissue involvement [Fry “Postoperative Fever”: NY, Raven Press 243, 1991]. Most wound infections are evident in 5 – 7 days post-op, with the exception of anaerobes which occur within the first few days – look for marked edema and crepitance, treat with debridement and penicillin.

Abscesses can also cause post-operative fevers, and septicemia will occur in 50% of these [Crit Care Clin 4: 345, 1988]. CT is 95% sensitive, and initial microbial therapy should be geared towards Gram-negative enterics including anaerobes. Treat definitively with drainage.

Other infections to think about (see below) – pneumonia, urinary tract, vascular catheters, and rare causes (sinusitis, acalculous cholecystitis, pseudomembranous colitis, etc.).

Venous Thromboembolism

Thromboembolism can produce fevers lasting up to a week [Am J Med 67: 232, 1979]. The leg exam can be normal in up to 50% of cases [NEJM 331: 1630, 1994]. Risk is highest in orthopedic procedures (esp. hip, knees) and lowest following CABG [Chest 99: 284, 1091]

Malignant Hyperthermia

Malignant hyperthermia occurs in 1:15,000 cases of generalized anesthesia and features fever, decreased consciousness, and usually (but not always) muscle rigidity (absent in 20% of cases [J Crit Illness 3: 13, 1988]). Rhabdomyolysis shortly follows. Treat with dantrolene 1 – 2 mg/kg IV bolus to start, repeat q15m.

Post-Procedural Fever

Febrile reactions following hemodialysis is usually attributed to endotoxin-contaminated equipment but bacteremia occurs on occasion [JAMA 260: 2106, 1988]. If toxicity is not suspected, continue dialysis and draw a blood culture, otherwise stop, culture, and start Abx (vanc/ceftaz, avoid aminoglycosides in these patients).

Bronchoscopy produces fever in 5% of cases, but pneumonia and bacteremia are rare and thus Abx are unneccessary unless sepsis (MS, hypotension) or asplenia [Chest 104: 1607, 1993]. Gastrointestinal endoscopy also rarely causes infection. Blood transfusions produce fever in 5% of patients, usually secondary to anti-leukocyte antibodies (platelets cause fever in up to 30%).

Always check heated mattresses and ventilators.

Common Infections

Pneumonia may be an overdiagnosed cause of fever in the ICU and may be the cause of only 19% of fevers in ICU patients with pulmonary infiltrates [Chest 106: 221, 1994] although more recent data dispute this. Pneumonia should be suspected clinically if you find A) a new infiltrate and B) two of the following – fever, leukocytosis, purulent sputum. To diagnose pneumonia you need quantitative cultures of deep specimens.

UTI are common but in chronically catheterized patients, bacteriuria does not equal UTI – one must see pyuria by gram stain or elevated leukocyte esterase.

Catheters should be a suspected source after 48 hours or when purulence is present. In patients who are not too sic, in the absence of purulence, the catheter can be left in place and quantitative cultures can be drawn and compared to peripheral sites. Do not start antibiotics unless an infection is confirmed or the patient becomes symptomatic (MS or hemodynamic changes).

Nosocomial Infections in SICU Patients
Pneumonia 33%
Sepsis 13%
ENT 4%
UTI 18%
Cardiovascular 4%
Skin/Soft Tissue 3%
Surgical Site 14%
GI tract 4%
Other 7%

Infect Control Hosp. Epidemiol. 21: 510, 2000

Note that 83% of pneumonia occur in intubated patients, 97% of UTI occur in patients with catheters, and 87% of sepsis occurs in patients with intravascular catheters [Infect Control Hosp. Epidemiol. 21: 510, 2000]

Wound infections occurs 5-7 days after surgery and can almost always be managed via debridement.

Rare Infections

NG and NT tubes can block draining ostia, leading to paranasal sinusitis in 15 – 20% of patients with nasal tubes. Interestingly, paranasal sinusitis is also found in patients intubated orally. Sinuses be viewed with plain films (although CT is often used) but as radiography is falsely positive in 1/3 of cases, puncture is required for diagnosis (> 103 CFU/mL) [Crit Care Med 21: 1132, 1993]. If Gram stain positive, start vanc and aztreonam/aminoglycoside and await cultures, remove nasal tubes if possible and consider antibacterial paste.

