- Email: [email protected]
Estimation of Contact Tympanic Membrane Temperature With a Noncontact Infrared Thermometer
PEDIATRICS/BRIEF
REPORT
Estimation of Contact Tympanic Membrane Temperature With a Noncontact Infrared Thermometer From the Departments of Emergency Medicine,* Pediatrics, ~ Physiology, ~ Otolaryngology and Communication Sciences,~land the Department of Nursing, ~ State University New York Health Science Center at Syracuse, Syracuse, NY.
Thomas E Terndrup, MD *~§
Studyobjective: To investigate the agreement between contact
Daniel J Crofton, MD, DDSq Anthony J Mortelliti, MD" Richard Kelley, MD p Joanne Rajk, RN~
tympanic membrane (TM) and noncontact infrared ear thermometers in children.
Receivedfor publication October 22, 1996. Revision received February 28, 1997. Accepted for publication March 12, 1997. Supported by Braun- Thermoscan, Incorporated, San Diego, CA. Dr Terndrup is a consultant for several infrared ear thermometer companies and has served on the American Society for Testing and Materials Infrared Fever Thermometry Task Group. Copyright © by the American College of Emergency Physicians.
Methods: Twenty-three children (ages .5 to 10 years) undergoing elective tympanostomy tube placement were studied. An assistant used standard technique to record temperature with an infrared ear thermometer before and after TM temperature was obtained with a bead thermistor placed against the anterior-inferior quadrant of the TM. Results: Mean temperatures were not significantly different: initial IR ear, 36.660+.33 ° C; TM, 36.71°+.42° C; final infrared ear, 36.570+.33° C. The mean bias (difference between initial individual IR ear and TM temperatures) of-.05°+.29 ° C and the 95% limits of agreement of +.53 ° to -.63 ° C indicate an acceptable confidence (error range within 3.2% of average TM temperature) for use of the initial infrared ear temperature as an estimate of TM temperature. Conclusion: The IR ear thermometer provides an accurate estimate of TM temperature in healthy children and may accurately reflect core body temperature. [Terndrup TE, Crofton BJ, Mortelliti A J, Kelley R, Rajk J: Estimation of contact tympanic membrane temperature with a noncontact infrared thermometer. Ann EmergMedAugust 1997;30:171-175.]
AUGUST 1997
30:2
ANNALS OF EMERGENCY MEDICINE
17I
TYMPANIC MEMBRANE Terndrup et al
INTRODUCTION
The clinical accuracy of infrared ear thermometers in children is controversial. 1,2 With one notable exception, in no study has the agreement between the actual temperature at the intended target site of these thermometers, the tympanic membrane (TM), and infrared ear readings been assessed. 1 Assessment of this agreement is critical to determine whether the errors and inaccuracies previously reported with use of infrared ear thermometers at nonear reference sites (eg, rectal) reflect body site temperature variation or an inherent inability of the infrared device to obtain an accurate measurement of TM temperature. The TM, especially the anterior-inferior quadrant, is an excellent indicator of core body temperature. 3,4 An assessment of the agreement between infrared ear and TM temperature may serve as a predictor of the ability of the infrared ear thermometer to assess core body temperature. 5 A previous study in critically ill children demonstrated good agreement between another recognized core body temperature site, the pulmonary artery, and infrared ear temperatures. 6 However, agreement has not previously been tested in a healthy cohort, whose body temperature homogeneity and physiology may be different from those of critically ill children, r The objective of this study was to investigate the accuracy of an infrared ear thermometer in estimating TM temperatures in otherwise healthy children undergoing tympanostomy tube placement. Clinical accuracy was assessed by calculating the differences between infrared ear and TM temperatures (ie, bias) in individual patients. The SD and the 95% limits of agreement (_+2 SD) were used to assess the variability of the agreement.8 We hypothesized that the infrared ear thermometer provides an acceptable estimate of TM temperature. MATERIALS AND METHODS
A convenience sample of children aged .5 to 10 years undergoing elective tympanostomy tube placement were enrolled. Patients were excluded for hemotympanum, cerebrospinal fluid otorrhea, or external otitis. Children were enrolled after informed consent by their parents, and this study was approved by the local investigational review board for the protection of human subjects. General anesthesia was induced with 2% to 2.5% inhaled halothane; no child received preoperative sedatives. Shortly after induction, the research temperature protocol was completed. A single, right-handed assistant obtained an initial infrared ear reading using a standardized, optimum technique in the right auditory canal. The infrared ear technique
1 72
used was as follows: the pinna of the right ear was grasped at the midpoint between the apex of the helix and inferior border of the lobule and pulled posteriorly while the probe was inserted into the auditory canal with a firm back and forth motion until the ear canal was fully sealed off by the lens cover. The tip of the ear probe was directed at the presumed midpoint of the contralateral temple between the eyebrow and the sideburn. This technique simulates that performed during routine childhood otoscopy in an attempt to visualize the TM consistently. An otolaryngologic surgeon then placed a metal speculum in the auditory canal and gently removed any excess cerumen with a cerumen spoon. A bead thermistor was placed against the anterior-inferior quadrant of the right TM under microscopic guidance. With care taken not to move the thermistor wire, the speculum was removed and the tragus was compressed gently with a fingertip to close the external auditory canal. The thermistor tip often produced a slightly erythematous lesion at its site of placement, verifying contact with the TM. After equilibration (usually 15 to 20 seconds), the wire was removed and a second infrared ear temperature reading was obtained. The first and final IR recordings were separated by approximately 1 minute. A