564390
research-article2014
SAXXXX10.1177/1079063214564390Sex AbuseTurner et al.
Article
Risk Assessment in Child Sexual Abusers Working With Children
Sexual Abuse: A Journal of Research and Treatment 1–25 © The Author(s) 2014 Reprints and permissions: sagepub.com/journalsPermissions.nav DOI: 10.1177/1079063214564390 sax.sagepub.com
Daniel Turner1, Martin Rettenberger1,2,3, Dahlnym Yoon1, Verena Klein1, Reinhard Eher3,4, and Peer Briken1
Abstract Child sexual abuse occurring in a child- or youth-serving institution or organization has attracted great public and scientific attention. In light of the particular personal and offense-related characteristics of men who have abused children within such an institution or organization, it is of special importance to evaluate the predictive performance of currently applied risk assessment instruments in this offender population. Therefore, the present study assessed the risk ratings and predictive performance of four risk assessment instruments and one instrument assessing protective factors concerning any, violent and sexual recidivism in child sexual abusers working with children (CSA-W) in comparison with extra-familial child sexual abusers (CSA-E) and intra-familial child sexual abusers (CSA-I). The results indicate that CSA-W mostly recidivate with a sexual offense. Although all included risk measures seem to function with CSA-W, the Static-99 seems to be the instrument that performs best in predicting sexual recidivism in CSA-W. CSA-W had the most protective factors measured with the Structured Assessment of PROtective Factors (SAPROF). While the SAPROF could not predict desistance from recidivism in CSA-W, it predicted desistance from any recidivism in all CSA. As CSA-W frequently hold many indicators for pedophilic sexual interests but only a few for antisocial tendencies, it can be suggested that CSA-W are at an increased risk for sexual
1Institute
for Sex Research and Forensic Psychiatry, University Medical Center Hamburg–Eppendorf, Germany 2Johannes Gutenberg-University, Mainz, Germany 3Federal Evaluation Center for Violent and Sexual Offenders, Vienna, Austria 4University of Ulm, Ulm, Germany Corresponding Author: Peer Briken, Institute for Sex Research and Forensic Psychiatry, University Medical Center HamburgEppendorf, Martinistrasse 52, 20246 Hamburg, Germany. Email: [email protected]
Downloaded from sax.sagepub.com at UNIV OF FLORIDA Smathers Libraries on November 15, 2015
2
Sexual Abuse
recidivism and thus risk measures especially designed for sexual recidivism work best in CSA-W. Nevertheless, CSA-W also hold many protective factors; however, their impact on CSA-W is not clear yet and needs further study. Keywords child sexual abusers, recidivism, risk factors, protective factors, risk assessment Previous research has shown that some men use their child- or youth-related professional or voluntary employment to create opportunities to abuse children with whom they work (child sexual abusers working with children [CSA-W]; Sullivan & Beech, 2004; Sullivan, Beech, Craig, & Gannon, 2011). Such offenses have repeatedly attracted great public and scientific attention (e.g., Boyle, 2014; Gallagher, 2000; Waterhouse, 2000). In this context, Gallagher (2000) reported a yearly incidence of 185 cases of institutional child sexual abuse in England and Wales between 1988 and 1992, accounting for 3% of all child sexual abuse cases. Community-based institutions, such as schools, voluntary clubs, and churches were among the institutions affected most frequently, and in about half of the cases two or more victims had been abused with a maximum of 30 victims in one incident (Gallagher, 2000). Concerning Catholic priests, the John Jay College study group revealed that 4% of all Catholic priests in the United States were accused of a sexual abuse offense between 1950 and 2002 (John Jay College of Criminal Justice, 2004). Furthermore, about half of the priests were accused of more than one sexual abuse offense and 3.5% had more than 10 accusations. In a representative sample of U.S. school students in 8th to 11th grade (n = 2,065), 9.6% indicated unwanted educator sexual misconduct and 6.7% reported about incidents involving physical sexual contact (Shakeshaft, 2004). Based on these data, Shakeshaft (2004) concluded that more than 4.5 million students in the United States are sexually victimized by an employee between kindergarten and 12th grade. In more recent studies, prevalence rates are obtained by interviewing the children’s caregiver. In this context, it was found that 5 per 1,000 children who were living in residential care facilities were sexually victimized in the Netherlands during the year 2010 (Euser, Alink, Tharner, van Ijzendoorn, & Bakermans-Kranenburg, 2013). Moreover, children living in residential care had a nine time greater risk of being sexually victimized than children from the general population (Euser et al., 2013). In light of these findings, it seems obvious that child sexual abuse in institutions is an internationally relevant problem and can occur in almost all youth-serving institutions. Although such incidents seem to account only for a small proportion of all child sexual abuse cases, their significance should not be underrated. Most cases stay unreported and detected cases frequently involve a large number of victims (Gallagher, 2000). Therefore, it seems to be of particular importance for policymakers and clinicians to evaluate the personal characteristics of CSA-W and to take these specific characteristics into consideration when evaluating and validating assessment and screening methods that can be used to identify men at risk for sexual offending as early as possible.
Downloaded from sax.sagepub.com at UNIV OF FLORIDA Smathers Libraries on November 15, 2015
3
Turner et al.
