The Pretrial Predictive Validity of the Public Safety Checklist (PSC) & Virginia Pretrial Risk Assessment Instrument (VPRAI) in Nine Oregon Counties

The Pretrial Predictive Validity of the Public Safety Checklist (PSC) & Virginia Pretrial Risk Assessment Instrument (VPRAI) in Nine Oregon Counties

Problem

The Problem: 

The study aimed to address the accuracy and validity of pretrial risk assessment tools (PSC and VPRAI) used in Oregon counties to predict failure to appear (FTA) and re-arrest rates during pretrial release.

Impact on the system and/or public: 

Pretrial detention practices directly impact community safety, fairness, and judicial efficiency. Validating these tools is critical to reduce racial/ethnic disparities and ensure just decision-making in release and detention.
 

Research Questions:

  1. How accurate are the pretrial risk tools in predicting failure to appear and re-arrest?
  2. How can these tools or protocols be improved for better predictive accuracy?

     

Method and Analysis

Program evaluated or gaps addressed: 

The study focused on evaluating the predictive accuracy of PSC and VPRAI, addressing gaps in tool validation and calibration across jurisdictions.
 

Data and Sample Size: 

  • Data were compiled from Oregon Judicial Department (OJD) and Law Enforcement Data System (LEDS).
  • Final sample: 24,438 cases from nine counties, representing diverse demographics.
     

Analysis Used:

  • Sensitivity and specificity analysis using Area Under the Curve (AUC) metrics to determine predictive validity.
  • Logistic regression models to control for external factors influencing outcomes.
  • Evaluation of racial/ethnic disparities in tool performance.
     

Outcome

Key Findings:

  • The PSC tools generally performed better than the non-validated VPRAI in predicting outcomes, particularly in identifying low-risk defendants.
  • VPRAI showed weak to poor performance in Oregon counties, likely due to uncalibrated use.
  • Variability in AUC values across counties highlighted the need for local calibration of risk tools.
  • Discrepancies in predictive accuracy for racial/ethnic groups indicate potential bias in the tools.
     

Implications or Recommendations: 

  • Validation and Calibration: Pretrial tools should be validated and recalibrated for local populations to improve accuracy and fairness.
  • Policy Alignment: Tools must align with judicial policies, including efforts to minimize racial and systemic biases.
  • Improved Data Integration: Enhance data systems to reduce sample loss and improve analysis reliability.
  • Periodic Review: Tools should be reviewed and revalidated regularly, as required by Oregon's Chief Justice Order.
  • Training and Awareness: Judicial and law enforcement personnel should receive training on the limitations and appropriate use of risk tools..

This research underscores the importance of data-driven, equitable pretrial decision-making processes and highlights the critical role of ongoing validation and recalibration of risk assessment tools.

Authors

Principal Investigator:
Christopher M. Campbell, Portland State University 

Co-Investigators:
Kris Henning, Ph.D., Portland State University
Arynn Infante, Ph.D., Portland State University
Ann Leymon, Ph.D., Portland State University
Mark (Harmon) Leymon, Ph.D., Portland State University
Mauri Matsuda, Ph.D., Portland State University
Brian Renauer, Ph.D., Portland State University
Curt Sobolewski, Ph.D., Portland State University
Katie Wuschke, Ph.D., Portland State University

 

Funding

Oregon Criminal Justice Commission, National Criminal Justice Reform Project, Arnold Ventures
 

Tags

Prosecution & Pretrial

 

Summary Report