ORIGINAL CONTRIBUTION fever, pediatric; thermometer
Comparison of Rectal, Axillary, and Tympanic Membrane Temperatures in Infants and Young Children Study objective: To evaluate the reliability of a tympanic membrane thermometer in detecting fever in young children presenting to the emergency department. Setting: Pediatric emergency department in an urban teaching hospital. Design/Measurement/Participants: Temperature measurements were obtained sequentially at three body sites in children less than 3 years old presenting to the pediatric ED. Axillary and rectal temperatures were obtained with an electronic thermistor probe (Diatek 500®), and tympanic membrane temperatures were obtained with a noncontact, infrared sensing device (FirstTEMP®). Patients were stratified by age, ear canal patency, presence of otitis media, and rectal temperature. Results: Of 224 patients enrolled, 87 (39%) were febrile. Overall correlation of axillary and tympanic membrane measurements to rectal for all strata was .75 (P = .001) and .81 (P = .00I), respectively. Sensitivity in detecting fever for axil]ary and tympanic membrane sites was .48 and. 55, respectively. Otitis media and ear patency did not influence correlation of tympanic membrane measurements. Low tympanic membrane temperature sensitivity m a y be a result of probe configuration. Conclusion: Tympanic membrane and axillary temperatures should be viewed with caution in children less than 3 years old as neither can detect fever reliably. [Muma BK, Treloar D J, Wurmlinger K, Peterson E, Vitae A: Comparison of rectal, axillary, and tympanic membrane temperatures in infants and young children.-Ann Emerg Med January 1991;20:41-44.]
INTRODUCTION
Bruce K Muma, MD* David J Treloar, MD, MSc* Karen Wurmlinger, RN* Edward Peterson, PhDt April Vitae, RN* Detroit, Michigan From the Division of Pediatric Emergency Medicine, Department of Pediatrics,* and the Department of Biostatistics and Clinical Epidemiology,t Henry Ford Hospital, Detroit, Michigan. Received for publication February 27, 1989. Revisions received September 15, 1989, and April 17, 1990. Accepted for publication June 13, 1990. Presented at the University Association for Emergency Medicine Annual Meeting in Cincinnati, Ohio, May 1988. Address for reprints: David J Trelear, MD, Division of Pediatric Emergency Medicine, Henry Ford Hospital, 2799 West Grand Bou!evard, Detroit, Michigan 48202.
Height of fever in infants and young children has been shown to be correlated with the incidence of bacterial disease,1 4 and emergency department personnel often record rectal and axillary temperatures of pediatric patients. Ideally, temperature determinations in an ED sl~ould be accurate reflections of core temperature (ie, mean temperature of the body's vital organs5), quick, easily performed, comfortable for the patient, and without complication. Recently, the validity of axillary temperatures has been questioned. 6 Rectal determinations are uncomfortable and time consuming and have been associated with rectal perforations in very young infants.h8 Core temperature measurements can be estimated from several sites, including the rectum, tympanic membrane, mouth, esophagus, and pulmonary artery. Although none is a perfect site, the distal esophagus is generally considered the most accurate available site.S, 9 Rectal temperatures tend to resist rapid changes in core and ambient temperature, s,9 whereas tympanic membrane temperatures are more subject to changes in ambient temperature, lo- 12 An infrared tympanic membrane thermometer (FirstTEMP ®, Intelligent Medical Systems, Carlsbad, California), purported to measure infrared output from the tympanic membrane, is now available and has been marketed for use in children. It is noninvasive, quick, and easy to use. Its use in young children in an ED has not been evaluate d. This study was conducted to determine correlation and sensitivity for both axillary and infrared tympanic membrane temperatures compared with rectal temperatures in infants and children less than 3 years old and
64/41
Annals of Emergency Medicine
20:1 January 1991
TEMPERATURES Muma et al
TABLE 1. Mean values
TABLE 2. Group correlations
Mean (SD) Age (mo) RT (C) AT (C) TMT (C)
12.4 38.00 36.48 37.29
(9.03) (0.94) (0.96) (1.05)
Temperature Differences Between Devices RT-AT RT-TMT AT-TMT
1,52 (0.67)* 0.71 (0.62)* 0.81 (0.74)*
RT, rectal temperature; A], axillary temperature; TMT, tympanfc membrane temperature. *P < .01
to d e t e r m i n e w h e t h e r t y m p a n i c membrane temperatures and axillary temperatures are useful screening tools for fever in this age group. MATERIALS AND METHODS
Rectal, a x i l l a r y , and t y m p a n i c membrane temperature measurements were performed on children less than 3 years old on presentation to the ED between April 1986 and April 1987. Informed consent was obtained. Children who were immunocompromised, were receiving chemotherapy, or had rectal trauma, infection, or a n o m a l i e s were excluded. Rectal and axillary temperatures were obtained with an electronic thermistor probe (Diatek 500 ®, Diatek Inc, San Diego, California). Calibration according to factory standards was performed daily for the first two weeks and biweekly thereafter. Tympanic membrane temperature was measured with a noncontact infrared-sensing device (FirstTEMP®). Its polyethylene, otoscopelike probe, measuring 8 m m in diameter, was inserted into the external auditory canal, and a digital readout was produced within three seconds. This readout represented a predicted rectal temperature based on one second of accumulated infrared energy when used in the rectal mode. The two devices were compared biweekly with a glass mercury therm o m e t e r to ensure accuracy. The presence of otitis media and ease in visualization of the tympanic membrane by otoscopy (ie, easily visualized, partially visualized, or not visualized) were evaluated by the researchers and pediatric housestaff as20:1 January 1991
RT Versus TMT r P
Group Total
N 223
RT (C) < 38.0 38 -- 39.5 > 39,5
136 67 20
.81"
.001
N 224
.12 .65* .24
.162 .001 .301
136 67 20
RT Versus AT r P .75"
.001
.03 .58* .22
.766 .001 .358
RT, rectal temperature; TMT, tympanic membrane temperature; AT, axillary temperature*P < .05.
