Andrew J. Harris Ph.D.
University of Massachusetts, Lowell
IN BOTH RESEARCH and practice, the past two decades have produced considerable developments in the field of specialized risk assessment for sexual offenders. Dozens of studies have contributed to a growing evidence base regarding recidivism risk factors and the potential efficacy of treatment interventions (Hanson & Bussiere, 1998; Hanson & Morton-Bourgon, 2004). On the basis of these findings, a broad array of specialized actuarial and guided clinical assessment instruments have been introduced and continue to be tested and refined (Doren, 2004a).
These developments have been fueled in part by resurgent policy attention to the issue of sex offender management during the past 15 years. Prompted in part by federal legislation, registration and community notification laws have been adopted nationwide, calling for effective systems of classifying levels of risk ( Adams, 2002). Since 1990, seventeen states have adopted civil commitment laws for sexual predators—policies predicated on predictions of future violence and increasingly requiring states to address the complex issues associated with an individual’s suitability for community release (Harris, 2005). The emergence of specialized models for community supervision of sexual offenders has demanded effective means of applying risk assessment in a multi-disciplinary context (English, Pullen, & Jones, 1997). The introduction of risk-based sentencing systems has produced unique demands for evidence-based decision tools that can both inform the sentencing process while ensuring due process (Kern & Farrar-Owens, 2004). As of this writing, legislative activity shows few signs of slowing, with the issue of sex offenders remaining at the top of state legislative crime control agendas (National Conference of State Legislatures, 2006).
While increased policy focus on these issues has produced both a significant expansion of the evidence base and the increased attention of researchers, it has also presented the burgeoning field of sex offender risk assessment with considerable challenges. The often overlooked heterogeneity of the population to be managed, coupled with the diverse range of organizational and programmatic contexts in which risk assessment is required, produces considerable potential for over-generalization of particular findings or the misapplication of particular tools or approaches.
As our policies have evolved, sex offender risk assessment has been called upon to respond to the needs of multiple stakeholders and to meet a wide range of legal, forensic, and clinical purposes. Treatment professionals use it to develop treatment plans or evaluate progress. Probation and parole personnel use it to establish suitability for community supervision, case management, and intervention. The courts apply it for purposes of civil commitment or criminal sentencing. Law enforcement has adopted it for purposes of profiling, investigation, or designation of sex offender risk levels for purposes of registration and community notification.
Considering this diversity of contexts, it would appear that one-dimensional “debates” over the relative merits of particular approaches to risk assessment (e.g. clinical vs. actuarial approaches) may miss an essential part of the picture. Indeed, the key to finding “middle ground” between competing viewpoints may rest primarily in achieving greater clarity of our goals and objectives, and in adapting our methods and practice to meet those goals (Dvoskin & Heilbrun, 2001).
Consistent with this view, this article aims to present the discussion regarding sexual offender risk assessment in a circumscribed programmatic context, with specific focus on the practice of community-based supervision of sexual offenders.
This article consists of two parts. The first reviews the current state of sex offender risk assessment, considering the factors known to be associated with sexual recidivism and the methods currently utilized to translate those factors into risk assessment practice. The article’s second section applies this understanding to the specific programmatic context of community-based supervision of sexual offenders, and sets forth a framework for integrating current risk assessment knowledge into systems of community-based supervision of sexual offenders.
Sexual Offender Risk Assessment—The State of the Field
As noted in the introduction, advances in the field of specialized sex offender risk assessment accelerated greatly beginning in the 1990s. While this development was spurred in part by advances in general violence risk assessment research, it was greatly facilitated by a range of critical developments in the realm of public policy surrounding sexual offenders (Hanson, 2005). On the legislative front, these developments included the spread of sex offender registration and community notification laws across the nation and the passage and implementation of civil commitment laws for sexually violent predators. Concurrently, with growing national emphasis on issues surrounding prisoner re-entry and community corrections, jurisdictions across the country were expanding and refining a range of specialized community supervision models for sexual offenders (English et al., 1997).
