An editorial error occurred in this article by Ray B Jones and others (7 April, pp 835-40). In the diagram of patient flow through the study (p 836) the number of patients in the computer only group who were lost to follow up after session 2 because of refusal should have been two (not 20). In addition, some superfluous arrows crept into the figure, wrongly linking the boxes of patients lost to follow up in the computer only group and in the community psychiatric nurse only group.