Objective To evaluate the modes of death and treatment offered in the last 24 h of life to patients dying in 10 Brazilian intensive care units (ICUs) over a period of 2 years. Design and setting Cross-sectional, multicentre, retrospective study based on medical chart review. The medical records of all patients that died in seven paediatric and three adult ICUs belonging to university and tertiary hospitals over a period of 2 years were included. Deaths in the first 24 h of (...) admission to the ICU and brain death were excluded. Intervention Two intensive care fellows of each ICU were trained in fulfilling a standard protocol (κ=0.9) to record demographic data and all medical management provided in the last 48 h of life. The Student t test, Mann–Whitney U test, χ2 test and RR were used for data comparison. Measurements and main results 1053 medical charts were included (59.4% adult patients). Life support limitation was more frequent in the adult group (86% vs 43.5%; p<0.001). A ‘do not resuscitate’ order was the most common life support limitation in both groups (75% and 66%), whereas withholding/withdrawing were more frequent in the paediatric group (33.9% vs 24.9%; p=0.02). The life support limitation was rarely reported in the medical chart in both groups (52.6% and 33.7%) with scarce family involvement in the decision making process (23.0% vs 8.7%; p<0.001). Conclusion Life support limitation decision making in Brazilian ICUs is predominantly centred on the medical perspective with scarce participation of the family, and consequently several non-coherent medical interventions are observed in patients with life support limitation. (shrink)