Clinical Assessment

In view of the heterogeneous nature of juveniles who have sexually offended, comprehensive assessments of individuals are needed to facilitate treatment and intervention strategies. This includes assessment of each juvenile's needs (psychological, social, cognitive, and medical), family relationships, risk factors, and risk management possibilities. To emphasize this point, Dougher (1995) began his chapter describing the process of assessing sex offenders with the subtitle "Comprehensive, In-Depth Assessment Is Prelude to Effective Treatment Planning and Implementation" and pointed out that the literature emphasizes the varied, complex, and multidetermined nature of sex offending. Dougher further emphasized that "[a]ccordingly, any attempt to explain or treat sexually offensive behavior must consider the specific factors pertinent to an individual's offense and the psychological characteristics of the individual offender" (p. 11.2). He added that because many, ". . . if not most, sex offenders tend to lie about their offenses and are unreliable and deceptive in their verbal reports, the value of a thorough assessment cannot be overemphasized" (p. 11.2).

Gathering Multiple Sources of Information

Morenz and Becker (1995) noted that parents or guardians of juveniles should be involved in the assessment and in the treatment process. Informed consent should be obtained from the juvenile and parent or guardian, and they should be clearly informed of the limits of confidentiality (Becker and Hunter, 1997).

Recommended procedures for comprehensive assessment of juveniles who commit sex offenses include clinical interviews with the juveniles and family members, psychological assessment, and, in certain cases (according to some), phallometric assessment (Bonner et al., 1998; Morenz and Becker, 1995). Structured clinical interviews (Morenz and Becker, 1995) and paper-and-pencil psychometric tests and questionnaires (Becker and Kaplan, 1993) also can be useful for assessing pertinent areas that may be related to sex offending, such as attitudes and values, social skills, psychological functioning, and sexual knowledge.

Becker and Hunter (1997) also noted that evaluators should review victim statements, juvenile court records, mental health reports, and school records as part of their assessments. Kraemer, Spielman, and Salisbury (1995) suggested that assessments should address the juvenile's beliefs regarding the sex-offending behaviors; issues of aggression, impulsivity, withdrawal, and depression; attitudes toward treatment; potential barriers to treatment; and approaches most likely to be effective. They also noted that objective measures to assess the prognosis for treatment outcomes are useful, citing as an example personality tests, which can help to identify individuals who are unlikely to succeed in treatment.

Using Psychological Tests

Psychological testing of sex offenders has a long history. Although not all of that history is positive, psychological testing is an important part of a comprehensive assessment (Dougher, 1995). In the past, testing was primarily used for identifying personality characteristics and psychological profiles of offenders; due to the heterogeneity of sex offenders, such attempts were not very successful. As Dougher pointed out, "Nevertheless, psychological tests can be useful in combination with other assessment procedures to create a clinical picture of an offender and to identify target areas for clinical interventions" (p. 11.7). In fact, psychological tests have been described as adding a "critical dimension" to comprehensive evaluations of juveniles who have sexually offended (Kraemer, Spielman, and Salisbury, 1995). What they add is "a norm-based reference that can assist in determining placement in an appropriate treatment modality, developing a viable treatment plan, and assessing treatment progress" (p. 11.2).

Bourke and Donohue (1996), in their article "Assessment and Treatment of Juvenile Sex Offenders: An Empirical Review," also observed that studies consistently reveal juvenile sex offenders to be a heterogeneous population. For example, they cited research findings that juvenile sex offending coexists with diagnoses of conduct disorders, attention deficit/hyperactivity disorders, antisocial personality disorders, narcissistic personality disorders, learning disabilities, affective disorders, posttraumatic stress disorders, and substance abuse. They concluded, "The high rate of comorbid diagnoses found within this population emphasizes the importance of utilizing sensitive, comprehensive, standardized methods when assessing and treating JSOs [juvenile sex offenders]" (p. 50).

