Content uploaded by Paul Jeffrey Young
Author content
All content in this area was uploaded by Paul Jeffrey Young on Sep 15, 2014
Content may be subject to copyright.
Available via license: CC BY-NC-ND 4.0
Content may be subject to copyright.
REVIEW
Critical Care and Resuscitation •Volume 13 Number 3 •September 2011194
Crit Care Resusc ISSN: 1441-2772 1 Septem-
ber 2011 13 3 194-199
©Crit Care Resusc 2011
www.jficm.anzca.edu.au/aaccm/journal/publi-
cations.htm
Review
Fever is common among critically ill patients, and its
detection has important implications for patient manage-
ment.1 Fever occurs as a response to tissue injury and
inflammation. The detection of a new fever may herald the
onset of an infection, often leading to additional investiga-
tions and changes in patient management. The presence of
persistent fever may indicate treatment failure, progressive
disease or disordered thermoregulation, and may be associ-
ated with poor prognosis.2 Accurate determination of body
temperature in the febrile range is not only important in
influencing intensive care unit practice, but also in the
setting of clinical research, where inclusion criteria or
interventions may be set at specific temperature thresholds.
Direct measurement of core temperature is regarded as
the most accurate method to determine body temperature,
as it is less influenced by variables such as ambient temper-
ature and peripheral perfusion. Core temperature is best
represented by the pulmonary artery catheter (PAC) ther-
mometer, the “gold standard” for clinical thermometry,
although other methods of core temperature assessment,
such as with oesophageal and bladder devices, are also
considered accurate.3 However, devices used to measure
core temperature are generally more expensive and more
invasive than peripheral thermometry, and may require a
skilled operator to position.
Over the past four decades, clinical thermometry has
developed to enable rapid, convenient core temperature
estimation using electronic devices at peripheral sites.4
These electronic thermometers often have functions that
convert temperatures measured at one site of the body to
an estimate of the temperature at another site (eg, core,
oral, rectal). Conversion algorithms for these functions vary
between thermometers and are determined by the manu-
facturer. For all modes other than “unadjusted” (or
“equal”) mode, a fixed number is automatically added to
the temperature taken. Clinical studies investigating the
accuracy of peripheral thermometers in determining core
temperature have reported differing accuracy at different
core temperatures,5,6 and recommendations have been
made for further investigation to take into account poten-
tial confounders, including temperature range variables.7
ABSTRACT
Background: There is uncertainty about the accuracy of
peripheral thermometers in measuring temperatures within
the febrile physiological range.
Objective: To determine the accuracy of peripheral
thermometers in detecting febrile core temperatures among
critically ill patients, and, if required, to determine a
standard conversion equation to improve accuracy.
Methods: A systematic search of MEDLINE, Embase, the
Cochrane Central Register of Controlled Trials and PubMed
was undertaken to identify clinical trials comparing
peripheral thermometry in critically ill adult patients with
core temperatures > 37.5°C. Our prespecified plan was to
perform a meta-analysis of the clinical accuracy of mean
peripheral thermometer temperature difference from core
temperature and calculation of limits of agreement.
Results: Systematic review identified three studies that
compared infrared tympanic, rectal or oral thermometer
readings with pulmonary artery catheter core temperature
readings among critically ill adults with fever. Studies were
heterogeneous and all failed to report appropriate
measurements of variation for the estimates of clinical
accuracy, which prevented meta-analysis and limited
interpretation of the results. Mean differences were within
±0.2°C in five of seven tympanic thermometer/mode/
temperature combinations and in the one oral thermometer
studied. All of three rectal thermometer/temperature
combinations studied reported mean differences outside
this range.
Conclusion: The identified studies suggest that in critically
ill patients, tympanic and oral thermometry provide, on
average, accurate measures of core temperatures within the
febrile range and can be recommended for this purpose.
Further studies with appropriate statistical methods are
required to assess the accuracy of peripheral thermometers
Crit Care Resusc 2011; 13: 194–199
among critically ill patients with fever.
A systematic review of the accuracy of peripheral
thermometry in estimating core temperatures among
febrile critically ill patients
Sarah Jefferies, Mark Weatherall,
Paul Young and Richard Beasley
REVIEW
Critical Care and Resuscitation •Volume 13 Number 3 •September 2011 195
In view of the uncertainty regarding the accuracy of
peripheral thermometry at elevated core temperatures, and
the importance of detecting fever in the ICU, we undertook
a systematic review with the intention of performing a
meta-analysis. We hypothesised that peripheral thermome-
try would tend to underestimate febrile core temperatures,
and that meta-analysis would determine a standard conver-
sion to calculate core temperature based on peripheral
temperature measurements.