The common thread running through each of these diverse policy strategies is the shared goal of reducing recidivism among individuals previously convicted of sexual offenses. Consistent with this goal, policymakers, researchers, and practitioners have focused increasing attention on three main areas:
- The identification of risk factors associated with sexual recidivism;
- The integration of those risk factors into structured assessment instruments; and
- The refinement of interventions aimed at reducing re-offense rates.
This section addresses the first two areas, setting the stage for a later discussion of the implications of risk assessment for guiding community supervision interventions.
Risk Factors for Sexual Re-Offense
In 1998, Hanson and Bussiere published a meta-analysis of 61 studies providing information on 28,972 sexual offenders and investigating risk factors associated with sexual recidivism (Hanson & Bussiere, 1998). This was followed in 2004 by an updated metaanalysis encompassing the initial studies plus additional research conducted between 1998 and 2003, accounting for a total of 95 studies and over 31,000 sexual offenders (Hanson & Morton-Bourgon, 2004).
The studies found aggregate sexual reoffense rates (based on average follow-up periods of 5-6 years) of 13.4 percent and 13.7 percent, respectively. The earlier study further differentiated sexual re-offense rates between child molesters (12.7 percent re-offending) and rapists (18.9 percent reoffending). The studies also reported on rates of re-offense related to non-sexual violent crimes, finding overall re-offense rates of 12.2 percent and 14 percent (9.9 percent of child molesters and 22.1 percent of rapists).
Identified risk Factors
While the later study included some notable additional findings, the two analyses were fairly consistent in their overall conclusions regarding the major predictors of long-term recidivism.
Topping the list in both studies was the presence of certain types of sexual deviancy, as measured by both phallometric assessment and deviant sexual preferences as measured by standardized tools or clinical records. The Hanson and Bussiere study found sexual interest in children (i.e., pedophilia) to be a strong predictive factor in child molesters, although it did not find sexual interest in rape to be a significant predictor among rapists. The later study confirmed this finding, adding the existence of other paraphilias (such as exhibitionism and voyeurism) as having additional predictive value. The study further cautioned that the lack of findings regarding an association between sexual reoffense and paraphilic interest may be due to a limited number of studies that investigated this association, and suggested further research in this area.
The second most dominant factor identified in both studies involved the presence of antisocial lifestyle and orientation, as characterized by “rule violations, poor employment history, and reckless, impulsive behavior.” (Hanson & Morton-Bourgon, 2004). Notably, in contrast with the sexual deviance variables, this factor has been consistently found to serve as a strong predictor of general recidivism in non-sexual criminals (cite Bonta et al. 1998). Considering this, some have questioned whether similar mechanisms are at work in sexual and nonsexual offenders, or if antisocial orientation interacts with other predictors to create a unique dynamic among sexual offenders.
Beyond these first two major factors—sexual deviance and antisocial orientation—the meta-analyses identified a range of additional factors established as having moderate predictive value. These factors included:
- Age (younger offenders presenting higher risk);
- Number of prior offenses;
- Single marital status;
- Treatment failure;
- Sexual preoccupations; and
- Intimacy deficits.
Despite the growing base of knowledge related to risk factors for sexual recidivism, research in this area has been constrained by a range of methodological issues. Most of these issues relate, in one way or another to the base rate of sexual offending—i.e. the proportion of individuals within the population who eventually re-offend.
As noted above, the aggregate base rate for sexual offender as established by studies included in meta-analyses is somewhere between 13 percent and 14 percent. For a variety of reasons, however, this figure most likely underestimates the “true” rate of sexual offending, and additionally does not effectively capture the range of variation in this rate across subsets of the sex offender population. Issues commonly associated with the base rate include:
Under-reporting — The vast majority of studies addressing the issue of sexual recidivism operationalize re-offense as incidents that are detected and lead to arrest and conviction. It is fairly well-established that only a limited proportion of sexual crimes—perhaps fewer than one in three—are reported to the police (Hart & Rennison, 2003). Accordingly, it is likely that actual re-offense rates may be substantially higher than those captured by recidivism researchers. A related confounding factor relates to the fact that the extent of this under-estimation may not be uniform across groups of offenders, considering that offenders with certain characteristics (such as higher intelligence) may simply be more adept at avoiding detection.
