Interpreting multiple risk scales for sex offenders: evidence for averaging

This study tested 3 decision rules for combining actuarial risk instruments for sex offenders into an overall evaluation of risk. Based on a 9-year follow-up of 940 adult male sex offenders, we found that Rapid Risk Assessment for Sex Offender Recidivism (RRASOR), Static-99R, and Static-2002R predicted sexual, violent, and general recidivism and provided incremental information for the prediction of all 3 outcomes. Consistent with previous findings, the incremental effect of RRASOR was positive for sexual recidivism but negative for violent and general recidivism. Averaging risk ratios was a promising approach to combining these risk scales, showing good calibration between predicted (E) and observed (O) recidivism rates (E/O index = 0.93, 95% CI [0.79, 1.09]) and good discrimination (area under the curve = 0.73, 95% CI [0.69, 0.77]) for sexual recidivism. As expected, choosing the lowest (least risky) risk tool resulted in underestimated sexual recidivism rates (E/O = 0.67, 95% CI [0.57, 0.79]) and choosing the highest (riskiest) resulted in overestimated risk (E/O = 1.37, 95% CI [1.17, 1.60]). For the prediction of violent and general recidivism, the combination rules provided similar or lower discrimination compared with relying solely on the Static-99R or Static-2002R. The current results support an averaging approach and underscore the importance of understanding the constructs assessed by violence risk measures.