REGULAR ARTICLE

Accuracy of tympanic and forehead thermometers in private paediatricpracticeJ Teller1*, M Ragazzi2*, GD Simonetti (giacomo.simonetti@insel.ch)2, SAG Lava2,3

1.Private Paediatric Practice, Langnau i.E., Switzerland2.Division of Paediatric Nephrology, University Children’s Hospital Bern, University of Bern, Bern, Switzerland3.Department of Paediatrics, Ospedale Regionale Bellinzona e Valli, Bellinzona and University of Bern, Bern, Switzerland

KeywordsChildren, Fever, Infrared forehead, Infrared tympanic,Thermometer

CorrespondenceGD Simonetti, MD, Division of PaediatricNephrology, University Children’s Hospital,Inselspital and University of Bern, Freiburgstrasse3010 Bern, Switzerland.Tel: +41 31 632 95 12 |Fax: +41 31 632 94 20 |Email: giacomo.simonetti@insel.ch

Received24 July 2013; revised 7 September 2013;accepted 10 October 2013.

DOI:10.1111/apa.12464

*These authors contributed equally to this work.

ABSTRACTAim: To compare infrared tympanic and infrared contact forehead thermometer

measurements with traditional rectal digital thermometers.

Methods: A total of 254 children (137 girls) aged one to 24 months (median 7 months)

consulting a private paediatric practice because of fever were prospectively recruited. Body

temperature was measured using the three different devices.

Results: The median and interquartile range for rectal, tympanic and forehead

thermometers were 37.6 (37.1–38.4)°C, 37.5 (37.0–38.1)°C and 37.5 (37.1–37.9)°C,respectively (p < 0.01). The limits of agreement in the Bland-Altman plots were �0.73 to

+1.04°C for the tympanic thermometer and �1.18 to +1.64°C for the forehead

thermometer. The specificity of both the tympanic and forehead thermometers for

detecting fever above 38°C was good, but sensitivity was low. Forehead measurements

were susceptible to the use of a radiant warmer.

Conclusion: Both the tympanic and forehead devices recorded lower temperatures than

the rectal thermometers. The limits of agreement were particularly wide for the forehead

thermometer and considerable for the tympanic thermometer. In the absence of valid

alternatives, because of the ease to use and little degree of discomfort, tympanic

thermometers can still be used with some reservations. Forehead thermometers should

not be used in paediatric practice.

INTRODUCTIONFever is a frequent symptom in childhood and one of themost common reasons for taking a child to the doctor (1).Ideally, temperature measurement techniques should besafe, easy to perform, noninvasive, cost-effective and time-efficient and should accurately reflect core body tempera-ture (2,3). Traditionally, body temperature has typicallybeen taken orally in older children and adults and rectallyin infants and young children. Rectal measurement, the goldstandard in young children (3), is reliable and accuratelyreflects the body’s core temperature. However, the proce-dure is invasive and not well accepted among children andparents. Therefore, alternative measurement methods areneeded (2,4).

In infants younger than 4 weeks, axillary measurement isextremely accurate (2,4). In older children and adults,however, this method, although easy to perform andgenerally well tolerated, showed variable sensitivity. Fur-thermore, although an axillary reading is generally 0.5°Clower than a rectal reading, an exact conversion factorcannot be established (2). Tympanic measurements usinginfrared thermometers have also been repeatedly investi-gated, with conflicting results (5). Furthermore, only one

study, performed more than 20 years ago (6), investigatedtympanic thermometers in the setting of a private paediatricpractice. More recently, infrared cutaneous contact ther-mometers that can assess forehead temperature haveappeared on the market. So far, only a few publishedstudies have examined the validity of these thermometersand none have been performed in a private paediatricpractice (2,3,7).

Key notes� Our study of 254 children presenting with fever in a

private paediatric practice compared readings frominfrared tympanic, infrared contact forehead and tradi-tional rectal digital thermometers.