single ear thermometer (Thermoscan Pro-1; BraunThermoscan), operated in the calibration or equal mode, was used to record infrared ear temperatures. In this mode of operation, no offset factor is added to the infrared readings obtained by the instrument. The calibration of the infrared ear thermometer was verified against a black body reference at 37 ° C (IR-3000; Thermoscan) with each daily use. Singleuse, bead thermistors (Mon-a-therm) were used to obtain TM temperatures. After use, these thermistors were placed in a plastic bag and calibrated against a stirred-water bath at approximately 37 ° C, using an immersed thermometer with calibration traceable to the National Institute of Standards and Technology. The TM thermistor calibrations were performed within 2 weeks of use. Corrections for calibration errors were made for each patient. For inclusion in the study, environmental conditions required an ambient temperature between 12 ° and 25 ° C and a relative humidity between 20% and 60%. Mean temperatures were compared by use of ANOVA. Overall clinical accuracy, or bias, was defined as the average of all 23 patients' infrared-to-TM temperature differences. To describe further whether the infrared ear thermometer provided an acceptable estimate of TM temperature, we calculated the error for the population of children studied as a function of the 95% limits of agreement for the average TM temperature. Clinical variability was defined as the SD of all infrared-to-TM temperature differences. The individual
ANNALS OF EMERGENCY MEDICINE
30:2
AUGUST 1997
TYMPANIC MEMBRANE Tern&up et al
TM temperatures were plotted against their biases to construct an agreement graph. Because in clinical practice a single temperature value is usually recorded and because the final infrared ear temperature reading may have been influenced by cleaning of the canal and temporary placement of the TM thermistor probe, we chose to graphically represent only the 95% limits of agreement between the initial infrared ear and TM temperatures. In the traditional Bland-Ahman format, the bias would be plotted against the average of the measurements obtained by the two methods being compared. We chose to plot the bias against the actual TM temperature because in this case the criterion standard (ie, TM temperature) was known. ANOVA was used to assess the effect of patient age on bias, and the Student t test was used to assess the influence of room temperature and humidity. Unless otherwise stated, values are reported as the mean~+SD, and P values less than .05 were considered significant. RESULTS
Twenty-three patients (11 girls and 12 boys), mean age 4.1+3.6 years (range, 1 to 10 years), were enrolled in the study over an 8-month period ending in February 1996. These children were all classified as American Society of Anesthesiology class I, and none was febrile or had taken an antipyretic medication in the 48 hours preceding study. Room temperature was relatively constant (20°+2.7 ° C; range, 13 ° to 24 ° C), and humidity varied between 20% and 53% (mean, 37%+_11%). Calibration verification of the infrared ear thermometer demonstrated all readings to be within .2 ° C of the calibration temperatures throughout the study period. Mean infrared ear thermometer readings were lower than the infrared black body calibration by. 15°_+.16 ° C. Mean TM bead thermistor readings were greater than the water bath temperature by .04°_+.19 ° C. The mean temperatures were not significantly different (Table). Clinical variability was greater for the final than for the initial infrared ear measurements; the final measurements were slightly lower (average, -.09 ° C) and their bias was slightly more negative (average, -.08 ° C). The estimated error for this population of children, obtained by dividing the 95% limits of agreement for bias (1.16 ° C) by the mean TM temperature (36.71 ° C), was 3.2%. To examine the effects of patient age, the mean TM temperature and bias values were calculated for three age groups: 0 to 3.5 years, 3.6 to 7 years, and older than 7 years. The mean bias was larger in the middle age group compared with the other two patient groups, but the difference was not
AUGUST 1997 3 0 : 2 ANNALS OF EMERGENCY MEDICINE
statistically significant (F=2.497, P=. ! 076; AN OVA); the mean TM temperature was slightly lower in the middle age group (mean, .2 ° C). A lower room temperature (<20 ° C) was associated with a smaller bias (.02 ° _+.25°, versus -. 11°_+.33° C for greater than 20 ° C), whereas room humidity did not appear to influence mean bias, and these effects were not statistically significant. DISCUSSION
This study demonstrates the close agreement between infrared ear and TM temperature measurements in otherwise healthy children undergoing tympanostomy tube placement using one thermometer model. Employing a specified technique, a single operator was able to dosely estimate the TM temperature of children, with a mean difference of .05 ° C. When studied under the relatively adynamic clinical conditions reported in this investigation, measurements with this IR ear thermometer closely approximated TM temperature, which has previously been shown to accurately measure core body temperature. 4 Only one previous study of the accuracy of IR ear thermometers in children has evaluated agreement with TM temperature. 1 Thirty-one newborns were studied and temperatures obtained with an older version of another infrared ear thermometer (First Temp 2000A; Intelligent Medical Systems) were compared with TM temperatures procured by the resistance-to-insertion method of localizing TM probe placement. Infrared ear temperature (in calibration-surface mode) averaged .20 ° C less than TM temperature, somewhat greater than the difference found in our study. This may reflect differences in the location of the TM probe, thermal protection of the TM by the cotton plug used to isolate the auditory canal, or differences in performance of the infrared ear thermometers used. In different infrared ear models, performance may vary as a function of size and shape of the probe, field of view of the infrared sensor, ambient temperature, and operator technique. The infrared ear thermometer Table.