signed to the ED before cleaning of the external auditory canal occurred. Correlation between rectal temperature and both tympanic membrane temperature and axillary temperature was estimated with Pearson's product-moment correlation coefficient for the entire group and various subgroups (stratified by age, rectal temperature, presence of otitis media, and visibility of the tympanic membrane). High correlations would indicate strong linear relationships between the temperatures. As an assessment of the ability of tympanic membrane temperature and axillary temperature to correctly classify subjects into fever and nonfever groups (based on rectal temperature of 38 C or less), the sensitivity, specificity, and predictive values were calculated for both axillary temperature and tympanic membrane temperature. The definition of fever was axillary temperature of more than 37.3 C and tympanic membrane temperature of 38 C or more. High sensitivities and specificities would indicate that tympanic m e m b r a n e temperatures (or axillary temperatures) were useful surrogate measures of rectal temperature. These measures can be low even when correlations are high and are better statistics for assessing the usefulness of axillary temperature and tympanic membrane temperature as screening tools.13 A s e c o n d set of m e a s u r e s for screening was generated after the data were collected using a new classification rule. This new definition of fever was based on the estimated value of tympanic membrane temperature or axillary temperature that predicted a rectal temperature value of 38 C using simple linear regression between the two devices. Sensitivity and specificity were estimated using these new fever definitions. Annals of Emergency Medicine
RESULTS
Two hundred twenty-four children were enrolled in the study. Of these, 87 were febrile. Mean values for age, temperature for each device, and the differences between the devices are s h o w n (Table 1). Eleven c h i l d r e n were admitted to the hospital; none was critically ill or bacteremic. Correlation of the infrared device and electronic t h e r m i s t o r to m e r c u r y thermometer was high (r = .89 for both). Data on external a u d i t o r y canal patency and otitis media were collected for 184 and 185 patients, respectively. Insertion of the tympanic membrane temperature probe was limited to only a few millimeters due to its relatively large diameter of 8 m m compared with the 3-ram pediatric ear speculum (Welch Allyn, disposable ear speculum, Skaneateles Falls, New York). Patients were stratified for height of rectal temperature. Correlation of t y m p a n i c m e m b r a n e temperature and axillary temperature to rectal temperature was calculated for each group (Table 2). Overall correlation was .81 for tympanic membrane temperature and .75 for axillary temperature. Correlation with axillary temperature was low in both afebrile and high-fever groups (.03 and .22, respectively). For tympanic membrane temperature, correlation with rectal temperature was also low in these subgroups (.12 and .24, respectively). The smaller correlations observed in the subgroups reflect the fact that when considering a smaller interval of observation, the general tendency for all temperatures to go up simultaneously is masked by the scatter in the data. The overall high correlation is a function of the general tendency of high temperatures from the three techniques to be grouped and low temperatures from the three techniques to be grouped. 42/65
TEMPERATURES M u m a et al
TABLE 3. Classification results for rectal versus tympanic membrane temperatures Afebrile (< 38C)
Rectal Temperature Low Fever High Fever Total (38 to 39.5C) (> 39.5C)
Tympanic Membrane Temperature Afebrile (< 38 C)
136
39
0
175
0
25
13
38
0 136
3 67
7 20
10 223
Low fever (38.0 to 39.5 C) High fever (> 39.5 C) Total
Sensitivity* =
48 48 + 39 = 55%
Specificity* =
36 136 + 0 = 100%
Positive predictive value*
48 48 + 0 = 100%
*StatisPc calculated from collapsed 2 x 2 table comparing afebrile with febrile group.