Methods
Search strategy
Four databases were used to identify clinical studies com-
paring measurements from peripheral thermometers with
those obtained simultaneously from core thermometers in
the critically ill. The databases used were MEDLINE (1950 to
present); Embase (1947 to present); the Cochrane Central
Register of Controlled Trials (1991 to present) and PubMed
(1950 to present): search date, 29 December 2010. Key-
words were: “thermometer*”; and “pulmonary artery” or
“bladder”; limited to human and clinical trials. Potentially
relevant articles that were not written in English were
translated. One of us (S J) examined each article’s title and
abstract and the full article if necessary. The reference lists
of all relevant articles were also reviewed, and additional
hand searching was carried out.
Inclusion criteria
Clinical trials that investigated the accuracy of peripheral
methods of temperature measurement compared with
pulmonary artery catheter or bladder thermometry among
adult patients with core temperatures > 37.5°C, and pre-
sented appropriate summary statistics, were included. For
the purpose of this systematic review, peripheral thermo-
metry included any method of thermometry other than
pulmonary artery, oesophageal or bladder thermometry.
Exclusion criteria
•Non-human studies.
•Studies that involved iatrogenic physical temperature
manipulation (eg, external warming or cooling), as this
may introduce thermal gradients that are distinct from
physiological fever.
•Studies with no data for febrile core temperature ranges.
•Studies that only presented data in graphical form, so
that use of data for analysis would have required estima-
tion of data points.
•Studies that examined temporal artery thermometry.
Data extraction and interpretation
Extraction of data was based on reported summary statis-
tics. These were the mean differences, reflecting bias, and
appropriate measurements of variance. In the original
publication by Bland and Altman, limits of agreement were
defined as plus or minus two times the standard deviation
of the difference in measurements of the same research
participants measured twice only, reflecting that about
95% of the difference between future paired measure-
ments would fall between the limits of agreement.8
We recognised that research designs in this area often
involve measuring the same patients’ temperatures on
multiple occasions, often with multiple different peripheral
thermometers. We therefore reviewed publications for
reports of appropriate measures of variance for the mean
difference that could be used to calculate approximate 95%
confidence intervals for an individual predicted value. In
particular, we attempted to find the reports of variance
components due to measurement error that took into
account multiple measurements on individual research par-
ticipants. We were planning on performing a meta-analysis
of the size of the mean bias and attempted to find reports
of variance measurements, or appropriately calculated con-
fidence intervals, appropriate to the mean difference.
Based on previous research, we were particularly interested
in clinical accuracy, defined as a mean peripheral thermome-
ter temperature difference from a febrile core temperature
within ±0.2°C, and limits of agreement within ±0.5°C.9-12
We had planned to perform an inverse variance weighted
meta-analysis in critically ill patients with febrile core temper-
atures for clinical accuracy and limits of agreement.
Results
Systematic review identified three studies reporting the
accuracy of peripheral thermometers in detecting febrile
core temperatures in critically ill patients (Figure 1).10,13,14 All
three studies included multiple measurements, often in a
poorly specified number of febrile patients, and the statisti-
cal methods either failed to account for repeated measures
on the same participants or did not report appropriate
measures of variation. As a result, we were unable to
conduct a meta-analysis.
Characteristics of included studies
All three included articles were prospective, non-experimen-
tal, observational studies in adult critical care patients,
undertaken over a decade ago (Table 1). In the three
studies, details of baseline characteristics of the febrile
patients, such as sex and diagnosis, were not provided.
Details of the number of febrile patients studied or the
number of measurements taken in each patient were also
lacking.
All articles used PAC as the method of core temperature
measurement. The type of peripheral electronic thermo-
REVIEW
Critical Care and Resuscitation •Volume 13 Number 3 •September 2011196
meter and modes selected varied between studies. Milewski
and colleagues studied both an infrared (IR) tympanic
thermometer in its unadjusted mode and a digital electronic
rectal thermometer.13 Rotello and colleagues investigated
three types of IR tympanic thermometer (one in adjusted-
to-oral mode, two in unadjusted mode) and one digital
electronic rectal thermometer.14 Giuliano and colleagues
investigated two types of IR tympanic thermometer (both in
the core mode setting) and one oral electronic thermome-
ter.10 In this study, most patients were intubated and oral
temperatures were not collected within 30 minutes of a
patient receiving mouth cares.10
In one study, the researcher operating the thermometers
was blinded.14 In two studies, more than one operator was
involved, with attempts made to evaluate interoperator
error.10,13 All study protocols provided operator device train-
ing, described adequate device calibration and accounted
for “draw down”, in which false reductions in tympanic
temperature occur when repeated readings are taken from
the same ear within 2 minutes.
