Actuarial risk assessment measures consist of a set of predetermined risk factors (sometimes statistically weighted) that are combined based on an algorithm, which produces a total score that is associated with a final risk estimate or probabilistic statement regarding the likelihood of future recidivism. When assessing risk of general and sexual violence, two commonly used actuarial risk assessment measures are the Violence Risk Appraisal Guide (VRAG) and the Static-99, respectively. Although there has been extensive development and validation of actuarial measures with adult offender populations, there has been limited uptake of purely actuarial measures among young offenders with measures such as the Youth Level of Service/Case Management Inventory representing an adjusted actuarial approach (i.e., lacking in the probabilistic estimate of risk).
This article begins by explaining the development of actuarial risk assessment. Next, it describes the strengths and criticisms of actuarial risk assessment. The entry then discusses the use of actuarial risk assessment within a risk management/ treatment paradigm and concludes with a brief discussion of future directions.
Development
While the initial selection of risk factors for consideration in an actuarial measure can be informed by the broader empirical literature (e.g., meta-analysis), final selection of the risk factors is typically based on their association with an outcome of interest within a development sample (e.g., violent recidivism) and their incremental contribution relative to the other risk factors in predicting outcome. Thus, actuarial measures represent an empirical approach to the risk assessment process. This is not to say, however, that all actuarial measures are void of psychological or behavioral theory. For instance, development of the Level of Service Inventory–Revised, an actuarial risk measure designed for assessing risk of general recidivism, was grounded within the Risk-Need-Responsivity principles and psychology of criminal conduct as developed by psychologist Don Andrews and his colleagues.
Strengths of Actuarial Risk Assessment
Actuarial risk assessment has consistently outperformed unstructured clinical judgment in predicting short- and long-term violent and sexual recidivism with predictive accuracy that is significantly better than chance. To date, several meta-analytic investigations have found actuarial risk assessment measures to be significant predictors of various forms of outcome such as violence, sexual violence, and general recidivism. Although the majority of the recidivism research has not focused on the nature and severity of violence, there is preliminary evidence to suggest that actuarial measures are significantly associated with such outcomes. This is not surprising, since several of the measure developers have pointed out that actuarial measures were designed to predict these types of outcomes.
Unlike clinical judgment, actuarial risk assessment is systematic, impartial, and transparent. Research conducted within applied settings has revealed moderate to high interrater agreement between raters and, in a few instances, between clinicians and researchers (e.g., the Static-99). Likewise, the predictive validity of actuarial risk assessment is similar whether it is scored within an applied (i.e., clinical) or research setting. Similarly, moderate to high predictive accuracy has been found when actuarial measures are scored by frontline staff (e.g., police and parole officers). This highlights an appealing aspect of actuarial risk assessment in that extensive clinical training is not required for optimal predictive accuracy to be achieved.
Although some may view the lower level of training required (e.g., 1-day workshop) for some actuarial risk measures as a potential limitation, there are several instances in which this is preferable. One example is the Ontario Domestic Assault Risk Assessment (ODARA), a 13-item actuarial risk assessment measure designed for use by police officers and frontline police personnel to predict male-to-female intimate partner violence recidivism. Police officers investigating incidents of intimate partner violence are often faced with situations in which quick decision-making is key. Such decisions may be related to whether the perpetrator should be released or detained following an incident of intimate partner violence (i.e., bail hearings). Actuarial measures such as the ODARA are ideal in situations where the ODARA was specifically designed to be administered quickly using information available to police officers (e.g., criminal record, victim interview). Moreover, with actuarial measures such as the ODARA, the provision of a final risk estimate further allows for easier interpretation of the results. Lastly, given the shorter administration time and lower requirements regarding education and training, actuarial measures such as the ODARA can be more cost-effective, without compromising predictive validity, relative to other in-depth psychologically based risk assessment measures, particularly when being administered by frontline personnel.
Criticisms of Actuarial Risk Assessment
Historically, given that the selection of risk factors is based on their association with an outcome of interest, actuarial risk assessment measures have predominantly comprised static risk factors or historical variables given their robust associations with outcome. Static risk factors remain relatively stable over time and are not amenable to treatment (e.g., criminal history variables such as age at first offense). This is problematic because the violence risk assessment process has moved away from a purely predictive model to a risk management model. Thus, the overreliance of some actuarial measures on static risk factors has limited their usefulness within a risk management/treatment paradigm. Granted, while an actuarial risk assessment measure comprising solely static risk factors may be of assistance in identifying who is at higher risk to reoffend, there remains a need to assess treatment targets (i.e., criminogenic needs or dynamic risk factors).
There are other important concerns that arise due to the procedures used for item selection. First, the number of risk factors may be limited and may ignore other risk factors that are equally important but whose validity is unknown (i.e., factors that were not coded within the development sample). Likewise, the weighting procedures, such as those associated with the VRAG, have led to criticisms regarding the appropriateness of items both on legal and ethical grounds. For instance, having a female victim and increased severity of victim injury were weighted negatively on the VRAG. In contrast, having a male victim and no victim contact resulted in a positive weighting. More recently, with the revision of the VRAG, items such as victim injury and female victim have been removed.
