on 08-Jun-2025 (Sun)

Though the signifi cance of fever was long recognized and Galileo had an instrument for measuring air temperature in 1600, the path- breaking work on medical thermometry was not done until around 1850 by syphilologist Felix von Bärensprung and famous clinician Ludwig Traube (see Chapter 17). Finally, the Hippocratic “cri- sis” and “critical days” were shown to correspond to temperature curves. Rapid defervescence occurred with “lysis” of the illness

Carl August Wunderlich then brought thermometry into clini- cal use. Over 15 years, all patients entering his clinic or attended by him in hospital had their temperature regularly measured and recorded. Wunderlich’s work was ridiculed as diffi cult and costly “fruitless fi ddling.” Indeed, the early thermometers were inconve- nient, measuring 25 cm long and so plump that the English said they had to be carried under the arm “as one might carry a rifl e.”

The temperature has traditionally been measured rectally or orally. The latter should be performed by placing the thermom- eter bulb under the tongue with the lips kept closed. With a mer- cury thermometer that has been well shaken down, equilibration requires about 3 minutes, while maximum accuracy requires 10 minutes (Blumenthal, 1992). With newer instruments using a thermistor, less than 60 seconds may be required.

The rectal temperature is usually about 1°F (0.55°C) higher than the oral temperature. In patients who are tachypneic (breath- ing more than 20 times per minute), the difference averages 1.67°F (0.93°C), increasing with the respiratory rate and becoming inde- pendent of whether the patient is mouth breathing or not (Tandberg and Sklar, 1983). Other causes for falsely low oral temperatures include recent ingestion of cold substances or failure to keep the lips closed.

Falsely high oral temperatures may result from failure to shake down the thermometer, ingestion of hot substances, or smok- ing. Patients should refrain from eating, drinking, or smoking for at least 15 minutes prior to an oral temperature measurement (Mack- owiak et al., 1992)

Axillary temperatures are said to be extremely inaccurate. Others claim it is as accurate as temperatures from other locations, if properly done: lie still, keep arm tight, and wait exactly 10 minutes (H. Nehrlich, personal communication, 2007)

With one batch of 24 cheap thermometers imported from China, the temperature reading of a 37.0°C water bath ranged from 36.7°C to 37.2°C (mean = 36.96°C, SD = 0.15).

Infrared tympanic thermometers, though widely used because of convenience and speed, have been shown by some to be unreliable in clinical practice. Sensitivity for fever detection in both children and adults is as low as 55% to 70%. It has been found that subjec- tive assessments by mothers were more sensitive for detecting fever than the infrared tympanic thermometers used in hospital emer- gency departments. Sequentially measured right and left ear tem- peratures have not uncommonly differed by as much as 1°C (Modell et al., 1998).

Ear canal occlusion by cerumen has been found in one study, surprisingly, to have no effect on the measured temperatures (Modell et al., 1998) and, in another, to make tympanic tempera- ture measurements even less reliable (Abolnik, 1999)

In nonintensive care–hospitalized adult patients, the 95% limits of agreement for oral electronic versus tympanic membrane tem- peratures were −2.11°F to +2.81°F (Manian and Griesenauer, 1998)

Others have reported more encouraging results. Excluding patients with complete occlusion of the ear canals with cerumen, Smitz et al. found “acceptable” sensitivity and specifi city for predict- ing rectal fever in older hospital inpatients. The correlation coef- fi cient between rectal and infrared emission detection (IRED) ear temperatures was 0.78. With 37.2°C (99°F) selected as the fever threshold for IRED temperatures, the sensitivity was 86%; the specifi city, 89%; the positive predictive value, 80%; and the nega- tive predictive value, 93%. The sensitivity dropped to 50% when the fever threshold was raised to 37.6°C. In 3 of 45 patients, the difference in IRED and rectal temperatures exceeded 1°C. The authors cautioned that selective brain cooling might occur dur- ing hyperthermia, especially in dehydrated patients (Smitz et al., 2000). It is important to remember that these investigators used calibrated thermometers, examined the ears, and used the highest of six measurements

With a single calibrated instrument, tympanic membrane tem- peratures in adult ICU patients correlated well with pulmonary artery temperatures (r = 0.909) (Klein et al., 1993)

This author suggests that students calibrate the instrument they always have with them—the back of their hand—against readings of known accuracy.

