Methodological considerations in the evaluation of offender interventions: the problem of attrition.

Attrition from offender interventions presents methodological problems when the effectiveness of the intervention is under evaluation. This article proposes a treatment-received (TR) design, which incorporates one-to-one matching on criminogenic variables. This design permits the comparison of completer, noncompleter, and nonstarter groups with their paired comparisons and, hence, allows a more accurate evaluation of program completion and noncompletion effects. A sample of 173 offenders, referred to one of two general offending behavior cognitive skills programs within community settings, was matched one-to-one by criminogenic variables to offenders from an appropriate comparison group. Reconviction analyses were undertaken using intention to treat (ITT) analysis, TR analysis using the full comparison group, and TR analysis using matched comparison groups. The ITT design found no impact of group membership on reconviction. The TR designs, however, provided evidence of moderate effects of completion: For every three completers, five of the matched comparison group members were reconvicted. The analysis also suggested a noncompletion effect: Noncompleters were twice as likely as their matched comparison group to be reconvicted.