Fever is a very common complaint that is encountered almost daily in any emergency department (20). Accurate temperature measurement is crucial in order to minimise risk of sepsis in infancy. Despite explicit research about the safety and accuracy of AT measurement method, limited published data about how accurate the AT method compared to RT in detecting fever is available for infants. Moreover, AT has been widely used for fever screening in paediatric emergency departments due to safety, hygiene and convenience for patients, families and the nursing staff. However, a concern regarding high false-negative rates (21) especially when the heart rate is elevated renders paediatricians to doubt the accuracy of AT and require the use of core temperature measurement such as the rectal method. The sensitivity of the AT compared to RT has not been investigated extensively in the infants age group. In addition, information about the association between infant age groups and level of AT sensitivity is minimal. Most published data were merely on newborns and neonates (10, 22-26). Our study explicitly investigates the accuracy of AT measurement compared to the gold standard RT. We believe that AT screening carries a substantial risk of missing the diagnosis of febrile infant. To our knowledge, this is the first study to delineate the importance of specifically using RT for any infants who presents to the emergency department with a recent history of fever.
The relationship between the RT and AT was assessed using the Bland-Altman analysis. It shows a mean difference of 0.8 °C with a 95% CI range of ± 1.5 °C which indicates a large difference between the two methods. The difference is considered significant due to a narrow range of normal temperature between 36.5-37.5 °C (18). The febrile infants would be more likely to be missed if the difference was up to 1.5 °C with such a narrow normal range for AT allowing them to succumb to complications. The first to show poor agreements between AT and RT was a meta-analysis done by Craig et al (27), pooled mean rectal-axillary temperature measurements difference was 0.25 °C for mercury thermometers (95% CI, -1.5-0.65) and 0.85 °C (95% CI, 0.19- 1.9) for the electronic thermometer. The latter reflects a similar result found in our study for the electronic thermometer. The majority of studies included in the pooled analysis of the electronic thermometer were performed on young children other than neonates allowing for more consistency to our findings. Although the mercury thermometer seems to provide a more accurate agreement between the two methods, it’s no longer being used due to fear of mercury toxicity and has been replaced by a more convenient and safer digital thermometer (28). Furthermore, a recent study by Teller et al(29) performed on children < 24 months presenting with fever also discovered a rectal-axillary mean difference of 1.1°C with 95% limit of agreement between 0.32-1.98 °C which confirm our findings.
Jones et al (30) conducted a study on 573 children under the age of five on Gambia, West Africa. They found 98% sensitivity of AT for the detection of fever. Although the sensitivity is extremely high, the authors argues that the reason for is due to the high prevalence of tropical infectious diseases such as malaria in Africa. Tropical infectious diseases are well known to present with a high-grade fever allowing for a higher likelihood of detection by the axillary method. On the contrary, several other studies performed on infants report a various sensitivity to our study (73% (2), 81% (31), 64% (28), 62% (32), and 49%(21)). The reason for this variation might be attributed to the measurement device that has been used, the different age groups enrolled, different ambient temperatures or children weight difference for height. It has been shown in multiple subgroup analysis that neonatal fever was detected with a high sensitivity by the axillary method resembling similar findings to our study (11, 32). Thus, a potential effect on accuracy is possible when considering the age factor.
Our analysis suggests that AT is less sensitive at detecting fever >35.4 °C upon patient triage in the emergency department. It may miss approximately 20% of infants with rectal hyperthermia. Shine et al (33) performed a comparison analysis exactly similar to our study and found nearly comparable results (29% false negative rate using AT). Subgroup analysis suggests that the AT is quite sensitive for infants <3 months as opposed to markedly decreasing sensitivity for infants in the older age groups. The latter finding was also reported in another studies (11, 33). An explanation for this difference may be related to body weight. Children with a low weight for height seems to have less wider range between the two methods compared to children with an average weight for heigh. Another explanation is a possible difference in the physiological mechanisms of thermoregulation for each age groups. Further investigation into this aspect should be sought for future studies.
To allow for a more convenient temperature measurement for families and healthcare providers, several lower than the normal cut-off value may be considered for AT measurement ranging from 37.2 °C to 37.4/ 37.5 °C. Lowering the cut-off value for the diagnosis of fever would substantially decrease the false negative rate. AT of 37.4/ 37.5 °C, 37.3 °C, 37.2°C showed a sensitivity of 79%, 83%, and 86% respectively.
Several limitations in our study should be taken into consideration. Almost all patients included in the study are febrile and the design of the study is prospective nested control. It may have an impact on increasing the sensitivity level. However, that was not observed in our analysis and the result reflects similar findings from other studies. Furthermore, the measurement of temperature may be affected by subjective techniques performed by the nursing staff. Some patients were excluded from the study due to a lack of rectal temperature readings.
In conclusion, the axillary temperature measurement represents a poor screening method and should not be considered accurate enough as an alternative to the rectal temperature method for the detection of fever in infants with a recent history of fever. The rectal method remains highly important for accurate and early diagnosis in the clinical context of suspected underlying infectious or inflammatory processes.