2022 Review and Revalidation of the First Step Act Risk Assessment Tool

• NCJ Number: 305720
• Author(s): Ryan Labrecque
• Date Published: March 2023
• Length: 45 pages
• Agencies: NIJ
• Publication Type: Report (Summary)

Annotation

This document is the third annual review and revalidation report on the Prisoner Assessment Tool Targeting Estimated Risk and Needs, evaluating for predictive accuracy, dynamic validity, and racial and ethnic neutrality, as mandated by the First Step Act of 2018.

Abstract

This review and revalidation report analyzes the Prisoner Assessment Tool Targeting Estimated Risk and Needs (PATTERN), according to the mandate of the First Step Act of 2018 (FSA). The report evaluates PATTERN for its predictive accuracy, dynamic validity, and racial and ethnic neutrality. It also expands upon prior analyses by including one-, two-, and three-year recidivism outcomes, assessing the proportion of change in risk scores and levels that is influenced by a person’s current age, and provides additional descriptive information on individual factors, risk scores and levels, and outcomes by race and ethnic group. The report also provides an update on actions taken by the National Institute of Justice (NIJ) and Department of Justice (DOJ) in the past year as well as ongoing efforts for review and improvement of PATTERN. The Fiscal Year (FY) 2018 cohort study findings continue to show that PATTERN is a strong and valid predictor of general and violent recidivism, with Area Under the Curve (AUC) statistics ranging from 0.76 to 0.78, and it predicted both general and violent recidivism at the three follow-up periods. Study findings continue to suggest that individuals can change their risk scores and levels during confinement; changes in risk were not driven exclusively by changes in age; and while PATTERN is predictively accurate across the five racial and ethnic groups analyzed, evidence suggests that the instrument overpredicts the risk of recidivism for some groups. Some efforts to refine and improve PATTERN include considerations of how recidivism is defined, which data inputs are used for scoring, and whether modeling changes might reduce racial and ethnic disparities in the tool.