Elsevier

Journal of Criminal Justice

Volume 34, Issue 3, May–June 2006, Pages 331-337
Journal of Criminal Justice

Exploring the validity of the Level of Service Inventory-Revised with Native American offenders

https://doi.org/10.1016/j.jcrimjus.2006.03.009Get rights and content

Abstract

The use of actuarial risk/need assessment tools is an increasingly important part of the correctional landscape. Actuarial tools ideally will provide a valid, dynamic assessment of an offender's overall risk/need level, and will identify their most prevalent criminogenic needs. What results is typically a number or score that can be used to assign an offender to a risk level that is associated with an assumed likelihood of recidivism. Testing the predictive validity of actuarial risk/need assessment tools is of paramount concern, particularly when they are utilized with new (and under-researched) populations. The current study assessed the predictive validity of the Level of Service Inventory-Revised using a sample of Native American and White offenders in a northern midwestern state. Results showed the instrument to have modest predictive validity utilizing the entire sample of offenders, with varying results for subsequent subgroups.

Introduction

Dynamic risk/need assessment processes have become an increasingly important part of virtually any correctional intervention. In previous decades, the most common method by which a correctional professional would get to know an offender would be through an unstructured (or semi-structured) interview, that may or may not involve information pertaining to relevant criminogenic factors. This ‘first generation’ of offender assessment is largely based on the prior experience and qualitative observations of the correctional professional. This method also possesses a high potential for bias to enter into decision-making (Bonta, 1996, Clear and Gallagher, 1985).

The practice of offender classification has undergone many changes via advances in research identifying the best predictors of antisocial behavior. Specifically, since the ‘first generation’ method of classification referenced above, correctional practice has involved the use of many different actuarial-based instruments involving both static (historic/unchanging) and dynamic (current/changeable) factors. Most of these tools provide a summary score that dictates a certain level of supervision or intervention (see for example the salient factors score, or the Wisconsin risk/need-CMC) (Baird et al., 1979, Hoffman, 1994).

Perhaps of most recent note, the Level of Service Inventory (LSI-R) has been used with increasing frequency over the last ten years (Jones, Johnson, Latessa, & Travis, 1999). Developed by Andrews and Bonta (1995a), the LSI-R is a fifty-four-item assessment tool that results in a composite score indicating an offender's overall level of risk, and criminogenic need. The composite score is rendered via the assessment of ten ‘domains,’ most of which are assessed in a dynamic way-meaning, as the domain exists in the offender's life (e.g., over the last twelve-month period the offender was in the community) (Andrews and Bonta, 1995b, Bonta and Motiuk, 1985). The ten domains covered via the LSI-R include: criminal history, education/employment, financial, family/marital circumstances, accommodations, leisure/recreation activities, companions, alcohol/drug use, emotional/personal, and attitudes/orientations (criminogenic thinking). In theory, assessing an offender using the LSI-R allows for a valid classification (e.g., high, medium, or low-risk), as well as the identification of the most prevalent criminogenic needs that may facilitate case planning and treatment intervention (Andrews & Bonta, 2001).

The predictive validity of the LSI-R had been demonstrated within several different correctional settings (Andrews, 1982, Andrews and Robinson, 1984, Bonta and Andrews, 1993, Bonta and Motiuk, 1985, Gendreau et al., 2002). In addition, there was evidence that the LSI-R had predictive validity for various subgroups of the offender population such as female offenders, and African-American offenders (Coulson et al., 1996, Lowenkamp et al., 2001, Lowenkamp and Latessa, 2002). Existing evidence notwithstanding, it is generally recommended that norming and validation processes take place whenever the instrument is implemented in new jurisdictions. Norming and validation processes become particularly important when a jurisdiction may have a higher than average representation of a particular subgroup of the general population.

Holsinger, Lowenkamp, and Latessa (2003) conducted a norming analysis using a sample of offenders who came from a northern midwestern state with a substantial representation of Native Americans. Significant differences between Native American and non-Native American offenders were revealed. The differences occurred not only regarding the overall composite risk score provided by the LSI-R (Native Americans scored significantly higher on the total score), but within several of the ten domains. At the time this norming study was published, the only data available were the initial LSI-R assessment scores, and basic demographic information. In addition, not enough time had elapsed to allow for a meaningful test of the tool's predictive validity. The current study incorporated follow-up data for the sample used in Holsinger et al. (2003). As such, the predictive validity of the LSI-R was tested for the entire sample, as well as several subgroups contained within the sample (specifically Native Americans, non-Native Americans, male offenders, and female offenders).