� We found that tympanic and forehead thermometersrecorded lower temperatures than digital rectal ther-mometers, and their limits of agreement were partic-ularly wide.

� The findings suggest that tympanic thermometers canbe used with some reservations, but forehead ther-mometers should not be used in paediatric practice.

e80 ©2013 Foundation Acta Pædiatrica. Published by John Wiley & Sons Ltd 2014 103, pp. e80–e83

Acta Pædiatrica ISSN 0803-5253

The aim of this noncommercially sponsored study was tocompare infrared tympanic and infrared contact foreheadthermometer measurements with those from traditionalrectal digital thermometers within a large sample of infantsand toddlers presenting at a private paediatric practice.

PATIENTS AND METHODSIn this prospective study, infants and toddlers between oneand 24 months of age with no underlying chronic illness orknown pathology and presenting to the private paediatricpractice because of fever were eligible to participate.Participants were consecutively recruited between Novem-ber 1, 2011 and April 2, 2012.

Age and gender were recorded for each patient, andtemperature measurements were collected as part of theroutine clinical diagnostic procedure. After explaining thepurpose of temperature measurement, informed consentwas obtained from the parents. No data are available onthose who declined to consent. After having created a quietatmosphere, during the physical examination, taking care ofavoiding any haste, a licensed paediatrician on the project(JT) took the three temperature measurements, each ofwhich was performed at room temperature (22–24°C). Allthree measurements were taken within 3 min.

Each patient’s rectal temperature was measured using theMicrolife MT1961TM (Microlife, Widnau, Switzerland) con-ventional digital thermometer. Infrared tympanic tempera-ture was measured using the ThermoScan 6022TM (Braun,Kronberg, Germany) thermometer, and infrared skin fore-head temperature was measured using the VoiceThermowdc 6603 BTM (Wellness-design-company, D€usseldorf,Germany) thermometer. All the thermometers (always thesame device for each type of thermometer) were usedstrictly according to the manufacturers’ manuals.

The use of an infrared radiant warmer was sometimesnecessary. Its use was annotated, and its influence onmeasurements was evaluated.

Power calculation was performed to determine samplesize. At least 198 children were needed to achieve a studypower of 80%, with a error = 0.05, estimating a potentialdifference between the two methods of 0.2°C and astandard deviation of 1°C (8). Normal distribution ofvariables was tested using the D’Agostino–Pearson omnibustest for normality. Not normally distributed data arepresented as median and interquartile range (IQR). Vari-ables were compared by the nonparametric Friedmananalysis of variance test, and the Dunn’s post-test procedurewas used to compare the different measurements. TheFisher exact test was used to compare dichotomousvariables; sensitivity, specificity, positive predictive valueand negative predictive value were calculated, using rectaltemperature of 38.0°C, as the generally used cut-off pointfor rectally measured fever (3,9). Moreover, receiver oper-ating characteristics (ROC) curve analysis was performed toestablish the best threshold in determining the presence ofrectally measured fever (≥38°C) for tympanic and foreheadthermometers. Concordance among the different methods

was examined using Bland–Altman plots, which display themean differences and 95% limits of agreement (10). Statis-tical significance was assigned at p < 0.05.

RESULTSOur final sample included 254 children (117 boys and 137girls) ranging in age from one to 24 months, with a medianof 7.0 (5.0–11.0) months.

Figure 1 shows the temperature readings from the threetypes of thermometers investigated. The medians and IQRfor the rectal, tympanic and forehead thermometers were37.6 (37.1–38.4)°C, 37.5 (37.0–38.1)°C and 37.5 (37.1–37.9)°C, respectively. The tympanic and forehead measure-ments were 0.1°C (�0.1–0.4, p < 0.01) and 0.15°C (�0.3–0.7; p < 0.05) lower than the rectal measurements, respec-tively (Fig. 1). However, tympanic and forehead measure-ments were not significantly different from each other(Fig. 1). Using Bland–Altman plots, the 95% limits ofagreement were �0.73 to +1.04°C (Fig. 2, upper panel) forthe tympanic measurements and �1.18 to +1.64°C (Fig. 2,lower panel) for the forehead measurements.