Agreement of temperature measurements obtained in 23 children undergoing tympanostomy tube placement.
Measurement Initial infrared ear TM Final infrared ear
Temperature (°C) [mean+SD] 36.66_+.33 36.71_+.42 36.57+.33
Bias (°C) [mean+SD] -.05+.29 NA -.13_+.42
Limits of Agreement +.53 to -.63 NA +.71 to -.97
NA, not applicable.
1 73
TYMPANIC
MEMBRANE
Terndrup et al
used in the previous study is now more than 10 years old and is no longer manufactured. In the same newborns (average postnatal age, 11 days), mean rectal thermistor probe temperature (36.97 ° C) was found to be almost identical to TM temperature (37.07 ° C). This lack of important differences between rectal and TM temperatures may provide additional confidence for challenging traditional practices of temperature assessment, which are driven by the desire to measure rectal temperatures. If rectal temperatures agree well with TM temperatures, and infrared ear thermometer readings agree with TM, then the ear infrared thermometer may also provide an indicator of equilibrated rectal thermistor temperatures. The combined results of the previous study and our present study demonstrate similarity between TM and infrared ear temperature measurements and extended these observations to somewhat older children. When non-TM body sites (eg, rectal) are used to evaluate the performance of an infrared ear thermometer, it is unknown whether they estimate the actual TM temperature, the intended measurement site for ear infrared thermometers. Additional factors of unknown significance in the use of non-TM sites to evaluate infrared ear thermometer perFigure.
Bias of individual infrared ear temperature readings and the 95% limits of agreement in relation to TM temperature readings in 23 children undergoing tympanostomy tube placement.
Bias (°C) 1.00.80.60.40.20.00- . . . . . . . . . . . . . . . . . . .
-------g . . . . . . . . . . . . . . . . . . . . . .
-.20-.40 -.60-.80-1.00
36
36.2 31~.4 36.8 36.8
37
Tympanic MembraneTernerature(°C) The bias is equal to infrared ear temperature minus TM temperature. One data point is obscured as a result of overlap. The horizontal lines define the mean bias (-.05 ° C; dashedline) and the limits of agreement (-+2 SB).