Kraemer, Spielman, and Salisbury (1995) described four primary domains that require assessment: intellectual and neurological, personality functioning and psychopathology, behavioral, and sexual deviance. In addition, Becker and Hunter (1997) pointed out that psychometric testing to assess intellectual functioning and reading ability is important to ensure that the juvenile is able to understand both paper-and-pencil tests and treatment experiences.

The Minnesota Multiphasic Personality Inventory-Adolescent (MMPI-A) has been described as the psychological test most widely used with juvenile sex offenders (Bourke and Donohue, 1996). Because of the heterogeneity of juveniles who have committed sex offenses, there is no MMPI sex offender profile that distinguishes these juveniles from others (Bourke and Donohue, 1996; Dougher, 1995). The MMPI-A's strengths include its validity scales, which help the evaluator assess a juvenile's attitude and approach to the evaluation. As Dougher pointed out, "The extent to which an offender is dishonest, defensive, or malingering has obvious implications for treatment amenability and prognosis" (p. 11.8). The MMPI-A also may be useful for gaining insight into a juvenile's personality and for assessing possible psychopathology (Bourke and Donohue, 1996; Dougher, 1995).

Bourke and Donohue (1996) also reviewed other psychological tests that have psychometric properties and that may be useful for identifying clinical issues and psychopathology relevant to the treatment of juvenile sex offenders. For example, the Multiphasic Sex Inventory (MSI) is an assessment instrument that is used with adult sex offenders to evaluate issues such as sexual interests, knowledge, fantasies, and behaviors. Bourke and Donohue expressed concern that only limited research has been conducted with the MSI; however, their review pertained to a 1984 version of the instrument. There is a juvenile version of the MSI, but it appears that even less research has been conducted with it than with the adult version. In their discussion of the adult MSI, Milner and Murphy (1995) also addressed the issue of limited validity data but stated that in spite of this important weakness, the MSI may have clinical utility for descriptive purposes in known offender groups. Milner and Murphy did not discuss the juvenile MSI.

A more comprehensive and promising approach involves computerized assessment with the Multidimensional Assessment of Sex and Aggression (MASA). The MASA, developed by Dr. Ray Knight, Dr. Robert Prentky, David Cerce, and Alison Martino, is a computerized, self-report inventory that covers multiple domains (Knight and Cerce, 1999; Knight, Prentky, and Cerce, 1994; Prentky and Edmunds, 1997). A juvenile version is currently being validated (R.A. Knight, personal communication, October 16, 2000). The questionnaire asks about attitudes and behaviors in many areas of an individual's life, including childhood experiences, family and social relationships, school and work experiences, alcohol and drug use, and sexual and aggressive behavior and fantasies. The questionnaire includes items that have been associated with different classifications of offenders and with recidivism and includes sophisticated methods for assessing response biases, random responding, and dissimulation.

Studies have demonstrated that psychopathy is a strong predictor of violent behavior in general among adult offenders (e.g., Harris, Rice, and Quinsey, 1993; Quinsey, Rice, and Harris, 1995; Salekin, Rogers, and Sewell, 1996; Serin, 1996) and juvenile offenders (Forth, Hart, and Hare, 1990; Hare, 1991; Forth and Burke, 1998). Studies also have documented an association between psychopathy and sexual violence among adult offenders (Quinsey, Rice, and Harris, 1995; Serin et al., 1994). Studies investigating psychopathy and juvenile sex offending are more limited.

Gretton et al. (as cited in Forth and Burke, 1998) found that adolescent sex offenders who were diagnosed as psychopathic used more threats and more severe violence during their sex offense than those adolescent sex offenders who were classified as nonpsychopathic. Although relatively small proportions of adult sex offenders have been found to be psychopathic (Serin et al., 1994), adult sex offenders who are diagnosed with psychopathy and phallometrically assessed sexual deviance have been described as particularly dangerous (Hare, 1996).

Most studies of psychopathy use the Psychopathy Check List-Revised, a reliable and valid psychometric instrument specifically designed to assess psychopathy. Publication of the juvenile version of the Psychopathy Check List is expected soon (Forth, Kosson, and Hare, in press).

