BACKGROUND AND OBJECTIVE: Clinical sedation assessment becomes insufficient in deeply sedated patients. Bispectral index as a processed electroencephalogram parameter provides a continuous and observer-independent value reported to change with sedation. The aim of this prospective observational study was to determine the reliability and possible confounding factors of the bispectral index to assess sedation in surgical intensive care patients. METHODS: Following major surgery, bispectral index, body temperature and electromyographic activity of 44 ventilated patients were recorded. Sedation levels were assessed with Ramsay sedation score. RESULTS: Although bispectral index correlated with Ramsay sedation score (-0.64; P < 0.01) we found that in deeply sedated patients temperature instability and electromyographic activity increased bispectral index. Bispectral index correlated significantly with electromyographic activity (0.80; P < 0.01) and with an increase of body temperature (0.55; P < 0.01) not only in all patients but also in clinically deeply sedated patients (0.57; P < 0.01 and 0.46; P < 0.05). CONCLUSIONS: Only under certain conditions, such as low muscular activity and body temperature stability, may the bispectral index be a useful addition to clinical scoring in the sedation assessment of critically ill patients.