While the specificity in detecting temperatures above38°C was good, the sensitivity was low when using eitherthe tympanic or the forehead thermometer (Table 1).Similarly, the positive predictive value was good (rangingfrom 88% to 95%), and the negative predictive value wasadequate (ranging from 71% to 84%, Table 1). Calculatingthe ROC curve, we found that the best threshold fordetecting rectal temperature ≥38°C was 37.55°C for boththe tympanic and the forehead thermometers. In particular,for a tympanic temperature of 37.55°C sensitivity was93.1% (95% CI: 86.4–97.2%) and specificity 83.5% (95% CI:76.7–89.0%) with an area under the curve of 0.94 (95% CI:

Figure 1 Box plot of temperatures measured with the 3 different devices. Theboundary of the box closest to zero indicates the 25th percentile, the line withinthe box marks the median, and the boundary of the box farthest from zeroindicates the 75th percentile. Whiskers (error bars) above and below the boxindicate the 90th and 10th percentiles. (ns = not significant).

©2013 Foundation Acta Pædiatrica. Published by John Wiley & Sons Ltd 2014 103, pp. e80–e83 e81

Teller et al. Accuracy of thermometers in paediatric practice

0.91–0.97, p < 0.0001). On the other side, for a foreheadtemperature of 37.55°C sensitivity was 80.4% (95% CI:71.4–87.6%) and specificity 73.0% (95% CI: 65.2–79.9%)with an area under the curve of 0.83 (95% CI: 0.78–0.88,p < 0.0001).

A radiant warmer was used in 117 children, but notdocumented in42 children. Subgroupanalyses indicated thatthe radiant warmer only influenced forehead measurements.

DISCUSSIONThis study demonstrates that both tympanic and foreheadthermometers devices record lower temperatures thanrectal thermometers. This can frequently underestimatethe core temperature. Although both devices showedexcellent specificity and positive predictive values, sensitiv-ity was low, and the negative predictive values weresuboptimal. With a lower threshold of 37.55°C for boththe tympanic and forehead thermometers, derived from aROC curve, sensitivity slightly improved, particularly for thetympanic thermometer.

The forehead thermometer showed clear inconsistency,with limits of agreement encompassing a range of almost3°C. While the tympanic thermometer showed a slightlybetter degree of agreement, its range of almost 2°C isclinically inappropriate (11,12). These results are consistentwith previous data (13–15). More importantly, a well-performed systematic review of studies comparing rectaland tympanic thermometers found acceptable mean differ-ences but similarly wide limits of agreement (5).

In addition, the present results indicate that foreheadmeasurements were susceptible to the use of a radiantwarmer, explaining at least in part their insufficient accu-racy, as well as supporting the conclusion that these devicesare not appropriate for use in a private paediatric practice.

There are two innovative aspects of this study. First, themeasurements were performed in the real-world setting of aprivate paediatric practice using modern devices. Childrenpresenting to a private paediatric practice represent the vastmajority of feverish children cared for by paediatricians.Nonetheless, only one study, performed more than 20 yearsago, analysed tympanic thermometer measurements in thissetting (6). Second, we assessed the influence of externalfactors on recorded temperatures, such as the use of aradiant warmer. Although the negative effect of radiantheaters on temperature measurement might appear evident,our data demonstrate that rectal and tympanic readings arenot affected by the presence of a radiant warmer and cantherefore be rated as robust. On the other side, foreheadthermometers are extremely dependent on external vari-ables and are therefore even more unreliable as alreadyshown by their wide limits of agreement.

Another significant strength of the present study is theinclusion of children with a febrile illness. In contrast, mostthermometer studies have utilised a mixed sample ofhealthy and sick patients, without a clear selection basedon the pretest probability of finding a feverish temperature.Moreover, infrared cutaneous contact thermometers haveonly been commercially available for a few years, and dataon their use in a private paediatric practice are lacking.