1 7 4
i
3~.2 37.4 37.8 37.8
formance include instrumentation differences, body site temperature variation, dynamic temperature changes, and inclusion of convenient but possibly nonrepresentative patient populations. It is well known that different body sites are highly variable in their temperature ranges; for example, oral temperature is usually .5 ° C less than a simultaneously obtained rectal temperature. What is less well appreciated is that these average body site differences may vary by several degrees Centigrade as a result of physiologic or environmental variation, lending additional uncertainty to the prospect of analyzing differences in the performance of thermometers. One study reported the limits of agreement for equilibrated rectal thermistor probe temperatures, compared with pulmonary artery temperatures, as .64 ° C. 6 These results were reported in 20 critically ill children with a median age of 3.6 years, similar to our patients' ages. Variability in the reproducibility of oral and axillary temperature measurements in children has also been demonstrated, with the limits of agreement in one study being .68 ° C and .74 ° C, respectively 9 The comparison in the present study reduced any error introduced by body site differences and was designed to test the agreement between two instruments attempting to measure the same temperature. Estimates of core body temperature may be obtained from thermistor catheters located in the pulmonary artery, bladder, esophagus, or TM. 3 Because these sites require invasive procedures, other, more convenient body sites have been selected as surrogates for measuring core body temperature. Oral, rectal, and axillary methods have been used, but concerns have been raised over their accuracy in estimating core body temperature. After comparisons with rectal temperature measurements, some investigators have concluded that infrared ear thermometers are inaccurate in estimating body temperature. Yet infrared ear temperatures more closely approximate measurements made at body sites that appear to more accurately estimate core temperature (eg, pulmonary artery), lo The limit of agreement between equilibrated rectal thermistor and pulmonary artery temperatures in critically ill children was .64 ° C, with a bias of .07 ° C, whereas the same values for the thermometer used m our study were .78 ° C and -. 13 ° C, respectively.6 These data and ours raise further concerns about the interpretation of accuracy when ear infrared temperature values are compared with a noncore standard such as rectal temperature. The limits of agreement between TM temperatures and those obtained with the infrared ear thermometer used in this study compare favorably with the results of variation studies performed at other body sites. There were some limitations in our study design. Because temperatures were obtained in a cool operating theater,
ANNALS OF EMERGENCY MEDICINE
30:2
AUGUST 1997
TYMPANIC
MEMBRANE
Terndrup et al
measured values were somewhat lower than at normal ambient temperature. In addition, the larger thermal gradient in the auditory canal under these conditions may have tended to increase the bias, compared with the gradient under warmer conditions. However, the mean bias was actually less in patients studied in rooms that were cooler than 20 ° C. No additional body site temperature measurements were obtained in these patients because that was not the purpose of this study, and experimental protocols in children are required to contain no additional subject risks, including any that may prolong, however briefly, the duration of general anesthesia. We removed excess cerumen because some authors have indicated a small but potentially important reduction in infrared ear temperature measurements associated with impacted cerumen. We found a small increase in bias (eg, decreased infrared ear readings) after cerumen removal and placement of the ear speculum and thermistor wire. This reduction may be related to the introduction of cooler air, often referred to as ~'drawdown," or contact of the canal with cool, metallic instruments. The accuracy and limits of agreement provided m this study may not apply to febrile patients or to those under dynamic clinical conditions. Because febrile patients are usually rescheduled for tympanostomy tube placement, it is unlikely that similar data can easily be obtained from a febrile population. Because infrared ear thermometers are operator dependent and vary in their design and IR data processing approaches, our results should not be generalized to other infrared ear thermometers before satisfactory validation is obtained. In conclusion, a noncontact infrared ear thermometer demonstrated good average agreement with TM temperatures in 23 children under short-term general anesthesia. The mean clinical bias (error) was -.05°+.29°C, with 95% limits of agreement of +.53 ° to -.63 ° C. These findings suggest that the infrared ear thermometer used in this study may accurately estimate core body temperature in children, as measured at the TM.
7. Nadel ER, HervathSM: Comparisonof tympanic membraneand deep bodytemperatures in man. Life Sciences 1970;9:869-875. 8. Bland JM, Altman DG: Measurementin medicine:The analysis of method comparisonstudies. The Statistician 1983;32:307-317. 9. ChamberlainJM, TemdrupTE, Alexander DT, et ah Determinationof normal ear temperature with an infrared emission detection thermometer.Ann Emerg Meal1995;25:15-20. 10. Rotello LC, Crawford L, TerndrupTE: Comparisonof infrared ear thermometerderived and equilibrated rectal temperaturesin estimated pulmonaryartery temperatures. Crit Care Med 1996;24:1-8.
Reprint no. 47/1/82890 Address for reprints: Thomas ETerndrup,MD Departmentof EmergencyMedicine SUNYHealthScienceCenter 750 EastAdamsStreet Syracuse,NY 13210 315464-4363 Fax315-464-6229 [email protected]
REFERENCES 1. Johnson KJ, Bhatia P, Bell EF: Infrared thermometryof newborn infants. Pediatrics1991;87:3438. 2. FreedGL, FraleyJK: Lackof agreementof tympanic membranetemperatureassessmentswith conventional methods in a private practice. Pediatrics 1992;89:384-386. 3. Bfinnel H, CabanacM: Tympanictemperature is a care temperaturein humans. J ThermBiol 1989;14:47-53. 4. Mayfield SR, NakamuraKT, Bhatia J, et el: Tympanic membranetemperatureof term and preterm neonates. Early Hum Oev 1£84;9:241-247. 5. GreenleafJE, Castle BL: Externalauditory canal temperatureas an estimate of core temperature. J Appl Physiol 1972;32:t94-I 98. 6.8amano MJ, FortanberryJD, Autrey E, et ah Infraredtympanic thermometryin the pediatric intensive care unit. Crit Care Med 1993;21:1181-1185.
