Developing the risk of administrative segregation tool to predict admissions to segregation

This study found that it was possible to develop an actuarial tool with a high degree of predictive accuracy in identifying offenders on their risk for an administrative segregation placement. The ability to determine the likelihood of being placed in administrative segregation within the first two years of incarceration could allow the Correctional Service of Canada (CSC) to develop strategies to intervene to reduce that risk before a segregation placement is needed. Reducing segregation placements could decrease disruption to institutional routine, free beds for those inmates in greatest need, and reduce costs.

The scale that was developed includes six static risk factors and can be scored quickly, without requiring detailed file reviews or interviews with the offenders. The ease of scoring this scale could allow for the identification of the highest risk offenders early in the intake period.

A simple six-item scale (the Risk of Administrative Segregation Tool), comprised of information generally available at intake, is highly predictive of an administrative segregation placement of six days or more within the first two years of incarceration. The six predictive items are: age at admission, number of prior convictions, previous admission to administrative segregation in prior federal sentences, sentence length, criminal versatility in current convictions, and prior violence.

Analyses were conducted to ensure that these same six items predicted equally well for Aboriginal men, non-Aboriginal men, Aboriginal women, and non-Aboriginal women. The tool predicted placements in segregation for both major reasons: inmate safety or jeopardizing the safety of the institution. Including dynamic predictors of segregation risk did not result in meaningful improvement in predictive accuracy.

Use of segregation has been of concern to CSC and other jurisdictions. Developing an effective screening tool to identify at admission inmates at risk of being placed in segregation is an important step forward in being able to intervene to reduce the number of segregation placements.

Computerized institutional records of inmates admitted to CSC (N = 16,701) were randomly divided into a development and a validation sample. Over 400 different variables were evaluated using operational concerns (e.g., ease of scoring, availability of information) and statistical analyses (e.g., overall predictive accuracy, unique information above other items) to determine the lowest number of items that best predicted a segregation placement. Special consideration was paid to differences across gender and Aboriginal ancestry subgroups. Results were subsequently tested on the validation sample to ensure generalization.

Helmus, L., Johnson, S., & Harris, A. J. R. (2014). Developing the Risk of Administrative Segregation Tool (RAST) to Predict Admissions to Segregation (Research Report R-325). Ottawa, ON: Correctional Service of Canada.

To obtain a PDF version of the full report, or for other inquiries, please e-mail the Research Branch or contact us by phone at (613) 995-3975.

You can also visit the website for a full list of research publications.