TABLE 4. Classification results for rectal versus axillary temperatures Afebrile (< 38C)
Rectal Temperature Low Fever High Fever Total (38 to 39.5C) (> 39.5C)
Axillary Temperature Afebrile (~< 37.3 C)
135
45
0
Febrile (> 37.3 C)
2
22
20
44
137
67
20
224
Total
Sensitivity* =
42 42 + 45
Specificity* =
135 135 + 2 = 96%
Positive predictive value*
~
180
48%
42
= 96%
*Statistic calculalod from collapsed 2 x 2 table comparing afebrfle with febrile group.
Correlation for t y m p a n i c membrane temperature did not vary with age (less than 2 months, .50; 2 to 18 months, .23; more than 18 months, .32; P > .373), canal patency (easy, .82; partial, .83; not seen, .76; P > .806), or presence of otitis media (yes, .76; no, .83; P > .433). Sensitivity and specificity for the prediction of fever (rectal temperature of 38 C or more) are shown (Tables 3 and 4). Sensitivity in detecting fever for both axillary temperature and tympanic membrane temperature was low. All false-negative tympanic membrane temperature or axillary t e m p e r a t u r e d e t e r m i n a t i o n s were in the low-temperature group. Moreover, of the 20 in the high-temperature group, 13 were categorized 66/43
into the low-temperature group by tympanic membrane temperature. DISCUSSION
Accurate temperature determination in febrile children is important in determining the presence and severity of disease. 1-4 Bacteremia is found in 3% to 15% of young children w h e n rectal t e m p e r a t u r e is more than 39.5 C, whereas meningitis is found in as many as 10% when rectal temperature is more than 41 C.1, ~4,1s Although axillary • temperature is often recorded in EDs, environmental and physiological factors confound its ability to accurately reflect core temperature. 6,7,16 Kresch reported that axillary temperatures did not reliably detect fever Annals of Emergency Medicine
in infants and children in an outpatient setting. 6 Shunting of blood from skin and periphery during fever onset or escalation or a low ambient temperature may falsely lower axillary and other surface temperatures.S-ZA 6 Axillary t e m p e r a t u r e has been shown to be useful in the stable temperature environment of a nursery or neonatal ICU,17, is but this relationship may be invalid when children are exposed to lower ambient temperatures, such as when being transported to the ED. Axillary temperature and tympanic m e m b r a n e temperature were poor predictors of fever in this study based on unacceptably low sensitivities. Although correlation between rectal and both axillary temperature and t y m p a n i c m e m b r a n e temperature was relatively high for the entire group, indicating a strong linear relationship, it was low in the high-fever group. The loss of correlation in the high-fever group may reflect greater physiologic shunting from the skin or simply small sample size. When the axillary temperature, defined as fever, decreased to 36.5 C (the best predictive value of fever using the data regression model), sensitivity increased to 80% and specificity decreased to 87%. These values are unacceptable because one of five fevers would be missed and false-positives would account for 13% of the axillary fevers. Conceptually, tympanic membrane temperature is a good indicator of core temperature because the membrane receives blood from the same vasculature as the hypothalamus.t0 Several investigators have used tympanic membrane temperature (FirstTEMP ®) in adult patients. Shinozaki 19 measured t y m p a n i c membrane temperature, pulmonary artery temperature, and rectal temperature in postoperative adult patients and reported a correlation of .98 between t y m p a n i c m e m b r a n e temperature and pulmonary artery temperature. The difference between pulmonary artery temperature and t y m p a n i c m e m b r a n e temperature, however, was larger than the difference between pulmonary artery temperature and rectal temperature. Ward et a120 compared rectal temperature and t y m p a n i c m e m b r a n e temperatures in adult ED patients and found a correlation of .94. Neither group, however, calculated sen20:1 January 1991
TEMPERATURES Mumaet al
sitivity or specificity for tympanic membrane temperature in detecting rectal temperature. The poor sensitivity for tympanic membrane temperature determinations in our study was surprising but may be due to the inability of the tympanic membrane temperature probe to adequately accumulate energyfrom the infant tympanic membrane. The 8-ram probe diameter, approximately twice the size of a pediatric ear speculum, limited probe insertion to only a few millimeters. This size discrepancy between probe tip and ear canal .may have produced readings representative of skin, external auditory canal, or cerumen. The inability of the probe to sense the tympanic m e m b r a n e may explain why correlations were unaffected by 0titis media or poor visibility. There are also data to suggest that the tympanic membrane and external auditory canal readings are more highly affected by ambient temperature, facial cooling, and wind conditions than rectal temperatures.lO, lz This m a y be more i m p o r t a n t in small children, whose temperature stability is less than that of adults. ~1 In addition to the low overall sensitivity, tympanic membrane temperature misclassified 13 of 20 high rectal t e m p e r a t u r e s i n t o the l o w - f e v e r group. The inability of t y m p a n i c membrane temperature and axillary temperature to correctly classify severity of fever c o m p r o m i s e s their value as screening tools in this setting.