C.dif colitis (on Abx, test for toxin x 2), endocarditis (prosthetic valves), meningitis (NSGY or HIV patients), and spontaneous bacterial peritonitis (cirrhosis, ascites) should be considered as appropriate.

Non-Infectious Causes of Fever

Drug fever is most commonly caused by amphotericin, cephalosporins, penicillins, phenytoin, procainamide, and quinidine (also cimetidine, carbamazepine, hydralazine, rifampin, streptokinase, and vancomycin to a lesser extent). Onset from a few yours to a few weeks. Rigors occur in 53%, with hypotension in 20% (can mimic severe illness). Also, myalgias, leukocytosis, eosinophilia, and rash, occur only ~ 20% each [Ann Intern Med 106: 728, 1987]. Diagnosis drug fever by excluding everything else, then d/c as many as possible and replace those necessary.

Acalculous cholecystitis is a serious disorder that occurs in up to 1.5% of ICU patients [J Intens Care Med 9: 235, 1994], particularly post-op, trauma, or on parenteral nutrition. Causes fever and RUQ pain, attempt to diagnose with U/S. Gallbladder may perforate within 48 hours to treat quickly with cholecystectomy.

Neuroleptic malignant syndrome, a variation of malignant hyperthermia should be considered in any patient on haldol.

SIRS can be triggered by tissue trauma (surgery) or translocation of endotoxin across bowel mucosa. Can lead to MOF and death.

Thryotoxicosis and adrenal insufficiency (secondary to anticoagulation, DIC, or hypotension) are two metabolic disorders known to cause fever.

Bowel infarction can cause fever but is notoriously difficult to diagnose – look for gas in the bowel wall or portal venous gas, but only laparatomy is definitive.

Pancreatitis and adrenal insufficiency are rare, but possible non-infectious causes of fever.

Early Management Decisions

Blood cultures should be obtained in every case of nosocomial fever that is not early-postoperative. Do not take more than one set from any one venipuncture site [AIM 106: 246, 1987]. Number of cultures is proportional to the likelihood of septicemia – two for pneumonia, three for vascular catheter infection, etc. Get four cultures if antibiotics are already started. Optimal volume is 20 – 30 mL per set of cultures [AIM 106: 728, 1987]

If an infectious site has not been determined, reserve empiric antibiotics for 1) high likelihood 2) evidence of septic shock or organ dysfunction or 3) for immunocompromised patients. If needed, Marino recommends starting with vanc and aztreonam, as there is no evidence that aminoglycosides are better than aztreonam for Gram negatives in immunocompetent individuals and their renal toxicity is worth avoiding [Marino]. A review article of RCTs studying antibiotic therapy in intraabdominal infections reported on the clinical success rate of for several antibiotic regimens in a larger population (cefotetan 92%; piperacillin/tazobactam ~ 90%; aztreonam + clindamycin 89%; meropenem 89%; cefoxitin 88%; ampicillin/sulbactam 87%; cefotaxime + metronidazole 87%; imipenem 85%; tobramycin + clindamycin 83%; moxalactam 83%; gentamicin + clindamycin 80%) but recommends use dependent on the expected pathogens and the resistance rate in a clinical setting [Eur J Med Res 30:6 277, 2001]

Antipyretic therapy should probably be avoided in the ICU. It may be a natural defense mechanism and has been shown to inhibit the growth of Pasturella Multocida [Science 203: 374, 1979]. A recent RCT of 82 trauma ICU patients showed a trend towards decreased mortality in patients allowed permissive pyrexia (treatment initiated at 40C as opposed to 38.5C, 1 vs. 7 deaths, p=0.06). They also had a statistically insignificant decrease in infections (3 vs. 4 per patients, p=0.26) and ICU stay (751 vs. 961 total days) [Surg Infect 6: 369, 2005]. Two other studies of 455 and 382 patients with severe sepsis show that the mortality rates are in febrile patients are half those of hypothermic patients [Crit Care Med 18: 801, 1990; Crit Care Med 27: 699, 1999] – the only exception to this is in the early period following ischemic brain injury, as fever has been shown to increase mortality in stroke patients [Stroke 31: 410, 2000]

Specific Patient Populations

Infection in the Neurosurgical Patient


Patients undergoing craniotomies have a high rate of nosocomial infection [PNAS 91: 2490, 1994]. Foreign bodies increase this risk, as do draining systems. Despite their prevalence, the efficacy of prophylactic antibiotics has yet to be proven [Heart Lung 19: 84, 1990; Heart Lung 24: 166, 1995]. S.aureus is the most common cause of pneumonia in ventilated patients s/p head trauma or neurosurgery [Chest 102: 525, 1992; J Infect Dis 160: 414, 1989]. Anaerobes must also be considered in the neurosurgical patient, as aspiration is often a risk.


