The APACHE II severity of illness scoring system was prospectively studied on 756 patients admitted to a general intensive care unit (ICU) from January 1986 to June 1988. Admission data were used. Individual and group risk of death were calculated for 3 diagnostic categories commonly seen in the A and E department and requiring admission to an ICU. The APACHE II score on admission tended to underestimate the risk of death following operative and non-operative trauma, and self-poisoning. This may have been related to the use of an analysis not yet validated against values obtained on admission to ICU. Such validation is urgently needed on a UK population if APACHE II scoring is to be of value in the A and E department.