20:1 January 1991
Further analysis of the data revealed that of the 55 febrile patients missed by either method, 33 were missed by both. This suggests that in some children, both thermometers may have been affected by a common physiologic or environmental factor such as ambient temperature or skin perfusion. Similarities in correlation and sensitivity between axillary temperature and t y m p a n i c m e m b r a n e t e m p e r a t u r e i m p l y that t y m p a n ic membrane temperature may reflect surface temperature in some children. CONCLUSION The presence and height of fever play a crucial role in guiding workup and subsequent therapy. Both axillary and infrared t y m p a n i c membrane t e m p e r a t u r e s (FirstTEMP ®) should be viewed with caution in children less than 3 years old who present to the ED as neither is able to reliably detect fever in this group.
REFERENCES 1. McCarthy PL, Dolan TF Jr: Hyperpyrexia in children: Eight-year emergency room experience. A m J Dis Child 1976;1;30:849-85l. 2. Caspe WB, Chamudes O, Louie B: The evaluation and treatment of the febrile infant. Pediatr Inf Dis 1983;2:131-135.
5. Elder PT: Accidental hypothermia, in Shoemaker WC, Thompson WL, Holbrook PR (eds}: Textbook of Critical Care Medicine. Philadelphia, WB Saunders, 1984, p 85-93. 6. Kresch MJ: Axillary temperature as a screening test for fever in children. ]" Pediatr 1984;104:596-599. 7. Togawa T: Body temperature measurement. Clin Phys PhysioI Meas 1985,6:83-108. 8. Wolfson JS: Rectal perforation in infants by thermometer. A m J Dis Child 1966;111:197-200. 9. Edwards RV: Core temperature measurement in man. A~qat Space Environ Med 1978~49:1289 1294. 10. McCaffrey TV, Geis GS, Chung JM, et ah Effect of isolated head heating and cooling *~n sweating in man. Aviat Space Environ Med 1975;46:1353-1357. 11. Morgans LF, Nunneley SA, gtribley RF: Influence of ambient and core temperatures on auditory canal temperature. Aviat Space Environ Med 1981;52:291-293. 12. Shiraki K, Konda N, Sagawa S: Esophageal and tyro panic temperature responses to core blood temperature changes during hyperthermia. J Appl PhysioI 1986; 61:98-102. 13. Feinstein AR: Clinical biostatistics: XXXI. On the sensitivity, specificity, and discrimination of diagnostic tests. Cfin Pharmacol Ther 1975~17:104-116. 14. Schwartz RH, Wientzen RL: Occult bacteremia in toxic appearing febrile infants. Pediatrics 1982;21:659. 15. Teele DW, Pehon SI, Grant MJ, et al: Bacteremia in febrile children under 2 years of age: Results of blood cultures of 600 consecutive febrile children in a "walk in" clinic, f Pediatrics 1975~87:221 300. 16. Jaffe DM, Tanz RR, Davis AT, et ah Antibiotic administration to treat possible occult bacteremia in febrile children. N Engl J Med 1987;317:1175-1180. 17. Guyton C: Guyton Textbook of Medical PhysioIogJa Philadelphia, WB Saunders, 1986, p 856-860. 18. Blainey CG: Site selection in taking body temperature. A m J Nuts 1974;74:1859-1861. 19. Shinozaki T, Deane R, Perkins FM: Infrared tympanic thermometer: Evaluation of a new clinical thermometer. Crit Care Med 1988116:148-180.
3. McCarthy PL, Jekel JF, Dolan TF Jr: Temperature greater than or equal to 40°C in children less than 24 m o n t h s of age: A p r o s p e c t i v e study. Pediatrics 1977;59:663-668.
20. Ward L, Kaplan RM, Paris PM: A comparison of tympanic and rectal temperatures in the emergency department (abstract}. Ann Emerg Med 1988;17:198.
4. McCarthy PL, Dolan TF: The serious implications of high fever in infants during their first three months: Six years' experience at Yale-New Haven Hospital emer~ gency room. Clin Pediatr 1976;15:794-796.
21. Livingston R: Visceral control mechanisms, in West JB (ed): Best and Taylor's Physiologic Bases of Medicine and Practice. Baltimore, Williams & Wilkins, 1984, p 1227-1230.