Population heterogeneity —Sexual offenders are an extremely diverse group. Beyond the fundamental distinction between rapists and child molesters, each of these groups includes a wide range of subtypes linked to victim choices, underlying motivations, behavioral patterns, and other factors (Knight & Prentky, 1990; Lanning, 1986). This heterogeneity has a range of implications for both research and practice. From a methodological vantage, failure to effectively distinguish between these subgroups in research designs complicates the capacity to conduct within-group analysis, especially with those groups that are under-represented in samples or those with low overall base rates. In terms of application, this diversity of the offender population is generally not acknowledged in commonly used actuarial tools, leading some to question the validity of these instruments as means of predicting violence in individual cases (Hart, Webster, & Menzies, 1993).
Timeframes —For reasons of resources and practicality, studies employ a wide range of follow-up periods in their assessment of recidivism. Although the studies included in Hanson and Bussiere’s meta-analysis involved an average follow-up period of five years, evidence suggests that the risk of re-offense may extend far beyond this threshold (Hanson, Stey, & Gauthier, 1993). Hence, the research time horizon must be viewed as a source of potential bias in the derivation of the base rate.
Statistical Significance —From a research standpoint, the most immediate implication of a relatively low observable base rate involves researchers’ reduced capacity to draw statistically significant conclusions from available data. While this may be mitigated in part by increasing the sample size, many studies are limited in their capacity to expand their samples due to resource or logistical constraints.
Diminished Predictive Value —The predictive capacity of actuarial risk assessment instruments is directly influenced by the base rate upon which that instrument has been based and validated—the lower the base rate, the higher the probability of error. In general, low base rates are most likely to increase the probability of “false alarms.”
Static and Dynamic Factors
Reviewing the major variables known to be most closely associated with long-range sexual recidivism, many have noted that the majority of these factors are either static or highly stable in nature. While we have developed a fairly good sense of these immutable case characteristics that might place certain individuals at higher risk of re-offense than others, we have a much more limited understanding of the influence of dynamic characteristics associated with sexual recidivism risk (Craissati & Beech, 2003; Hanson & Harris, 2000b).
Certainly, the use of static variables in an applied context carries some clear advantages. Beyond their long-range predictive value, they are comparatively easily accessible through official records, and generally involve little or no subjective judgment. Yet over time, these static predictors gradually lose their utility for the majority of offenders under community supervision. Hence, effective systems of community supervision may begin with an understanding of an individual’s general risk as predicted by static variables, but ultimately depend on our capacity to identify and respond to changes in risk levels over time.
Dynamic variables may be viewed in terms of stable and acute factors (Gendreau, Little, & Goggin, 1996; Hanson & Harris, 2000b). Stable factors are those mutable characteristics of the individual that may change over time, but are not generally subject to short-term fluctuations. Key stable dynamic factors include variables such as cognitions, insight, treatment compliance, and attitudes related to offending behaviors. Acute factors reflect case characteristics that may change over more limited periods of times—in some cases weeks, days, or even hours. These factors might include both short-term life changes in domains such as employment, residence, or relationships, and immediate conditions such as intoxication or circumstances that may provide access to potential victims.
Research regarding the effects of dynamic variables on sex offense recidivism remains in a developmental state. While methodological limitations have constrained much research in this area, key dynamic factors that appear to be related to recidivism include social adjustment, attitudes towards victims, self-awareness regarding risk to recidivate, victim access, and cooperation with supervision and treatment (Hanson & Harris, 2000b). The results of the Dynamic Supervision Project, a five-year longitudinal study currently following 1,000 offenders under community supervision in Canada, Alaska, and Iowa, may eventually provide further perspective on these factors (Harris & Hanson, 2003).
Approaches to Sex Offender risk Assessment
Hanson (2002) cites three potential approaches to sexual offender risk assessment— pure actuarial approaches, which make predictions based on survey instruments that leave no room for subjective interpretation; guided clinical approaches, which rely on the systematic professional judgment of qualified professionals based on empirically-derived instruments; and adjusted actuarial approaches in which professional judgment is superimposed on actuarial scores. To these options, we might add a fourth plausible approach—the use of unstructured clinical judgment to determine risk.