AUGUST 1997
30:2
ANNALS OF EMERGENCY MEDICINE
175
REPORT
Estimation of Contact Tympanic Membrane Temperature With a Noncontact Infrared Thermometer From the Departments of Emergency Medicine,* Pediatrics, ~ Physiology, ~ Otolaryngology and Communication Sciences,~land the Department of Nursing, ~ State University New York Health Science Center at Syracuse, Syracuse, NY.
Thomas E Terndrup, MD *~§
Studyobjective: To investigate the agreement between contact
Daniel J Crofton, MD, DDSq Anthony J Mortelliti, MD" Richard Kelley, MD p Joanne Rajk, RN~
tympanic membrane (TM) and noncontact infrared ear thermometers in children.
Receivedfor publication October 22, 1996. Revision received February 28, 1997. Accepted for publication March 12, 1997. Supported by Braun- Thermoscan, Incorporated, San Diego, CA. Dr Terndrup is a consultant for several infrared ear thermometer companies and has served on the American Society for Testing and Materials Infrared Fever Thermometry Task Group. Copyright © by the American College of Emergency Physicians.
Methods: Twenty-three children (ages .5 to 10 years) undergoing elective tympanostomy tube placement were studied. An assistant used standard technique to record temperature with an infrared ear thermometer before and after TM temperature was obtained with a bead thermistor placed against the anterior-inferior quadrant of the TM. Results: Mean temperatures were not significantly different: initial IR ear, 36.660+.33 ° C; TM, 36.71°+.42° C; final infrared ear, 36.570+.33° C. The mean bias (difference between initial individual IR ear and TM temperatures) of-.05°+.29 ° C and the 95% limits of agreement of +.53 ° to -.63 ° C indicate an acceptable confidence (error range within 3.2% of average TM temperature) for use of the initial infrared ear temperature as an estimate of TM temperature. Conclusion: The IR ear thermometer provides an accurate estimate of TM temperature in healthy children and may accurately reflect core body temperature. [Terndrup TE, Crofton BJ, Mortelliti A J, Kelley R, Rajk J: Estimation of contact tympanic membrane temperature with a noncontact infrared thermometer. Ann EmergMedAugust 1997;30:171-175.]
AUGUST 1997
30:2
ANNALS OF EMERGENCY MEDICINE
17I
TYMPANIC MEMBRANE Terndrup et al
INTRODUCTION
The clinical accuracy of infrared ear thermometers in children is controversial. 1,2 With one notable exception, in no study has the agreement between the actual temperature at the intended target site of these thermometers, the tympanic membrane (TM), and infrared ear readings been assessed. 1 Assessment of this agreement is critical to determine whether the errors and inaccuracies previously reported with use of infrared ear thermometers at nonear reference sites (eg, rectal) reflect body site temperature variation or an inherent inability of the infrared device to obtain an accurate measurement of TM temperature. The TM, especially the anterior-inferior quadrant, is an excellent indicator of core body temperature. 3,4 An assessment of the agreement between infrared ear and TM temperature may serve as a predictor of the ability of the infrared ear thermometer to assess core body temperature. 5 A previous study in critically ill children demonstrated good agreement between another recognized core body temperature site, the pulmonary artery, and infrared ear temperatures. 6 However, agreement has not previously been tested in a healthy cohort, whose body temperature homogeneity and physiology may be different from those of critically ill children, r The objective of this study was to investigate the accuracy of an infrared ear thermometer in estimating TM temperatures in otherwise healthy children undergoing tympanostomy tube placement. Clinical accuracy was assessed by calculating the differences between infrared ear and TM temperatures (ie, bias) in individual patients. The SD and the 95% limits of agreement (_+2 SD) were used to assess the variability of the agreement.8 We hypothesized that the infrared ear thermometer provides an acceptable estimate of TM temperature. MATERIALS AND METHODS
A convenience sample of children aged .5 to 10 years undergoing elective tympanostomy tube placement were enrolled. Patients were excluded for hemotympanum, cerebrospinal fluid otorrhea, or external otitis. Children were enrolled after informed consent by their parents, and this study was approved by the local investigational review board for the protection of human subjects. General anesthesia was induced with 2% to 2.5% inhaled halothane; no child received preoperative sedatives. Shortly after induction, the research temperature protocol was completed. A single, right-handed assistant obtained an initial infrared ear reading using a standardized, optimum technique in the right auditory canal. The infrared ear technique
1 72
used was as follows: the pinna of the right ear was grasped at the midpoint between the apex of the helix and inferior border of the lobule and pulled posteriorly while the probe was inserted into the auditory canal with a firm back and forth motion until the ear canal was fully sealed off by the lens cover. The tip of the ear probe was directed at the presumed midpoint of the contralateral temple between the eyebrow and the sideburn. This technique simulates that performed during routine childhood otoscopy in an attempt to visualize the TM consistently. An otolaryngologic surgeon then placed a metal speculum in the auditory canal and gently removed any excess cerumen with a cerumen spoon. A bead thermistor was placed against the anterior-inferior quadrant of the right TM under microscopic guidance. With care taken not to move the thermistor wire, the speculum was removed and the tragus was compressed gently with a fingertip to close the external auditory canal. The thermistor tip often produced a slightly erythematous lesion at its site of placement, verifying contact with the TM. After equilibration (usually 15 to 20 seconds), the wire was removed and a second infrared ear temperature reading was obtained. The first and final IR recordings were separated by approximately 1 minute. A single ear thermometer (Thermoscan