Characteristics of CSA-W and Other Child Sexual Abuser Subgroups Although CSA-W are by definition a particular subgroup of extra-familial child sexual abusers (CSA-E), there exists scientific evidence that CSA-W are distinguishable from CSA-E who do not work with children and also from intra-familial child sexual abusers (CSA-I; McAlinden, 2006; Moulden, Firestone, & Wexler, 2007; Sullivan & Beech, 2002; Sullivan et al., 2011). In a previous study (also using parts of the sample assessed within the current study), we found that CSA-W tend to have a higher number of victims, more often boy victims, a higher score on the Screening Scale for Pedophilic Interests (SSPI; Seto & Lalumière, 2001), a higher prevalence of a pedophilia diagnosis, and a higher prevalence of exclusive pedophilic sexual interests compared with CSA-E and CSA-I (Turner, Rettenberger, Lohmann, Eher, & Briken, 2014; cf. Firestone, Moulden, & Wexler, 2009; Loftus & Camargo, 1993, for comparable findings in clergy CSA). Moreover, CSA-W compared with CSA-E and CSA-I had a lower prevalence of an antisocial personality disorder diagnosis, less often previous problems with alcohol or illegal drugs, and less often previous convictions with any or a violent offense (Turner et al., 2014; cf. Haywood, Kravitz, Grossman, Wasyliw, & Hardy, 1996; Loftus & Camargo, 1993; Moulden et al., 2007, for comparable findings in clergy and child care providers). Based on our results and the findings of other research groups, we argued that CSA-W often present indicators for pedophilic sexual interests, but only seldom present general antisocial behaviors (Turner et al., 2014; Sullivan et al., 2011).
Recidivism in Sexual Offenders, CSA, and Relevant CSA Subgroups With average follow-up periods ranging from 4 to 6 years, different meta-analyses found sexual recidivism rates of convicted CSA between 13% and 14% and of about 36% concerning any recidivism (Hanson & Bussière, 1998; Hanson & MortonBourgon, 2005). However, recidivism rates seem to vary between different CSA subgroups. CSA-I are consistently described to have lower rates for sexual reoffending ranging from 5% to 11%, whereas CSA-E are reported to have sexual recidivism rates between 9% and 20% (Bartosh, Garby, Lewis, & Gray, 2003; Firestone et al., 2000; Firestone et al., 1999; Greenberg, Bradford, Firestone, & Curry, 2000). To our knowledge, only one previous study exists that evaluated recidivism rates in CSA-W. In a sample of 337 Catholic priests and friars (a CSA-W subpopulation), a sexual recidivism rate of 6.2% was reported (Montana et al., 2012). The mean follow-up period was 16.05 years (SD = 5.12 years), and follow-up assessment started after completion of a 6-month in-patient psychotherapeutic treatment program, which consisted of a combination of cognitive-behavioral, psychoeducational, and psychodynamic approaches (Montana et al., 2012). However, in the Montana et al. (2012) study, recidivism was defined quite broadly as any self-reported sexual contacts with minors or the use of child pornography, while most other studies defined new convictions as recidivism.
Downloaded from sax.sagepub.com at UNIV OF FLORIDA Smathers Libraries on November 15, 2015
4
Sexual Abuse
Risk Assessment in Sexual Offenders, CSA, and Relevant CSA Subgroups Various researchers have claimed that the varying absolute recidivism rates reported for different groups of sexual offenders might influence the predictive performance of risk assessment instruments (Bartosh et al., 2003; Hanson & Bussière, 1998; Langton et al., 2007; Rettenberger & Eher, 2007; Rettenberger, Matthes, Boer, & Eher, 2010). Furthermore, the differing psychological and criminological characteristics of different sexual offender subgroups and the capability of different risk assessment instruments to capture the unique risk-relevant characteristics of each subgroup might also influence the predictive performance. Within the present study, the following risk assessment instruments were used: Static-99 (Hanson & Thornton, 2000), Sexual Offender Risk Appraisal Guide (SORAG; Quinsey, Harris, Rice, & Cormier, 2006), Sexual Violence Risk-20 (SVR20; Boer, Hart, Kropp, & Webster, 1997), and Psychopathy Checklist–Revised (PCLR; Hare, 2003). Only a limited number of studies have assessed predictive performance of these risk assessment instruments in CSA-only samples so far. Moreover, due to differences in the study populations, study methods (e.g., duration of follow-up), or divergent conceptualizations of recidivism, these studies have reported differing predictive validity values. For the Static-99, area under the curve (AUC) values between 0.55 and 0.92 have been found concerning the prediction of sexual recidivism, AUCs between 0.58 and 0.75 for the prediction of violent recidivism, and AUCs between 0.62 and 0.74 for any or general recidivism (Allan, Grace, Rutherford, & Hudson, 2007; Beech, Friendship, Erikson, & Hanson, 2002; Beggs & Grace, 2008; Ducro & Pham, 2006; Hanson & Thornton, 2000; G. T. Harris et al., 2003; Helmus & Hanson, 2007; Looman & Abracen, 2010; Parent, Guay, & Knight, 2010; Sjöstedt & Långström, 2001; Thornton, 2002). Concerning the SORAG, AUC values between 0.65 and 0.70 were found for the prediction of sexual recidivism, AUCs between 0.68 and 0.72 for violent recidivism, and AUCs between 0.66 and 0.67 for any or general recidivism (Ducro & Pham, 2006; G. T. Harris et al., 2003; Parent et al., 2010). The predictive performance found for the SVR-20 was between 0.68 and 0.77 for sexual recidivism, between 0.57 and 0.75 for violent recidivism, and between 0.62 and 0.75 for any or general recidivism (Parent et al., 2010; Rettenberger, Boer, & Eher, 2011; note: Rettenberger et al., 2011, also used parts of the present sample). Although not a risk assessment instrument per se, the PCL-R was also used to predict recidivism in CSA revealing AUCs between 0.50 and 0.82 for sexual recidivism, AUCs between 0.67 and 0.73 for violent recidivism, and AUCs between 0.68 and 0.76 for any or general recidivism (Beggs & Grace, 2008; Eher & Rettenberger, 2009; Olver & Wong, 2006; Parent et al., 2010; note: Eher & Rettenberger, 2009, also used parts of the present sample).1 These findings imply that most studies found moderate to large AUC values according to a commonly used classification (AUC ≤ 0.63 = small, AUC between 0.64 and
Downloaded from sax.sagepub.com at UNIV OF FLORIDA Smathers Libraries on November 15, 2015
5
Turner et al.
