OBJECTIVE To compare three devices that measure core body temperature at the bedside in ICU patients. METHODS Prospective, consecutive sample. METHODS Voluntary community teaching hospital. METHODS Fifteen patients, 78 +/- 6 (SD) yrs of age, admitted to the medical ICU over a 5-month period who had pulmonary artery catheters inserted as part of their routine care were studied. Thirteen patients were studied once, one patient twice, and one patient six times for a total of 21 sets of measurements. METHODS All patients had urinary bladder thermistor catheters inserted just before pulmonary artery catheterization. Simultaneous core temperatures were measured for the duration of pulmonary artery catheterization every 4 hrs by the pulmonary artery thermistor catheter, the bladder thermistor catheter, and by a tympanic membrane infrared probe set on its core temperature setting. The three devices were then compared with each other in vitro using a specialized constant water bath setup. Finally, two of the tympanic membrane infrared probes were compared with each other in 20 ambulatory emergency department patients. RESULTS Over 32 hrs of pulmonary artery catheterization, the pulmonary artery thermistor catheters and bladder thermistor catheters showed excellent agreement, with a bias of only -0.04 degrees C between the two. However, the bias comparing the tympanic membrane infrared probe with the pulmonary artery thermistor catheter was -0.38 degrees C, and the bias was -0.34 degrees C comparing the tympanic membrane infrared probe with the bladder thermistor catheter. The tympanic membrane infrared probe readings remained significantly higher than the pulmonary artery thermistor catheter or bladder thermistor catheter readings over the entire 32-hr period. The two tympanic membrane infrared probes compared with each other in emergency room patients showed excellent agreement (p less than .001). In the in vivo water bath setup, the tympanic membrane infrared probe on most of its settings again registered significantly (p less than .01) higher than both the pulmonary artery thermistor catheter and the bladder thermistor catheter. CONCLUSIONS Pulmonary artery thermistor catheters and bladder thermistor catheters appear to give consistent, highly reliable bedside measurements of core body temperature in ICU patients. The currently available device that measures core body temperature from the tympanic membrane appears to give erroneously high readings, and should be used with caution.