Annals of Emergency Medicine
44/67
Comparison of Rectal, Axillary, and Tympanic Membrane Temperatures in Infants and Young Children Study objective: To evaluate the reliability of a tympanic membrane thermometer in detecting fever in young children presenting to the emergency department. Setting: Pediatric emergency department in an urban teaching hospital. Design/Measurement/Participants: Temperature measurements were obtained sequentially at three body sites in children less than 3 years old presenting to the pediatric ED. Axillary and rectal temperatures were obtained with an electronic thermistor probe (Diatek 500®), and tympanic membrane temperatures were obtained with a noncontact, infrared sensing device (FirstTEMP®). Patients were stratified by age, ear canal patency, presence of otitis media, and rectal temperature. Results: Of 224 patients enrolled, 87 (39%) were febrile. Overall correlation of axillary and tympanic membrane measurements to rectal for all strata was .75 (P = .001) and .81 (P = .00I), respectively. Sensitivity in detecting fever for axil]ary and tympanic membrane sites was .48 and. 55, respectively. Otitis media and ear patency did not influence correlation of tympanic membrane measurements. Low tympanic membrane temperature sensitivity m a y be a result of probe configuration. Conclusion: Tympanic membrane and axillary temperatures should be viewed with caution in children less than 3 years old as neither can detect fever reliably. [Muma BK, Treloar D J, Wurmlinger K, Peterson E, Vitae A: Comparison of rectal, axillary, and tympanic membrane temperatures in infants and young children.-Ann Emerg Med January 1991;20:41-44.]
INTRODUCTION
Bruce K Muma, MD* David J Treloar, MD, MSc* Karen Wurmlinger, RN* Edward Peterson, PhDt April Vitae, RN* Detroit, Michigan From the Division of Pediatric Emergency Medicine, Department of Pediatrics,* and the Department of Biostatistics and Clinical Epidemiology,t Henry Ford Hospital, Detroit, Michigan. Received for publication February 27, 1989. Revisions received September 15, 1989, and April 17, 1990. Accepted for publication June 13, 1990. Presented at the University Association for Emergency Medicine Annual Meeting in Cincinnati, Ohio, May 1988. Address for reprints: David J Trelear, MD, Division of Pediatric Emergency Medicine, Henry Ford Hospital, 2799 West Grand Bou!evard, Detroit, Michigan 48202.
Height of fever in infants and young children has been shown to be correlated with the incidence of bacterial disease,1 4 and emergency department personnel often record rectal and axillary temperatures of pediatric patients. Ideally, temperature determinations in an ED sl~ould be accurate reflections of core temperature (ie, mean temperature of the body's vital organs5), quick, easily performed, comfortable for the patient, and without complication. Recently, the validity of axillary temperatures has been questioned. 6 Rectal determinations are uncomfortable and time consuming and have been associated with rectal perforations in very young infants.h8 Core temperature measurements can be estimated from several sites, including the rectum, tympanic membrane, mouth, esophagus, and pulmonary artery. Although none is a perfect site, the distal esophagus is generally considered the most accurate available site.S, 9 Rectal temperatures tend to resist rapid changes in core and ambient temperature, s,9 whereas tympanic membrane temperatures are more subject to changes in ambient temperature, lo- 12 An infrared tympanic membrane thermometer (FirstTEMP ®, Intelligent Medical Systems, Carlsbad, California), purported to measure infrared output from the tympanic membrane, is now available and has been marketed for use in children. It is noninvasive, quick, and easy to use. Its use in young children in an ED has not been evaluate d. This study was conducted to determine correlation and sensitivity for both axillary and infrared tympanic membrane temperatures compared with rectal temperatures in infants and children less than 3 years old and
64/41
Annals of Emergency Medicine
20:1 January 1991
TEMPERATURES Muma et al
TABLE 1. Mean values
TABLE 2. Group correlations
Mean (SD) Age (mo) RT (C) AT (C) TMT (C)
12.4 38.00 36.48 37.29
(9.03) (0.94) (0.96) (1.05)
Temperature Differences Between Devices RT-AT RT-TMT AT-TMT
1,52 (0.67)* 0.71 (0.62)* 0.81 (0.74)*
RT, rectal temperature; A], axillary temperature; TMT, tympanfc membrane temperature. *P < .01
to d e t e r m i n e w h e t h e r t y m p a n i c membrane temperatures and axillary temperatures are useful screening tools for fever in this age group. MATERIALS AND METHODS
Rectal, a x i l l a r y , and t y m p a n i c membrane temperature measurements were performed on children less than 3 years old on presentation to the ED between April 1986 and April 1987. Informed consent was obtained. Children who were immunocompromised, were receiving chemotherapy, or had rectal trauma, infection, or a n o m a l i e s were excluded. Rectal and axillary temperatures were obtained with an electronic thermistor probe (Diatek 500 ®, Diatek Inc, San Diego, California). Calibration according to factory standards was performed daily for the first two weeks and biweekly thereafter. Tympanic membrane temperature was measured with a noncontact infrared-sensing device (FirstTEMP®). Its polyethylene, otoscopelike probe, measuring 8 m m in diameter, was inserted into the external auditory canal, and a digital readout was produced within three seconds. This readout represented a predicted rectal temperature based on one second of accumulated infrared energy when used in the rectal mode. The two devices were compared biweekly with a glass mercury therm o m e t e r to ensure accuracy. The presence of otitis media and ease in visualization of the tympanic membrane by otoscopy (ie, easily visualized, partially visualized, or not visualized) were evaluated by the researchers and pediatric housestaff as20:1 January 1991
RT Versus TMT r P
Group Total
N 223
RT (C) < 38.0 38 -- 39.5 > 39,5
136 67 20
.81"
.001
N 224
.12 .65* .24
.162 .001 .301
136 67 20
RT Versus AT r P .75"
.001
.03 .58* .22
.766 .001 .358
RT, rectal temperature; TMT, tympanic membrane temperature; AT, axillary temperature*P < .05.