Blood stream infections are common in neurosurgical patients, with 90% or more caused by central venous catheters [Maki, D: Hospital Infections]. PICC lines have lower infection rates than centrally inserted catheters [AIM 153: 1791, 1993], and there is some evidence that application of antimicrobial agents at the insertion site can decrease the incidence of infection [Crit Care Clin 14: 339, 1998]

Subgaleal abscess

Between the galea and pericarnium, usually staphylococcus, streptococci, and anaerobes. Scalp tenderness and swelling. Treat via surgical drainage and debridement plus antibiotics.


Diagnosis requires clinical findings, cultures, and ESR. Imaging can confirm. Staphylococcus, anaerobes, and pseudomonas are common. Treat with drainage, debridement of infected bone, and 6 weeks of antibiotics.

Incidence after lumbar discectomy is 0.8%, which increases to 6-8% after the addition of hardware [Ortho Clin North Am 27: 87, 1987]. Presentation in spine patients is usually associated with muscle spasm (neurologic signs are late). MRI is the most accurate diagnostic technique. Staphyloccocus is the most common pathogen, and most cases are treated successfully with 4-6 weeks of antibiotics plus immobilization. Open surgery is indicated in patients who develop neurologic deficits or dangerous deformities (10-20% – anterior approaches are usually indicated, as direct access to the infected tissue can be obtained (involvement of the posterior elements are rare). ESR can be followed to gauge response to therapy.

Epidural Abscess

Usually due to extension of osteomyelitis, rarely due to trauma or operative contamination. Treat with drainage, debridement, and systemic antibiotics. Spinal epidural abscesses are an emergency – S.aureus is most common. MRI is the diagnostic modality of choice. Treat with urgent decompression and 4-6 weeks of antibiotic therapy.

Subdural Empyema

Makes up 15-20% of all intracranial infections. Causes hydrocephalus and cerebral edema, eventually leading to benous thrombosis, parenchymal infarction, and thus has a mortality rate 10-20%. Lumbar puncture is forbidden in these patients. Treatment is neurosurgical evacuation – studies suggest that primary craniotomy with wide exposure, aggressive subdural exploration, and careful debridement yield better results than simple burr hole evacuation [Stroke 21: 122, 1990; J Neurol Neurosurg Psych 50: 1136, 1987] although for small, asymptomatic collections, empiric antibiotics may suffice [Neurosurg Clin N Am 3: 359, 1992]

Brain abscess

Accounts for 1-2% of intracranial space occupying lesions. Headache is present in 70%, but fever is present in only 50%. Seizures occur in 30-50%. Lumbar puncture is contraindicated as the sensitivity is < 10% [Adv Intern Med 43: 403, 1998]. Note that 30-60% of cases will be polymicrobial. Indications for surgery and adjuvant steroids are controversial. Medical therapy may be appropriate in high risk patients or those with early stage cerebritis. In all other patients, surgical therapy followed by 4 weeks of antibiotics is recommended [Surg Neurol 17: 338, 1982]. Still, surgery will only lead to identification of the organism in 60-80% of cases.

Post-operative infection

Neurosurgical wounds are different from other wounds in that the dura protects the CSF from the environment. Neurosurgical wounds can therefore not be drained unless the dura is known to be sealed. Post-craniotomy meningitis incidence is 0.34%. The most common causes are staphylococcus and gram negatives (the latter of which carries a 70% mortality rate).


