Comparison of rectal and infrared ear temperatures in older hospital inpatients

To assess the agreement between infrared emission detection (IRED) ear and rectal temperatures and to determine the validity of IRED ear thermometry in detecting rectal fever.

Prospective, convenience sample, unblinded study.

An acute geriatric unit (teaching hospital) and a multidisciplinary intensive care unit.

The study included 45 inpatients (26 women and 19 men), aged 78.3+/-6.9 years, admitted over a 4-month period. Twelve of the patients were definitely infected.

Sequential rectal (RT) and ear temperature (ET) measurements were performed using mercury-in-glass and IRED ear thermometers, respectively. IRED ear temperatures were measured at both ears (unadjusted mode), with the highest of six ear temperatures considered the true value.

Mean RT (37.39 degrees C +/- 0.52 degrees C) was significantly (P<.001) higher than mean ET (36.89 degrees C +/-0.59 degrees C). A highly significant positive correlation was found between RT and ET (slope = 0.69; 95% CI, 0.52-0.86; P<.001; r = 0.78). The mean bias (mean of the differences) between RT and ET was 0.50 degrees C +/-0.37 degrees C (95% CI, 0.41 degrees C-0.59 degrees C), and the 95% limits of agreement -0.22 degrees C and 1.23 degrees C (95% CI, -0.38 degrees C to 1.39 degrees C). According to the standard criterion (RT > or =37.6 degrees C), 14 patients were febrile. Using an optimum IRED ear fever threshold (37.2 degrees C), the sensitivity and specificity of IRED ear thermometry for predicting rectal fever were 86% and 89%, respectively (positive predictive value, 80%; negative predictive value, 93%).

The degree of agreement between rectal temperature and the highest of six IRED ear temperatures was acceptable. Using an optimal IRED ear fever threshold of 37.2 degrees C (99 degrees F), IRED ear thermometry had acceptable sensitivity and specificity for predicting rectal fever.