Two of the three studies did not describe the exclusion
of ear pathological features, which can affect the accu-
racy of tympanic thermometry;10,13 and none described
device cleaning, replacement of IR thermometer probe
caps with each reading or the control of other variables,
such as recent warming of the ear by head position on
pillow.9,15 The use of the “ear tug” technique, which may
influence tympanic measurements, varied in the three
studies.
Clinical accuracy data
Table 2 summarises the mean differences reported in the
included studies which compared tympanic, oral and rectal
thermometry with core temperature measurements. Meta-
analysis could not be performed, as all studies reported
error data without accounting for the effects of repeated
measures due to multiple measurements among study
patients.
An assessment of clinical accuracy could therefore only
be made based on the reported mean differences. Five of
seven different tympanic thermometer/mode/core tempera-
ture range combinations among the studies were clinically
accurate, with a mean difference within ±0.2°C of core
febrile temperatures (Table 2). The two tympanic thermo-
meter/mode/core temperature range combinations that
exceeded this limit were the Thermoscan Pro-1 in unad-
justed mode for core temperatures 37.6–38.0°C (mean
difference, − 0.22°C from core), although clinical accuracy
was demonstrated in this device/mode at >38.0°C, and the
Thermoscan HM-1 in oral mode (mean difference from core
0.24°C). Oral thermometry was used in one study and was
clinically accurate with a mean difference from core of
Figure 1. Flow diagram of search strategy
Results of “thermometer*”; and “pulmonary
artery” or “bladder”; limited to human and
clinical trials (n = 167)
Clinical trials comparing the accuracy of
peripheral thermometry with core
temperatures in critically ill adult patients
with fever (n = 3)
Exclusion of duplicates and
articles not potentially relevant
(n = 143)
Potentially relevant
from reference lists
and hand searching
(n = 50)
n = 24
n = 74
n = 62
n = 46
n = 43
n = 10
Excluded if not a clinical trial
(n = 12):
Review (n = 11)
Laboratory experiment (n = 1)
Excluded if inappropriate type
of thermometer (n = 16):
No comparison with core
temperature (n = 8)
Peripheral device is temporal
artery thermometer (n = 8)
Excluded if sample population
paediatric patients (n = 3)
Excluded if no data for febrile
core temperature ranges or
febrile data points presented in
an unclear graphical form
(n = 33)
Excluded if sample population
included patients subject to
iatrogenic physical temperature
manipulation (n = 7)
REVIEW
Critical Care and Resuscitation •Volume 13 Number 3 •September 2011 197
0.18°C.10 Rectal thermometry demonstrated clinical inaccu-
racy, overestimating core temperature measured by PAC
(Table 2).
Discussion
Our systematic review suggests that for critically ill patients,
on average, tympanic and oral thermometry provide accu-
rate estimations of core temperatures within the febrile
range, but that rectal thermometry is clinically inaccurate.
However, because of limitations in study reports, we were
unable to present a pooled estimate of clinical accuracy,
limits of agreement or appropriate confidence limits.
To the best of our knowledge, no other systematic review
has evaluated the accuracy of peripheral thermometry in
the febrile critically ill adult patient population. Hooper and
colleagues reviewed the accuracy of peripheral thermo-
metry among adult critical care patients, including several
studies that did not meet the criteria for our review.7 They
found variable reports of accuracy among a highly hetero-
geneous set of studies, and concluded that there was a
requirement for further investigation to account for differ-
ences in accuracy across sex, ethnicity, age and temperature
range variables.7 In another adult critical medicine review,
O’Grady and colleagues concluded that the accuracy of
tympanic thermometry was “consistently poor” and ranked
their recommendations for peripheral thermometry in the
order of rectal, oral, and lastly tympanic.16 They did not
discuss the flaws in statistical methods or reporting, nor the
heterogeneity of the studies upon which they based this
advice.
Two studies that did not meet the criteria for our review
further demonstrate the variability in the existing evidence.
