Forensic Psychology
Page 83
18.4.1 The Offender Group Reconviction Scale (OGRS)
(Copas & Marshall, 1998; OGRS2: Taylor, 1999; OGRS3: Howard, Francis, Soothill & Humphries, 2009; OGRS4: Howard, 2015a)
The Offender Group Reconviction Scale (OGRS; Copas & Marshall, 1998) is an actuarial risk of reconviction instrument developed within England and Wales. Early versions of OGRS utilised a small number of static datapoints – age, gender and criminal history variables (such as type of offence and number of previous sanctions) – to provide estimates of reconviction (for any offence) within the next two years. The latest version, OGRS4 (Howard, 2015a), however, allows the user to input similar data to obtain one-year and two-year predictions of both general and non-sexual violent proven reoffending. A further improvement of OGRS4 over its predecessors is that it takes into consideration “offence-free time” in the calculation of risk; previous versions of OGRS provided fixed OGRS scores – the implication being that the probability of reconviction remains static through time. In reality, however, an offender’s probability of proven offending falls the longer their offence-free period after sentencing or discharge from custody (Howard, 2015a). The incorporation of this information into OGRS4, therefore, increases the predictive utility of the tool across the sentence.
The development of OGRS4 was based on a study of more than 1,809,300 offenders discharged from prison or sentenced to a community penalty between March 2005 and March 2008. The amount of “offence-free time” prior to March 2008 was calculated for each individual and the sample’s subsequent two-year (from March 2008 to March 2010) reconviction history was elicited from the Police National Computer. The proven reoffending information, both general and violent, was used to assess the predictive utility of a range of static data points. Those data points that proved to be efficient in predicting reoffending were incorporated into the OGRS4 algorithm, which produces a risk of proven reoffending score of between 0 and 1.
The OGRS instrument is widely used within correctional services within the UK; the ease and speed of the tool has ensured its practicality. The risk assessment tool has been criticised, however, for focusing solely on static and historic criminal history and demographic variables at the expense of more dynamic social and behavioural items. As such the tool acts exclusively as a risk of proven reoffending estimator and hence does not provide any assessment of offender need. Despite this, the accuracy of the tool is good: the authors report Area Under the Curve statistics (AUC: a measure of predictive validity) of close to 80% for both general and violent proven reoffending estimates.
18.4.2 Level of Service Inventory-Revised (LSI-R)
(Andrews & Bonta, 1995)
The LSI-R is a 54-item risk and need assessment instrument originally designed utilising Canadian data (Andrews, 1982) and has been used extensively with a variety of offender samples within Europe and North America. This tool not only provides an assessment of the risk of reoffending but also information relating to the treatment needs of the assessed offender. The items of the LSI-R, which comprise both static and dynamic factors, produce scores relating to 10 subcomponents. These subcomponents provide information relating to the offender’s needs within the domains of criminal history, education/employment, finances, family/marital, accommodation, leisure/recreation, companions, alcohol/drug problems, emotional/personal, and attitude/orientations. These domain scores are also weighted according to how efficiently previous research has shown them to predict reconviction to produce a composite score. The composite score hence represents the likelihood of future offending. Lower scores represent a lower prevalence of criminogenic risk factors and hence a lower risk of future offending.
Hollin and Palmer (2003) have described the LSI-R as an effective and efficient assessment instrument for needs-risk assessment with offenders. Indeed research findings attest to the validity and reliability of the assessment tool with: the general offender population (Raynor, 2007); in relation to females (Coulson, Nutbrown & Giulekas, 1993); minority ethnic groups (Schlager & Simourd, 2007); violent offenders (Hollin & Palmer, 2003), sexual offenders (Simourd & Malcolm, 1998); and young offenders (Shields, 1993; Shields & Simourd, 1991).
