OBJECTIVE There is an increasing recognition of the necessity to establish the predictive validity of risk assessment scores within specific population subgroups, particularly those (including Indigenous peoples) who are overrepresented in the criminal justice system. I compared measures of discrimination and calibration of the Youth Level of Service/Case Management Inventory (YLS/CMI) in Indigenous and non-Indigenous youth probationers in Ontario, Canada. HYPOTHESES Compared with non-Indigenous youth, Indigenous youth would have higher risk scores and reoffense rates. The YLS/CMI would predict reoffending and time to reoffense significantly and comparably for Indigenous and non-Indigenous youth, but there would be group difference discrimination (sensitivity, specificity) and calibration (positive predictive value, negative predictive value). METHOD Justice ministry-supplied data on 400 Indigenous and non-Indigenous youth (330 male, 70 female) individually matched on key background variables were analyzed to provide measures of discrimination and calibration of the YLS/CMI, with 3-year recidivism as the primary outcome. RESULTS Indigenous youth were assessed at significantly higher risk than non-Indigenous youth (d = .60); 70% of Indigenous youth and 46% of non-Indigenous youth reoffended (ϕ = .24). Overall measures of discrimination (area under the curve) and calibration (logistic regression) were significant and did not differ across groups. Cross-area under the curve results indicated that the YLS/CMI discriminated Indigenous recidivists from non-Indigenous nonrecidivists but differentiated Indigenous nonrecidivists from non-Indigenous recidivists at chance level. In addition, recidivism was underestimated for low-risk Indigenous youth compared with non-Indigenous youth, but specificity was also low; only 28% of Indigenous youth who did not reoffend were assessed as low risk. Results were largely consistent across male and female youth. CONCLUSIONS Examining subgroup predictive validity using multiple indices provides important information that should inform policy and practice discussions regarding fair use of risk assessment tools. (PsycInfo Database Record (c) 2024 APA, all rights reserved).

中文翻译：

---

比较加拿大土著和非土著青年服务水平/案例管理量表分数的预测效度。


目的 人们越来越认识到有必要在特定人口亚群体中建立风险评估分数的预测有效性，特别是那些在刑事司法系统中占比过高的人（包括土著人民）。我比较了加拿大安大略省土著和非土著青年缓刑人员的歧视措施和青年服务水平/案例管理量表 （YLS/CMI） 的校准。假设 与非土著青年相比，土著青年的风险评分和再犯罪率更高。YLS/CMI 将显着且可比地预测土著和非土著青年的再犯罪和再犯罪时间，但会存在群体差异区分（敏感性、特异性）和校准（阳性预测值、阴性预测值）。方法 分析了司法部提供的 400 名土著和非土著青年（330 名男性，70 名女性）在关键背景变量上单独匹配的数据，以提供歧视的测量和 YLS/CMI 的校准，以 3 年累犯为主要结果。结果 土著青年被评估为风险显著高于非土著青年 （d = .60）;70% 的原住民青年和 46% 的非原住民青年再次犯罪 （φ = .24）。区分 （曲线下面积） 和校准 （logistic 回归） 的总体测量是显着的，并且在组间没有差异。曲线下跨面积结果表明，YLS/CMI 区分了土著累犯和非土著非累犯，但在机会水平上区分了土著非累犯和非土著累犯。 此外，与非土著青年相比，低风险土著青年的累犯率被低估，但特异性也很低;只有 28% 没有再犯罪的土著青年被评估为低风险。男性和女性青年的结果基本一致。结论 使用多个指数检查亚组预测效度提供了重要信息，这些信息应为有关公平使用风险评估工具的政策和实践讨论提供信息。（PsycInfo 数据库记录 （c） 2024 APA，保留所有权利）。