Classifying Inmates for Strategic Programming

• NCJ Number: 188891
• Author(s): Eileen Sullivan Ph.D.; Carmen Cirincione Ph.D.; Katherine Nelson Ph.D.; Jennifer Wallis M.A.
• Date Published: March 2001
• Length: 70 pages
• Grant Number(s): 98-CE-VX-0010
• Sponsoring Agency: National Institute of Justice (NIJ)
• Corporate Author: Vera Institute of Justice
• Sale Source: Vera Institute of Justice; NCJRS Photocopy Services
• Publication Format: Document
• Publication Type: Report (Study/Research)
• Language: English
• Country: United States of America
• Note: Dataset may be archived by the NIJ Data Resources Program at the National Archive of Criminal Justice Data

Annotation

This paper reports on a study designed to improve the New York City Department of Corrections' (DOC) methods of selecting jail inmates appropriate for drug treatment.

Abstract

Researchers designed and tested statistical models that could predict inmates' approximate length of stay in jail. Specifically, the DOC wanted to identify inmates headed either for prison or the community who could complete a treatment program of 45 days or more. Researchers constructed two models by using logistic regression and survival analysis and applied them to two sets of data. The first set contained data from DOC on a group of inmates eligible for admission to the treatment program, including demographic characteristics and information on the current case. The second set included the same records from DOC supplemented by data from other criminal justice agencies such as additional criminal history and community ties. Out of 100 new inmates entering jail, the researchers could correctly predict the outcome -- prison or community -- about 60 percent of the time and community-bound inmates about 90 percent of the time. Once researchers had determined outcome, they found that the vast majority of prison-bound inmates, nearly 85 percent, remained in jail at least 51 days. These inmates were ideal candidates for a long-term therapeutic program. Among inmates bound for the community, the researchers were able to predict which people were likely to stay in jail for at least 37 days and who could gain more from a concentrated program that helped them connect with resources in the community. Neither statistical model, however, could predict which community-bound inmates would stay for certain periods beyond 37 days and who could complete a 2-month treatment program. Despite this uncertainty, DOC officials can still effectively distinguish prison and community-bound inmates and match them with the most appropriate version of its drug treatment programs. To offset any imprecision associated with the length of stay of particular community-bound inmates, the researchers suggest that DOC modify the structure of its treatment curriculum. Overall, jail-based treatment programs can adapt to their changeable environment by creating self-contained curricular modules that can stand independently. 6 tables, 31 references, and appended supplementary data