18.4.3 The Offender Assessment System (OASys)
The turn of the twenty-first century saw the ambitious development and subsequent implementation of a new risk and needs assessment tool within the prison and probation services of England and Wales. Prior to the development of the Offender Assessment System (OASys), the prison and probation services within England and Wales used a range of risk prediction tools, including ACE (Assessment, Case management and Evaluation instrument; Roberts, Burnett, Kirby, & Hamill, 1996), OGRS, and LSI-R. The provision of a common tool that could be used across prison and probation environments, it was argued, would allow for continued assessment and evaluation throughout NOMS.
The data required to complete OASys, which comprise both static and dynamic risk factors, are collected through file review and offender interviews. As such OASys is a structured clinical assessment tool – it relies on both actuarial methods and practitioner interpretation of the offender’s circumstances in the formation of the final risk assessment. The main criticism of OASys, however, is the time required to complete the assessment. The Home Office (2003, cited in Raynor, 2007) has estimated that a full OASys assessment can take two and a half hours compared to the LSI-R, which can take as little as 10 minutes (Raynor, 1997, cited in Raynor, 2007).
The payback for this resource outlay, however, is the richness of data that OASys produces. These data are contained within the five components of OASys, which are summarised in Table 18.2. The first of these components, which comprises the main section of OASys, consists of the assessment of 12 different offending related factors (see Table 18.2). Similar to the LSI-R, the scores obtained for each section are weighted and summed to provide risk scores. The most recent version of OASys provides a general risk of reconviction score (OGP) and a risk of violent reconviction score (OVP) – an improvement on the previous version, which provided a general risk of reconviction score only. The OVP score predicts homicide and assaults, threats and harassment, possession of offensive weapons, public order offences, criminal damage and violent acquisitive crime, whereas OGP predicts other reoffending (not including sex offending).
Table 18.2 The components of the OASys risk and need assessment tool (Moore, 2006)
1. Risk of reconviction and offending related factors Offending information
Analysis of offences
Accommodation
Education, training and employability
Financial management and income
Relationships
Lifestyle and associates
Drug misuse
Alcohol misuse
Emotional well-being including anxiety and depression
Thinking and behaviour
Attitudes towards offending and supervision
2. Risk of serious harm, risks to the individual and other risks
3. OASys summary sheet
4. Sentence planning
5. Self-assessment
As can be seen from Table 18.2, however, the OASys tool provides more than just a risk of reconviction score. The second section provides an analysis of risk of serious harm, defined as “a risk which is life-threatening and/or traumatic, and from which recovery, whether physical or psychological, can be expected to be difficult or impossible” (OASys manual, cited in Howard, Clark, & Garnham, 2006, p. 7), through the evaluation of actuarial, social and personal factors. The output from this section, therefore, is intended to inform the employment of risk management procedures. The summary sheet and sentence planning sections of OASys are also designed to inform the management and supervision of the offender. Finally, OASys has provision for a self-assessment questionnaire, which provides the opportunity for practitioners to gain an insight into the offenders’ own views of their needs. An evaluation of the self-assessment questionnaires of more than 100,000 offenders concluded that offen
ders tended to be more optimistic about their chances of desisting from crime than their OASys assessment would predict and less likely than practitioners to identify problem areas in their lives (Moore, 2007).
The development of OASys has received positive acclaim: “OASys is a very comprehensive and strongly research-based assessment instrument, informed by detailed study of others including LSI-R and ACE” (Raynor, 2007, p. 135). Indeed OASys, which has been adopted for use outside England and Wales (Bunton & Morphew, 2007), has been subject to an extensive piloting and review process that has refined and improved the tool in line with the ongoing research. For example, research into the predictive utility of previous versions of OASys led to the introduction of OVP, the refinement of the prediction of general offending (OGP) and more recently the development of the more accurate predictors, OVP2 and OGP2. These evidence-based amendments have improved the validity of the tool; both OGP2 and OVP2, which, like OGRS4, include “offence-free” elements, outperform OVP and OGP respectively (Howard, 2015a).