Assessing Deviant Sexual Arousal

To adequately assess individuals who appear to evidence deviant arousal, Weinrott (1998a) stressed the importance of using direct measurement of an individual's sexual arousal, through phallometric assessment (penile plethysmography). Becker et al. (as cited in Becker and Kaplan, 1993) reported preliminary research findings involving phallometric assessment that suggested deviant erectile responding was common in adolescents who had abused young boys and who had been sexually abused themselves.

Others, however, have discussed potential ethical concerns related to using phallometry with juveniles who have committed sex offenses (Bourke and Donohue, 1996; Cellini, 1995). Concerns include the invasive nature of phallometric assessment, possible exposure of juveniles to sexual material beyond their experience, and limited research documenting the validity of phallometric assessment with juveniles. The limited research regarding the utility of phallometric assessment with juveniles who have committed sex offenses is partly due to ethical issues related to obtaining control groups. Weinrott (1998a) noted, however, that many of these issues could be addressed by using stimuli that are less sexually explicit in detail and language, because adolescents typically have strong responses to most sexually explicit stimuli. In addition, less explicit stimuli may increase the validity of phallometric assessment with juveniles.

Another psychophysiological assessment measure used with juveniles who have sexually offended is the Abel Assessment for Interest in Paraphilias (Abel Screening, Inc., 1996). The Abel Assessment is a computer-driven assessment approach that provides an evaluation of a juvenile's sexual interest patterns based on his or her reaction times when viewing slides of potentially sexually evocative stimuli. This methodology is significantly less invasive than phallometric assessment. Good reliability and significant correlations with diagnoses and self-reported arousal patterns have been reported (Abel Screening, Inc., 1996). However, an independent study of the Abel Assessment's reliability and validity raised questions about the use of this assessment approach with juveniles at this time (Smith and Fischer, 1999). The Abel Assessment is relatively new, and additional independent, published research is needed.

Using Other Assessment Strategies

Substance abuse assessment. In addition to the assessment of personality functioning and deviant arousal, it is also important to assess whether the individual has a substance abuse problem and, if so, whether it is a risk factor for that individual's sex offending. The importance of using valid and reliable assessment tools to screen for substance abuse difficulties has been emphasized (Becker and Hunter, 1997; Lightfoot and Barbaree, 1993).

Polygraph tests. Although controversial, the use of polygraph tests in treatment programs for juveniles who have been sexually abusive is increasing (National Adolescent Perpetrator Network [NAPN], 1993). The polygraph is used with some juveniles to facilitate more complete disclosures of sexually abusive behaviors and to monitor compliance with treatment. The National Task Force on Juvenile Sexual Offending noted that "[i]t is critical that submissions to polygraph examinations be voluntary and with full informed consent of the youth, parent, or guardian" (NAPN, 1993, p. 85). When disclosures during polygraph testing reveal previously unreported information, additional investigations can result. Furthermore, the Task Force pointed out that some professional organizations' ethical requirements preclude the use of instruments without empirical evidence of reliability and validity. Research regarding the reliability and validity of the polygraph for assessing juvenile sex offenders is very limited (Hunter and Lexier, 1998). Some researchers have seriously questioned the validity of the polygraph (Cross and Saxe, as cited in Bonner et al., 1998; Saxe, Dougherty, and Cross, as cited in Bonner et al., 1998). Both false positives and false negatives occur, and the emotional impact of polygraph administration on juveniles and the resulting effects on the therapeutic process remain unknown (Hunter and Lexier, 1998).

Risk Assessment

Few empirical studies have investigated sexual reoffense rates among juveniles or risk factors associated with recidivism. Two retrospective studies that investigated the frequency of offenses prior to the referral offense found relatively high offense rates. Awad and Saunders (1991) investigated the sex offense histories of 49 juveniles who sexually assaulted peer or adult females and 45 juveniles who sexually abused younger children. They reported that 61 percent of those who sexually assaulted peers or adults had histories of prior sex offenses and that 40 percent of those who abused younger children had histories of prior molestation. Fehrenbach et al. (1986) found that 57.6 percent of the 297 juvenile sex offenders in their sample had perpetrated other sex offenses prior to their referral offense.