Schmitz and colleagues compared peripheral with core
methods of thermometry specifically among critically ill
patients who were febrile at baseline.6 However, as their
analysis incorporated readings taken while the patients
were afebrile, their research was excluded from our ana-
lysis. They reported that tympanic thermometry overesti-
mated core temperature at PAC temperatures < 38.3°C and
underestimated at PAC temperatures ⭓38.3°C. This sug-
gests that the comparable performance of tympanic ther-
mometry may vary across the physiological febrile range.
In a more recent study by Moran and colleagues, the
average temperature difference between PAC and tympanic
thermometry was − 0.36 (SD, 0.47), with reported limits of
agreement of 0.56 and − 1.28°C, representing a greater
bias and variability from core temperature than that found
with axillary thermometry.17 In this article, the results were
derived from patients who experienced a range of core
temperatures, from hypothermia to hyperthermia, and
therefore it did not meet the criteria for our analysis.
However, the authors further report that regression analyses
examining temperature differences between PAC and tym-
panic, and PAC and axillary measurements, across the range
of patient temperatures, were not statistically significant.17
Table 1. Characteristics of the included studies
Study
Partici-
pants; no.
Peripheral
thermometer
(mode) Protocol
No. of
operators
(blinding)
Operator
training
Use of
ear tug
technique
Avoid
draw-
down
Otoscopic
exclusion
of ear
pathology
Calibration
of devices
Evaluation
of other
variables
Milewski et
al, 199113
Adult ICU;
febrile n=?
total n=9
Tympanic TM1
(unadjusted); rectal RT1
(monitor/non-predictive)
1–2-hourly single
measurements for up
to 48 h; RT read at
3 min from insertion
>1 (not
blinded)
YR ear, no
tug
Y N Y Individual
temperature
trends
Rotello et
al, 199614
Adult ICU;
febrile n=9,
total n=20
Tympanic TM2 (oral);
tympanic TM3
(unadjusted);
tympanic TM4
(unadjusted); rectal RT2
(monitor/non-predictive)
3 readings with each
TM at 2 min intervals
and a RT2 reading
with the third TM, the
RT2 having been
placed for 6 min
1 (single
blinding)
Y R ear, with
tug
Y Y Y Clinical
repeatability
between the
TM devices
Giuliano et
al, 200010
Adult ICU;
febrile n=?
total n=72
TM5 (core); TM6 (core);
oral OT1 (not specified)
1 reading per device
(one TM per ear) over
1 min. Repeated up
to 3 times at least
20 min apart
3 (not
blinded)
YAlternating
ears, ear tug
use not
specified
Y N Y Clinician–
thermometer
interaction
ICU = intensive care unit. TM = tympanic ther mometer. RT = rectal thermo meter. OT = oral thermo meter. TM1 = Thermoscan Pro-1 (Thermoscan Inc, San Diego, Calif, USA). TM2 = Thermoscan
HM-1 (Thermoscan Inc). TM3= Thermoscan Pro-1 (Thermoscan Inc). TM4 =Thermoscan Pro-LT (Thermoscan Inc). TM5 = FirstTemp Genius II (model 3000A, Sherwood Medical, Carlsbad,
Calif, USA). TM6 =ThermoScan Ear Pro-1 (model IR-1, Thermoscan Inc). RT1= Electronic rectal thermometer (Diatek, San Diego, Calif, USA). RT2 =electronic rectal thermometer (Diatek).
OT1 =SureTemp oral thermometer (model 678, Welch Allyn, San Diego, Calif, USA).
REVIEW
Critical Care and Resuscitation •Volume 13 Number 3 •September 2011198
This suggests that the bias between PAC and these two
thermometry measurements does not change in the febrile
range. When we considered the reported limits of agree-
ment by inspection of the Bland–Altman plots for tympanic
or axillary thermometry versus PAC, there appeared to be
greater variability in differences at mean temperatures
<38°C. However, this is difficult to verify without analysis
of the febrile-range participants.
Inspection of the plots also strongly suggested that the
limits of agreement calculated by the authors were based
on the standard deviation of all the paired readings and do
not take into account the multiple readings on the same
participants, and so are likely to be too narrow. This is less
likely to have a significant effect on the calculated mean
bias. Moran and colleagues concluded that urinary ther-
mometry was superior in accuracy to tympanic thermo-
metry and questioned the use of tympanic thermometers in
critically ill patients.17
This review is significant for several reasons. Firstly, it
summarises the limited available data comparing peripheral
and core temperatures among febrile critically ill patients,
but indicates that tympanic and oral thermometry may be
used to accurately detect febrile core temperatures among
these patients. Secondly, it highlights the lack of published
data on the performance of some of the brands of
peripheral thermometers currently used in clinical practice.