Doren (2004a) identifies over 20 instruments that have been applied in the assessment of sex offender risk. These instruments are varied—some have been developed as a means of evaluating the potential for general violence risk (Quinsey, Harris, Rice, & Cormier, 1998; Webster, Douglas, Eaves, & Hart, 1997), while others have been geared towards identifying the risk specifically for sexual offenders (Epperson et al., 1999; Hanson, 1997; Hanson & Harris, 2000a; Hanson & Thornton, 1999; Hart, Kropp, & Laws, 2004). Some are pure actuarial tools that present a fairly one-dimensional perspective on an individual’s relative risk level (Hanson, 1997) while others are designed to be utilized as support systems for more comprehensive clinical determinations (Hart et al., 2004). Some rely solely on static variables (Hanson & Thornton, 1999), while others integrate dynamic predictors on a limited (Epperson et al., 1999) or exclusive (Hanson & Harris, 2000a) basis.
Following a brief review of some of these instruments, we will consider the relative utility of the various approaches.
Several specialized actuarial instruments for the prediction of sexual re-offense have emerged during the past decade. The actuarial approach, in a nutshell, gathers a series of variables believed to have predictive validity, applies relative weights to each variable, and combines these data into an aggregated risk score and classification.
One widely used instrument is the Rapid Risk Assessment for Sexual Offense Recidivism, known as the RRASOR (Hanson, 1997). The RRASOR is notable for its brevity and ease of use—it consists of only four variables, all of which can be easily pulled from official records. These four factors —prior sexual offenses, extra-familial victims, offender age under 25, and male child victims—were identified for use in the scale based on research indicating a strong correlation between these factors and risk of re-offense. While demonstrating moderate predictive accuracy, the RRASOR omits several variables shown to have particularly high correlations with re-offense risk, including deviant sexual preferences, antisocial orientation, and treatment compliance.
A second commonly used tool, the Static-99, addresses some of these shortcomings by combining the RRASOR with a second scale, the Structured Anchored Clinical Judgement-Minimum. Beyond the variables contained in the RRASOR, the Static-99 considers a range of additional factors including sexual deviance, range of available victims, persistence, and a pattern of antisocial behaviors (Hanson & Thornton, 1999; Hanson & Thornton, 2000). In a comparative review, the Static-99 has been demonstrated to add to the predictive accuracy of the RRASOR in the measurement of long-term risk potential (Hanson & Thornton, 2000).
A third instrument—the Sex Offender Risk Appraisal Guide (SORAG) (Quinsey et al., 1998)—measures a different, although likely closely related group of factors compared to the Static-99. This scale, adapted from a general violence prediction tool known as the VRAG, is notable for its integration of psychiatric and psychological variables, including psychopathy and mental illness diagnoses. Its relative predictive value appears comparable to the Static-99 in the prediction of sexual recidivism, and appears to more effectively predict non-sexual violent recidivism (Hanson & Thornton, 2000).
Beyond these instruments designed for general use, some states have developed customized instruments, generally under the auspices of a state agency, designed for specific uses. The Minnesota Sex Offender Screening Tool (MnSOST) was originally developed in the early 1990s by the Minnesota Department of Corrections as a means of codifying factors viewed to place an individual at high risk for re-offense (Huot, 1999). Revised to the MnSOST-R in 1996 (Epperson et al., 1999), the tool was explicitly designed to be used by non-clinical staff. Research on the MnSOST has demonstrated moderate predictive capacity, comparable to other commonly used actuarial instruments (Barbaree, Seto, Langton, & Peacock, 2001; Hanson & Morton-Bourgon, 2004).