0.71 = moderate, and AUC ≥ .72 = large effect size; Cohen, 1992; Dahle, Schneider, & Ziethen, 2007); however, there still exists considerable variation in some instruments based on the offender’s risk and the recidivism category. Studies assessing the predictive performance of risk assessment instruments in CSA-W, as we defined this group above, are completely missing so far. Montana et al. (2012) assessed the predictive validity of the Static-99 for sexual recidivism in clergy CSA, a distinct CSA-W subgroup, reporting an AUC of 0.67.
Protective Factors in Sexual Offenders It was suggested that CSA-W hold more prosocial attitudes and show more prosocial behavior compared with other CSA groups (Langevin, Curnoe, & Bain, 2000; Moulden, Firestone, Kingston, & Wexler, 2010), factors that might protect individuals from recidivating. Therefore, it could be suggested that the risk-reducing effects of protective factors might be of particular importance for the risk management in this special offender population. Regarding the assessment of personal and environmental resources, the Structured Assessment of PROtective Factors (SAPROF) for violence risk (de Vogel, de Ruiter, Bouman, & de Vries Robbé, 2007, 2009) was developed as an instrument focusing on protective factors in violent offenders and following the structured professional judgment (SPJ) approach (de Vogel, de Vries Robbé, de Ruiter, & Bouman, 2011; Douglas & Reeves, 2010). Within the SPJ approach, a final risk judgment is not derived from summing up the item scores of the instrument but is based on the personal judgment of a clinician or another evaluator. The evaluator’s decision, however, is usually based on the item scores. Concerning the SAPROF, one has to bear in mind though that it was not primarily developed as a risk assessment instrument but rather as a tool to support treatment planning and evaluation processes. In an assessment of 30 sexual offenders on probation or with parole supervision, the SAPROF correlated negatively with the SVR-20, while no correlation with the Static-99 could be found (Yoon, Spehr, & Briken, 2011). Similarly, in juvenile sexual offenders, the SAPROF showed a negative correlation with risk factors and aggressive or antisocial behaviors suggesting its usefulness for sexual offender evaluations as well (Klein et al., 2012). First results indicate that the SAPROF can also predict desistance from recidivism with a sexual or violent offense in adult sexual offender samples; however, it has to be noted that of the assessed 83 sexual offenders only 25% could be classified as CSA, while all others had adult victims only (de Vries Robbé, de Vogel, Koster, & Bogaerts, 2014). It was found that the SAPROF total score could significantly predict desistance from contact sexual recidivism 3 years after release from a forensic psychiatric hospital (AUC = 0.76) and also after 15 years (AUC = 0.71). Furthermore, the SAPROF total score could significantly predict desistance from violent recidivism 3 years post release (AUC = 0.77) and also 15 years post release (AUC = 0.74; de Vries Robbé et al., 2014). Studies evaluating the predictive validity of the SAPROF in CSA only and also in different CSA subgroups are completely missing so far.
Downloaded from sax.sagepub.com at UNIV OF FLORIDA Smathers Libraries on November 15, 2015
6
Sexual Abuse
Aims Many CSA-W seem to hold strong pedophilic sexual interests, while only few appear to be characterized by antisocial personality traits (Turner et al., 2014). Deviant sexual interests as well as antisocial characteristics were described as those risk factors with the highest predictive power concerning recidivism (Hanson & Morton-Bourgon, 2005). This is also reflected within the currently applied risk assessment instruments as all of them consist of at least some items that assess these constructs. However, some instruments have a focus on sexual deviance, whereas others on antisocial behaviors. In light of the CSA-W’s specific characteristics and offending behavior, it seems to be of particular importance to evaluate and compare the predictive performance of frequently used risk assessment instruments to identify the instruments most suitable for CSA-W. Thus, the aims of the present study were threefold: (a) assessment of the rates of any, violent, and sexual recidivism of CSA-W in comparison with CSA-I and CSA-E; (b) assessment and comparison of the mean risk ratings of four risk assessment instruments (Static-99, SVR-20, SORAG, PCL-R) and one instrument assessing protective factors (SAPROF) for CSA-W, CSA-I, and CSA-E; and (c) assessment and comparison of the predictive performance of the five instruments among CSA-W, CSA-I, and CSA-E.
