This study tests the accuracy of clinicians' predictions of future patient violence using a new model to depict both clinician assessment and patient behavior. This model reflects the findings of our current research showing that the prediction of dangerousness is not a logical, straightforward task of generating a likelihood estimate. It is instead a task that is done within, and shaped by, the need for continuous case management of violent patients. The proposal adopts the position that clinicians assessments of dangerousness are tied to broader schemas or scripts about the patient and his or her situation. These scripts are represented and tested in this study as a judgment of the clinician about what type of violence will occur under what conditions. We hypothesize that when one represents the clinical judgment and patient behavior in these conditional terms clinicians' ability to predict and manage dangerousness will be found to be higher than commonly assumed. The research will collect a detailed conditional assessment of dangerousness on 1600 patients from the two emergency room clinicians who assess the patient and, if the patient is admitted, the unit attending physician at discharge. A projected 400 patients judged dangerous, and 400 matched controls will be followed through three interviews over a six-month period and reviews of clinical and arrest records for these patients will be done. The results will allow us both to determine the predictive validity of various clinical predictions and to suggest other conditions which might aid in the management of these patients. These results will both prompt and focus policy debate about the role of dangerousness predictions.
Showing the most recent 10 out of 16 publications
