We selected a stratified random sample of 32 hospitals, four from each former health region ( and see bmj.com
). At each hospital we interviewed two to five key emergency and psychiatric staff about hospital service structures and made arrangements with them to start audits of the processes of care. We assessed each hospital on 21 recommended self harm service standards (see table A on bmj.com
In 2001-2 each hospital did a prospective eight week audit of their management of self harm (see bmj.com
). Trust staff used emergency department, medical, and mental health records if audit data were incomplete.
Variation in management of self harm patients across 32 English hospitals
A designated self harm or liaison service was available at 23 of the 32 hospitals. At 11 hospitals, more than half of the 21 recommended service structures were not in place (median score 12; range 7 to 20). The most commonly available aspects of service were guidelines for medical management (at 31 hospitals) and 24 hour access to specialist psychosocial assessments (at 30 hospitals) (see table A on bmj.com
Guidelines for assessing the risk of suicide for use by staff in emergency departments were available at 17 hospitals. Only 14 hospitals had self harm service planning meetings with mental health services, emergency department, or medical staff. Routine contact with patients' general practitioners within 24 hours of discharge from emergency departments happened at only half of the hospitals. Service scale scores were weakly associated with hospital size (rank correlation 0.20, P = 0.28).
During the eight week audit, staff identified 4222 episodes of self harm. Hospitals varied widely in the proportion of attendances leading to a psychosocial assessment (median 55%; range 36% to 82%), hospital admission (42%; 22% to 83%), psychiatric admission (9.5%; 2.5% to 23.8%), and mental health follow up (51%; 35% to 82%). Using metaregression techniques, we found no significant difference in the proportion of assessments (55% v 58%; odds ratio 0.88; 95% confidence interval 0.56 to 1.38; P = 0.57), admissions (42% v 52%; 0.65; 0.37 to 1.13; P = 0.13), psychiatric admissions (10.5% v 11.4%; 0.89; 0.59 to 1.37; P = 0.61), or arrangements for follow up (53% v 56%; 0.91; 0.66 to 1.25, P = 0.54) between hospitals with and without a designated service. However, at hospitals with a designated service, assessments were considerably less likely to be undertaken by junior (training grade) psychiatrists alone (22% v 75%; 0.04; 0.01 to 0.14; P < 0.01).