Policies, guidelines and current practices

The Department of Corrections' Psychological Service does not have specific guidelines for risk assessment of female offenders and consequently neither for violent or sexual female offenders. In both the Psychological Service Manual and the training manual "Psychological Assessment of Risk" (Department of Corrections, 2003c) no distinction is made between male and female offenders  "most likely by default" given the higher rates of crime and convictions by men than by women. The policy document to enhance the effectiveness of Integrated Offender Management for women offenders (Department of Corrections, 2003a) states that "risk prediction tools are effective in predicting risk of re-offending, but may be significantly more effective for women if gender-specific factors were researched, developed and incorporated. There is no agreement to date on what those factors might be" (p. 3).

Risk assessment tools that incorporate systematically both actuarial (static) and dynamic factors are recognised as more accurate than clinical judgement or pure static risk prediction. In discussing risk prediction Bakker, O'Malley and Riley (1999) report that research on statistical or actuarial scales "consistently outperform the judgements of experts in almost every investigation comparing these two approaches to risk assessment … irrespective of the experience and professional training of those making judgements?" (p. 7). In making predictions, the authors suggest to clarify the context of the prediction: general re-offending, the type of re-offending or "to facilitate a decision based on the probability of serious re-offending or re-imprisonment" (p. 10). The Department of Corrections adopted the models of re-conviction (RoC) and re-imprisonment (RoI) developed by Bakker, O'Malley and Riley (1999), which are based on the criminal offending histories of 133,000 male and female individuals and tested on at least 8000 subjects. Gender was one of the predictor variables i.e. being male is predictive of violent reconviction. The models can predict:

  • whether a further conviction would occur during a five-year follow-up
  • if a conviction did occur, whether the offence would be at a low, medium or high level of seriousness
  • whether an individual would be imprisoned
  • if the individual was imprisoned, whether they would be sentenced to a short, medium or long prison term" (p. 3).

The RoC*RoI are two separate calculated algorithms which combined provide the probability of offending occurring during a five year follow-up period which would be serious enough to result in prison sentence. For example, a high RoI does not necessarily mean the offender is at high risk of re-offending violently. The RoC*RoI was found to work well for female offenders too. Its predictive accuracy is approximately 2% less when applied to female offenders as a subgroup than when applied to the entire research population (D. Riley, personal communication, November 29, 2005). If a specific predictive model for female offenders was developed from the outset, different variables may have been used. This may have resulted in somewhat different relationships between the predictive variables and re-offending (King, 2004), but not necessarily in a significant improvement of its predictive accuracy. It has been suggested for the RoC*RoI to incorporate gender-specific factors to increase its accuracy when applied to female offenders (Department of Corrections, 2003a). The RoC*RoI has not been proven to be a predictive measure of misconducts (internal risk factor relating to security classification) for incarcerated women (Department of Corrections, 2002).

All Psychological Service offices received a request for information from staff that had assessed female offenders. The current practice of assessing violent and sexual female offenders within Psychological Service was reviewed, based on information and 16 reports provided by 10 psychologists who responded to the request.

For violent females, most staff used the RoC*RoI score although for some offenders this score was unavailable. Static and dynamic factors for violent recidivism were usually considered based on the Violence Risk Scale (VRS 1) or the Historical Clinical Risk Scale (HCR-20 2). The Level of Service/Case Management Inventory (LS/CMI 3) and the Psychopathy Checklist-Revised or Psychopathy Checklist:Screening Version (PCL-R or PCL:SV 4) were administered twice, because of availability of norms or validity data for female offenders. One psychologist commented the HCR-20 was used as an aide-memoir only and another psychologist stated that multiple use of risk tools is recommended. Other lists of dynamic factors in the reviewed reports could not be clearly associated with any of the known risk assessment tools. Risk factors for general recidivism were also mentioned, but often it was unclear what they were based on. In some cases, these factors clearly referred to Andrews and Bonta's "big eight 5" or to the criminogenic needs of the Criminogenic Needs Inventory ( CNI 6). Some staff members made specific reference to acute dynamic factors unlikely to be replicated after release and whether treatment was received or declined. 

Information about the risk assessment of four female sexual offenders was available from four different psychologists. Three people stated that no actuarial measures normed for female sexual offenders are available or that research in this area is limited. The limitations of the RoC*RoI were mentioned (i.e. it does not accurately predict risk in relation to sex offending and to women). Only one psychologist appeared to have used research on female sexual offenders and female prisoners in the risk assessment section. Reference was made to criminogenic needs according to the CNI, to static and dynamic risk factors specific to the person's situation and to static and dynamic factors identified in the SONAR but with clear reference to its male-based validation. In one case the PCL:SV was administered to confirm clinical assessment that psychopathic personality features were absent and to endorse low risk of future violent and non-violent offending.

In conclusion, in the absence of specific departmental guidelines for risk assessment of re-offending of serious female offenders, there appears to be some inconsistency amongst Department of Corrections' staff's assessment of female violent offenders, using a combination of actuarial and clinical assessment, although rarely specific tools are named or limitations are pointed out in their application to females. For female sexual offending, the limits of existing risk assessment tools are recognised but the choice of which risk factors to assess appears inconsistent.


1 For discussion and reference see page 35.

2 For discussion and reference see page 47.

3 For discussion and reference see page 69.

4 For discussion and reference see page 41.

5 For discussion and reference see page 26.

6 For discussion and reference see page 54.