Finally, the SONAR (Sex Offender Needs Assessment Rating) was designed in 2001 as an actuarial tool based on dynamic variables (Hanson & Harris, 2000a). Viewed as an adjunct to actuarial instruments based on static factors, the SONAR captures information across both stable and acute dimensions. Stable factors include intimacy deficits, negative social influences, attitudes toward sex offending, and self-regulation. Acute factors include substance abuse, negative moods, anger, and victim access. The SONAR was subsequently adapted into two scales—the STABLE 2000 and the ACUTE 2000. These scales, combined with the Static-99, form the basis for a blended approach toward community supervision designed to capture long-term, intermediate, and short-term factors associated with sexual recidivism (Harris & Hanson, 2003).
Structured Clinical Decision Tools
In contrast with actuarial instruments, which contain explicit rules for weighting each variable, structured clinical assessment guides the evaluator to consider a range of empirically validated risk factors, which the evaluator then assesses for a general estimate of risk.
One example of a structured clinical decision tool is the Sexual Violence Rating Scale (SVR-20) (Boer, Hart, Kropp, & Webster, 1997). Applying a similar approach to the HCR-20—a tool used to structure clinical decisions regarding the risk of general violence (Webster et al., 1997)—the SVR-20 encompasses twenty variables that are distributed into three broad domains. These domains include psychosocial adjustment (encompassing factors such as sexual deviance, history of childhood sexual abuse, psychopathy, relationship problems, employment instability, and offending history); the nature of sexual offending (such as levels of violence employed, escalation in offense severity, and attitudes toward offending behaviors); and future plans (i.e., responses to interventions). One notable characteristic of tools such as the SVR-20 lies in their potential to capture and integrate the individual’s responses and reactions to treatments and interventions. The SVR-20 has recently been modified into a new instrument known as the Risk for Sexual Violence Protocol (RSVP) (Hart et al., 2004).
Comparing the Approaches
The research on sexual offense recidivism has focused primarily on issues of long-term risk. Bounded by certain methodological limitations, this research has been highly focused on static factors that have been demonstrated to be associated with the probability of future sexual offenses.
In the context of these circumstances, it is not terribly surprising that actuarial assessment has carried the day. Research on the comparative ability of these approaches to predict general recidivism in a population of sexual offenders has found actuarial assessments to be most accurate, followed by guided clinical approaches, then by unstructured clinical judgment (Hanson & Bussiere, 1998; Hanson & Morton-Bourgon, 2004).
Regarding adjusted actuarial methods, actuarial “purists” argue that the track record of clinical judgment is so poor that scores on validated instruments should not be tainted with any subjective interpretation (Quinsey et al., 1998). However, some have suggested that the proven superiority of such adjusted approaches in other domains (notably weather forecasting) suggests that the adjusted actuarial approach may represent a promising approach (Monahan & Steadman, 1996; Swets, Dawes, & Monahan, 2000). To date, however, little or no empirical evidence has emerged testing this premise (Hanson & Morton-Bourgon, 2004).
Yet before concluding that actuarial methods represent a uniformly superior means of risk assessment, one must also recognize certain limitations to the actuarial approach.
First, on a purely conceptual level, many have questioned the validity of making case-level determinations on the basis of tools that have been derived solely from population-based probabilities. This issue has arisen particularly in the context of sexually violent predator civil commitment decisions. Accordingly, some have argued that sole reliance on actuarial instruments risks missing important clinical information that can aid significantly in prediction (Hart, 1998).
Second, the most widely employed actuarial scales are implicitly one-dimensional in nature. Conceivably, individuals with entirely different constellations of risk factors may be classified at similar levels of risk—a fact that may obscure important distinctions for purposes of service planning and supervision. In this sense, actuarial scales’ reliance on measures of cumulative risk obscure important case characteristics that may indicate an elevated risk. While some have called for multi-dimensional models drawing from multiple actuarial instruments (Doren, 2004b), evidence to date has not lent support to such an approach (Seto, 2005).
Ultimately, the relative superiority of one method or another is highly dependent on the questions that we are asking. If our primary concern deals with the aggregated long-term risk posed by a group of individuals, actuarial instruments almost certainly provide the most valid means of assessing such risk. If we are concerned with setting forth the relative probability that a particular individual will re-offend at some undetermined point in the future, actuarial instruments provide a moderate degree of accuracy, albeit one prone to errors.
