Sensitivity and Specificity of the Static-2002 in a Diverse Population

Assessments designed to estimate the probability of sexual offense recidivism have gained popularity due to research that suggests these instruments are psychometrically sound for this purpose. The Static-2002 (Hanson & Thornton, 2003), an actuarial instrument of sexual recidivism, has been validated in specific populations, but in the absence of multicultural populations. It is necessary to establish the validity of the Static-2002 for a multicultural population, especially considering the increasing impetus to mandate measures of sexual recidivism. This study was designed to assess the predictive validity of the Static-2002 as an actuarial instrument that estimates the probability of sexual, violent, and general recidivism within a multicultural population.Recent criminal histories and de-identified archival records of 103 adult sex offenders were used to complete demographic questionnaires and to calculate retroactive predictive accuracy scores for the Static-2002. Records were obtained from a private practice office in South Florida, which provides outpatient treatment to conditionally released offenders. Hypothesis testing utilized a receiver operating characteristic area under the curve (ROC AUC) analysis to examine the discriminating power of the Static-2002 and to identify optimal cutoff values for each risk category. These cutoff values were then compared to the established Static-2002 risk categories to evaluate the cutoff scores' congruency.For this sample population, resulting ROC AUC values indicated the predictive accuracy of the Static-2002 to be moderate for any (including general; .67) offense recidivism, and high for violent recidivism (.74). However, the Static-2002 was ineffective for predicting sexual recidivism among this sample population since the predictive accuracy of the ROC area for sexual reoffenses was not statistically significant above chance (p = .07). The four Static-2002 risk categories derived for this study are: low, moderate-low, moderate-high, and high. When compared with the established Static-2002 risk categories, this study's risk categories were found to be mostly concurrent with the established categories.Similar to previous research, this study found that when the sample population differs racially/ethnically from the normed population (60% of this study's participants were Hispanic), the Static-2002 was ineffective for accurately predicting sexual reoffending. However, the Static-2002 effectively predicted violent and general recidivism with this sample population.