Statistics from Altmetric.com
Body temperature is an important prognostic factor in acute stroke, and it is a marker of life threatening infections such as pneumonia.1 In clinical practice rectal temperature is regarded as a valid and reliable indicator of body temperature, but in the past 15 years tympanic infrared thermometers have been introduced into widespread clinical use. Tympanic temperature measurement is easier, faster, and less invasive than rectal measurement. However, concerns remain about the reliability and validity of this method.2
The product manual of the tympanic thermometer warns against overestimation of body temperature when the patient has been lying on one ear. Many stroke patients are hemiparalytic or hemiparetic and therefore may lie on one ear. The aim of our study was to investigate the error in tympanic temperature measurements in patients who have had a stroke and have been lying on one ear in this way.
For this observational study we included patients who had had an ischaemic or haemorrhagic stroke and were admitted to the stroke unit of a university medical centre or to the stroke rehabilitation unit of an affiliated nursing home, both in an urban area. Exclusion criteria were inability to lie on one ear and absence of a rectal cavity. We used the NIH stroke scale (NIHSS) to assess stroke severity.3 No follow up was conducted.
Tympanic temperature was measured in both ears after the patient had been lying on one ear of choice for at least 15 minutes (first measurement). The measurements were carried out in patients on waking in the morning or after an afternoon nap. The ear the patient had been lying on and the other ear will be further referred to as the lower ear and the upper ear, respectively. The second measurement was conducted by a different investigator—who was not aware of the results of the first measurement—approximately 10 minutes after the patient had no longer been lying on one ear, and consisted of tympanic and rectal temperature measurements.
To assess reliability we computed the difference between lower and upper ear temperature in consecutive tympanic measurements and analysed this difference graphically, as described by Bland and Altman.4 Validity was assessed by comparing the mean tympanic temperatures with mean rectal temperature.
We studied 30 patients (nine male, 21 female). Their mean age was 69.5 years, and their median NIHSS score was 11, ranging from 0 to 38. The mean of the first tympanic temperature taken from the lower ear was 37.6°C. The mean tympanic temperatures taken from the upper ear were both 37.2°C and the second measurement from the lower ear had a mean of 37.3°C. Mean rectal temperature was also 37.3°C. The mean difference between the two ears was 0.39°C (95% confidence interval, 0.22 to 0.56). This difference ranged from −0.4°C to 1.7°C (fig 1). The mean difference between the first measurement taken from the lower ear and the rectal temperature was 0.29°C (0.13 to 0.45).
Our study showed a clinically significant difference between tympanic temperature measurements in the two ears after a stroke patient had been lying on one ear. This difference disappeared after a while when the patient was no longer lying on one ear. There was no relation between actual body temperature and the size of the measurement error. Although the manufacturer of the tympanic thermometer cautions against heating of the auditory canal when the patient has been lying on one ear, especially in children, neither this effect nor its magnitude is well known.
In our opinion, a systematic error of 0.4°C on average is not acceptable from a clinical point of view. It may lead to unnecessary investigations and treatment with antibiotics in a considerable number of patients; moreover, an error of this size could have decreased the statistical power of clinical trials of temperature lowering treatment in acute stroke patients, if tympanic temperature was used without attention being paid to the side the patient was lying on.5
The support of the nursing staff of the stroke units of the Erasmus Medical Centre (René Rook) and nursing home Antonius Binnenweg (Hannie van Gijsen) is gratefully acknowledged.
Competing interests: None declared