Method Participants The present sample is part of a larger sample of a still ongoing research project and included 277 convicted and incacerated sexual offenders who were reported to the Federal Evaluation Centre for Violent and Sexual Offenders (FECVSO; Eher, Matthes, Schilling, Haubner-Maclean, & Rettenberger, 2012) between the years 2000 and 2008 because of a child sexual abuse offense. Parts of the sample were also investigated in other studies (e.g., Eher, Matthes et al., 2012; Rettenberger et al., 2010; Turner et al., 2014). The FECVSO is a department of the Austrian Ministry of Justice and collects and evaluates data about every convicted and incacerated sexual offender in Austria to make correctional plans (Rettenberger et al., 2010). All offenders had been investigated psychiatrically and psychologically during a 2-week assessment period. The current sample consisted of 71 child sexual abusers (25.6%) who had abused children exclusively outside of the family (CSA-E), but not in their professional or voluntary employment, 136 child sexual abusers (49.1%) who had abused children exclusively within their families (CSA-I), and 39 child sexual abusers (14.1%) who had abused children with whom they had contact during their professional or voluntary employment at a youth-serving institution (CSA-W). The three groups were mutually exclusive and group allocation was based on the index offense. Seventeen (6.1%) CSA had extra- and intra-familial victims in the index offense and 14 (5.1%) CSA were convicted for non-contact offenses only. To ensure homogenous
Downloaded from sax.sagepub.com at UNIV OF FLORIDA Smathers Libraries on November 15, 2015
7
Turner et al.
subgroups, these two groups were excluded from all further analyses. Due to missing data in 17 cases, all analyses were conducted with a final sample size of 229 CSA consisting of 126 CSA-I (55.0%), 66 CSA-E (28.8%), and 37 CSA-W (16.2%). One should take into consideration that all CSA-W abused their victims outside of their own family as well and are thus a particular subgroup within CSA-E. To be included in the CSA-W group and not in the CSA-E group, however, the offender had to be employed at an institution or an organization that is involved with the care of children, irrespective of whether this position was carried out as a paid profession or on a voluntary basis. Furthermore, all CSA-W established contact with their victims through the specific youth-serving institution they were employed at, meaning that offenders as well as victims were in all cases involved in the same institution or organization. Table 1 provides an overview of sociodemographic, offense-related, and psychiatric variables relevant for the sample. Of the included CSA-W, 19 had paid employment, 15 had a voluntary position, and 3 had multiple paid and voluntary positions. The offenders were employed as teachers or a related position at a school (n = 10), as priests or related positions within the church (n = 7), as supervisors in children’s homes (n = 3), or as a supervisor in a daycare center (n = 1). Among the voluntary positions were caretakers in youth organization (e.g., boy scouts, youth fire department; n = 9), coaches at a youth sports club (n = 6), caretakers in school excursions (n = 5), and foster-parents in a registered children’s foster care organization (n = 2).
Data Collection The SVR-20, Static-99, SORAG, and PCL-R were part of the initial diagnostic and risk assessment process and were thus rated prospectively during the offenders’ stay at the FECVSO. The information obtained from the initial diagnostic process at the FECVSO is subsequently summarized in forensic-sexological reports. The reports do not only provide recommendations for future strategies to reduce the offenders’ risk of recidivism but also include detailed information about the offender’s personal history, his previous employments and his social environment, about previous offenses and a detailed description and analysis of the index offense. The reports also contain information about the offender’s personality, about his biographical and sexual development, a comprehensive risk assessment section, and information about psychiatric as well as sexual disorders. The scores of the single items and the total scores concerning the Static-99, SORAG, SVR-20, and the PCL-R were retrieved from the FECVSO database. Because the SAPROF is not part of the initial diagnostic and risk assessment process at the FECVSO, it had to be rated retrospectively. The ratings were performed by three of the authors (D.T., D.Y., V.K.) based on the forensic-sexological reports. All reports contained sufficient information for coding the SAPROF. All three raters are trained for using the SAPROF and, moreover, one rater is an official SAPROF trainer. The raters were blind concerning recidivism of the participants.
Downloaded from sax.sagepub.com at UNIV OF FLORIDA Smathers Libraries on November 15, 2015
8
Sexual Abuse
Table 1. Sociodemographic, Offense-Related, and Psychiatric Characteristics of the Sample Divided by Subgroup. CSA-W (n = 37)
CSA-E (n = 66)
CSA-I (n = 126)
(SD)
M
(SD)
M
(SD)
F
46.59
(13.21)
41.83
(13.39)
41.85
(11.17)
2.353
.097
34.84a,b
(25.60)
28.29a
(18.78)
36.95b
(21.57)
3.535
.031
48.75
(13.22)
43.73
(13.66)
44.63
(11.24)
2.127
.122
n
(%)
n
n
(%)
χ2
p
Any previous conviction Previous conviction for violent offense Previous conviction for sexual offense Under influence of alcohol or drugs during index offense Multiple offenders during index offense
7a 2a
(18.9) (5.4)
38b 25b
(57.6) (37.9)
46c 30c
(36.5) (23.8)
16.015 13.548
research-article2014
SAXXXX10.1177/1079063214564390Sex AbuseTurner et al.
