Relevance of interrater agreement to violence risk assessment

This study considered whether assessments of violence risk in which 2 clinicians reach similar conclusions are more accurate than the conclusions of either clinician aIone when their assessments disagree. One hundred ten physicians and 44 nurses estimated the probability of physical assault of 478 patients admitted to a shea-term locked psychiatric inpatient unit. The level of assessed risk showed a substantial correspondence with the likelihood of later violence when the physician and nurse ratings were highly concordant. As the extent of agreement between the physician and nurse ratings decreased, the strength of the association between the risk assessments and the occurrence of violence decreased accordingly.