signed to the ED before cleaning of the external auditory canal occurred. Correlation between rectal temperature and both tympanic membrane temperature and axillary temperature was estimated with Pearson's product-moment correlation coefficient for the entire group and various subgroups (stratified by age, rectal temperature, presence of otitis media, and visibility of the tympanic membrane). High correlations would indicate strong linear relationships between the temperatures. As an assessment of the ability of tympanic membrane temperature and axillary temperature to correctly classify subjects into fever and nonfever groups (based on rectal temperature of 38 C or less), the sensitivity, specificity, and predictive values were calculated for both axillary temperature and tympanic membrane temperature. The definition of fever was axillary temperature of more than 37.3 C and tympanic membrane temperature of 38 C or more. High sensitivities and specificities would indicate that tympanic m e m b r a n e temperatures (or axillary temperatures) were useful surrogate measures of rectal temperature. These measures can be low even when correlations are high and are better statistics for assessing the usefulness of axillary temperature and tympanic membrane temperature as screening tools.13 A s e c o n d set of m e a s u r e s for screening was generated after the data were collected using a new classification rule. This new definition of fever was based on the estimated value of tympanic membrane temperature or axillary temperature that predicted a rectal temperature value of 38 C using simple linear regression between the two devices. Sensitivity and specificity were estimated using these new fever definitions. Annals of Emergency Medicine
RESULTS
Two hundred twenty-four children were enrolled in the study. Of these, 87 were febrile. Mean values for age, temperature for each device, and the differences between the devices are s h o w n (Table 1). Eleven c h i l d r e n were admitted to the hospital; none was critically ill or bacteremic. Correlation of the infrared device and electronic t h e r m i s t o r to m e r c u r y thermometer was high (r = .89 for both). Data on external a u d i t o r y canal patency and otitis media were collected for 184 and 185 patients, respectively. Insertion of the tympanic membrane temperature probe was limited to only a few millimeters due to its relatively large diameter of 8 m m compared with the 3-ram pediatric ear speculum (Welch Allyn, disposable ear speculum, Skaneateles Falls, New York). Patients were stratified for height of rectal temperature. Correlation of t y m p a n i c m e m b r a n e temperature and axillary temperature to rectal temperature was calculated for each group (Table 2). Overall correlation was .81 for tympanic membrane temperature and .75 for axillary temperature. Correlation with axillary temperature was low in both afebrile and high-fever groups (.03 and .22, respectively). For tympanic membrane temperature, correlation with rectal temperature was also low in these subgroups (.12 and .24, respectively). The smaller correlations observed in the subgroups reflect the fact that when considering a smaller interval of observation, the general tendency for all temperatures to go up simultaneously is masked by the scatter in the data. The overall high correlation is a function of the general tendency of high temperatures from the three techniques to be grouped and low temperatures from the three techniques to be grouped. 42/65
TEMPERATURES M u m a et al
TABLE 3. Classification results for rectal versus tympanic membrane temperatures Afebrile (< 38C)
Rectal Temperature Low Fever High Fever Total (38 to 39.5C) (> 39.5C)
Tympanic Membrane Temperature Afebrile (< 38 C)
136
39
0
175
0
25
13
38
0 136
3 67
7 20
10 223
Low fever (38.0 to 39.5 C) High fever (> 39.5 C) Total
Sensitivity* =
48 48 + 39 = 55%
Specificity* =
36 136 + 0 = 100%
Positive predictive value*
48 48 + 0 = 100%
*StatisPc calculated from collapsed 2 x 2 table comparing afebrile with febrile group.