Pro-1; BraunThermoscan), operated in the calibration or equal mode, was used to record infrared ear temperatures. In this mode of operation, no offset factor is added to the infrared readings obtained by the instrument. The calibration of the infrared ear thermometer was verified against a black body reference at 37 ° C (IR-3000; Thermoscan) with each daily use. Singleuse, bead thermistors (Mon-a-therm) were used to obtain TM temperatures. After use, these thermistors were placed in a plastic bag and calibrated against a stirred-water bath at approximately 37 ° C, using an immersed thermometer with calibration traceable to the National Institute of Standards and Technology. The TM thermistor calibrations were performed within 2 weeks of use. Corrections for calibration errors were made for each patient. For inclusion in the study, environmental conditions required an ambient temperature between 12 ° and 25 ° C and a relative humidity between 20% and 60%. Mean temperatures were compared by use of ANOVA. Overall clinical accuracy, or bias, was defined as the average of all 23 patients' infrared-to-TM temperature differences. To describe further whether the infrared ear thermometer provided an acceptable estimate of TM temperature, we calculated the error for the population of children studied as a function of the 95% limits of agreement for the average TM temperature. Clinical variability was defined as the SD of all infrared-to-TM temperature differences. The individual
ANNALS OF EMERGENCY MEDICINE
30:2
AUGUST 1997
TYMPANIC MEMBRANE Tern&up et al
TM temperatures were plotted against their biases to construct an agreement graph. Because in clinical practice a single temperature value is usually recorded and because the final infrared ear temperature reading may have been influenced by cleaning of the canal and temporary placement of the TM thermistor probe, we chose to graphically represent only the 95% limits of agreement between the initial infrared ear and TM temperatures. In the traditional Bland-Ahman format, the bias would be plotted against the average of the measurements obtained by the two methods being compared. We chose to plot the bias against the actual TM temperature because in this case the criterion standard (ie, TM temperature) was known. ANOVA was used to assess the effect of patient age on bias, and the Student t test was used to assess the influence of room temperature and humidity. Unless otherwise stated, values are reported as the mean~+SD, and P values less than .05 were considered significant. RESULTS
Twenty-three patients (11 girls and 12 boys), mean age 4.1+3.6 years (range, 1 to 10 years), were enrolled in the study over an 8-month period ending in February 1996. These children were all classified as American Society of Anesthesiology class I, and none was febrile or had taken an antipyretic medication in the 48 hours preceding study. Room temperature was relatively constant (20°+2.7 ° C; range, 13 ° to 24 ° C), and humidity varied between 20% and 53% (mean, 37%+_11%). Calibration verification of the infrared ear thermometer demonstrated all readings to be within .2 ° C of the calibration temperatures throughout the study period. Mean infrared ear thermometer readings were lower than the infrared black body calibration by. 15°_+.16 ° C. Mean TM bead thermistor readings were greater than the water bath temperature by .04°_+.19 ° C. The mean temperatures were not significantly different (Table). Clinical variability was greater for the final than for the initial infrared ear measurements; the final measurements were slightly lower (average, -.09 ° C) and their bias was slightly more negative (average, -.08 ° C). The estimated error for this population of children, obtained by dividing the 95% limits of agreement for bias (1.16 ° C) by the mean TM temperature (36.71 ° C), was 3.2%. To examine the effects of patient age, the mean TM temperature and bias values were calculated for three age groups: 0 to 3.5 years, 3.6 to 7 years, and older than 7 years. The mean bias was larger in the middle age group compared with the other two patient groups, but the difference was not
AUGUST 1997 3 0 : 2 ANNALS OF EMERGENCY MEDICINE
statistically significant (F=2.497, P=. ! 076; AN OVA); the mean TM temperature was slightly lower in the middle age group (mean, .2 ° C). A lower room temperature (<20 ° C) was associated with a smaller bias (.02 ° _+.25°, versus -. 11°_+.33° C for greater than 20 ° C), whereas room humidity did not appear to influence mean bias, and these effects were not statistically significant. DISCUSSION
This study demonstrates the close agreement between infrared ear and TM temperature measurements in otherwise healthy children undergoing tympanostomy tube placement using one thermometer model. Employing a specified technique, a single operator was able to dosely estimate the TM temperature of children, with a mean difference of .05 ° C. When studied under the relatively adynamic clinical conditions reported in this investigation, measurements with this IR ear thermometer closely approximated TM temperature, which has previously been shown to accurately measure core body temperature. 4 Only one previous study of the accuracy of IR ear thermometers in children has evaluated agreement with TM temperature. 1 Thirty-one newborns were studied and temperatures obtained with an older version of another infrared ear thermometer (First Temp 2000A; Intelligent Medical Systems) were compared with TM temperatures procured by the resistance-to-insertion method of localizing TM probe placement. Infrared ear temperature (in calibration-surface mode) averaged .20 ° C less than TM temperature, somewhat greater than the difference found in our study. This may reflect differences in the location of the TM probe, thermal protection of the TM by the cotton plug used to isolate the auditory canal, or differences in performance of the infrared ear thermometers used. In different infrared ear models, performance may vary as a function of size and shape of the probe, field of view of the infrared sensor, ambient temperature, and operator technique. The infrared ear thermometer Table.
