Background Quality ICU end-of-life-care has been found to be related to good communication. Handover is one form of communication that can be problematic due to lost or omitted information. A first step in improving care is to measure and describe it. Objective The objective of this study was to describe the quality of ICU nurse handover related to end-of-life care and to compare the practices of different ICUs in three different countries. Design This was a descriptive comparative study. Settings The study was conducted in seven ICUs in three countries: Australia (1 unit), Israel (3 units) and the UK (3 units). Participants A convenience sample of 157 handovers was studied. Methods Handover quality was rated based on the ICU End-of-Life Handover tool, developed by the authors. Results The highest levels of handover quality were in the areas of goals of care and pain management while lowest levels were for legal issues (proxy and advanced directives) related to end of life. Significant differences were found between countries and units in the total handover score (country: F(2,154) = 25.97, p = <.001; unit: F(6,150) = 58.24, p = <.001), for the end of life subscale (country: F(2, 154) = 28.23, p < .001; unit: F(6,150) = 25.25, p = <.001), the family communication subscale (country: F(2,154) = 15.04, p = <.001; unit: F(6,150) = 27.38, p = <.001), the family needs subscale (F(2,154) = 22.33, p = <.001; unit: F(6,150) = 42.45, p = <.001) but only for units on the process subscale (F(6,150) = 8.98, p = <.001. The total handover score was higher if the oncoming RN did not know the patient (F(1,155) = 6.51, p = <.05), if the patient was expected to die during the shift (F(1,155) = 89.67, p = <.01) and if the family were present (F(1,155) = 25.81, p = <.01). Conclusions Practices of end-of-life-handover communication vary greatly between units. However, room for improvement exists in all areas in all of the units studied. The total score was higher when quality of care might be deemed at greater risk (if the nurses did not know the patient or the patient was expected to die), indicating that nurses were exercising some form of discretionary decision making around handover communication; thus validating the measurement tool.