Yet as soon as we turn to different types of questions, the relative utility of currently available actuarial instruments dissipates considerably. Under what circumstances would this person be most likely to reoffend? What is the probable timeframe of re-offense? How has this person’s re-offense risk been mitigated by our interventions? What is the probable impact of treatment and supervision? While work continues on actuarial approaches that might eventually answer some of these questions, these issues simply cannot be adequately addressed by current actuarial methods.
Considering these factors, the remainder of this article is grounded on three fundamental premises regarding the clinical-actuarial distinction—first, that any discussion regarding the relative merits of clinical vs. actuarial approaches cannot occur in a vacuum, and must be placed in its appropriate programmatic and operational context; second, that, while the clinical-actuarial distinction is important from a theoretical perspective, and while some circumstances call for orthodox adherence to one of the two methods, the majority of sex offender management practice calls for operating on a “middle ground” that draws from both approaches; and third, that the clinical-actuarial continuum is only one dimension within a broader practical framework that integrates a range of related constructs.
Risk Assessment and Community Supervision Practice: A Framework
Having reviewed the existing state of sexual offender risk assessment knowledge and practice, we now turn to the fundamental question presented at the outset of this article—how can risk assessment systems and methods be effectively aligned with the specific goals and challenges of community supervision practice?
As noted earlier, the risk assessment methods to be applied in a given situation are highly dependent on the specific questions that need to be answered. The variability of these questions may be viewed in terms of organizational demands, in accordance with the distinct information needs of central management, unit supervision, and line staff; in terms of case level demands, recognizing the significant heterogeneity of the sex offender population in terms of offense type and severity, motivations, and associated levels of risk; and in terms of temporal demands, noting that the challenges associated with managing a particular case change over time.
Figure 1 presents a multi-dimensional framework that aims to address some of these sources of variability, integrating the risk assessment concepts described earlier with the range of challenges associated with the community supervision of sexual offenders.
The five noted dimensions, described below, are not intended to be categorical in nature—rather, each should be viewed as a part of a continuous spectrum of choices that must be made in conjunction with the process of community supervision.
Dimension #1: Primary Orientation
The model begins by framing the distinction between a nomothetic (i.e. population-based) approach and an idiographic (i.e. individual-based) approach as a framework with which to understand the appropriate (and in turn the inappropriate) application of risk assessment to sexual offenders. Under a nomothetic orientation, decisions are driven exclusively by population-based probabilities based on empirically validated systems of evaluating risk. Evidence-based policy and practice in this arena involves the assessment of the populations of concern using actuarial methods that rely primarily on static or highly stable predictors of re-offense. Under an idiographic orientation, decisions are based on case-specific attributes based on circumstances presented during a particular point in time. Evidence-based practice in this instance depends on the far less developed area of research into dynamic predictors of sexual recidivism, and generally employs informed practitioner judgment.
Dimension #2: Risk Emphasis
Heilbrun (1997) distinguishes between prediction-oriented styles and management-oriented styles of communicating risk. Prediction-oriented risk assessments are appropriately applied in contexts that explicitly call for understanding general likelihood of an event occurring at some undefined point in the future. Conversely, management-oriented approaches are more suited to the ongoing task of understanding and managing risk at the case-level. Dvoskin & Heilbrun (2001) associate the predictive orientation with actuarial methods and the management orientation with clinical methods, suggesting this distinction as a means of bridging the divide in the actuarial vs. clinical debate.
Dimension #3: Risk Factors
The framework presents risk factors as a spectrum covering three general domains—static factors, stable (dynamic) factors, and acute (dynamic) factors (Hanson & Harris, 1998). As a matter of practice, there is a strong rationale for viewing these factors as a continuum rather than as discrete categories. While some factors such as historical variables are by definition immutable, other “static” variables might straddle the domains. For example, whether psychopathy falls in the category of a static, immutable category or a highly stable (but ultimately changeable) personality characteristic remains open for debate. The precise boundaries between stable and acute dynamic factors, often loosely defined by issues of timing or magnitude, may be similarly unclear.
