Heterogeneity Among Juvenile Sexual Abusers: An Exploration of Latent Classes Emerging from Sexual Crime Characteristics (WITHDRAWN)

Background and Purpose: Juveniles represent approximately 20% of those arrested for sexual crimes in the United States each year and more than a third of those arrested for sexual crimes involving juvenile victims. Although researchers have determined this population to be heterogeneous, subtypes of juvenile sexual abusers (JSAs) in the literature are typically based upon a priori assumptions of what sexual offense characteristics (e.g. victim age) might mean in relationship to other factors (e.g. personality, mental health). Latent class analysis, on the other hand, provides a person-centered approach to classifying JSAs using sexual crime characteristics by letting the data determine how classes emerge. Such classes might be useful in identifying more targeted treatments, which currently vary widely and lack a strong evidence-base, as well as policies, which tend to be one-size-fits-all and have failed to be effective in reducing sexual offenses committed by JSAs.

Methods: Data were collected using a cross-sectional survey design from youths adjudicated for a sexual offense and residing in locked residential facilities in two states (N = 573). Approximately half the sample was White (47%). The mean age of the sample was 16.8 years, and most of the youths were in the 10^th grade. Six commonly researched sexual crime characteristics in the JSA literature were included in the latent class analysis, including items from the Self-Report Sexual Aggression Scale (SERSAS), a seven point perpetration scale used in other JSA research.

Absolute fit and relative fit were successfully estimated using SAS 9.3 with PROC LCA and confirmed with a bootstrap likelihood ratio test (BLRT). Model interpretation was agreed upon by three MSW/Ph.D scholar-clinicians who specialize in the field of JSAs. Classification accuracy was measured using the average posterior class probabilities of best class assignments, with all four classes meeting or exceeding the minimum recommended threshold of .70.

Results: The best fitting latent class model was for a four class solution: Curious/Bad Boundaries (49%), Impulsive/Antisocial (20%), Sexually Focused/Covert (17%), and Early Starter/Opportunist (15%).

Conclusions and Implications: Consistent with the clinical experience of the author, clusters of sexual crime characteristics revealed distinct classes of JSAs that presented promising directions for research, treatment, and policy. While this research was exploratory, authors of future investigations should consider that correlative properties of latent characteristics might be more effective in understanding recidivism risk than the current practice of using additive properties of sexual crime characteristics and delinquency risk factors. Treatment implications include the need to reconsider how JSAs in group treatment (the most common form of service delivery) are assigned to treatment groups. Rather than separating JSAs by victim age, for example, it might make more sense to consider how latent typologies of JSAs help identify shared treatment goals. Finally, policy implications include the need to allocate funds toward targeting the different treatment needs of the majority of JSAs, while continuing policies aimed at mitigating risk-management for a very small proportion of this population.