Thirdly, and perhaps most importantly, the analysis high-
lights a major statistical flaw in the existing published
literature, discussed further below.
This systematic review was performed with a sensitive
search of four major databases, with further hand searches
and review of reference lists. We are confident that all
relevant articles were identified by this strategy. However,
the studies included in the systematic review were hetero-
geneous by design and did not use statistical analysis to
adequately account for variation due to repeated measures.
The original description of the Bland–Altman method
assumes that each research participant in an agreement
study is measured once with each measurement device.8
The purpose of the method is to derive approximately 95%
confidence intervals for a future individual predicted value
of bias. For the simple design where participants are only
measured twice, this is based on plus or minus two times
the standard deviation of the differences and the two
methods of measurement agree if the limits of agreement
are acceptable on subject-matter grounds. For comparisons
between measurements of temperature, ±0.5°C has been
proposed as acceptable limits of agreement.10-12
The situation is more complex when participants in a
study of agreement are measured more than twice. This is
because there are two or more sources of variation to
consider: between patients, between measurement instru-
ments, if more than one is used, and the “leftover varia-
tion”. The leftover variation is the measurement error that is
appropriate to assess agreement. In the Bland–Altman
procedure, with only two measurements per participant,
the between-patient component is automatically accounted
for by working with the differences only. The standard
method of accounting for repeated, more than two, meas-
urements on the same patients are mixed linear models, of
which repeated-measures analysis of variance is a simple
example. These techniques use a variety of methods, of
which the most used is restricted maximum likelihood, to
estimate the variation due to measurement error. Failure to
Table 2. Summary data presented by the included studies of mean differences from PAC for febrile core
temperature ranges in critically ill patients
Mean difference from PAC, °C* (n)†
Study Febrile range(s), °C Tympanic Oral Rectal
Milewski et al, 199113 37.6–38.0 TM1 − 0.22 (n= 41) —RT1 0.46 (n= 35)
38.1–40.0 TM1 − 0.01 (n= 40) RT1 0.53 (n= 36)
Rotello et al, 199614 37.6–39.1 TM2 0.24 (n= 27) — RT2 0.35 (n= 27)
TM3 0.14 (n= 27)
TM4 0.01 (n= 27)
Giuliano et al, 200010 38.1–39.3 TM5 − 0.17 (n= 50) OT1 0.18 (n= 48) —
TM6 − 0.05 (n= 50)
PAC= pulmonary artery catheter. TM =tympanic thermometer. RT= rectal thermometer. OT= oral thermometer. TM1 = Thermoscan Pro-1 (Thermoscan Inc,
San Diego, Calif, USA). TM2 = Thermoscan HM-1 (Thermoscan Inc). TM3 = Thermoscan Pro-1 (Thermoscan Inc). TM4 = Thermoscan Pro-LT (Thermoscan Inc).
TM5 =FirstTemp Genius II (model 3000A, Sherwood Medical, Carlsbad, Calif, USA). TM6 = ThermoScan Ear Pro-1 (model IR-1, Thermoscan Inc). RT1 =
electronic predictive (Diatek, San Diego, Calif, USA). RT2 = Electronic predictive (Diatek). OT1 = SureTemp (model 678, Welch Allyn, San Diego, Calif, USA).
* Peripheral thermometer mean − core thermometer mean; negative numbers therefore represent a mean underestimate of core temperatures. †No. of
measurements.
REVIEW
Critical Care and Resuscitation •Volume 13 Number 3 •September 2011 199
account for repeatedly measuring the same research parti-
cipants, or that different research participants are measured
a different number of times, will give inappropriate esti-
mates of variation (eg, standard deviations or standard error
of the mean, and confidence intervals) for differences
between different measuring techniques. This difference is
likely to mean that confidence intervals are inappropriately
narrow. Simply taking the standard deviation of all the
differences for repeated measurements does not appropri-
ately estimate the correct element of variation to judge
agreement.
It is likely that the estimates of the mean values of bias,
shown in Table 2, are close to those obtained from analysis
that properly accounts for components of variation, if each
participant’s temperature was measured about the same
number of times. However, if some participants’ tempera-
tures were measured far more than those of others, these
participants are inappropriately weighted in calculating the
mean and the mean difference may itself be biased.