Dimension #4: Primary Methods
The actuarial-clinical dimension, as the other dimensions contained within the framework, is viewed as a spectrum of alternatives, rather than an “either-or” proposition. At the far ends of this spectrum, the specified method is framed as the predominant (although not necessarily exclusive) means of gathering salient and valid information. In the middle of this spectrum, the framework considers blended approaches integrating both actuarial and bounded practitioner judgment as the most effective means of assessing risk.
Dimension #5: Frequency of Assessment
The final dimension—the frequency of risk assessment processes—represents a critical operational issue related to planning and implementation of community supervision systems. Baseline assessments based on exclusively static variables, by definition, tend not to require repeated administration, presenting minimal burden on operations and resources. Similarly, periodic structured assessments to gauge gradual change in relatively stable case characteristics can be integrated into regular work processes with predictable impact. The ongoing demands associated with identifying and responding to imminent risk, however, present a wide range of operational challenges associated with issues such as communication, surveillance systems, and staff workloads. This factor may also be viewed as a significant potential operational impediment to the introduction of actuarial methods as a means of assessing acute risk.
Policy and Practice Domains
Having briefly considered the general dimensions and the relationships between them, our next step is to apply these dimensions to the specific challenges associated with the community supervision of sexual offenders. The figure’s primary columns, denoted by the shaded boxes, divide these challenges into four general domains—policy and management, baseline planning, case management, and acute intervention.
Policy and Management
The policy and management domain encompasses the actions and decisions of organizational leadership within parole and probation agencies. It may also, under certain circumstances, encompass the actions of legislators charged with the crafting of public policies associated with community-based sex offender management.
Although actors within this domain generally operate independently of case-level decisions, reliable and valid data regarding the risk levels presented by individuals under agency supervision emerge as vital management indicators—indicators that affect such matters as the formulation of policies and procedures, the allocation of resources, organizational strategy, quality management, and program design.
At the policy and management level, effective decision-making depends largely on the maintenance of a nomothetic perspective centered upon population-based indices and patterns. The information required to make key decisions in this domain emphasizes the prediction of general risk within the population, rather than the specific risk posed by individual cases. Consistent with these goals, actuarial assessments driven by static characteristics within the population generally provide appropriate levels of information to inform decisions regarding resources and organizational strategy.
At the individual case level, one of the initial tasks faced by community supervision agencies involves the establishment of baseline levels of risk. This assessment may occur as part of a Pre-Sentence Investigation (PSI) process, as part of a prisoner re-entry plan, or as part of the agency’s classification process, often in collaboration with law enforcement, correctional authorities, and treatment providers.
Within this domain, line managers and staff are required to make a range of decisions associated both with the initial terms of probation or parole and with the allocation of often limited resources. Who is appropriate for lifetime or intensive supervision? What special conditions and restrictions need to be placed on each individual? What are each individual’s treatment needs and potential responsiveness to treatment? Whose risk might be mitigated with access to ancillary services such as substance abuse treatment, employment programming, mental health services, or residential programming?
Within this domain, prediction-oriented assessments provide case managers with a baseline assessment of an individual’s general risk. Such general predictions may inform such decisions as development of initial case plans and the resource-intensity of supervision. Over time, however, prediction-oriented notions of risk gradually lose their relative utility to those charged with individual supervision, giving way to a significant demand for management-oriented approaches.
In contrast with the baseline planning stage, the case management domain shifts the emphasis from the realm of prediction into the realm of management. While baseline risk levels provide highly relevant context to ongoing service planning and risk management, the greater concern becomes the flow of information regarding changes in the offender’s psychological, social, or environmental status. Are insights and attitudinal adjustments being gained in treatment? Has the individual managed to maintain relationships, employment, and housing? Does the process of community integration seem to be succeeding?
The answers to these types of questions carry a range of implications for both the agency and for the individual case. At the agency level, they help to prioritize the assignment and allocation of resources, and provide potentially valuable information to managers regarding the efficacy of interventions. At the case level, they provide supervision staff with vital data relevant to adjustments in service plans, expansion or contraction of terms and conditions, or identification of emergent needs.