The present study has some limitations. First, only onemodel for each type of thermometer was analysed; thus, ourresults cannot be generalised to other models. Second, onlychildren 24 months of age or less were included. Thus, noconclusions for older children can be drawn. Thirdly,sensitivity, specificity, positive predictive value and negativepredictive value are computed from children presenting

Figure 2 Bland–Altman plots, which display the difference between twomeasurements against their average. The data plotted should ideally scatteraround a difference of zero across the range of means: the smaller the scatter isaway from the mean (solid line), the greater the agreement. The scatter isquantified by calculating the limits of agreement (dotted lines), which areestimated by the mean difference and 2 standard deviations of the difference.Upper panel: Bland–Altman plot of tympanic measurements. The meandifference is �0.15°C, and the 95% limits of agreement range from �0.73 to+1.04°C. Lower panel: Bland–Altman plot of the forehead thermometer. Themean difference is �0.23°C, and the 95% limits of agreement range from �1.18to +1.64°C.

Table 1 Sensitivity, specificity, positive (PPV) and negative predictive values (NPV)with 95% confidence interval for detecting rectal fever (≥38.0°C) with the tympanicand infrared skin thermometers

Thermometer Sensitivity Specificity PPV NPV

Tympanic 0.72

(0.62–0.80)

0.97

(0.93–0.99)

0.95

(0.87–0.99)

0.84

(0.77–0.89)

Infrared skin 0.42

(0.32–0.52)

0.96

(0.92–0.99)

0.88

(0.75–0.95)

0.71

(0.64–0.77)

e82 ©2013 Foundation Acta Pædiatrica. Published by John Wiley & Sons Ltd 2014 103, pp. e80–e83

Accuracy of thermometers in paediatric practice Teller et al.

with fever; therefore, these data should be interpreted withrespect to the study design. Fourthly, we did not evaluatethe reproducibility of measurements performed by eachtype of thermometer. However, the high number of partic-ipants included in the study should have minimised theimpact of measurement imprecision. Finally, our resultsmight not be automatically generalised to inpatients or tochildren presenting to a paediatric emergency department.

In conclusion, according to the results of the presentstudy, a reliable and valid alternative to invasive rectaltemperature measurement is still lacking. Forehead ther-mometers similar to that investigated in the present studyshould not be used in private paediatric practice because oftheir wide limits of agreement along with their vulnerabilityto the influence of external factors such as a radiantwarmer. In the absence of valid alternatives, and consider-ing that tympanic thermometers are easy to use, create littlediscomfort and are the preferred method of physicians andnurses at several institutions, such thermometers can beemployed but with some reservations. Their lack of sensi-tivity and negative predictive value, together with their widelimits of agreement, render them suboptimal as a screeningtool, in contrast to some recently proposed data (16).According to other studies, axillary measurement mightrepresent a better screening tool (15). This notion is alsosupported by the recently published update of the ItalianGuidelines on Fever Management (17). Finally, veryrecently, a temporal artery thermometer was also shownto perform very well, with limits of agreement in a range ofapproximately 0.5°C (18).

Patient history and a careful clinical examination remainthe most important criteria for the evaluation of a feverishchild. The clinical assessment and suspicion should guidediagnostic and therapeutic decisions, rather than tempera-ture readings (19). As already suggested more than 40 yearsago by Smith, it is the child that should be treated, not thethermometer (20).

CONFLICT OF INTERESTNone.

ACKNOWLEDGEMENTSNo support from any organisation for the submitted work.

References

1. Diethelm M, Largo RH. Prospektive Studie €uber die T€atigkeitdes praktizierenden Kinderarztes – II. Besuchsgr€unde undDiagnosen. Helv Paediatr Acta 1986; 41: 29–52.