Rates of Recidivism

One prospective study followed juvenile offenders (19 who had committed sex offenses and 58 who had committed other types of offenses) into adulthood (Rubinstein et al., as cited in Sipe, Jensen, and Everett, 1998). Findings revealed that 37 percent of those who had committed sex offenses as juveniles went on to have criminal records for sexual assaults as adults, in contrast to 10 percent of those who had committed other types of offenses as juveniles. A weakness of this study was the relatively small sample size for sex offenders. A strength of the study was its relatively long followup period of 8 years. It may not be possible to generalize the study's findings to other juveniles who have committed sex offenses, not only because the sample of sex offenders was small but also because it included juveniles described as very assaultive (a trait not representative of many juvenile sex offenders).

In contrast to the Rubinstein et al. study (as cited in Sipe, Jensen, and Everett, 1998), most studies have suggested that once a juvenile's sex offending has been officially recognized, subsequent detected sexual recidivism is relatively infrequent (Bremer, 1992; Hagan, King, and Patros, as cited in Kramer et al., 1997; Kramer et al., 1997; Miner, Siekert, and Ackland, 1997; Rasmussen, 1999; Sipe, Jensen, and Everett, 1998; Weinrott, 1996). Sipe, Jensen, and Everett (1998) found that only 9.7 percent of their sample of 124 juveniles who had committed "nonviolent" sex offenses against children under 16 years old were subsequently arrested for a sex offense as an adult. Interestingly, 3 percent of a sample who had committed nonsexual offenses as juveniles were also arrested for a sex offense as an adult. Both groups were more likely to be arrested for nonsexual offenses as adults (16.1 percent of the juvenile sex offenders and 32.6 percent of the other juvenile offenders). Followup periods in the study ranged from 1 to 14 years, with an average of 6 years.

Smith and Monastersky (1986) examined the juvenile justice records of 112 juvenile sex offenders. During a 17-month period of time when they had the opportunity to commit an offense while in the community, 16 (14.3 percent) committed another sex offense and 39 (34.8 percent) committed a nonsexual offense.

Schram, Milloy, and Rowe (1991) followed 197 juvenile sex offenders after they completed 1 of 10 different treatment programs. The followup period ranged from 2 to 7 years. The study found that 37 percent had no new arrests. Of the 63 percent who had new arrests, only 12 percent were arrested for a sexual offense. Similarly, relatively few were subsequently arrested for violent felonies (15 percent). Most rearrests were for either nonviolent felonies (40 percent) or misdemeanors (53 percent). (The offense categories were not mutually exclusive, and the juveniles may have been rearrested for more than one type of offense.) The 2 years immediately following discharge from treatment represented the period of highest risk, especially for those treated in institutions. Although some of the juveniles may have offended later, results suggested that most reoffending occurred when the subjects studied were still juveniles. A very small subset of offenders (seven, or 4 percent of the sample) were deemed, for the purpose of the study, to be "chronic" offenders (defined as having two or more sex offense arrests after the referral offense or one prior and one subsequent sex offense arrest). The researchers found that most of the juveniles in their sample desisted from sex offending after their first sex offense arrest, adjudication, and treatment. They concluded that very few who commit sex offenses as juveniles go on to commit such offenses as young adults. This finding is consistent with that of Sipe, Jensen, and Everett (1998), who, as noted above, found that only 9.7 percent of their juvenile sex offender sample were arrested for sex offenses as adults.

Kahn and Chambers (1991) described 221 juvenile sex offenders identified by Schram and Rowe (as cited in Kahn and Chambers, 1991), who only included 197 in the study reported above (Schram, Milloy, and Rowe, 1991). The subjects in this sample were in the community with the opportunity to reoffend for an average time of 20.4 months. Not surprisingly, recidivism rates were similar to those found by Schram, Milloy, and Rowe (1991). Nearly 45 percent of the 221 juveniles in this sample were convicted of one or more subsequent offenses. Of those who recidivated, only 6.6 percent had new convictions for nonsexual violent crimes and only 7.5 percent had convictions for sex crimes.