Another method is to use an electronic thermometer to measure the temperature of the urine as it is voided. Data and nomograms are available (Murray et al., 1977). If the oral temperature is 38°C, the expected urine temperature is 37.3°C (lower 99% confi dence limit = 36.15°C). For an oral temperature of 39°C, the comparable fi gures are 38.15°C and 36.95°C. Thus, if the patient’s measured oral temperature is 40°C and the urine temperature is only 37.7°C, one should suspect factitious fever.

Factitious fever is to be distinguished from factitious infection in which fever is a secondary but genuine phenomenon. As mercury thermometers vanish, so too should this condition.

The fi gure of 98.6°F (37°C) as the upper limit of normal for temperature is credited to the 19th century investigator Carl Wunderlich, who reportedly analyzed more than 1 million axil- lary temperature readings from 25,000 individuals. He found a mean temperature of 98.6°F (37°C), with a range from 97.2°F (36.2°C) to 99.5°F (37.5°C). The minimum occurred between 2:00 and 8:00 AM and the maximum between 4:00 and 9:00 PM. He considered temperature readings in excess of 100.4°F (38°C) to be “suspicious” and “probably febrile.” He thought that women had slightly higher temperatures and exhibited greater and more sudden changes than men, that there might be racial differences in body temperature, and that old people had a temperature about 0.9°F (0.5°C) less than young persons. Recent tests conducted with one of Wunderlich’s instruments suggest that it may have been calibrated as much as 1.4°C to 2.2°C (2.6°F to 4.0°F) higher than today’s instruments (Mackowiak, 1998), casting doubt on some cherished dictums about the special signifi cance of 98.6°F or 37°C

In the 1930s and 1940s, Barnes, in exploring the hypothesis that basal temperature mirrored the basal metabolic rate, found that the normal axillary temperature ranges between 97.8°F and 98.2°F (Barnes, 1942)

A modern study of 700 temperature readings in 148 healthy v o l u n t e e r s f o u n d a m e a n o r a l t e m p e r a t u r e o f 9 8 . 2 ° F ( 3 7 ° C ) a n d s u g - gested that the upper limit of normal for the oral temperature be regarded as 98.9°F (37.2°C) in the morning and 99.9°F (37.7°C) overall. Investigators corroborated Wunderlich’s opinion that women have a slightly higher normal temperature (Mackowiak et al., 1992).

Wunderlich wrote that the diurnal variation in normal tem- perature is 0.9°F (0.5°C). This was recently corroborated by one investigator who took his own oral temperature every 5 minutes for a day, fi nding that it fl uctuated over a range of 97.3°F to 97.9°F (N. Robinson, unpublished observations, 2003). Mackowiak found daily oscillations as wide as 2.4°F (1.3°C)

variation in ovulating women, with the early morning temperature (taken with the patient still in bed, just after awakening) dropping slightly just before the onset of menstruation. There may be a fur- ther drop just prior to ovulation, followed by a rise coincident with ovulation. Thereafter, the temperature remains at that level until just prior to the next menstruation.

An elevation in temperature is usually a sign of an infectious or infl ammatory condition, although it may also result from thy- rotoxicosis, heat stroke, neoplasia, drugs, and many other entities.

Relapsing fevers have afebrile days alternating with days of fever. Diseases characterized by relapsing fevers include familial Mediterranean fever; brucellosis in which the fever is associated with physical activity and disappears on days of bed rest; Hodgkin disease; Borrelia infections; tuberculo- sis, especially extrapulmonary; and malaria.

The Pel–Ebstein relapsing fever of Hodgkin disease occurs in about 16% of cases and is quite variable. It may last for hours to days, followed by days or weeks without fever

A double quotidian fever occurs twice a day and is seen in about half the cases of gonococcal endocarditis

A double-humped (“saddle-backed”) fever curve (with two peaks during 1 week and a low-grade fever in the valley) is seen in some viral diseases such as West Nile fever, dengue, and Bornholm disease

Tertian fevers return on the third day and so have a periodicity of 48 hours. They signify infection with Plasmodium vivax or P. ovale. Quartan fever is one that returns on the fourth day and so has a 72-hour periodicity. It is seen with P. malariae.

