Accuracy of non-invasive body temperature measurement methods in critically ill patients: a prospective, bicentric, observational study

Objective: The accuracy of different non-invasive body temperature measurement methods in intensive care unit (ICU) patients is uncertain. We aimed to study the accuracy of three commonly used methods. Design: Prospective observational study. Setting: ICUs of two tertiary Australian hospitals. Participants: Critically ill patients admitted to the ICU. Interventions: Invasive (intravascular and intra-urinary bladder catheter) and non-invasive (axillary chemical dot, tympanic infrared, and temporal scanner) body temperature measurements were taken at study inclusion and every 4 hours for the following 72 hours. Main outcome measures: Accuracy of non-invasive body temperature measurement methods was assessed by the Bland–Altman approach, accounting for repeated measurements and significant explanatory variables that were identified by regression analysis. Clinical adequacy was set at limits of agreement (LoA) of 1C compared with core temperature. Results: We studied 50 consecutive critically ill patients who were mainly admitted to the ICU after cardiac surgery. From over 375 observations, invasive core temperature (mostly pulmonary artery catheter) ranged from 33.9C to 39C. On average, the LoA between invasive and non-invasive measurements methods were about 3C. The temporal scanner showed the worst performance in estimating core temperature (bias, 0.66C; LoA, 1.23C, +2.55C), followed by tympanic infrared (bias, 0.44C; LoA, 1.73C, +2.61C) and axillary chemical dot methods (bias, 0.32°C; LoA, 1.64C, +2.28C). No methods achieved clinical adequacy even accounting for significant explanatory variables. Conclusions: The axillary chemical dot, tympanic infrared and temporal scanner methods are inaccurate measures of core temperature in ICU patients. These non-invasive methods appeared unreliable for use in ICU patients.