Article
Risk Assessment in Child Sexual Abusers Working With Children
Sexual Abuse: A Journal of Research and Treatment 1–25 © The Author(s) 2014 Reprints and permissions: sagepub.com/journalsPermissions.nav DOI: 10.1177/1079063214564390 sax.sagepub.com
Daniel Turner1, Martin Rettenberger1,2,3, Dahlnym Yoon1, Verena Klein1, Reinhard Eher3,4, and Peer Briken1
Abstract Child sexual abuse occurring in a child- or youth-serving institution or organization has attracted great public and scientific attention. In light of the particular personal and offense-related characteristics of men who have abused children within such an institution or organization, it is of special importance to evaluate the predictive performance of currently applied risk assessment instruments in this offender population. Therefore, the present study assessed the risk ratings and predictive performance of four risk assessment instruments and one instrument assessing protective factors concerning any, violent and sexual recidivism in child sexual abusers working with children (CSA-W) in comparison with extra-familial child sexual abusers (CSA-E) and intra-familial child sexual abusers (CSA-I). The results indicate that CSA-W mostly recidivate with a sexual offense. Although all included risk measures seem to function with CSA-W, the Static-99 seems to be the instrument that performs best in predicting sexual recidivism in CSA-W. CSA-W had the most protective factors measured with the Structured Assessment of PROtective Factors (SAPROF). While the SAPROF could not predict desistance from recidivism in CSA-W, it predicted desistance from any recidivism in all CSA. As CSA-W frequently hold many indicators for pedophilic sexual interests but only a few for antisocial tendencies, it can be suggested that CSA-W are at an increased risk for sexual
1Institute
for Sex Research and Forensic Psychiatry, University Medical Center Hamburg–Eppendorf, Germany 2Johannes Gutenberg-University, Mainz, Germany 3Federal Evaluation Center for Violent and Sexual Offenders, Vienna, Austria 4University of Ulm, Ulm, Germany Corresponding Author: Peer Briken, Institute for Sex Research and Forensic Psychiatry, University Medical Center HamburgEppendorf, Martinistrasse 52, 20246 Hamburg, Germany. Email: [email protected]
Downloaded from sax.sagepub.com at UNIV OF FLORIDA Smathers Libraries on November 15, 2015
2
Sexual Abuse
recidivism and thus risk measures especially designed for sexual recidivism work best in CSA-W. Nevertheless, CSA-W also hold many protective factors; however, their impact on CSA-W is not clear yet and needs further study. Keywords child sexual abusers, recidivism, risk factors, protective factors, risk assessment Previous research has shown that some men use their child- or youth-related professional or voluntary employment to create opportunities to abuse children with whom they work (child sexual abusers working with children [CSA-W]; Sullivan & Beech, 2004; Sullivan, Beech, Craig, & Gannon, 2011). Such offenses have repeatedly attracted great public and scientific attention (e.g., Boyle, 2014; Gallagher, 2000; Waterhouse, 2000). In this context, Gallagher (2000) reported a yearly incidence of 185 cases of institutional child sexual abuse in England and Wales between 1988 and 1992, accounting for 3% of all child sexual abuse cases. Community-based institutions, such as schools, voluntary clubs, and churches were among the institutions affected most frequently, and in about half of the cases two or more victims had been abused with a maximum of 30 victims in one incident (Gallagher, 2000). Concerning Catholic priests, the John Jay College study group revealed that 4% of all Catholic priests in the United States were accused of a sexual abuse offense between 1950 and 2002 (John Jay College of Criminal Justice, 2004). Furthermore, about half of the priests were accused of more than one sexual abuse offense and 3.5% had more than 10 accusations. In a representative sample of U.S. school students in 8th to 11th grade (n = 2,065), 9.6% indicated unwanted educator sexual misconduct and 6.7% reported about incidents involving physical sexual contact (Shakeshaft, 2004). Based on these data, Shakeshaft (2004) concluded that more than 4.5 million students in the United States are sexually victimized by an employee between kindergarten and 12th grade. In more recent studies, prevalence rates are obtained by interviewing the children’s caregiver. In this context, it was found that 5 per 1,000 children who were living in residential care facilities were sexually victimized in the Netherlands during the year 2010 (Euser, Alink, Tharner, van Ijzendoorn, & Bakermans-Kranenburg, 2013). Moreover, children living in residential care had a nine time greater risk of being sexually victimized than children from the general population (Euser et al., 2013). In light of these findings, it seems obvious that child sexual abuse in institutions is an internationally relevant problem and can occur in almost all youth-serving institutions. Although such incidents seem to account only for a small proportion of all child sexual abuse cases, their significance should not be underrated. Most cases stay unreported and detected cases frequently involve a large number of victims (Gallagher, 2000). Therefore, it seems to be of particular importance for policymakers and clinicians to evaluate the personal characteristics of CSA-W and to take these specific characteristics into consideration when evaluating and validating assessment and screening methods that can be used to identify men at risk for sexual offending as early as possible.
Downloaded from sax.sagepub.com at UNIV OF FLORIDA Smathers Libraries on November 15, 2015
3
Turner et al.