18.4.4 Risk and Need Instruments for Violent Offenders
According to the Crime Survey for England and Wales, there were approximately 1.34 million violent offences against adults during the year to December 2016, representing approximately 22% of all crime during this period (Office for National Statistics, 2017).
When a violent offender is reprimanded it is helpful to be able to assess, first, how likely they are to reoffend in a violent manner and, second, their level of criminogenic need in order to inform decisions relating to their sentencing and management. For example, an offender who has a history of violent offending and harbours strong negative views against a certain sector of society would be managed in a different way to an individual who, after consuming large amounts of alcohol, offended violently towards a partner. It is hence necessary to assess, for each individual, the presence of factors that indicate repeat violent behaviour, and the triggers of that potential behaviour. We have seen that the OASys and OGRS4 tools incorporate an assessment of violent risk. Another such tool is the HCR-20.
18.4.5 Historical, Clinical Risk Management-20 (HCR-20V3)
(Douglas, Hart, Webster, & Belfrage, 2013)
The HCR-20 is a risk assessment instrument that incorporates structured clinical judgements to provide an assessment of future violent behaviour amongst civil psychiatric, forensic and criminal justice populations. Cooke (2000) has stated that the HCR-20 is “the best known, and best researched, empirically-based guide to risk assessment” (p. 155). The tool consists of 20 items: 10 historical, 5 clinical, and 5 risk management factors (see Table 18.3). Indeed the tool takes its name from the initial letter of each of these domains – HCR – and the number of items that comprise the tool. The HCR-20 currently in use is the third version, which was developed following careful consideration of the empirical literature and subsequently amended in the light of evidence from the clinical experience of early trials. The use of consultation with forensic clinicians in the development and refinement of the tool is seen as positive: “as such, the HCR-20 is an attempt to merge science and practice by offering an instrument that can be integrated into clinical practice but also is empirically based and testable: (Douglas, Guy, & Weir, 2006, p. 4).
Table 18.3 The components of the HCR-20 (Douglas, Hart, Webster & Belfrage, 2013)
Historical items Clinical scale (recent problems)
1. Violence 1. Insight
2. Other antisocial behaviour 2. Violent ideation or Intent
3. Relationships 3. Symptoms of major mental disorder
4. Employment 4. Instability
5. Substance abuse 5. Treatment or supervision response
6. Major mental disorder Other considerations
7. Personality disorder Risk Management (future problems)
8. Traumatic experiences 1. Professional service and plans
9. Violent attitudes 2. Living situation
10. Treatment or supervision response 3. Personal support
Other considerations 4. Treatment or supervision response
5. Stress or coping
Other considerations
The HCR-20 tool has been shown to be predictive of inpatient violence by civil psychiatric clients (Klassen, 1999) as well as violent crime by the same population on release from hospital (Douglas & Webster, 1999). Research on forensic populations, both community and prison-based, also attests to its predictive ability in relation to violence (Douglas & Webster, 1999; Strand, Belfrage, Frannson, & Levander, 1999; Wintrup, 1996). Indeed, the authors of the HCR-20 tool periodically maintain an annotated list of research papers that assess the utility of the HCR-20 with a range of populations both civil and criminal. The latest update contained in the region of 100 studies reporting on the usefulness of this violence risk and need tool (Douglas, Guy, & Weir, 2006).
18.4.6 Risk and Need Instruments for Sex Offenders
Sex offending represents a small but significant proportion of all crime. In England and Wales, in the year to December 2016, out of a total of 6.1 million crimes, 116,012 crimes of a sexual nature were reported to the police (Office for National Statistics, 2017). It is apt to remember, however, that official figures are considered to underestimate the true prevalence of sex offences – some individuals are reluctant, feel too ashamed or feel unable to report such crimes. Indeed the British Crime Survey indicated that only 11% of victims of serious sexual assault tell the police about their victimisation (Povey, Coleman, Kaiza, & Roe, 2009). This figure may have increased in recent years, however; the year ending December 2016 saw an increase of 115% in police recorded sexual offences compared to statistics from 2011 (Office for National Statistics, 2017). Taking all available statistics together, it seems that sex offences represent a small fraction of all crimes committed. Nevertheless, correctional services still need to be equipped with the tools to assess the risk and need of the perpetrators of this type of offence.
