We found, that as a screening tool to detect fever ≥38.0 °C, the tympanic thermometer was significantly better than the temporal thermometer. However, the 6 months to 5 years age group was the only one in which the sensitivity of the ear thermometer was above 90%. If the tympanic device was used as a screening tool for fever and the tympanic measurement ≥37.8 °C was regarded as a positive test of fever, this meant that more than 95% of all febrile children would be detected at the screening, that only 44% of the children needed to continue with a rectal measurement and among these children 83% would be truly febrile. The forehead and ear temperature measurements were both poor to estimate the exact rectal temperature, with a 95% limits of agreement from around − 1.5 to + 1.2 °C for the forehead measurements and − 1 to + 1 °C for the ear measurements.
The search for alternatives to a rectal measurement of the temperature in children has been going on for decades. Several reviews have reported disappointing results [2, 10, 11]. The instruments used in our study were based on current technologies and developed for professional use. These devices have now been evaluated in a range of studies with mixed results.
The temporal device, TAT-5000 IR, was found to be reliable in children in studies from the US of 47 children, India 50 children, and Turkey with 218 children. These studies only found minor temperature differences to the rectal temperature and sensitivities above 80% to detect fever [4,5,6].
In contrast, other research groups found that the temporal device was still too inaccurate to be recommended. In Belgium, 294 children had a deviation from the rectal temperature from − 1.32 to + 1.33 °C and the thermometer had sensitivity of 41% to detect fever [8]. Four studies from the US of 44, 52, 147, and 474 children found too large mean differences in temperature and sensitivities from 53 to 70% to detect fever, [12,13,14,15] which are similar to findings among 156 children from Nigeria and 205 children from Argentina [16, 17]. These studies are weakened by small numbers, inadequate statistical methods or no sub-analysis within age groups. Our study does not have these limitations and concludes that temporal measurement is not able to measure the temperature within a clinically safe limits of agreement range nor is it useful as a screening tool for fever.
The infrared ear thermometer, ThermoScan PRO 4000, was examined in a single study of 205 children from Argentina, which also compared it against the same temporal device as in our study: the TAT-5000 IR. The mean temperature difference to rectal temperature was 0.001 °C and the 95% limits of agreement from − 0.77 to + 0.77 °C. The sensitivity to detect fever was 92% [17]. These findings are quite similar to our findings. We conclude that the tympanic thermometer is better than the temporal thermometer at measuring temperature but still has a wide limit of agreement range, and almost every tenth child with a fever will go undetected using this method.
These conclusions are valid for the devices used in the study. However, a recent study from Switzerland reached similar results, using other devices which were based on the same technologies [18].
Our findings have a number of clinical implications. Firstly, temporal devices have no place as a substitute or screening tool for temperature measurements. Secondly, the tympanic thermometer has reached a level of accuracy where it can be used as a screening tool for detecting fever, with an AUC of 0.97 and more than 95% of all children with fever will be detected if the cut point for tympanic measurement is 37.8 °C. It should be considered, however, that there will still be 5% of the children who are not detected. This means, that an approach of screening all children with the tympanic device and continue with a rectal measurement if the screening is positive would result in that less than half of the children would need a rectal measurement. Thirdly, the usefulness of tympanic temperature to measure the exact temperature is less convincing and the clinicians must accept a limits of agreement range of around +/− 1 °C, which is too wide for clinical use, where a range of less than +/− 0.5 °C has been considered acceptable [13].
The strength of the current study is that it is a comprehensive real-life pragmatic study with sub-analyses in different age groups. Furthermore, in contrast to most of the published studies, it was analyzed with appropriate statistical methods, including Bland-Altman analysis [19, 20]. The study is weakened by the fact that the patients were not consecutively included, only when time allowed. Only 39% had fever, which limits the analysis, even though this is partially compensated by the high number of participants. Since the study was performed anonymously, no comparison was possible between the participating and non-participating children to assess selection bias. Furthermore, the order of temperature measurements was the same throughout the study: temporal, tympanic and rectal. We did so to reduce the risk of crying which could influence the temporal measurements. The procedure might introduce a bias however, if the temperature lowered during the examination. We tried to avoid this by undressing the child partially and only for a short time after the two other measurements had been performed. Finally, in accordance with the vast majority of other studies, and since our clinical guidelines and everyday clinical practice base on rectal measurements, we chose the rectal temperature as the gold standard to represent the core temperature. It has been questioned whether the rectal temperature represents the true core temperature better than the tympanic temperature [1]. Our study design does not address this question, instead merely examining whether the alternatives could replace the well-established rectal method.
The examined ear device is useful for screening purposes, but there is still room for improvement. We encourage the continued search for and development of new methods and technologies to replace the inconvenience of the rectal thermometer. For each new device developed, it is crucial to evaluate the accuracy in a clinical setting by using sufficient numbers of patients and appropriate methods for comparison.