A remittent fever is one that falls each day but not quite to nor- mal, remaining at 99.2°F (37.3°C) or above

An intermittent fever falls to normal or below each day.

In remittent or intermittent fever, the excursion in temperature is more than 0.3°C (0.5°F) and less than 1.4°C (2.5°F).

A sustained fever has a less than 0.3°C (0.5°F) fl uctuation during a 24-hour period

A hectic fever is a remittent or intermittent fever with a difference of 1.4°C (2.5°F) or more between peak and trough

The latter four fever patterns are of minimal diagnostic value, as shown in a study of 200 patients with single episodes of fever (Musher et al., 1979). A sustained fever occurred frequently enough in cases of Gram-negative pneumonia and of central nervous system damage (and infrequently enough in other conditions) to suggest, but not establish, one of these two diagnoses. The absence of the usual diurnal effect (maximum temperature between 4:00 PM and midnight) supported but did not prove a noninfectious etiology, especially central nervous system damage. Impairment of hypotha- lamic control of temperature was thought to be the mechanism of the fever associated with central nervous system damage

Lesions in the posterior portion of the hypothalamus may be attended by hypothermia or poikilothermia, the latter possibly passing unnoticed unless the patient’s temperature is taken after changing the room temperature. Somnolence and hypotension may also be associated with such lesions (Adams and Victor, 1981)

The most common hypothalamic fever is part of the hypnotic– sedative withdrawal syndrome. It appears only during acute absti- nence and abates with treatment

Extremely high fevers (in excess of 106°F or 41.2°C) are rarely attributable to infection, with the exception of central nervous system infections such as bacterial meningitis or viral encephalitis. In my own experience, heat stroke has been the most common cause. The temperature obtained with a regular thermom- eter may be a dangerous underestimate. Such patients should be monitored immediately with a thermistor placed high in the rec- tum, as aggressive cooling has begun

Some central nervous system lesions cause a disturbance in the body’s temperature regulation. After operations in the region of the fl oor of the third ventricle, the temperature may rise to 106°F or higher and remain there until death, hours or days later. Icy cold- ness of the extremities, dry skin, tachycardia, and tachypnea are also present (Adams and Victor, 1981)

Fever is usually accompanied by an increase in the pulse rate of approximately 10 beats per minute per degree Fahrenheit. With some infections, the pulse is characteristically slower than expected: salmonelloses (including typhoid fever), tularemia, brucellosis, bacterial meningitis complicated by increased intracranial pres- sure, mycoplasmal pneumonia, rickettsialpox, Legionella pneumo- nia, mumps, infectious hepatitis, Colorado tick fever, and dengue. A disproportionately low pulse may also be seen in factitious fever or in patients taking digitalis glycosides or beta-blockers

For any rapidly changing variable, which moves widely about a mean (such as platelet count or temperature), a cusum (cumulative sums) plot will help you determine when a real change has taken place. To make such a plot, arbitrarily defi ne a standard value, such as 100°F for the temperature. Make a list of the temperatures and of their deviations from the standard value and calculate the cusum, which is the cumulative sum of the deviations (Table 6.2). Plot the cusums on a graph (Fig. 6-9). The change point is the point at which there is a change in the slope. This may help you determine which antibiotic made a difference in the fever. Another technique is to record the maximum temperature for each 24-hour period

In fever, as opposed to a failure of thermoregulatory homeostasis, the core temperature rise is mediated by cytokines and accompanied by the release of acute phase reactants and the activation of numerous immunologic and endocrinologic systems (Mackowiak, 1998). An inability to mount a fever when appropri- ate may signify a poor prognosis

When examining or reporting on the course of a fever, be sure to consider whether antipyretics have been given. The proper use of such agents has been the subject of considerable controversy in recent years

Hypothermia is defi ned as an oral temperature of less than 95°F (35°C). Because routine thermometers do not read temperatures this low, it may be missed unless the examiner suspects it and checks the temperature with a thermistor.