Characteristics of CSA-W and Other Child Sexual Abuser Subgroups Although CSA-W are by definition a particular subgroup of extra-familial child sexual abusers (CSA-E), there exists scientific evidence that CSA-W are distinguishable from CSA-E who do not work with children and also from intra-familial child sexual abusers (CSA-I; McAlinden, 2006; Moulden, Firestone, & Wexler, 2007; Sullivan & Beech, 2002; Sullivan et al., 2011). In a previous study (also using parts of the sample assessed within the current study), we found that CSA-W tend to have a higher number of victims, more often boy victims, a higher score on the Screening Scale for Pedophilic Interests (SSPI; Seto & Lalumière, 2001), a higher prevalence of a pedophilia diagnosis, and a higher prevalence of exclusive pedophilic sexual interests compared with CSA-E and CSA-I (Turner, Rettenberger, Lohmann, Eher, & Briken, 2014; cf. Firestone, Moulden, & Wexler, 2009; Loftus & Camargo, 1993, for comparable findings in clergy CSA). Moreover, CSA-W compared with CSA-E and CSA-I had a lower prevalence of an antisocial personality disorder diagnosis, less often previous problems with alcohol or illegal drugs, and less often previous convictions with any or a violent offense (Turner et al., 2014; cf. Haywood, Kravitz, Grossman, Wasyliw, & Hardy, 1996; Loftus & Camargo, 1993; Moulden et al., 2007, for comparable findings in clergy and child care providers). Based on our results and the findings of other research groups, we argued that CSA-W often present indicators for pedophilic sexual interests, but only seldom present general antisocial behaviors (Turner et al., 2014; Sullivan et al., 2011).
Recidivism in Sexual Offenders, CSA, and Relevant CSA Subgroups With average follow-up periods ranging from 4 to 6 years, different meta-analyses found sexual recidivism rates of convicted CSA between 13% and 14% and of about 36% concerning any recidivism (Hanson & Bussière, 1998; Hanson & MortonBourgon, 2005). However, recidivism rates seem to vary between different CSA subgroups. CSA-I are consistently described to have lower rates for sexual reoffending ranging from 5% to 11%, whereas CSA-E are reported to have sexual recidivism rates between 9% and 20% (Bartosh, Garby, Lewis, & Gray, 2003; Firestone et al., 2000; Firestone et al., 1999; Greenberg, Bradford, Firestone, & Curry, 2000). To our knowledge, only one previous study exists that evaluated recidivism rates in CSA-W. In a sample of 337 Catholic priests and friars (a CSA-W subpopulation), a sexual recidivism rate of 6.2% was reported (Montana et al., 2012). The mean follow-up period was 16.05 years (SD = 5.12 years), and follow-up assessment started after completion of a 6-month in-patient psychotherapeutic treatment program, which consisted of a combination of cognitive-behavioral, psychoeducational, and psychodynamic approaches (Montana et al., 2012). However, in the Montana et al. (2012) study, recidivism was defined quite broadly as any self-reported sexual contacts with minors or the use of child pornography, while most other studies defined new convictions as recidivism.
Downloaded from sax.sagepub.com at UNIV OF FLORIDA Smathers Libraries on November 15, 2015
4
Sexual Abuse
Risk Assessment in Sexual Offenders, CSA, and Relevant CSA Subgroups Various researchers have claimed that the varying absolute recidivism rates reported for different groups of sexual offenders might influence the predictive performance of risk assessment instruments (Bartosh et al., 2003; Hanson & Bussière, 1998; Langton et al., 2007; Rettenberger & Eher, 2007; Rettenberger, Matthes, Boer, & Eher, 2010). Furthermore, the differing psychological and criminological characteristics of different sexual offender subgroups and the capability of different risk assessment instruments to capture the unique risk-relevant characteristics of each subgroup might also influence the predictive performance. Within the present study, the following risk assessment instruments were used: Static-99 (Hanson & Thornton, 2000), Sexual Offender Risk Appraisal Guide (SORAG; Quinsey, Harris, Rice, & Cormier, 2006), Sexual Violence Risk-20 (SVR20; Boer, Hart, Kropp, & Webster, 1997), and Psychopathy Checklist–Revised (PCLR; Hare, 2003). Only a limited number of studies have assessed predictive performance of these risk assessment instruments in CSA-only samples so far. Moreover, due to differences in the study populations, study methods (e.g., duration of follow-up), or divergent conceptualizations of recidivism, these studies have reported differing predictive validity values. For the Static-99, area under the curve (AUC) values between 0.55 and 0.92 have been found concerning the prediction of sexual recidivism, AUCs between 0.58 and 0.75 for the prediction of violent recidivism, and AUCs between 0.62 and 0.74 for any or general recidivism (Allan, Grace, Rutherford, & Hudson, 2007; Beech, Friendship, Erikson, & Hanson, 2002; Beggs & Grace, 2008; Ducro & Pham, 2006; Hanson & Thornton, 2000; G. T. Harris et al., 2003; Helmus & Hanson, 2007; Looman & Abracen, 2010; Parent, Guay, & Knight, 2010; Sjöstedt & Långström, 2001; Thornton, 2002). Concerning the SORAG, AUC values between 0.65 and 0.70 were found for the prediction of sexual recidivism, AUCs between 0.68 and 0.72 for violent recidivism, and AUCs between 0.66 and 0.67 for any or general recidivism (Ducro & Pham, 2006; G. T. Harris et al., 2003; Parent et al., 2010). The predictive performance found for the SVR-20 was between 0.68 and 0.77 for sexual recidivism, between 0.57 and 0.75 for violent recidivism, and between 0.62 and 0.75 for any or general recidivism (Parent et al., 2010; Rettenberger, Boer, & Eher, 2011; note: Rettenberger et al., 2011, also used parts of the present sample). Although not a risk assessment instrument per se, the PCL-R was also used to predict recidivism in CSA revealing AUCs between 0.50 and 0.82 for sexual recidivism, AUCs between 0.67 and 0.73 for violent recidivism, and AUCs between 0.68 and 0.76 for any or general recidivism (Beggs & Grace, 2008; Eher & Rettenberger, 2009; Olver & Wong, 2006; Parent et al., 2010; note: Eher & Rettenberger, 2009, also used parts of the present sample).1 These findings imply that most studies found moderate to large AUC values according to a commonly used classification (AUC ≤ 0.63 = small, AUC between 0.64 and
Downloaded from sax.sagepub.com at UNIV OF FLORIDA Smathers Libraries on November 15, 2015
5
Turner et al.
