RESUMEN
The brain's temperature measurements (TB) in patients with severe brain damage are important, in order to offer the optimal treatment. The purpose of this research is the creation of mathematical models for the TB's prediction, based on the temperatures in the bladder (TBL), femoral artery (TFA), ear canal (TΕC), and axilla (TA), without the need for placement of intracranial catheter, contributing significantly to the research of the human thermoregulatory system.The research involved 18 patients (13 men and 5 women), who were hospitalized in the adult intensive care units (ICU) of Larissa's two hospitals, with severe brain injury. An intracranial catheter with a thermistor was used to continuously measure TB and other parameters. The TB's measurements, and simultaneously one or more of TBL, TFA, TEC, and TA, were recorded every 1 h.To create TB predicting models, the data of each measurement was separated into (a) model sample (measurements' 80%) and (b) validation sample (measurements' 20%). Multivariate linear regression analysis demonstrated that it is possible to predict brain's temperature (PrTB), using independent variables (R2 was TBL = 0.73, TFA = 0.80, TEC = 0.27, and TA = 0.17, p < 0.05). Significant linear associations were found, statistically, and no difference in means between TB and PrTB of each prediction model. Also, the 95% limits of agreement and the percent coefficient of variation showed sufficient agreement between the TB and PrTB in each prediction model.In conclusion, brain's temperature prediction models based on TBL, TFA, TEC, and TA were successful. Its determination contributes to the improvement of clinical decision-making.