Six specifi c causes of severe hypothermia are reversible and may require emergency treatment: hypoglycemia, hypothy- roidism, hypoadrenocorticism, overwhelming infection (Bryant et al., 1971), intoxications, and exposure

With a severe depression in body temperature, metabolic pro- cesses slow, and the patient may resemble a person with myxedema or may even appear to be dead. Hypothermia prolongs the time that a person may survive anoxia

Thermometers are used to measure the temperature of the patient’s oral cavity, rectum, axilla, tympanic membrane, or forehead (i.e., temporal artery). Because of potential toxicity from mercury exposure, the time-honored mercury thermometer has been replaced by electronic thermometers with thermistors (oral, rectal, and axillary mea- surements) and infrared thermometers (tympanic or forehead measurements)

Normal body temperature varies widely, depending in part on the site mea- sured. Rectal readings are on average 0.4 to 0.6°C higher than oral ones, which are 0.1 to 0.2°C higher than axillary readings.5-8

Temporal (forehead) measure- ments typically fall between rectal and oral readings.7,9

Tympanic readings are the most variable, with some studies showing them to be systematically higher than rectal readings10 and others showing them to be systematically lower than oral readings.11

Even so, these studies, which are designed to detect systematic differences between instruments, do not reflect the variability observed in individual patients. For example, comparisons of sequential rectal and oral readings measured in large numbers of patients reveal the rectal-minus-oral difference to be 0.6 ± 0.5°C.10 This indicates that on average rectal readings are 0.6°C greater than oral read- ings (i.e., the systematic difference), but it also indicates that the rectal reading of a particular patient may vary from as much as 0.4°C lower than the oral reading to 1.6°C higher than the oral reading.*

A better question is how well different instruments detect infection. In one study of elderly patients presenting to an emergency department, three different techniques—rectal, temporal, and tympanic measurements—had similar diagnostic accuracy for infection (likelihood ratios [LRs] 4.2 to 8.5; EBM Box 18.1), although each instrument had a different definition of fever (rectal T >37.8°C; forehead T >37.9°C; tympanic T >37.5°C).9

EATING AND SMOKING5,12-14 The oral temperature measurement increases about 0.3°C after sustained chewing and stays elevated for up to 20 minutes, probably because of increased blood flow to the muscles of mastication. Drinking hot liquids also increases oral readings about 0.6 to 0.9°C, for up to 15 to 25 minutes, and smoking a cigarette increases oral read- ings about 0.2°C for 30 minutes. Drinking ice water causes the oral reading to fall 0.2 to 1.2°C, a reduction lasting about 10 to 15 minutes

TACHYPNEA Tachypnea reduces the oral temperature reading about 0.5° C for every 10 breaths/ minute increase in the respiratory rate.15,16 This phenomenon probably explains why marathon runners, at the end of their race, often have a large discrepancy between normal oral temperatures and high rectal temperatures.17 In contrast, the administration of oxygen by nasal cannula does not affect oral temperature.18

CERUMEN Cerumen lowers tympanic temperature readings by obstructing the radiation of heat from the tympanic membrane.5

HEMIPARESIS In patients with hemiparesis, axillary temperature readings are about 0.5°C lower on the weak side compared with the healthy side. The discrepancy between the two sides correlates poorly with the severity of the patient’s weakness, suggesting that it is not due to difficulty holding the thermometer under the arm, but instead to other factors, such as differences in cutaneous blood flow between the two sides.19

MUCOSITIS Oral mucositis, a complication of chemotherapy, increases oral readings on average by 0.7°C,20 even without fever. This increase in temperature likely reflects inflam- matory vasodilation of the oral membranes.

Most studies show that a temperature greater than 37.8°C with any instrument is abnormal (and therefore indicative of fever).6

In the early days of clinical thermometry, clinicians observed that prolonged fevers could be categorized into one of four fever patterns—sustained, intermittent, remit- tent, and relapsing (Fig. 18.1).3,22-24

(1) Sustained fever. In this pattern the fever varies little from day to day (the modern definition is variation ≤0.3°C [≤0.5°F] each day)

(2) Intermittent fever. In this pattern the temperature returns to normal between exac- erbations. If the exacerbations occur daily, the fever is quotidian; if they occur every 48 hours, it is tertian (i.e., they appear again on the third day); and if they occur every 72 hours, it is quartan (i.e., they appear again on the fourth day).