Section snippets

Methods

The following section defines the participants included in the sample, the procedures for data collection and sample selection, and a review of the measures utilized.

Results

Table 1 presents the results for correlation analyses between the LSI-R composite score and any new arrest while at risk in the community. Overall, for the entire sample a correlation of .18 (N = 403) was observed. While the relationship was not of a great magnitude regarding strength, it was in the expected (positive) direction, and was statistically significant. The 95 percent confidence intervals are reported for each correlation as well. For all the cases in the analysis, the 95 percent

Discussion

The analyses above offered support for the LSI-R as a composite tool on the entire sample of offenders, but mixed support when the results were disaggregated by certain subgroups of the sample. In terms of predictive validity, the instrument appeared to show the most promise for White offenders overall, males offenders overall, and White males and White females in particular.

Using any new arrest as the only measure of recidivism may provide one explanation for the non-intuitive results

References (25)

  • J. Bonta et al.

    The level of supervision inventory: An overview

    IARCA Journal

    (1993)
  • J. Bonta et al.

    Risk prediction and re-offending: Aboriginal and non-Aboriginal offenders

    Canadian Journal of Criminology

    (1997)
  • Cited by (72)

    • The Cultural Agency-Model of Criminal Behavior

      2021, Aggression and Violent Behavior
      Citation Excerpt :

      However, when it comes to cross-cultural transferability, these DRF appear to be less able to validly detect crime-related problems in Non-Euro-American contexts. Several studies (e.g., Bechtel et al., 2007; Dahle & Schmidt, 2014; Fass et al., 2008; Holsinger et al., 2006; Lowder et al., 2019; McGrath et al., 2018; Schlager & Simourd, 2007; Schmidt et al., 2018; Shepherd et al., 2015; Wormith et al., 2015) and meta-analyses (Gutierrez et al., 2013; Olver et al., 2009, 2014) have found inferior predictive validity for culturally diverse samples compared to the samples within which these factors were initially identified. For example, Schlager and Simourd (2007) found no LSI-R scale to be able to differentiate between recidivists and non-recidivists within a 2-year follow up among Hispanic individuals who offend and only the Family/Marital and Educational subscale to be able to do so among African Americans.

    • Exposure to fetal testosterone, aggression, and violent behavior: A meta-analysis of the 2D:4D digit ratio

      2017, Aggression and Violent Behavior
      Citation Excerpt :

      Indeed, virtually all of the complex risk assessment instruments currently in use, from the Minnesota Multiphasic Personality Inventory (MMPI), to the Psychopathy Checklist-Revised (PCL-R), to the Level of Service Inventory-Revised (LSI-R), each require hours of training to learn how to use, and hours to ultimately administer (Neumann, Johansson, & Hare, 2013; Prince & Butters, 2014; Ready & Veague, 2014). Even then, their predictive capacity is still limited, where their scores correlate reasonably well for some groups of offenders (e.g., white males, relatively non-serious offenders), yet their accuracy tends to break down when it comes to members of racial/ethnic minority groups (e.g., African-Americans, Hispanics, and Native Americans) and women (Hanson & Morton-Bourgon, 2009; Holsinger, Lowenkamp, & Latessa, 2006; Holtfreter, Reisig, & Morash, 2004; Schlager & Simourd, 2007). Collectively, these findings point to a rather sobering conclusion: that predicting antisocial behavior (including aggression and violence) appears to be sufficiently complex that even when we know a lot of things about people, we are still going to be wrong a lot more often than we would like to be.

    • Exploring the relationship between criminogenic risk assessment and mental health court program completion

      2016, International Journal of Law and Psychiatry
      Citation Excerpt :

      The process of assessing criminogenic risk has evolved from an examination based strictly on professional judgment and opinion to standard tools used to assess static and dynamic areas of criminogenic risk and guide service delivery (Andrews, Bonta & Wormith, 2006; Gendreau, et al., 1996). The Level of Services Inventory—Revised (LSI-R) is a widely used criminogenic risk/need classification instrument with demonstrated reliability and validity (Andrews & Bonta, 2000; Bonta, 2002; Holsinger, Lowenkamp & Latessa, 2006). The LSI-R is not designed to guide delivery of services as the revised Level of Service/Case Management Inventory (LS/CMI), does.

    View all citing articles on Scopus
    View full text