More recently, Miner, Siekert, and Ackland (1997) followed 96 juveniles who participated in the Minnesota Department of Correction Juvenile Sex Offender Program. The average time at risk for the followup was 19.3 months. During the followup period, 27.2 percent were arrested for a crime that did not involve a person, 10.4 percent were arrested for a new crime against a person, and only 8.3 percent were arrested for a new sex offense.

Rasmussen (1999) also recently reported findings on factors related to recidivism rates among first-time juvenile sex offenders. Rasmussen's results were consistent with previous research in that 54.1 percent (N=92) of the sample committed a new nonsexual offense, whereas only 14.1 percent (N=24) committed a new sex offense. The relatively higher reoffense rates may reflect the comparatively long followup period of 5 years.

Notes: a M=mean time at risk in the community. b Percentages have been rounded off to the nearest whole number. c N=the total number of subjects in the group sample. d M=mean number of months or years followed by the study. e R=range. f Offense categories were not mutually exclusive, and the juveniles may have been rearrested for more than one type of offense.

Table 2 summarizes results of the recidivism studies reviewed above. Two of the studies included comparison groups of juveniles who apparently committed only nonsexual offenses. As the table indicates, recidivism involving nonsexual offenses was consistently and significantly higher than recidivism involving sex offenses, for both juvenile sex offenders and comparison groups. Weinrott (1996) provided a more extensive review of studies investigating recidivism rates among juvenile sex offenders. The findings summarized herein are consistent with Weinrott's overall findings. Methodological variations clearly influence recidivism rates (Prentky et al., 1997). These variations include issues such as the definition of recidivism (i.e., a new arrest versus a new adjudication), the adequacy of delinquency or criminal records, and the duration of the followup period (Prentky et al., 2000). Yet, as Weinrott noted:

What virtually all of the studies show, contrary to popular opinion, is that relatively few JSOs [juvenile sex offenders] are charged with a subsequent sex crime. Whether this is due to deterrence, humiliation, lack of opportunity, clinical treatment, increased surveillance, or inadequate research methodology is difficult to ascertain. (p. 67)

Factors Associated With Recidivism

Becker (as cited in Friedrich, 1990) suggested that adolescent sex offenders were probably more likely to reoffend if one or more of the following factors were present: initial offending was pleasurable, consequences for the offense were minimal, the deviant sexual behavior was reinforced through masturbation or fantasy, and/or the offender had social skills deficits. These factors appear to have good face validity but require additional assessment.

As noted above, Smith and Monastersky (1986) examined the juvenile justice records of 112 juvenile sex offenders. They found that very few selected predictor variables were associated with reoffending. Offenders described as having "unhealthy" attitudes regarding sexuality (i.e., those who naively denied normal adolescent sexual behavior) were less likely to reoffend by committing a sex offense and somewhat less likely to reoffend by committing a nonsexual offense. The only other statistically significant findings involved nonsexual reoffenses. Offenders who appeared to understand the exploitive nature of their sex offenses were less likely to reoffend nonsexually, and those who were unable to identify their personal strengths were more likely to reoffend nonsexually. Interestingly, a lack of depression and a willingness to explore the referral sex offense nondefensively were both marginally related to an increased rate of sexual reoffending and a reduced rate of nonsexual reoffending.

In the Smith and Monastersky study, some offense characteristics also were marginally associated with reoffending in general. Rapists were less likely to reoffend than those who committed seemingly less serious crimes. Those who offended against substantially younger victims (4 or more years younger than the offender) were less likely to reoffend. In contrast, those who committed offenses against strangers were more likely to reoffend sexually (and less likely to reoffend nonsexually) than those who offended against relatives or acquaintances. Lastly, those who had at least one recent offense against boys were described as "somewhat" more likely to reoffend sexually than those who offended only against girls.