TABLE 4. Classification results for rectal versus axillary temperatures Afebrile (< 38C)
Rectal Temperature Low Fever High Fever Total (38 to 39.5C) (> 39.5C)
Axillary Temperature Afebrile (~< 37.3 C)
135
45
0
Febrile (> 37.3 C)
2
22
20
44
137
67
20
224
Total
Sensitivity* =
42 42 + 45
Specificity* =
135 135 + 2 = 96%
Positive predictive value*
~
180
48%
42
= 96%
*Statistic calculalod from collapsed 2 x 2 table comparing afebrfle with febrile group.
Correlation for t y m p a n i c membrane temperature did not vary with age (less than 2 months, .50; 2 to 18 months, .23; more than 18 months, .32; P > .373), canal patency (easy, .82; partial, .83; not seen, .76; P > .806), or presence of otitis media (yes, .76; no, .83; P > .433). Sensitivity and specificity for the prediction of fever (rectal temperature of 38 C or more) are shown (Tables 3 and 4). Sensitivity in detecting fever for both axillary temperature and tympanic membrane temperature was low. All false-negative tympanic membrane temperature or axillary t e m p e r a t u r e d e t e r m i n a t i o n s were in the low-temperature group. Moreover, of the 20 in the high-temperature group, 13 were categorized 66/43
into the low-temperature group by tympanic membrane temperature. DISCUSSION
Accurate temperature determination in febrile children is important in determining the presence and severity of disease. 1-4 Bacteremia is found in 3% to 15% of young children w h e n rectal t e m p e r a t u r e is more than 39.5 C, whereas meningitis is found in as many as 10% when rectal temperature is more than 41 C.1, ~4,1s Although axillary • temperature is often recorded in EDs, environmental and physiological factors confound its ability to accurately reflect core temperature. 6,7,16 Kresch reported that axillary temperatures did not reliably detect fever Annals of Emergency Medicine
in infants and children in an outpatient setting. 6 Shunting of blood from skin and periphery during fever onset or escalation or a low ambient temperature may falsely lower axillary and other surface temperatures.S-ZA 6 Axillary t e m p e r a t u r e has been shown to be useful in the stable temperature environment of a nursery or neonatal ICU,17, is but this relationship may be invalid when children are exposed to lower ambient temperatures, such as when being transported to the ED. Axillary temperature and tympanic m e m b r a n e temperature were poor predictors of fever in this study based on unacceptably low sensitivities. Although correlation between rectal and both axillary temperature and t y m p a n i c m e m b r a n e temperature was relatively high for the entire group, indicating a strong linear relationship, it was low in the high-fever group. The loss of correlation in the high-fever group may reflect greater physiologic shunting from the skin or simply small sample size. When the axillary temperature, defined as fever, decreased to 36.5 C (the best predictive value of fever using the data regression model), sensitivity increased to 80% and specificity decreased to 87%. These values are unacceptable because one of five fevers would be missed and false-positives would account for 13% of the axillary fevers. Conceptually, tympanic membrane temperature is a good indicator of core temperature because the membrane receives blood from the same vasculature as the hypothalamus.t0 Several investigators have used tympanic membrane temperature (FirstTEMP ®) in adult patients. Shinozaki 19 measured t y m p a n i c membrane temperature, pulmonary artery temperature, and rectal temperature in postoperative adult patients and reported a correlation of .98 between t y m p a n i c m e m b r a n e temperature and pulmonary artery temperature. The difference between pulmonary artery temperature and t y m p a n i c m e m b r a n e temperature, however, was larger than the difference between pulmonary artery temperature and rectal temperature. Ward et a120 compared rectal temperature and t y m p a n i c m e m b r a n e temperatures in adult ED patients and found a correlation of .94. Neither group, however, calculated sen20:1 January 1991
TEMPERATURES Mumaet al
sitivity or specificity for tympanic membrane temperature in detecting rectal temperature. The poor sensitivity for tympanic membrane temperature determinations in our study was surprising but may be due to the inability of the tympanic membrane temperature probe to adequately accumulate energyfrom the infant tympanic membrane. The 8-ram probe diameter, approximately twice the size of a pediatric ear speculum, limited probe insertion to only a few millimeters. This size discrepancy between probe tip and ear canal .may have produced readings representative of skin, external auditory canal, or cerumen. The inability of the probe to sense the tympanic m e m b r a n e may explain why correlations were unaffected by 0titis media or poor visibility. There are also data to suggest that the tympanic membrane and external auditory canal readings are more highly affected by ambient temperature, facial cooling, and wind conditions than rectal temperatures.lO, lz This m a y be more i m p o r t a n t in small children, whose temperature stability is less than that of adults. ~1 In addition to the low overall sensitivity, tympanic membrane temperature misclassified 13 of 20 high rectal t e m p e r a t u r e s i n t o the l o w - f e v e r group. The inability of t y m p a n i c membrane temperature and axillary temperature to correctly classify severity of fever c o m p r o m i s e s their value as screening tools in this setting.