Another criticism of actuarial risk assessment is the issue of generalizability of the final risk estimates or absolute recidivism rates (referred to as calibration). A meta-analytic examination of the absolute recidivism rates of the Static-99 (and its variants) found significant variation across samples. This highlights the need for recalibration of the absolute recidivism rates when an actuarial measure is applied to a population that lies outside that of the development sample. Similarly, with the passage of time, rates in the population will fluctuate and legislative changes may alter what constitutes a violent act. As such, the absolute recidivism rates must be adjusted accordingly to ensure that the actuarial measure is up-to-date; however, it is the responsibility of the assessor to determine whether or not an actuarial risk assessment measure is an appropriate fit for the client being assessed. In addition, some researchers have argued that the need for recalibration is not a criticism per se; rather, it is a necessary condition of good science.
To date, of all the criticisms that have been leveled against actuarial risk assessment, none have been as controversial as the argument that absolute recidivism rates derived from group data cannot be applied to the individual. However, despite these claims being refuted on statistical grounds, this criticism has advocates within the literature. This has, in part, led to the development and use of other metrics to communicate actuarial risk (e.g., percentile ranks, risk ratios).
Moving Beyond a Purely Actuarial Approach
With the advancement in assessment technologies and the development of several actuarially based risk assessment measures, describing these measures against a purely actuarial framework is no longer sufficient. Jeremy Mills and his colleagues offer examples of actuarial risk assessment, which fit well within a risk management/treatment paradigm: integrated-actuarial risk assessment and dynamic-actuarial risk assessment.
Integrated-Actuarial Risk Assessment
The integrated-actuarial approach is a form of actuarial risk assessment that incorporates actuarial risk estimates (AREs), potentially dynamic risk factors, intervention/treatment recommendations, and risk management strategies. This form of actuarial risk assessment is exemplified by the Two-Tiered Violence Risk Estimates, which represents a two-tiered approach to risk assessment. Embedded within the measure are the ARE that comprises 10 items (e.g., childhood antisocial behavior, prior incarcerations) and the Risk Management Indicators that assess 13 areas of functioning that have been empirically linked with violence (e.g., employment status, family instability). The Risk Management Indicators, which is an aid in explaining the underlying risk level, is designed to identify areas in need of risk management and to monitor potential changes in risk. Measures such as the Two-Tiered Violence Risk Estimates eliminate the need for multiple risk assessment measures and overcome many of the criticisms concerning purely actuarial risk assessment while allowing for the assessment to be anchored by the ARE.
Dynamic-Actuarial Risk Assessment
The dynamic-actuarial approach is similar to the integrated-actuarial approach in that it incorporates AREs and potentially dynamic risk factors; however, within the dynamic-actuarial approach, the reassessment of dynamic risk has the potential to alter the underlying ARE. An example of this form of actuarial risk assessment is The Dynamic Supervision Project that incorporates the Static-99, Stable-2007, and Acute-2007. Within this framework, changes in dynamic risk on the Stable-2007 can alter the risk level of the Static-99, and changes on the Acute-2007 can impact the combined Static-99/Stable-2007 risk level. These changes can lead to differences in absolute recidivism rates across multiple assessments.
Future Directions
Actuarial risk assessment remains an integral component of the risk assessment process. Because criticisms are associated with purely actuarial risk assessments, in combination with advances in risk assessment technologies, the area has made considerable gains that have resulted in the development of integrated- and dynamic-actuarial approaches.
References:
- Hanson, R. K. (2009). The psychological assessment of risk for crime and violence. Canadian Psychology, 50, 172–182. doi:10.1037/a0015726
- Hanson, R. K., Harris, A. J. R., Scott, T.-L., & Helmus, L. (2007). Assessing the risk of sexual offenders on community supervision: The dynamic supervision project (User Report No. 2007-05). Ottawa, Canada: Public Safety Canada.
- Helmus, L., Hanson, R. K., Thornton, D., Babchishin, K. M., & Harris, A. J. R. (2012). Absolute recidivism rates predicted by Static-99R and Static-2002R sex offender risk assessment tools vary across samples: A meta-analysis. Criminal Justice and Behavior, 39, 1148–1171. doi:10.1177/0093854812443648
- Imrey, P. B., & Dawid, A. P. (2015). A commentary on statistical assessment of violence recidivism risk. Statistics and Public Policy, 2, 1–18. doi:10.1080/2330443X.2015.1029338
- Mills, J. F. (2017). Violence risk assessment: A brief review, current issues and future directions. Canadian Psychology, 58, 40–49. doi:10.1037/cap0000100
- Mills, J. F., & Gray, A. L. (2013). Two-Tiered Violence Risk Estimates: A validation study of an integrated-actuarial risk assessment instrument. Psychological Services, 10, 361–371. doi:10.1037/a0032608
- Mills, J. F., Kroner, D. G., & Morgan, R. D. (2011). Clinician’s guide to violence risk assessment. New York, NY: Guilford Press.