(3) Remittent. Remittent fevers vary at least 0.3°C (0.5°F) each day but do not return to normal.

Hectic fevers are intermittent or remittent fevers with wide swings in temperature, usually greater than 1.4°C (2.5°F) each day.

(4) Relapsing fevers. These fevers are characterized by periods of fever lasting days interspersed by equally long afebrile periods

Each of these patterns was associated with prototypic diseases: sustained fever was associated with lobar pneumonia (lasting 7 days until it disappeared abruptly by crisis or gradually by lysis); intermittent fever with malarial infection; remit- tent fever with typhoid fever (causing several days of ascending remittent fever, whose curve resembles climbing steps before becoming sustained); hectic fever with chronic tuberculosis or pyogenic abscesses; and relapsing fever with relapse of a previous infection (e.g., typhoid fever).

Other causes of relapsing fever are the Pel- Ebstein fever of Hodgkin disease,25 rat-bite fever (Spirillum minus or Streptobacillus moniliformis),26 and Borrelia infections.27

Despite these etiologic associations, early clinicians recognized that the diagnostic significance of fever patterns was limited.28 Instead, they used these labels more often to communicate a specific observation at the bedside rather than imply a specific diagnosis, much like we use the words “systolic murmur” or “lung crackle” today

Over 80% of patients with bacterial infections have specific focal signs or symptoms that point the clinician to the correct diagnosis.29

One potentially misleading focal sign, however, is jaundice. Although fever and jaundice are often due to hepatitis or cholangitis, jaundice is also a nonspecific complication of bacterial infection distant to the liver, occurring in 1% of all bac- teremias.30,31 This reactive hepatopathy of bacteremia was recognized over a century ago by Osler, who wrote that jaundice appeared in pneumococcal pneumonia with curious irregularity in different outbreaks.28

Relative bradycardia, a traditional sign of intracellular bacterial infections (e.g., typhoid fever), refers to a pulse rate that is inappropriately slow for the patient’s temperature. One definition is a pulse rate that is lower than the 95% confidence limit for the patient’s temperature, which can be estimated by multiplying the patient temperature in degrees Celsius times 10 and then subtracting 323.32 For example, if the patient’s temperature is 39°C, relative bradycardia would refer to pulse rates below 67/minute (i.e., 390 − 323).†

† This formula combines separate formulas for women (<ts

Classically, patients with heat stroke have “bone-dry skin,” but most modern studies show that anhidrosis appears very late in the course and has a sensitivity of only 3% to 60%.33-35

In contrast, 91% of patients with heat stroke have significant pyrexia (exceeding 40°C), and 100% have abnormal mental status.

Muscle rigidity suggests the diagnosis of neuroleptic malignant syndrome (a febrile complication from dopamine antagonists) or serotonin syndrome (from proseroto- nergic drugs).36,37

Two findings increase the probability of fever: the patient’s subjective report of fever (LR = 5.3) and the clinician’s perception that the patient’s skin is abnormally warm (LR = 2.8; EBM Box 18.2). When either of these findings is absent, the prob- ability of fever decreases (LR = 0.2 to 0.3)

In patients hospitalized with fever, 8% to 37% will have documented bactere- mia,43,44,46,47,49,50,54,57,58 a finding associated with an increased hospital mortality.59

Of all the bedside findings that help diagnose bacteremia, the most important are the patient’s underlying disorders, in particular the presence of renal failure (LR = 4.6; EBM Box 18.3), hospitalization for trauma (LR = 3), and poor functional status (i.e., bedridden or requiring attendance; LR = 3.6).‡

One study even showed that the amount of food consumed by a febrile hospitalized patient was predictive of bacteremia: low food consumption (i.e., less than half of the meal served just before the blood culture) increased the probability of bacteremia (LR = 2.3), whereas high food consumption (more than 80% consumed) decreased it (LR = 0.2).66

A few physical findings also modestly increase the probability of bacteremia: presence of an indwelling urinary catheter (LR = 2.7), presence of a central venous catheter (LR = 2.4), and hypotension (LR = 2.3). The only finding significantly decreasing the probability of bacteremia is age under 50 years (LR = 0.3)