The term “sex offending” encompasses a wide variety of different behaviours with varying levels of severity – from stranger rape and child sexual assault, through the collection of child abuse images and grooming behaviour, to indecent exposure and voyeurism. As such, those individuals deemed by their behaviour to be “sex offenders” are not homogenous in nature but will present to correctional agencies with very different risk and need profiles. Given the destructive effects of sex offending, the reduction and prevention of such violence should be the ultimate goal of risk and need assessment. The identification of dynamic risk factors, which inform targets for treatment and the content of management plans, and the assessment sexual recidivism risk, are therefore the aims of sex offender risk and need tools. One such tool, the Structured Assessment of Risk and Need (SARN), was developed empirically using the sex offender literature and is used alongside the England and Wales Prison Service’s Sex Offender Treatment Programme (SOTP) to inform the intervention work of those on the programme.
18.4.7 Structured Assessment of Risk and Need (SARN) and Risk Matrix 2000 (RM2000)
(SARN: Thornton, 2002; RM2000: Thornton, et al., 2003)
The SARN initially measures static risk through the use of the Risk Matrix 2000, a purely actuarial risk tool (Thornton, et al., 2003). Based on the earlier Structured Anchored Clinical Judgement (Thornton, 1997) and Static 99 (Hanson & Thornton, 1999) tools, the Risk Matrix 2000 uses a two-step process to determine the offender’s judged risk banding (low, medium, high or very high). Step one provides an initial assessment of risk level through the evaluation of the number of sex offence sentences, the number of any criminal offence sentences, and the offender’s age on release from prison. The second step considers the presence of aggravating factors and adjusts the initial assessed risk level, if needed. The presence of two or more factors results in the increase of the initial risk level by one unit. Level two risk factors are: a male victim in any previous sex offence, a stranger victim in any previous sex offence, any non-contact sex offence, never having been married.
<
br /> Once the Risk Matrix 2000 banding has been determined, the second step of the SARN assesses the presence or absence of 16 dynamic risk factors that, through empirical research, have been linked to sex offender recidivism (Thornton, 2002). These risk factors are grouped across four domains: sexual interests; distorted attitudes; social and emotional functioning; and self-management. The 16 factors are scored twice: first in relation to its relevance to the offender’s offence chain and second in relation to their life generally. Utilising psychometric data, personal history information, interview material and file review, each factor is rated as 0 (not present), 1 (present but not a central characteristic) or 2 (a central characteristic). Any factor that receives a score of 2 both for the offence chain and the offender’s general life is designated as a relevant treatment need for that particular offender. This tool, therefore, provides a risk score in conjunction with a needs profile that can be utilised within treatment planning.
Webster et al. (2006), mindful that the clinical scoring of the 16 risk factors may vary between assessors, have conducted an evaluation of the inter-rater reliability of the SARN. Using two pools of assessors, the authors found that experienced, or “expert”, raters showed greater levels of reliability than less experienced raters. The authors argue that in the field of sex offending, where the offenders often display denial and minimisation to confuse and manipulate assessors, very high levels of reliability are perhaps unattainable. The finding that reliability is further weakened when less experienced or poorly trained assessors are used brings the authors to conclude that:
it is not appropriate to use the SARN framework to make quantitative predictions about risk of reoffending. It is our view, however, that the studies reported in this paper support the use of the SARN as an instrument for guiding clinical assessment of treatment need. Provisos for its use to this end are as follows: users of the SARN should have demonstrated reasonable inter-rater reliability before conducting assessments without supervision, the SARN should be applied wherever possible by experienced psychologists rather than trainee-level staff, and the use of the SARN should continue to be monitored and evaluated.