0.71 = moderate, and AUC ≥ .72 = large effect size; Cohen, 1992; Dahle, Schneider, & Ziethen, 2007); however, there still exists considerable variation in some instruments based on the offender’s risk and the recidivism category. Studies assessing the predictive performance of risk assessment instruments in CSA-W, as we defined this group above, are completely missing so far. Montana et al. (2012) assessed the predictive validity of the Static-99 for sexual recidivism in clergy CSA, a distinct CSA-W subgroup, reporting an AUC of 0.67.
Protective Factors in Sexual Offenders It was suggested that CSA-W hold more prosocial attitudes and show more prosocial behavior compared with other CSA groups (Langevin, Curnoe, & Bain, 2000; Moulden, Firestone, Kingston, & Wexler, 2010), factors that might protect individuals from recidivating. Therefore, it could be suggested that the risk-reducing effects of protective factors might be of particular importance for the risk management in this special offender population. Regarding the assessment of personal and environmental resources, the Structured Assessment of PROtective Factors (SAPROF) for violence risk (de Vogel, de Ruiter, Bouman, & de Vries Robbé, 2007, 2009) was developed as an instrument focusing on protective factors in violent offenders and following the structured professional judgment (SPJ) approach (de Vogel, de Vries Robbé, de Ruiter, & Bouman, 2011; Douglas & Reeves, 2010). Within the SPJ approach, a final risk judgment is not derived from summing up the item scores of the instrument but is based on the personal judgment of a clinician or another evaluator. The evaluator’s decision, however, is usually based on the item scores. Concerning the SAPROF, one has to bear in mind though that it was not primarily developed as a risk assessment instrument but rather as a tool to support treatment planning and evaluation processes. In an assessment of 30 sexual offenders on probation or with parole supervision, the SAPROF correlated negatively with the SVR-20, while no correlation with the Static-99 could be found (Yoon, Spehr, & Briken, 2011). Similarly, in juvenile sexual offenders, the SAPROF showed a negative correlation with risk factors and aggressive or antisocial behaviors suggesting its usefulness for sexual offender evaluations as well (Klein et al., 2012). First results indicate that the SAPROF can also predict desistance from recidivism with a sexual or violent offense in adult sexual offender samples; however, it has to be noted that of the assessed 83 sexual offenders only 25% could be classified as CSA, while all others had adult victims only (de Vries Robbé, de Vogel, Koster, & Bogaerts, 2014). It was found that the SAPROF total score could significantly predict desistance from contact sexual recidivism 3 years after release from a forensic psychiatric hospital (AUC = 0.76) and also after 15 years (AUC = 0.71). Furthermore, the SAPROF total score could significantly predict desistance from violent recidivism 3 years post release (AUC = 0.77) and also 15 years post release (AUC = 0.74; de Vries Robbé et al., 2014). Studies evaluating the predictive validity of the SAPROF in CSA only and also in different CSA subgroups are completely missing so far.
Downloaded from sax.sagepub.com at UNIV OF FLORIDA Smathers Libraries on November 15, 2015
6
Sexual Abuse
Aims Many CSA-W seem to hold strong pedophilic sexual interests, while only few appear to be characterized by antisocial personality traits (Turner et al., 2014). Deviant sexual interests as well as antisocial characteristics were described as those risk factors with the highest predictive power concerning recidivism (Hanson & Morton-Bourgon, 2005). This is also reflected within the currently applied risk assessment instruments as all of them consist of at least some items that assess these constructs. However, some instruments have a focus on sexual deviance, whereas others on antisocial behaviors. In light of the CSA-W’s specific characteristics and offending behavior, it seems to be of particular importance to evaluate and compare the predictive performance of frequently used risk assessment instruments to identify the instruments most suitable for CSA-W. Thus, the aims of the present study were threefold: (a) assessment of the rates of any, violent, and sexual recidivism of CSA-W in comparison with CSA-I and CSA-E; (b) assessment and comparison of the mean risk ratings of four risk assessment instruments (Static-99, SVR-20, SORAG, PCL-R) and one instrument assessing protective factors (SAPROF) for CSA-W, CSA-I, and CSA-E; and (c) assessment and comparison of the predictive performance of the five instruments among CSA-W, CSA-I, and CSA-E.
