When statistical models fail: problems in the prediction of escape and absconding behaviour from high-security hospitals

The issue of the prediction of problematic behaviour in psychiatric settings has generated a large body of literature over the last decade. There appears to be a clear consensus that statistical methods of prediction are more accurate and defensible than those based solely upon clinical judgement. The present study explores the prediction of absconding behaviour from maximum-security hospitals. A number of difficulties with standard statistical methods are identified and it is concluded that there is a strong case for idiographic assessment of absconding risk.