In 11 studies of over 6000 patients with fever, the presence of chills modestly increased the probability of bacteremia (sensitivity 24% to 95%; specificity 45% to 88%; positive LR = 1.9; negative LR = 0.7).44,47,49-55,67,68 If chills are instead pro- spectively defined as shaking chills (i.e., the patient feels so cold that his or her body involuntarily shakes even under thick clothing or blanket), the finding of shaking chills accurately detects bacteremia (sensitivity 45% to 90%, specificity 74% to 90%, positive LR = 3.7).52,69

The presence of toxic appearance fails to discriminate serious infection from trivial illness.29,70

Extreme pyrexia (i.e., temperature exceeding 41.1°C [106°F]) has diagnostic sig- nificance because the cause is usually gram-negative bacteremia or problems with temperature regulation (heat stroke, intracranial hemorrhage, severe burns).35

In a wide variety of disorders, the finding of a very high or low temperature indicates a worse prognosis.71,72 For example, temperatures greater than 39°C are associated with an increased risk of death in patients with pontine hemorrhage (LR = 23.7; EBM Box 18.4). Very low temperatures are associated with an increased risk of death in patients hospitalized with congestive heart failure (LR = 6.7), pneumo- nia (LR = 3.5), and bacteremia (LR = 3.3).

Most fevers today, whether infectious or noninfectious in origin, are intermittent or remittent and lack any other characteristic feature.73,74

Antibiotic medications have changed many traditional fever patterns. For example, the fever of lobar pneu- monia, which in the preantibiotic era was sustained and lasted 7 days, now lasts only 2 to 3 days.75,76 The double quotidian fever pattern (i.e., 2 daily fever spikes), a feature of gonococcal endocarditis present in 50% of cases during the preantibiotic era, is consistently absent in reported cases from the modern era.77 The character- istic tertian or quartan intermittent fever of malaria infection also is uncommon today, because most patients are treated before the characteristic synchronization of the malaria cycle.78

Nonetheless, although traditional fever patterns may be less common, they still have significance. In tropical countries, the presence of the stepladder remittent pattern of fever is highly specific for the diagnosis of typhoid fever (LR = 177.4).79 Also, among travelers with malarial infection who reported a tertian pattern, most are infected with Plasmodium vivax (traditionally the most common cause of this pattern).80

Moreover, the antibiotic era has given fever patterns a new significance, because once antibiotics have been started, the finding of an unusually prolonged fever is an important sign indicating either that the diagnosis of infection was incorrect (e.g., the patient instead has a connective tissue disorder or neoplasm) or that the patient has one of several complications, such as resistant organisms, superinfection, drug fever, or an abscess requiring surgical drainage

‡ For comparison, the LRs of these findings are superior to those for traditional laboratory signs of bacteremia, such as leukocytosis and bandemia. In detecting bacteremia, a WBC greater than 15,000 has an LR of only 1.6,29,43,49,60 whereas a band count greater than 1500 has an LR of 2.6.29,43,50

Clinical studies demonstrate that some infections, such as intracellular bacterial infections (e.g., typhoid fever and Legionnaire disease) and arboviral infections (e.g., sandfly fever and dengue fever) do produce less tachycardia than other infec- tions, but few patients with these infections actually have a relative bradycardia as defined earlier in the Findings section. Nonetheless, in one study of 100 febrile patients admitted to a Singapore hospital, a pulse rate of 90/minute or less increased the probability of dengue infection (LR = 3.3) and a pulse rate of 80/minute or less increased the probability even more (LR = 5.3).81

Fever of unknown origin (FUO) is defined as a febrile illness lasting at least 3 weeks without an explanation after at least 1 week of investigation. Most etiologies of FUO are noninfectious, particularly malignancies and noninfectious inflammatory disorders.

In three studies of almost 300 patients with FUO, two physical findings modestly increased the probability that a bone marrow examination would be diag- nostic (usually of a hematologic malignancy): splenomegaly (sensitivity 35% to 53%; specificity 82% to 89%; LR = 2.9) and peripheral lymphadenopathy (sensitiv- ity 21% to 30%; specificity 83% to 90%; LR 1.9).82-84