2. Chiappini E, Principi N, Longhi R, Tovo PA, Becherucci P,Bonsignori F, et al. Management of fever in children: summary

of the Italian Pediatric Society guidelines. Clin Ther 2009; 31:1826–43.

3. Paes BF, Vermeulen K, Brohet RM, van der Ploeg T, de WinterJP. Accuracy of tympanic and infrared skin thermometers inchildren. Arch Dis Child 2010; 95: 974–8.

4. Richardson M, Lakhanpaul M. Assessment and initialmanagement of feverish illness in children younger than 5years: summary of NICE guidance. BMJ 2007; 334: 1163–4.

5. Craig JV, Lancaster GA, Taylor S, Williamson PR, Smyth RL.Infrared ear thermometry compared with rectal thermometry inchildren: a systematic review. Lancet 2002; 360: 603–9.

6. Freed GL, Fraley JK. Lack of agreement of tympanic membranetemperature assessments with conventional methods in aprivate practice setting. Pediatrics 1992; 89: 384–6.

7. De Curtis M, Calzolari F, Marciano A, Cardilli V, Barba G.Comparison between rectal and infrared skin temperaturein the newborn. Arch Dis Child Fetal Neonatal Ed 2008; 93:F55–7.

8. Chiappini E, Sollai S, Longhi R, Morandini L, Laghi A, OsioCE, et al. Performance of non-contact infrared thermometerfor detecting febrile children in hospital and ambulatorysettings. J Clin Nurs 2011; 20: 1311–8.

9. Avner JR, Baker MD. Management of fever in infants andchildren. Emerg Med Clin North Am 2002; 20: 49–67.

10. Bland JM, Altman DG. Agreed statistics - measurement methodcomparison. Anesthesiology 2011; 116: 182–5.

11. Duncan AL, Bell AJ, Chu K, Greenslade JH. Can a non-contactinfrared thermometer be used interchangeably with otherthermometers in an adult Emergency Department? AustralasEmerg Nurs J 2008; 11: 130–4.

12. Rajee M, Sultana RV. NexTemp thermometer can be usedinterchangeably with tympanic or mercury thermometers foremergency department use. Emerg Med Australas 2006; 18:245–51.

13. Brennan DF, Falk JL, Rothrock SG, Kerr RB. Reliability ofinfrared tympanic thermometry in the detection of rectal feverin children. Ann Emerg Med 1995; 25: 21–30.

14. Fadzil FM, Choon D, Arumugam K. A comparative study onthe accuracy of noninvasive thermometers. Aust FamPhysician 2010; 39: 237–9.

15. Shann F, Mackenzie A. Comparison of rectal, axillary, andforehead temperatures. Arch Pediatr Adolesc Med 1996; 150:74–8.

16. Edelu BO, Ojinnaka NC, Ikefuna AN. Fever detection in under5 children in a tertiary health facility using the infrared tympanicthermometer in the oral mode. Ital J Pediatr 2011; 37: 8.

17. Chiappini E, Venturini E, Principi N, Longhi R, Tovo PA,Becherucci P, et al. Update of the 2009 Italian PediatricSociety Guidelines about management of fever in children. ClinTher 2012; 34: 1648–53 e3.

18. Batra P, Goyal S. Comparison of rectal, axillary, tympanic, andtemporal artery thermometry in the pediatric emergency room.Pediatr Emerg Care 2013; 29: 63–6.

19. Lava SA, Simonetti GD, Ferrarini A, Ramelli GP, BianchettiMG. Regional differences in symptomatic fever managementamong paediatricians in Switzerland: the results of across-sectional Web-based survey. Br J Clin Pharmacol 2013;75: 236–43.

20. Smith DS. Fever and the pediatrician. J Pediatr 1970; 77:935–6.

©2013 Foundation Acta Pædiatrica. Published by John Wiley & Sons Ltd 2014 103, pp. e80–e83 e83

Teller et al. Accuracy of thermometers in paediatric practice