Development of an Empirically-Based Risk Assessment Instrument, Final Report

• NCJ Number: 222802
• Author(s): Laura Winterfield; Mark Coggeshall; Adele Harrell
• Date Published: April 2003
• Length: 81 pages
• Sponsoring Agency: District of Columbia Pretrial Services Agency
• Corporate Author: The Urban Institute
• Sale Source: The Urban Institute
• Publication Format: Document
• Publication Type: Program/Project Description
• Language: English
• Country: United States of America
• Note: Downloaded May 22, 2008

Annotation

This is the final report on the second and final phase of the development of a risk-assessment instrument for the District of Columbia Pretrial Services Agency to use in recommending conditions for defendants' pretrial release.

Abstract

The analysis of instrument performance found that its overall accuracy in predicting failure under pretrial release reached a maximum of approximately 80 percent on both the Appearance Scale (risk of failing to appear for trial) and the Safety Risk Scale (risk of rearrest while on pretrial release). These results indicate that the instrument can be used to assist pretrial-release decisionmaking by standardizing the process; however, because it could only use existing information, it does only a fair job of prediction. This report recommends that a prospective validation of the instrument be conducted by implementing it on a trial basis and reanalyzing the validity of the current set of predictors as well as any additional predictors collected by the new computer system, Pretrial Real-time Information System Manager (PRISM). The Risk Prediction Instrument is composed of 22 items divided into 2 subscales, the Safety Risk Scale and the Appearance Risk Scale. The instrument is designed to predict two outcomes, the risk of failure-to-appear for trial (indicated by issuance of a bench warrant for failure-to-appear) and risk of rearrest, which includes either a new arrest record or a citation. The data include information about the criminal history, demographics, health, employment, and drug use of defendants. Scores on the two subscales are based on weights developed to maximize the correct prediction of risk. The instrument is being submitted by the researchers in the form of a Microsoft Excel spreadsheet that can be used to compute risk scores based on answers to the questions input into the appropriate cells. 18 tables, 16 figures, 10 references, and appended information on data processing, construction of variables, statistical models, and assessment of model performance