The reliability and validity of dangerous behavior predictions.

Taken generally, the literature over the past ten or fifteen years casts considerable doubt on the ability of forensic psychiatrists to predict dangerous behavior of their patients.' Much of this work has been based on large-scale studies of persons found not guilty by reason of insanity or incompetent to stand trial. • Such studies, though extremely important, are limited because they deal only with patients who are alleged to have committed extremely serious crimes. Other problems are that in these previous ventures clinicians have not recorded their predictions in a reasonably standard fashion and that interclinician reliabilities of predictions have been lacking. In an important recent study Mullen and Reinehr put the matter well when, after having found essentially no relationship between clinical predictions of dangerousness and outcome at four years, they stated: "In point of fact, dangerousness has never been demonstrated to be an identifiable personality dimension" and go on to provide the challenge that gives impetus to the present article: "No investigator has been able to show agreement between judges or other instruments which purport to predict it" (p. 230). In a previous study published in this journal, 5 we offered data based on a sample of 598 persons assessed within the Brief Assessment Unit (BAU) at METFORS. The group-based, court-ordered assessment procedures used in the BAU are described in a recent book.6 Four psychiatrists made predictions about future dangerousness on a four-point scale, and all patients were followed via searches of records after an interval of two years. The methodology employed in the previous study was similar to that described below for the present project. We discovered in the previous study that aside from previous offense patterns, most demographic variables (age, sex, previous psychiatric history, and so forth) could not be used to predict outcomes but that clinical opinion was reasonably effective (though with many false positives).