Effect of age-at-release on long term sexual re-offense rates in civilly committed sexual offenders

A cohort of 136 rapists and 115 child molesters civilly committed to a prison in Massachusetts and followed for 25 years (see Prentky, Lee, Knight, & Cerce, 1997) was examined for the effect of age at time of release on sexual recidivism. The present study (1) examined the recidivism rates for each of five age-at-release groups, separately for rapists and child molesters, (2) tested the fit of linear and quadratic models for 5, 10, 15, 20, and 25-year windows using Cox regression analysis, (3) presented the predicted failure rates for rapists (up to five years post-release) and child molesters (out 21 years post-release), and (4) provided a computational formula for estimating the sexual recidivism rate given an individual's age and number of years post-release. For rapists, a linear model extending 5 years best captured our data (LR=5.62, p<.02). Going out any further than 5 years did not enhance the predictive efficacy of the model. By contrast, a quadratic model extending the full duration of the study (25 years) provided the best fit (LR=6.30, p<.04) for child molesters. Our data supported the general conclusion that risk of sexual recidivism diminishes as a function of increasing age at time of release for rapists. We found marked differences, however, in the re-offense patterns of rapists and child molesters, with the latter group evidencing a distinct quadratic, rather than linear, pattern. Since these findings derive from a population screened for civil commitment by virtue of their presumptive dangerousness, they may not be generalizable to samples of sex offenders drawn from the general prison population.