Betting Against the Odds: The Mysterious Case of the Clinical Override in Risk Assessment of Adult Convicted Offenders

• Published date: 01 July 2023
• DOI: http://doi.org/10.1177/0306624X211049181
• Author: Julien Frechette,Patrick Lussier
• Date: 01 July 2023
• Subject Matter: Articles

https://doi.org/10.1177/0306624X211049181

International Journal of

Offender Therapy and

Comparative Criminology

2023, Vol. 67(9) 887 –909

© The Author(s) 2021

Article reuse guidelines:

sagepub.com/journals-permissions

DOI: 10.1177/0306624X211049181

journals.sagepub.com/home/ijo

Article

Betting Against the Odds:

The Mysterious Case of

the Clinical Override in

Risk Assessment of Adult

Convicted Offenders

Julien Frechette1 and Patrick Lussier1,2

Abstract

Various tools were designed to guide practitioners in the risk assessment of offenders,

including the Level of Service and Case Management Inventory (LS/CMI). This

instrument is based on risk assessment principles prioritizing the actuarial approach to

clinical judgment. However, the tool’s architects allowed subjective judgment from the

practitioners—referred to as clinical override—to modify an offender’s risk category

under certain circumstances. Few studies, however, have examined these circumstances.

Therefore, the current study used decision tree analyses among a quasi-population

of Quebec offenders (n = 15,744) to identify whether there are offenders more likely

to be subjected to this discretion based on their characteristics. The results suggest

that, although the override is rare, it occurred under few specific combinations of

circumstances. More precisely, these findings propose that the utilization of the clinical

override stems from a perceived discrepancy between risk prediction and management.

Keywords

clinical override, risk assessment, corrections, decision tree algorithms, level of

service and case management inventory, violent offenders

In Canada, individuals sentenced to imprisonment or probation are systematically sub-

jected to a risk-and-needs assessment that fulfills several classification and prediction

objectives. One of the key objectives is to gather systematic information that will

1Université Laval, Quebec City, QC, Canada

2Centre International de Criminologie Comparée, Montreal, QC, Canada

Corresponding Author:

Patrick Lussier, School of Social Work and Criminology, Université Laval, Pavillon Charles-De Koninck,

1030 Avenue des Sciences-Humaines, Quebec City, QC G1V 0A6, Canada.

Email: patrick.lussier@tsc.ulaval.ca

1049181IJOXXX10.1177/0306624X211049181International Journal of Offender Therapy and Comparative CriminologyFrechette and Lussier

research-article2021

888 International Journal of Offender Therapy and Comparative Criminology 67(9)

inform the criminal justice practitioner (CJP) responsible for developing and propos-

ing intervention programs tailored to the offender’s risk and needs. In Canada, these

risk-and-needs assessments are generally aligned with the principles and guidelines

set by the Risk-Needs-Responsivity model (RNR; Andrews et al., 1990). This model

was specifically designed for corrections to guide the level of service offered to offend-

ers. To support CJPs, tools based on the RNR model and principles have been devel-

oped, such as the Level of Service and Case Management Inventory (LS/CMI;

Andrews et al., 2004). This instrument is based on risk assessment principles prioritiz-

ing the actuarial assessment approach. Despite the instrument’s strong actuarial com-

ponent, it also allows, to some extent, the CJP’s subjective judgment. This discretion,

known as the clinical override, grants CJPs the possibility to reconsider an offender’s

actuarial risk category in an upward or downward fashion. This form of discretionary

power—focus of the current investigation—is, therefore, a compromise between actu-

arial and clinical approaches to risk.

Literature Review

The clinical override is an important but largely understudied discretionary power

attributed to CJPs working in corrections. This feature is the product of several years

of research and debates about the relative superiority between actuarial-based (also,

mechanical, statistical) risk assessment and those based on human judgment (also,

clinical, unstructured clinical). While the results of meta-analyses comparing actuarial

and clinical approaches appeared to have dismissed the latter because of its low pre-

dictive validity (e.g., Hanson & Morton-Bourgon, 2009; Meehl, 1954), recent studies

have somewhat revived the debate. Indeed, despite its relative predictive superiority,

several years of research on actuarial risk assessment have highlighted several short-

comings related to prediction and treatment (e.g., Hart et al., 2007; Lussier et al.,

2011). The limitations of early actuarial assessments led to various innovations, such

as the inclusion of dynamic, potentially changeable risk factors (e.g., Gendreau et al.,

1996) as well as the inclusion of criminogenic dynamic risk factors that can be tar-

geted by treatment programs and interventions (e.g., Andrews et al., 2006). These

innovations were not enough for some researchers who proposed a more contrasting

alternative to the actuarial approach that was framed on structured clinical judgment

(also, professional judgment; e.g., Webster et al., 1997).

Proponents of the actuarial approach who also recognized the importance of clini-

cal judgment in risk assessment have proposed another alternative to common risk

assessment principles. Acknowledging the limitations of a strict actuarial and clini-

cally unstructured method, Andrews et al. (1990) proposed a somewhat middle-ground

approach by adding a clinical override component to an existing actuarial instrument,

the Level of Service Inventory-Revised (LSI-R; Andrews & Bonta, 1995). The LSI-R

eventually paved the way for the LS/CMI (Andrews et al., 2004), a fourth-generation

risk assessment tool endorsing the most recent advancements in research (at the time

of its development). Although the LS/CMI is an upgraded and modified version of the

LSI-R, the former still allows the use of the clinical override feature. This feature