Schram, Milloy, and Rowe (1991) found that juvenile sexual recidivists had higher rates of truancy, higher rates of thinking errors (erroneous perceptions, ideas, and beliefs that justify abusive behavior—e.g., blaming victims), and at least one prior sex offense. They also were much more likely to have deviant sexual arousal patterns, although this was not assessed with physiological measures. Sexual recidivism was not related to the type of referral sex offense, treatment location, or type of treatment received.

In Schram, Milloy, and Rowe (1991), those who did not reoffend generally were older, had less previous contact with the juvenile justice system, and were less likely to have school behavior problems or truancy. They also were significantly less likely to have been sexually abused or have a sibling who was abused. They were more likely to have social skills deficits and were significantly less likely to blame their victims or exhibit deviant arousal patterns.

Kahn and Chambers (1991) found (in Schram and Rowe's sample, as cited in Kahn and Chambers, 1991) that only two variables were significantly positively associated with sexual reoffending: using verbal threats during the commission of the offense and blaming the victim for the crime. Surprisingly, denial of the offense was negatively associated with reoffense rates: none of the eight offenders who completely denied their offenses sexually reoffended. Although offenders with "therapist-identified" deviant arousal (i.e., assessed by clinical judgment) reoffended at a higher rate than those without deviant arousal (13 percent versus 6 percent), this difference was not statistically significant. Similarly, although offenders who victimized a child they knew but were not related to were more likely to be adjudicated delinquent for a new sex offense than those who were related to their victims, the difference was not statistically significant. It also is important to note that more than 50 percent of the adolescent sex offenders in this study had histories of nonsexual criminal offenses.

In their 1997 study of juvenile sex offenders, Miner, Siekert, and Ackland indicated that predictors of reoffending included penetrating the victim during the original sex offense and coming from an unstable home. In Rasmussen's 1999 study of juvenile sex offenders, multivariate analyses revealed that sexual recidivism was associated with perpetrating sex offenses against multiple female victims; i.e., juveniles with a history of multiple female victims, as contrasted with a single female victim or multiple male victims, were more likely to sexually reoffend. This finding is contrary to Smith and Monastersky's (1986) finding suggesting that juvenile sex offenders who sexually abused male victims may pose a higher risk of sexual reoffending. In addition, Rasmussen (1999) found that nonsexual recidivism among juvenile sex offenders was related to having a relatively high rate of previous nonsexual offenses and to not completing treatment.

In spite of the various descriptions of characteristics identified in juveniles who have sexually offended, Weinrott (1998b) reported that very few characteristics have been empirically associated with sexual recidivism. He noted that these characteristics include the following (Weinrott, 1998b, p. 1):

• Psychopathy.
• Deviant arousal.
• Cognitive distortions.
• Truancy.
• A prior (known) sex offense.
• Blaming the victim.
• Use of threat/force.

For example, a recent study (Hunter and Figueredo, 1999) found that nearly 75 percent of the juveniles who did not evidence any denial of their sex offenses when beginning treatment successfully complied with treatment requirements during the 12-month period under study. In contrast, only 25 percent of the juveniles who evidenced full denial complied with treatment during the same period. The authors reported that "attitudes of openness and accountability proved to be the best predictors of a positive treatment outcome" (p. 65). It is important to note that adjudication may have been a confounding variable in this study, given that most of the juveniles who were adjudicated completely acknowledged their offense and relatively few of those who had not been adjudicated did so. Thus, the adjudication process and its consequences may have contributed significantly to treatment compliance. Furthermore, this study did not investigate whether openness and accountability were related to reduced recidivism rates.

Prediction of Recidivism

In a recent commentary, Chaffin and Bonner (1998) pointed out that there are no true experimental studies comparing untreated and treated juvenile sex offenders and no prospective studies evaluating risk factors or the natural course of sexual offending. As noted above, empirically based typologies have received some attention; however, an actuarial risk assessment schedule with adequate empirical validation is lacking (NAPN, 1993).