In contrast with the case management domain, in which programmatic adjustments are made based on gradual evolution of circumstances, decisions in this domain are concerned with short-term changes in psychological, social, or environmental conditions that might presage offending behavior. This domain’s primary concern is based on one central question—namely, when is an individual at imminent risk of re-offending?
By necessity, this domain focuses on the unique characteristics of the individual case, and accordingly falls at the idiographic end of the spectrum. While knowledge of the individual’s general risk level might provide useful context, general predictions of the person’s probability of re-offense are far less salient than information that will identify factors associated with pending re-offense and, in turn, inform appropriate intervention. Accordingly, the static risk factors that might have contributed to this individual’s baseline risk assessment carry relatively little practical utility when compared to time-specific situational factors such as access to potential victims (i.e., opportunity), relapse into drug or alcohol use, lapses in compliance with terms of supervision, and stressors such as the loss of a job, home, or relationship.
In contrast with domains towards the other end of the spectrum, assessment methods within this domain remain highly dependent on bounded professional judgment. While actuarial systems for evaluating changes in acute dynamic factors among sexual offenders remain under development (Hanson & Harris, 2001), their efficacy and utility have not been fully explored. Accordingly, given the current state of knowledge, the assessment of acute dynamic factors remains largely dependent on practitioner judgment supported by effective training, protocols, and systems of communication.
Addressing Population Variation
The framework presented above suggests that risk assessment methods must adapt to variation across organizational processes and functions. Equally important, however, these methods must respond to another critical source of variation, specifically that related to the population under supervision.
The heterogeneity of the sex offender population is well established, and has been delineated in a range of typologies developed in both the clinical and law enforcement context. These typologies have identified significant areas of divergence within populations of rapists (Knight, 1999), child molesters (Knight, Carter, & Prentky, 1989; Lanning, 1986) and even sexual murderers (Schlesinger, 2004). Key dimensions associated with this variation include primary motivation, intelligence, underlying sexual deviance, anger, opportunity, victim relationship, level of force, and a broad range of other factors.
Within sex offender typologies, the various constellations of these factors (some of which may be closely related) produce a wide range of potential offender subtypes, each associated with distinctive levels of risk. Further, and perhaps more critically, the triggers for re-offense may be markedly different across these subtypes. This factor presents significant challenges to the development of uniform methodologies for assessing dynamic risk.
Of particular importance to community supervision practice is the fact that, to date, the evidence base has been weighted towards higher-risk offenders, particularly those who have been released following a prison sentence. Accordingly, the dynamic factors that may trigger re-offense among probationers who fall into lower long-term risk categories are far less understood, and represent a critical area for future research (Hepburn & Griffin, 2004; Meloy, 2005).
Approximately 60 percent of sex offenders under correctional supervision in the United States are sent to serve their sentences in the community (Greenfeld, 1997). Moreover, sexual offenders comprise approximately 5-6 percent of individuals released to parole agencies, with an estimated 30,000 under parole supervision on a given day (Hughes, Wilson, & Beck, 2001). Considering these figures, the development of effective systems for community-based supervision looms large in our overall approach to sex offender management.
As more and more jurisdictions develop specialized capacity to manage sex offenders in the community (English et al., 1997), the demands for effective risk assessment have continued to expand. As such, it remains vital that the role of risk assessment, and consequently the methods that are employed, be bounded by the specific challenges faced by probation, parole, and community corrections agencies. This requires recognizing and adapting to the range of variation both in organizational-programmatic goals and within the population under supervision.
On a final note, the role of risk assessment in community supervision practice cannot be divorced from the unique social and political context in which our society views sexual crime, its perpetrators, and its victims. With the issue of sexual offending remaining at the forefront of legislative agendas, and with persistent public misconceptions surrounding the nature of sexual crime, community supervision agencies operate in a political environment with a “zero tolerance” approach to errors, in which one tragic case can lead to widespread calls for system reform. In this environment, the imperative of targeted, adaptable, and responsive means of risk assessment should be evident.