This article describes a systematic review of the psychometric properties of the Short-Term Assessment of Risk and Treatability (START) and a meta-analysis to assess its predictive efficacy for the 7 risk domains identified in the manual (violence to others, self-harm, suicide, substance abuse, victimization, unauthorized leave, and self-neglect) among institutionalized patients with mental disorder and/or personality disorder. Comprehensive terms were used to search 5 electronic databases up to January 2013. Additional articles were located by examining references lists and hand-searching. Twenty-three papers were selected to include in the narrative review of START's properties, whereas 9 studies involving 543 participants were included in the meta-analysis. Studies about the feasibility and utility of the tool had positive results but lacked comparators. START ratings demonstrated high internal consistency, interrater reliability, and convergent validity with other risk measures. There was a lack of information about the variability of START ratings over time. Its use in an intervention to reduce violence in forensic psychiatric outpatients was not better than standard care. START risk estimates demonstrated strong predictive validity for various aggressive outcomes and good predictive validity for self-harm. Predictive validity for self-neglect and victimization was no better than chance, whereas evidence for the remaining outcomes is derived from a single, small study. Only 3 of the studies included in the meta-analysis were rated to be at a low risk of bias. Future research should aim to investigate the predictive validity of the START for the full range of adverse outcomes, using well-designed methodologies, and validated outcome tools.