Two studies have investigated the accuracy of recidivism predictions by program staff. Schram, Milloy, and Rowe (1991) found that treatment staff members very accurately identified offenders who presented a low risk for sexual reoffending, but some of these juveniles reoffended in other ways. In contrast, only 18 percent of juveniles who were identified by program staff as "at risk" or "dangerous" sexually reoffended during the period under study. It is possible, of course, that some of these at-risk or dangerous offenders actually reoffended but were not detected. This finding is consistent with others (e.g., Smith and Monastersky, 1986) and suggests that treatment providers may tend to overpredict sexual recidivism (and therefore keep offenders in treatment) rather than risk the dire consequences associated with failing to predict recidivism that comes to pass.

There are a number of explanations for the relatively poor accuracy of attempts to predict sexual (and violent) recidivism (Smith and Monastersky, 1986). Sex offending is a relatively infrequent event. Predicting any low-frequency event is difficult. The hidden nature of sexual abuse may contribute to low reoffense rates because reoffending may tend to go undetected; however, juveniles who have already been identified as sex offenders may be followed more closely and have less opportunity to reoffend. Too short followup periods also may account for low predictive accuracy; some offenders may offend sometime in the future, but after the study period. Further, as Smith and Monastersky observed, "It may be that the low rate of sexual reoffending is due to lasting changes in the offender and/or his family as a result of being identified, evaluated, treated, adjudicated, and/or sentenced" (p. 135).

Other problems associated with poor predictive accuracy include the absence of pertinent information needed for decisionmaking and clinicians' overreliance on inadequate predictors (MacArthur Violence Risk Assessment Study, 1996). An additional confounding factor is conservative decisionmaking that occurs to avoid predicting that someone will not reoffend, when in fact they might (Smith and Monastersky, 1986). As researchers have noted (Smith and Monastersky; Webster et al., 1997), to enhance predictive accuracy, professionals should balance historical and actuarial information with clinical and situational information. Assessment of risk should address a variety of factors that pertain to the individual juvenile and the juvenile's environment and situational factors that could increase or reduce risk.

Ageton and her colleagues (as cited in Prentky et al., 2000) investigated the predictive utility of several measures and found that four variables correctly classified 77 percent of the juveniles who reoffended sexually: involvement with delinquent peers, history of crimes against persons, attitudes toward rape and sexual assault, and family normlessness. Subsequent analysis revealed that only one variable—involvement with delinquent peers -- was necessary to correctly classify 76 percent of the cases.

Prentky et al. (2000) have developed and conducted initial testing of an actuarial risk assessment schedule designed to assess the risk of reoffending among juvenile sex offenders. The schedule includes four factors: sexual drive/preoccupation, impulsive/antisocial, clinical/treatment, and community adjustment. The factors and individual items are based on literature reviews of studies pertaining to juvenile sex offenders, adult sex offenders, and juvenile delinquents in general. The risk assessment schedule was evaluated by following 96 juvenile sex offenders who had received treatment on an outpatient basis. The followup period was 12 months. Only 11 percent of the offenders who were studied committed any type of criminal offense during the followup period, and only three of these juveniles (4 percent of the sample) committed another sex offense. In evaluating the validity of the risk assessment schedule, Prentky et al. reported that, overall, the interrater reliabilities for the items (which indicate consistency in scoring between individual raters) were good to excellent and the scale alphas (which provide a conservative estimate of a measure's reliability) were quite good. Because of the very low base rate of sexual recidivism, the researchers were unable to evaluate predictive validity. Overall, however, the findings were encouraging. Data collection continues at a number of different sites to gather sufficient cases to permit a proper look at the usefulness of the schedule for assessing risk and predicting reoffending.

As Epps (1994) noted, potential problems in using risk assessment tools to predict juvenile sex offenders' likelihood of reoffending include difficulties in gathering reliable and valid information on which to base such instruments. Sufficient staff training and supervision also are important to ensure appropriate and reliable risk assessment.