Bengtson Susanne, Långström Niklas
Department of Psychology, University of Aarhus, Jens Chr. Skous Vej 4, 8000, Arhus, Denmark.
Sex Abuse. 2007 Jun;19(2):135-53. doi: 10.1177/107906320701900205. Epub 2007 May 30.
Meta-analyses suggest that actuarial risk assessments outperform unguided clinical judgment for prediction of recidivism in criminal offenders. However, there is a lack of direct comparisons of the predictive accuracy of clinical judgment and actuarial risk scales for sexual offenders. We followed up 121 male sex offenders (> or =18 years) subjected to pre-trial forensic psychiatric assessment in Denmark in 1978-1992 (mean post-detainment time = 16.4 years) to compare the predictive validity of unstructured clinical judgment of recidivism risk with that of the well-established Static-99 (Hanson and Thornton, Law and Human Behavior 24:119-136, 2000) and an extension of the Static-99, the Static-2002 (Hanson and Thornton, Notes on the development of Static-2002 (Rep. No. 2003-01), Public Safety and Emergency Preparedness Canada, Ottawa, Canada, 2003). The predictive accuracy of unguided judgment did not exceed chance for any sexual, severe sexual or any violent (sexual or non-sexual) reconviction (AUCs of the ROC curve = 0.52, 95%CI = 0.41-0.63; 0.50, 95%CI = 0.34-0.67; and 0.57, 95%CI = 0.40-0.73, respectively). In contrast, all three outcomes were predicted significantly better than chance by the Static-99 (AUC = 0.62, 95%CI = 0.52-0.72; 0.72, 95%CI = 0.59-0.84; and 0.71, 95%CI = 0.56-0.86) and the Static-2002 (AUC = 0.67, 95%CI = 0.57-0.77; 0.69, 95%CI = 0.56-0.83; and 0.70, 95%CI = 0.55-0.86). Static-99 outperformed clinical judgment for sexual recidivision (chi(2) = 5.11, df = 1, p < .05). The Static-2002 was significantly more accurate for the prediction of any sexual recidivism as compared to unguided clinical judgment but its advantage fell just short of statistical significance for severe sexual recidivism (chi(2) = 3.56, df = 1, p = 0.06). When tested for recidivism within 2 years, none of the three prediction methods yielded results significantly better than chance for any outcome. This direct trial of the unguided clinical method argues against its continued use for risk assessment of sexual offenders.
荟萃分析表明，在预测刑事罪犯的再犯情况方面，精算风险评估优于无指导的临床判断。然而，对于性犯罪者，临床判断和精算风险量表的预测准确性缺乏直接比较。我们对1978年至1992年在丹麦接受审前法医精神病学评估的121名男性性犯罪者（年龄≥18岁）进行了随访（平均拘留后时间 = 16.4年），以比较无结构化临床判断再犯风险与成熟的Static - 99（汉森和桑顿，《法律与人类行为》24:119 - 136，2000年）以及Static - 99的扩展版Static - 2002（汉森和桑顿，《关于Static - 2002的发展说明》（报告编号2003 - 01），加拿大公共安全与应急准备部，渥太华，加拿大，2003年）的预测效度。对于任何性犯罪、严重性犯罪或任何暴力（性或非性）再定罪，无指导判断的预测准确性未超过随机水平（ROC曲线的AUC分别为 = 0.52，95%CI = 0.41 - 0.63；0.50，95%CI = 0.34 - 0.67；以及0.57，95%CI = 0.40 - 0.73）。相比之下，Static - 99（AUC = 0.62，95%CI = 0.52 - 0.72；0.72，95%CI = 0.59 - 0.84；以及0.71，95%CI = 0.56 - 0.86）和Static - 2002（AUC = 0.67，95%CI = 0.57 - 0.77；0.69，95%CI = 0.56 - 0.83；以及0.70，95%CI = 0.55 - 0.86）对所有三种结果的预测均显著优于随机水平。对于性再犯，Static - 99优于临床判断（χ² = 5.11，df = 1，p <.05）。与无指导的临床判断相比，Static - 2002在预测任何性再犯方面显著更准确，但在预测严重性再犯方面其优势略低于统计学显著性（χ² = 3.56，df = 1，p = 0.06）。当测试2年内的再犯情况时，三种预测方法对于任何结果均未产生显著优于随机水平的结果。这种对无指导临床方法的直接试验反对其继续用于性犯罪者的风险评估。