Method Participants The present sample is part of a larger sample of a still ongoing research project and included 277 convicted and incacerated sexual offenders who were reported to the Federal Evaluation Centre for Violent and Sexual Offenders (FECVSO; Eher, Matthes, Schilling, Haubner-Maclean, & Rettenberger, 2012) between the years 2000 and 2008 because of a child sexual abuse offense. Parts of the sample were also investigated in other studies (e.g., Eher, Matthes et al., 2012; Rettenberger et al., 2010; Turner et al., 2014). The FECVSO is a department of the Austrian Ministry of Justice and collects and evaluates data about every convicted and incacerated sexual offender in Austria to make correctional plans (Rettenberger et al., 2010). All offenders had been investigated psychiatrically and psychologically during a 2-week assessment period. The current sample consisted of 71 child sexual abusers (25.6%) who had abused children exclusively outside of the family (CSA-E), but not in their professional or voluntary employment, 136 child sexual abusers (49.1%) who had abused children exclusively within their families (CSA-I), and 39 child sexual abusers (14.1%) who had abused children with whom they had contact during their professional or voluntary employment at a youth-serving institution (CSA-W). The three groups were mutually exclusive and group allocation was based on the index offense. Seventeen (6.1%) CSA had extra- and intra-familial victims in the index offense and 14 (5.1%) CSA were convicted for non-contact offenses only. To ensure homogenous
Downloaded from sax.sagepub.com at UNIV OF FLORIDA Smathers Libraries on November 15, 2015
7
Turner et al.
subgroups, these two groups were excluded from all further analyses. Due to missing data in 17 cases, all analyses were conducted with a final sample size of 229 CSA consisting of 126 CSA-I (55.0%), 66 CSA-E (28.8%), and 37 CSA-W (16.2%). One should take into consideration that all CSA-W abused their victims outside of their own family as well and are thus a particular subgroup within CSA-E. To be included in the CSA-W group and not in the CSA-E group, however, the offender had to be employed at an institution or an organization that is involved with the care of children, irrespective of whether this position was carried out as a paid profession or on a voluntary basis. Furthermore, all CSA-W established contact with their victims through the specific youth-serving institution they were employed at, meaning that offenders as well as victims were in all cases involved in the same institution or organization. Table 1 provides an overview of sociodemographic, offense-related, and psychiatric variables relevant for the sample. Of the included CSA-W, 19 had paid employment, 15 had a voluntary position, and 3 had multiple paid and voluntary positions. The offenders were employed as teachers or a related position at a school (n = 10), as priests or related positions within the church (n = 7), as supervisors in children’s homes (n = 3), or as a supervisor in a daycare center (n = 1). Among the voluntary positions were caretakers in youth organization (e.g., boy scouts, youth fire department; n = 9), coaches at a youth sports club (n = 6), caretakers in school excursions (n = 5), and foster-parents in a registered children’s foster care organization (n = 2).
Data Collection The SVR-20, Static-99, SORAG, and PCL-R were part of the initial diagnostic and risk assessment process and were thus rated prospectively during the offenders’ stay at the FECVSO. The information obtained from the initial diagnostic process at the FECVSO is subsequently summarized in forensic-sexological reports. The reports do not only provide recommendations for future strategies to reduce the offenders’ risk of recidivism but also include detailed information about the offender’s personal history, his previous employments and his social environment, about previous offenses and a detailed description and analysis of the index offense. The reports also contain information about the offender’s personality, about his biographical and sexual development, a comprehensive risk assessment section, and information about psychiatric as well as sexual disorders. The scores of the single items and the total scores concerning the Static-99, SORAG, SVR-20, and the PCL-R were retrieved from the FECVSO database. Because the SAPROF is not part of the initial diagnostic and risk assessment process at the FECVSO, it had to be rated retrospectively. The ratings were performed by three of the authors (D.T., D.Y., V.K.) based on the forensic-sexological reports. All reports contained sufficient information for coding the SAPROF. All three raters are trained for using the SAPROF and, moreover, one rater is an official SAPROF trainer. The raters were blind concerning recidivism of the participants.
Downloaded from sax.sagepub.com at UNIV OF FLORIDA Smathers Libraries on November 15, 2015
8
Sexual Abuse
Table 1. Sociodemographic, Offense-Related, and Psychiatric Characteristics of the Sample Divided by Subgroup. CSA-W (n = 37)
CSA-E (n = 66)
CSA-I (n = 126)
(SD)
M
(SD)
M
(SD)
F
46.59
(13.21)
41.83
(13.39)
41.85
(11.17)
2.353
.097
34.84a,b
(25.60)
28.29a
(18.78)
36.95b
(21.57)
3.535
.031
48.75
(13.22)
43.73
(13.66)
44.63
(11.24)
2.127
.122
n
(%)
n
n
(%)
χ2
p
Any previous conviction Previous conviction for violent offense Previous conviction for sexual offense Under influence of alcohol or drugs during index offense Multiple offenders during index offense
7a 2a
(18.9) (5.4)
38b 25b
(57.6) (37.9)
46c 30c
(36.5) (23.8)
16.015 13.548
