Heat or miss: Peripheral thermometers underestimate body temperature

Peripheral thermometers, such as those used on the ear and armpits, are specific but not sensitive, and may be insufficient to rule out a fever, according to a recent meta-analysis.

“Most thermometers are afflicted with substantial random error,” the researchers said. “The under-appreciation of the uncertainty in measuring temperature while practicing medicine might have serious consequences; the limited accuracy and reproducibility of thermometers may translate into weak decision-making, a huge waste of resources, and suboptimal patient and population health outcomes.”

Forty-six studies were considered eligible for the meta-analysis, totalling 12,602 patients. Mercury-in-glass or digital thermometers used at the rectal site were set as the reference standard method for temperature measurement, and 38 °C was defined as the cut-off for fever.

Temporal artery infrared thermometry was employed in nine studies, including 2,533 participants, of whom 885 had a fever, according to rectal measurements. Specificity was generally high, ranging from 0.85 to 1.00, but sensitivity was highly varied, running from a maximum value of 0.91 to a minimum of 0.41. The respective pooled estimates were 0.96 and 0.76. [Intern Emerg Med 2020;doi:10.1007/s11739-020-02556-0]

“This means that adopting a 38 °C index test threshold, there are very few false positives, even at the relatively high prevalence of fever at 30 percent, but there are several false negatives, so the test is useful to rule in the disease when positive,” the researchers explained, pointing out that the certainty of evidence was moderate at best.

A similar pattern was reported for the use of tympanic infrared thermometry. The pooled estimates were 0.98 for specificity and 0.77 for sensitivity, and the quality of evidence likewise ranged from low to moderate.

For the other devices assessed, including forehead and digital axillary thermometers, there were three or fewer studies available for pooled analysis, thus making meta-analysis impossible.

Pairwise comparisons of different thermometer devices after network meta-analysis showed that relative to the standard rectal mercury-in-glass measurements, most other devices tended to underestimate body temperature.

For example, axillary digital thermometers returned temperature measurements that were 0.67 °C lower than the standard, and the difference was statistically significant. The same was true for axillary mercury-in-glass thermometry, which found temperatures 0.55 °C cooler.

“Our meta-analysis showed that alternative peripheral thermometers were not always accurate at estimating central core temperature, with a tendency to underestimate it up to 1 °C,” the researchers said, also noting that these devices tended to come with random errors.

“The interplay between the fixed and random error originating by the use of different thermometers might generate, in the worst case, clinically relevant differences in the order of 2 °C,” they added.