Agreement of temperature measurements obtained in 23 children undergoing tympanostomy tube placement.
Measurement Initial infrared ear TM Final infrared ear
Temperature (°C) [mean+SD] 36.66_+.33 36.71_+.42 36.57+.33
Bias (°C) [mean+SD] -.05+.29 NA -.13_+.42
Limits of Agreement +.53 to -.63 NA +.71 to -.97
NA, not applicable.
1 73
TYMPANIC
MEMBRANE
Terndrup et al
used in the previous study is now more than 10 years old and is no longer manufactured. In the same newborns (average postnatal age, 11 days), mean rectal thermistor probe temperature (36.97 ° C) was found to be almost identical to TM temperature (37.07 ° C). This lack of important differences between rectal and TM temperatures may provide additional confidence for challenging traditional practices of temperature assessment, which are driven by the desire to measure rectal temperatures. If rectal temperatures agree well with TM temperatures, and infrared ear thermometer readings agree with TM, then the ear infrared thermometer may also provide an indicator of equilibrated rectal thermistor temperatures. The combined results of the previous study and our present study demonstrate similarity between TM and infrared ear temperature measurements and extended these observations to somewhat older children. When non-TM body sites (eg, rectal) are used to evaluate the performance of an infrared ear thermometer, it is unknown whether they estimate the actual TM temperature, the intended measurement site for ear infrared thermometers. Additional factors of unknown significance in the use of non-TM sites to evaluate infrared ear thermometer perFigure.
Bias of individual infrared ear temperature readings and the 95% limits of agreement in relation to TM temperature readings in 23 children undergoing tympanostomy tube placement.
Bias (°C) 1.00.80.60.40.20.00- . . . . . . . . . . . . . . . . . . .
-------g . . . . . . . . . . . . . . . . . . . . . .
-.20-.40 -.60-.80-1.00
36
36.2 31~.4 36.8 36.8
37
Tympanic MembraneTernerature(°C) The bias is equal to infrared ear temperature minus TM temperature. One data point is obscured as a result of overlap. The horizontal lines define the mean bias (-.05 ° C; dashedline) and the limits of agreement (-+2 SB).