We recommend that further studies with appropriate
statistical methods be conducted to properly assess the
accuracy of peripheral thermometers currently being used
in critically ill adult patients with fever. Such studies need to
account for the components of variation to estimate meas-
urement error. This will require analysis that explicitly
accounts for repeated measurements together with appro-
priate estimates of bias and precision derived from, for
example, mixed linear models, with appropriate specifica-
tion of random and fixed effects. For the purposes of
ongoing clinical practice and trials using oral, tympanic or
rectal thermometry, we advise that, on the basis of the
limited available evidence, tympanic and oral thermometry
methods should be regarded as equivalent to core tempera-
ture and that rectal thermometry should not be used.
Competing interests
None declared.
Author details
Sarah Jefferies, Medical Research Fellow,1 and Medical Registrar2
Mark Weatherall, Physician,2 and Associate Professor3
Paul Young, Intensivist2
Richard Beasley, Director,1 and Physician2
1 Medical Research Institute of New Zealand, Wellington, New
Zealand.
2 Capital and Coast District Health Board, Wellington, New Zealand.
3 University of Otago Wellington, Wellington, New Zealand.
Correspondence: Sarah.Jefferie s@mrinz.ac.nz
References
1 Laupland KB, Shahpori R, Kirkpatrick AW, et al. Occurrence and
outcome of fever in critically ill adults. Crit Care Med 2008; 36:
1531-5.
2 Circiumaru B, Baldock G, Cohen J. A prospective study of fever in
the intensive care unit. Intensive Care Med 1999; 25: 668-73.
3 Lefrant JY, Muller L, de La Coussaye JE, et al. Temperature
measurement in intensive care patients: comparison of urinary
bladder, oesophageal, rectal, axillary, and inguinal methods versus
pulmonary artery core method. Intensive Care Med 2003; 29:
414-8.
4 Crawford DC, Hicks B, Thompson MJ. Which thermometer? Factors
influencing best choice for intermittent clinical temperature assess-
ment. J Med Eng Technol 2006; 30: 199-211.
5 Erickson RS, Kirklin SK. Comparison of ear-based, bladder, oral, and
axillary methods for core temperature measurement. Crit Care Med
1993; 21: 1528-34.
6 Schmitz T, Bair N, Falk M, Levine C. A comparison of five methods
of temperature measurement in febrile intensive care patients. Am
J Crit Care 1995; 4: 286-92.
7 Hooper VD, Andrews JO. Accuracy of noninvasive core temperature
measurement in acutely ill adults: the state of the science. Biol Res
Nurs 2006; 8: 24-34.
8 Bland JM, Altman DG. Comparing methods of measurement: why
plotting difference against standard method is misleading. Lancet
1995; 346: 1085-7.
9 Erickson RS, Meyer LT. Accuracy of infrared ear thermometry and
other temperature methods in adults. Am J Crit Care 1994; 3: 40-
54.
10 Giuliano KK, Giuliano AJ, Scott SS, et al. Temperature measurement
in critically ill adults: a comparison of tympanic and oral methods.
Am J Crit Care 2000; 9: 254-61.
11 Cattaneo CG, Frank SM, Hesel TW, et al. The accuracy and
precision of body temperature monitoring methods during regional
and general anesthesia. Anesth Analg 2000; 90: 938-45.
12 Lawson L, Bridges EJ, Ballou I, et al. Accuracy and precision of
noninvasive temperature measurement in adult intensive care
patients. Am J Crit Care 2007; 16: 485-96.
13 Milewski A, Ferguson KL, Terndrup TE. Comparison of pulmonary
artery, rectal, and tympanic membrane temperatures in adult
intensive care unit patients. Clin Pediatr (Phila) 1991; 30: 13-6;
discussion 34-5.
14 Rotello LC, Crawford L, Terndrup TE. Comparison of infrared ear
thermometer derived and equilibrated rectal temperatures in esti-
mating pulmonary artery temperatures. Crit Care Med 1996; 24:
1501-6.
15 Davie A, Amoore J. Best practice in the measurement of body
temperature. Nurs Stand 24: 42-9.
16 O’Grady NP, Barie PS, Bartlett JG, et al. Guidelines for evaluation of
new fever in critically ill adult patients: 2008 update from the
American College of Critical Care Medicine and the Infectious
Diseases Society of America. Crit Care Med 2008; 36: 1330-49.
17 Moran JL, Peter JV, Solomon PJ, et al. Tympanic temperature
measurements: are they reliable in the critically ill? A clinical study
of measures of agreement. Crit Care Med 2007; 35: 155-64. ❏