Recommendations and guidelines

It has been well documented that actuarial measures of risk are more accurate than clinical judgements alone (Bonta, 2002). However, using actuarial instruments has limitations as Beech, Fisher and Thornton (2003) list, for instance its misleading potential when applied to "unusual individuals" not well represented in the research samples. They note the tendency to follow the assessment guidelines for male sexual offenders for female sex offenders but "how valid it is to do is not known at the present time" (p. 347). "Most of the measurement tools, or evaluations in the fields of anger management, assessment of risk and need, risk prediction, institutional violence and assault, are based on men, or relate to male populations" (Shaw & Dubois, 1995, Measurement and tools, 1). Actuarial predictors of recidivism are generally developed on particular (and large) populations. Offenders who do not fit the picture may not benefit from such actuarial predictor, such as female offenders. In that case detailed clinical assessment is necessary (Ditchfield, 1997). Blanchette (2001) agrees that assessment "is the cornerstone to effective correctional intervention" (p. 16) since risk assessment classification measures are lacking predictive validity for female offenders. Motiuk (2000) also supports the importance of assessment at the admission stage, "critical to the ability to gauge accurately risk during the later phases of the sentence, when decisions as to possible release are taken" (p. 11). He adds that "multiple methods of assessment are preferred" (p. 17; also Bonta, 2002) over multiple domains. Risk assessment tools that incorporate systematically actuarial/static and dynamic risk factors are more accurate than clinical judgement or only static risk prediction. The combined assessment of both risks and needs also improves the prediction of re-offending (Motiuk, 1993). In addition, the present discussion focussed on prediction of risk with offenders, but many research studies have dealt with psychiatric or forensic populations or with psychological instruments not developed for risk prediction but for psychological instability such as mental health problems.

In sum, many caveats exist when considering using risk prediction tools evaluated on male populations for female offenders. Based on this literature review no one single actuarial risk assessment tool can be accepted as valid, reliable and normed on female offenders.

The Level of Service/Case Management Inventory or LS/CMI (a revision and refining of the Level of Service Inventory-Revised), described as "a comprehensive measure of risk and need factors, as well as a fully functional case management tool" (Andrews, Bonta, & Wormith, 2004a, p. xiii) is appropriate for use with male and female offenders 16 years and older. For young people under 16 the Youth Level of Service/CMI is used. The LS/CMI contains items assessing all eight predictors of recidivism ("the big eight"). In addition the tool measures major criminogenic needs, responsivity issues and potential strengths. The LS/CMI has gender and population based norms for different countries (New Zealand is not included). The authors state that "the relationship between increased risk level and increased recidivism is consistent, without exception, for all of the offender groups examined (adults, youth offenders, males, females, mentally disordered and nondisordered, violent and non-violent)" (Andrews, Bonta, & Wormith, 2004a, p. 118). The exception was that female offenders" recidivism rates in the medium and high/very high risk category were one half (i.e. down) of the population rate.

The different sections of the LS/CMI cover general need/risk factors but also specific risk/need factors that may not apply to the general offender population, prison experience, other health and mental health issues, responsivity barriers (including gender-specific issues such as health for females, mothering concerns, victimization and cultural issues) and a case management plan based on the principles of risk, need and responsivity. The authors acknowledge the importance of non-criminogenic needs as "they may have an impact on the potential effectiveness of other interventions that do target criminogenic needs" (Andrews, Bonta, & Wormith, 2004b, p. 32). The LS/CMI appears an acceptable alternative to use with female offenders in general, keeping in mind some of its limitations 1 : it does not measure potential criminogenic needs related to Maori and norms for New Zealand offender populations are unavailable (Coebergh, Bakker, Anstiss, Maynard, & Percy, 2001).

McLean (1995) has focused on the process of psychological assessment of female offenders. She argues for a comprehensive assessment of criminogenic and non-criminogenic needs (in particular history of suicide and self-injury, depression, psychological difficulties) as the last may be paramount in the woman's adjustment and stability whilst in prison and after release, similar to Sorbello, Eccleston, Ward and Jones" (2002) proposal of using an enhancement model rather than a risk management model with female offenders. It is believed that McLean's statement "psychologists have only restricted ability to make predictive statements about reoffence in women and what will lower risk" (1995, p. 45) is still valid a decade later. The Code of Ethics (Psychologists Board, 2002) for psychologists working in New Zealand states:

2.1.3. "Psychologists who conduct psychological assessments select appropriate procedures and instruments and are able to justify their use and interpretation.… Any reservations concerning the validity or reliability of an assessment procedure, arising from its administration, norms, or domain-reference, should be made explicit in any report."

2.2.4. "Psychologists utilise and rely on scientifically and professionally derived knowledge, and are able to justify their professional decisions and activities in the light of current psychological knowledge and standards of practice."

The Department of Corrections also acknowledges that "psychometric tools should be normed and validated on the populations upon which they will be administered" (Department of Corrections, 2003c, p. 23), which applies to New Zealand, criminal populations (many studies have been validated on psychiatric or forensic populations) and subsequently to female offenders. Bonta (2002) affirms that psychologists need to be able to explain the proper use of a test and the empirical support for it and reminds the reader that "we are not at the point where we can achieve a level of prediction that is free from error" (p. 375).

Risk assessment is a dynamic area that evolves because of new research, knowledge and standards of practice. One needs to remember that risk assessment is not only assessment of the probability of re-offending but also assessment of how risk can be managed.

Recommendations (Hart, 2000; McLean, 1995):

  1. When using risk assessment instruments developed for male offenders, it is unsafe to assume validity and reliability for female offenders. It is not recommended that instruments without female norms are used.
  2. Use the limited research on female offenders to formulate a risk statement incorporating static and dynamic risk factors, rather than (altering) male-based risk assessment tools and programmes based on experience and feedback from participants. See Appendix A for details on specific research with female offenders and risk of general recidivism, Appendix B on females and risk of violent re-offending and Appendix C on females and risk of sexual re-offending.
  3.  When using instruments that have been partially validated on female offenders, be cautious. Always make limitations explicit and do not rely on existing cut-off scores. Be aware that the criminogenic needs of women are still not fully understood, that the base rate of serious offending by females is low and that originating factors to offending differ from maintaining factors.
  4. Consider using the Level of Service/Case Management Inventory to assess general recidivism by females as an alternative to existing departmental risk and need assessment measures.
  5. When assessing and interpreting risk factors that are similar for both sexes, include knowledge of the gender differences surrounding the offence (see Steffensmeier and Allan's suggestions on page 7) and the risk factor (e.g. substance abuse and anger), different motives and different understanding of constructs (e.g. violence and psychopathy).
  6. Become familiar with specific factors relevant to female offending from psychological literature (e.g. domestic violence, child abuse sequelae, substance abuse, connectedness) and non-psychological literature (e.g. social work) and keep up to date with research.
  7. Do not offer opinion on how understandable the offender's actions were.


  8. Be aware of bias such as blaming the victim, perceiving the women as a passive victim and pathologising female offenders.
  9. Avoid gender-based assumptions of causality that rely on philosophy alone.

    And finally:

  10. Assessment of criminogenic and non-criminogenic needs and responsivity issues, at different time intervals for risk and rehabilitation purposes, is the key in applying a holistic approach to helping female offenders reducing re-offending.

1 In the LSI data were not disaggregated by gender for analysis. Consequently "data of minority groups [such as women] becomes lost in that of the majority" (Blanchette, 2001, p. 50). It is unknown whether this was addressed in the development of gender norms for the LS/CMI.