20:1 January 1991
Further analysis of the data revealed that of the 55 febrile patients missed by either method, 33 were missed by both. This suggests that in some children, both thermometers may have been affected by a common physiologic or environmental factor such as ambient temperature or skin perfusion. Similarities in correlation and sensitivity between axillary temperature and t y m p a n i c m e m b r a n e t e m p e r a t u r e i m p l y that t y m p a n ic membrane temperature may reflect surface temperature in some children. CONCLUSION The presence and height of fever play a crucial role in guiding workup and subsequent therapy. Both axillary and infrared t y m p a n i c membrane t e m p e r a t u r e s (FirstTEMP ®) should be viewed with caution in children less than 3 years old who present to the ED as neither is able to reliably detect fever in this group.
REFERENCES 1. McCarthy PL, Dolan TF Jr: Hyperpyrexia in children: Eight-year emergency room experience. A m J Dis Child 1976;1;30:849-85l. 2. Caspe WB, Chamudes O, Louie B: The evaluation and treatment of the febrile infant. Pediatr Inf Dis 1983;2:131-135.
5. Elder PT: Accidental hypothermia, in Shoemaker WC, Thompson WL, Holbrook PR (eds}: Textbook of Critical Care Medicine. Philadelphia, WB Saunders, 1984, p 85-93. 6. Kresch MJ: Axillary temperature as a screening test for fever in children. ]" Pediatr 1984;104:596-599. 7. Togawa T: Body temperature measurement. Clin Phys PhysioI Meas 1985,6:83-108. 8. Wolfson JS: Rectal perforation in infants by thermometer. A m J Dis Child 1966;111:197-200. 9. Edwards RV: Core temperature measurement in man. A~qat Space Environ Med 1978~49:1289 1294. 10. McCaffrey TV, Geis GS, Chung JM, et ah Effect of isolated head heating and cooling *~n sweating in man. Aviat Space Environ Med 1975;46:1353-1357. 11. Morgans LF, Nunneley SA, gtribley RF: Influence of ambient and core temperatures on auditory canal temperature. Aviat Space Environ Med 1981;52:291-293. 12. Shiraki K, Konda N, Sagawa S: Esophageal and tyro panic temperature responses to core blood temperature changes during hyperthermia. J Appl PhysioI 1986; 61:98-102. 13. Feinstein AR: Clinical biostatistics: XXXI. On the sensitivity, specificity, and discrimination of diagnostic tests. Cfin Pharmacol Ther 1975~17:104-116. 14. Schwartz RH, Wientzen RL: Occult bacteremia in toxic appearing febrile infants. Pediatrics 1982;21:659. 15. Teele DW, Pehon SI, Grant MJ, et al: Bacteremia in febrile children under 2 years of age: Results of blood cultures of 600 consecutive febrile children in a "walk in" clinic, f Pediatrics 1975~87:221 300. 16. Jaffe DM, Tanz RR, Davis AT, et ah Antibiotic administration to treat possible occult bacteremia in febrile children. N Engl J Med 1987;317:1175-1180. 17. Guyton C: Guyton Textbook of Medical PhysioIogJa Philadelphia, WB Saunders, 1986, p 856-860. 18. Blainey CG: Site selection in taking body temperature. A m J Nuts 1974;74:1859-1861. 19. Shinozaki T, Deane R, Perkins FM: Infrared tympanic thermometer: Evaluation of a new clinical thermometer. Crit Care Med 1988116:148-180.
3. McCarthy PL, Jekel JF, Dolan TF Jr: Temperature greater than or equal to 40°C in children less than 24 m o n t h s of age: A p r o s p e c t i v e study. Pediatrics 1977;59:663-668.
20. Ward L, Kaplan RM, Paris PM: A comparison of tympanic and rectal temperatures in the emergency department (abstract}. Ann Emerg Med 1988;17:198.
4. McCarthy PL, Dolan TF: The serious implications of high fever in infants during their first three months: Six years' experience at Yale-New Haven Hospital emer~ gency room. Clin Pediatr 1976;15:794-796.
21. Livingston R: Visceral control mechanisms, in West JB (ed): Best and Taylor's Physiologic Bases of Medicine and Practice. Baltimore, Williams & Wilkins, 1984, p 1227-1230.
Annals of Emergency Medicine
44/67