In patients with bacterial meningitis, the sooner antibiotics are administered, the better the outcome [AIM 129: 862, 1998]. To review the risks of lumbar puncture and the benefits of obtaining an prior CT scan, a group prospectively reviewed 301 ED patients with suspected meningitis. 235 received CT prior to LP. Baseline clinical features associated with a high risk for abnormal findings on CT were age greater than 60, immunocompromise, history of a CNS lesion, a seizure within 1 week before presentation, an abnormal level of consciousness, and abnormal focal signs on examination (using the modified NIH stroke scale). Of 96 patients undergoing initial CT who had no risk factors for abnormal CT, only 3 had abnormal CT scans; none of the 3 had complications of the tap. Of 11 patients with mass effect on CT, 10 were identified based on the presence of 1 or more risk factors at baseline. Four patients had abnormal CT findings that precluded lumbar puncture; all 4 had at least 1 of the baseline risk factors. Lumbar puncture was performed an average of 2 hours later in the group that underwent initial CT than in the group that did not first undergo CT [NEJM 345: 1727, 2001]

In adults without the aforementioned risk factors, immediate lumbar puncture and CSF/blood cultures followed by antibiotics are the best management. In the presence of risk factors, management should be: 1. blood culture (which may be positive even when spinal cultures are negative; see [NEJM 328: 21, 1993]), 2. IV antibiotics, 3. CT scan, followed by lumbar puncture unless the CT scan indicates a contraindication.

Common etiologies are treated for 7-14 days, whereas gram negative etiologies are often treated for 21 days [AIM 112: 610, 1990]. Housemates of infected patients should receive Rifampin prophylaxis. Dexamethosone may reduce neurologic disability and/or mortality by ~ 50% [NEJM 347: 1549, 2002]

Empirical Treatment of Meningitis

Normal Adult
Ceftriaxone 1g IV a12h
Vancomycin 1 g IV q12h
(Ampicillin if you suspect Listeria)
Dexamethazone 10 mg IV q6h x 4 days
B-lactam Allergic Adult
Chlormamphenico + Bactrim + Vancomycin
Ampicillin + ceftazidime +/- vancomycin + acyclovir
S/p neurosurgery, head trauma, or shunt
Vancomycin + ceftazidime

Shunt infections

Occur in 5-15% of shunt procedures and in 15-25% of patients with hydrocephalus who receive shunts [Childs Nerv Syst 3: 106, 1987; J Neurosurg 61: 1072, 1984; J Neurosurg 60: 1014, 1984; Medicine 69: 244, 1990]. Staphylococcus are the most common culprits, however in infants, Enterobacteriaceae is also common. In general, shunt removal is necessary, although in occasional instances shunts can be salvaged by using a combination of systemic and intrathecal antibiotics [J Neurosurg 60: 354, 1984; Neurosurgery 24: 96, 1989]

Antibiotic prophyaxis

Has not been validated, because this incidence of infection is low (3-4%) in neurosurgical cases, thus a statistically significant study would require a huge enrollment [J Antimicrob Chemother 31 Suppl. B: 49, 1993]. More recent studies have suggested that the postoperative infection rate is decreased by antibiotic prophylaxis [J Neurosurg 69: 216, 1991; J Neurosurg 69: 687, 1998; J Neurosurg 73: 383, 1990; Scand J Infect Dis 20: 633, 1998; J Neurosurg 66: 701, 1987], however most of these studies were too small to reach a statistically significant conclusion [J Antimicrob Chemother 31 Suppl. B: 49, 1993]. [Scand J Infect Dis 20: 633, 1998;], a study of 310 patients given cloxacillin, did reach statistical significance in favor of antibiotics. Given these studies and trends, the use of first or second generation cephalosporins is now broadly supported [Clin Ther 18: 84, 1996; Orthop Clin N Am 27: 47, 1996; Neurosurgery 9: 142, 1981; J Neurosurg 62: 243, 1985]

In neurosurgical implant procedures, the incidence of infection is higher (10-15%). In CSF leaks secondary to trauma, the infection rate is 11-25% [Lancet 1: 1013, 1973; Br J Surg 42: 1, 1954] and the primary pathogen is S.pneumonia.

Ampicillin and penicillin G achieve good CSF levels in infected patients. Cefotaxime, ceftriaxone, and metronidazole also have good CSF penetration.

Rare infections

Cryptococcus is the most common fungal meningitis. Coccidiodies can also cause meningitis but is difficult to diagnose as meningeal signs are present in only 30% and cultures are positive in 20-50%. Candidiasis usually presents as multiple microabscesses.