Predictive validity of the Short-Term Assessment of Risk and Treatability for violent behavior in outpatient forensic psychiatric patients.

It remains unclear whether prediction of violence based on historical factors can be improved by adding dynamic risks, protective strengths, selection of person-specific key strengths or critical vulnerabilities, and structured professional judgment (SPJ). We examine this in outpatient forensic psychiatry with the Short-Term Assessment of Risk and Treatability (START) at 3 and 6 months follow-up. An incident occurred during 33 (13%) out of 252 3-month and 44 (21%) out of 211 6-month follow-up periods (n = 188 unique clients). Pearson correlations for all predictor variables were in the expected directions. Prediction of recidivism based on historical factor ratings (odds ratio [OR] = 1.10) could not be improved through the addition of dynamic risk, protective strength, or key or critical factor scores (all ORs ns). The addition of the SPJ improved the model to modest accuracy (area under the curve [AUC] = .64) but made no independent significant contribution (OR = 1.55, p = .21) for the 3-month follow-up. For the 6-month follow-up, SPJ scores also increased predictive accuracy to modest (AUC = .67) and made a significant independent contribution to the prediction of the outcome (OR = 1.98, p = .04). Multicollinearity limits were unviolated. Limitations apply, however, results are similar to those from clinical, researcher rated samples and are discussed in the light of setting specific characteristics. Although it is too early to advocate implementing risk assessment instruments in clinical practice, we can conclude that clinicians in a heterogeneous outpatient forensic psychiatric setting can achieve similar results with the START as clinicians and research staff in more homogeneous inpatient settings. (c) 2015 APA, all rights reserved).

RCT Entities:   Population Interventions Outcomes