|Home | About | Journals | Submit | Contact Us | Français|
Measurement of body temperature using the tympanic method (TTM) represents a standard non-invasive method which best approximates core temperature (CTM). However, published evidence increasingly casts doubt on the agreement between TTM and CTM. The aim of this study was to assess the agreement of these two methods of temperature measurement in critically ill patients.
We recruited 20 consecutive critically ill patients, who required indwelling thermistor-tipped arterial catheters for haemodynamic monitoring. For each patient we simultaneously collected the core temperature, obtained from the arterial catheter, and tympanic temperatures in the left and right ears, in the supine position. Tympanic temperature was recorded using the ear setting (TTMe) and the core setting tympanic temperature (TTMc) in accordance with the manufacturer's instructions (Genius™ 2; Kendall, Tyco Healthcare, MA, USA). Local ethics committee approval was waived and the study was conducted as a service improvement audit. Bland-Altman analysis was used to measure agreement between the two set of measurements.
From the 20 patients we made 102 paired measurements. Comparison of the TTM in the right and left ears resulted in high variability in the recorded temperature with a percentage error of 34.4%. In view of this observed large variability, the mean TTM between the two ears was used for subsequent comparison with core temperature. Comparison of core measurements against tympanic measurements using ear settings (TTMe) showed a small positive mean bias (limits of agreement (LOA); percentage error) of 0.3°C (-0.4 to 1.1°C, 2.4%). However, when the CTM and TTM were compared using the core temperature setting of the thermometer, the bias (LOA) increased to -0.97°C (-1.6 to -0.4°C), creating a larger bias than the unadjusted TTM.
Tympanic temperature measurements showed a large random variability compared with core temperatures and suggest bilateral averaged tympanic measurements may be necessary. In our series, using the core setting increased error and increased the bias between the two methods. Our findings raise the possibility that the diagnosis of fever and the need for further investigation may be affected by the choice of temperature measurement site and device setting used.