Table 6.
• Twenty-eight studies investigated the psychometric properties of the SARA measures, most of them taking place in North America. • Slightly over half the studies (54%) had the SARA coded in a research context, while only seven were coded by professionals in a clinical context. • Only three studies looked at the psychometric properties of the SARA-V3 and found them to be inferior to the SARA-V2. • Studies did not consistently report reliability coefficients. Those that did report found variable results. Overall, internal consistency was found to be moderate, mean inter-item correlation to be poor, and corrected item totals ranged from very poor to excellent depending on the item. • Overall, the SARA measures demonstrated good IRR when scored in an actuarial manner but much lower reliability when SRR were used to communicate risk. IRR at the item level showed a great amount of variability, with multiple items being less than adequate. • The SARA measures showed strong convergent validity with multiple IPV measures, most notably the ODARA, but also the DVSI, the DVRAG, DA, and the B-SAFER. The SARA measures also showed good convergent validity with general and violence measures such as the PCL, HCR-20, and LSI-R. • The SARA was found to have acceptable if somewhat modest, predictive accuracy when predicting IPV recidivism (weighted summary AUC = 0.63 for actuarially summed total score, AUC = 0.65 for SRR). Predictive accuracy was slightly higher for violent (AUC = 0.67) and general (AUC = 0.72) recidivism. • Only one study investigated the predictive accuracy of the SARA-V3. The SARA-V3 predicted violence (AUC = 0.68–0.70) and general recidivism (AUC = 0.75–0.77) with moderate accuracy. • Few studies analyzed validity at the item level. Those who did found that most of the SARA’s items do not directly inform IPV recidivism risk assessment. |
Note. AUC = area under the curve; B-SAFER = Brief-Spousal Assault Form for the Evaluation of Risk; DA = Danger Assessment; DVRAG = Domestic Violence Risk Assessment Guide; DVSI = Domestic Violence Risk Instrument; HCR-20 = Historical Clinical Risk Management-20 – Historical subscale; IRR = Inter-rater reliability; LSI-R = Level of Service Inventory-Revised; ODARA = Ontario Domestic Assault Risk Assessment; PCL = Psychopathy Checklist-Revised or Screening Version; SARA = Spousal Assault Risk Assessment; SRR = summary risk ratings.