1 7 4
i
3~.2 37.4 37.8 37.8
formance include instrumentation differences, body site temperature variation, dynamic temperature changes, and inclusion of convenient but possibly nonrepresentative patient populations. It is well known that different body sites are highly variable in their temperature ranges; for example, oral temperature is usually .5 ° C less than a simultaneously obtained rectal temperature. What is less well appreciated is that these average body site differences may vary by several degrees Centigrade as a result of physiologic or environmental variation, lending additional uncertainty to the prospect of analyzing differences in the performance of thermometers. One study reported the limits of agreement for equilibrated rectal thermistor probe temperatures, compared with pulmonary artery temperatures, as .64 ° C. 6 These results were reported in 20 critically ill children with a median age of 3.6 years, similar to our patients' ages. Variability in the reproducibility of oral and axillary temperature measurements in children has also been demonstrated, with the limits of agreement in one study being .68 ° C and .74 ° C, respectively 9 The comparison in the present study reduced any error introduced by body site differences and was designed to test the agreement between two instruments attempting to measure the same temperature. Estimates of core body temperature may be obtained from thermistor catheters located in the pulmonary artery, bladder, esophagus, or TM. 3 Because these sites require invasive procedures, other, more convenient body sites have been selected as surrogates for measuring core body temperature. Oral, rectal, and axillary methods have been used, but concerns have been raised over their accuracy in estimating core body temperature. After comparisons with rectal temperature measurements, some investigators have concluded that infrared ear thermometers are inaccurate in estimating body temperature. Yet infrared ear temperatures more closely approximate measurements made at body sites that appear to more accurately estimate core temperature (eg, pulmonary artery), lo The limit of agreement between equilibrated rectal thermistor and pulmonary artery temperatures in critically ill children was .64 ° C, with a bias of .07 ° C, whereas the same values for the thermometer used m our study were .78 ° C and -. 13 ° C, respectively.6 These data and ours raise further concerns about the interpretation of accuracy when ear infrared temperature values are compared with a noncore standard such as rectal temperature. The limits of agreement between TM temperatures and those obtained with the infrared ear thermometer used in this study compare favorably with the results of variation studies performed at other body sites. There were some limitations in our study design. Because temperatures were obtained in a cool operating theater,
ANNALS OF EMERGENCY MEDICINE
30:2
AUGUST 1997
TYMPANIC
MEMBRANE
Terndrup et al
measured values were somewhat lower than at normal ambient temperature. In addition, the larger thermal gradient in the auditory canal under these conditions may have tended to increase the bias, compared with the gradient under warmer conditions. However, the mean bias was actually less in patients studied in rooms that were cooler than 20 ° C. No additional body site temperature measurements were obtained in these patients because that was not the purpose of this study, and experimental protocols in children are required to contain no additional subject risks, including any that may prolong, however briefly, the duration of general anesthesia. We removed excess cerumen because some authors have indicated a small but potentially important reduction in infrared ear temperature measurements associated with impacted cerumen. We found a small increase in bias (eg, decreased infrared ear readings) after cerumen removal and placement of the ear speculum and thermistor wire. This reduction may be related to the introduction of cooler air, often referred to as ~'drawdown," or contact of the canal with cool, metallic instruments. The accuracy and limits of agreement provided m this study may not apply to febrile patients or to those under dynamic clinical conditions. Because febrile patients are usually rescheduled for tympanostomy tube placement, it is unlikely that similar data can easily be obtained from a febrile population. Because infrared ear thermometers are operator dependent and vary in their design and IR data processing approaches, our results should not be generalized to other infrared ear thermometers before satisfactory validation is obtained. In conclusion, a noncontact infrared ear thermometer demonstrated good average agreement with TM temperatures in 23 children under short-term general anesthesia. The mean clinical bias (error) was -.05°+.29°C, with 95% limits of agreement of +.53 ° to -.63 ° C. These findings suggest that the infrared ear thermometer used in this study may accurately estimate core body temperature in children, as measured at the TM.
7. Nadel ER, HervathSM: Comparisonof tympanic membraneand deep bodytemperatures in man. Life Sciences 1970;9:869-875. 8. Bland JM, Altman DG: Measurementin medicine:The analysis of method comparisonstudies. The Statistician 1983;32:307-317. 9. ChamberlainJM, TemdrupTE, Alexander DT, et ah Determinationof normal ear temperature with an infrared emission detection thermometer.Ann Emerg Meal1995;25:15-20. 10. Rotello LC, Crawford L, TerndrupTE: Comparisonof infrared ear thermometerderived and equilibrated rectal temperaturesin estimated pulmonaryartery temperatures. Crit Care Med 1996;24:1-8.
Reprint no. 47/1/82890 Address for reprints: Thomas ETerndrup,MD Departmentof EmergencyMedicine SUNYHealthScienceCenter 750 EastAdamsStreet Syracuse,NY 13210 315464-4363 Fax315-464-6229 [email protected]
REFERENCES 1. Johnson KJ, Bhatia P, Bell EF: Infrared thermometryof newborn infants. Pediatrics1991;87:3438. 2. FreedGL, FraleyJK: Lackof agreementof tympanic membranetemperatureassessmentswith conventional methods in a private practice. Pediatrics 1992;89:384-386. 3. Bfinnel H, CabanacM: Tympanictemperature is a care temperaturein humans. J ThermBiol 1989;14:47-53. 4. Mayfield SR, NakamuraKT, Bhatia J, et el: Tympanic membranetemperatureof term and preterm neonates. Early Hum Oev 1£84;9:241-247. 5. GreenleafJE, Castle BL: Externalauditory canal temperatureas an estimate of core temperature. J Appl Physiol 1972;32:t94-I 98. 6.8amano MJ, FortanberryJD, Autrey E, et ah Infraredtympanic thermometryin the pediatric intensive care unit. Crit Care Med 1993;21:1181-1185.
AUGUST 1997
30:2
ANNALS OF EMERGENCY MEDICINE
175