Prioritizing Police Investigations of Intimate Partner Violence Using Actuarial Risk Assessment

Problem

The problem addressed: 

Police departments often use risk assessment tools to allocate resources for intimate partner violence (IPV) investigations, but these tools have limitations, including questionable predictive validity and resource-intensive data requirements. This study explored whether IPV recidivism could be predicted using variables already available in police records, eliminating the need for additional data collection.

General impact on the system and/or public: 

IPV represents a substantial portion of police workload and poses significant harm to victims and families. Efficient resource allocation for investigating high-risk cases is critical to mitigating harm, ensuring safety, and reducing IPV recurrence.
 

Research Questions Answered:

1. Can IPV recidivism be predicted using data from law enforcement records systems (RMS)?
2. What is the predictive value of victim characteristics and supplemental data collected during police interviews?
3. Can an effective, automated IPV risk assessment tool be developed using existing police records?
   
   
    

Method and Analysis

Program Evaluated or Gaps Addressed: 

The study evaluated existing gaps in risk assessment tools by focusing on predictive variables available in NIBRS-compliant RMS data and tested the necessity of supplemental victim interviews for improving predictive accuracy.
 

Data and Sample Size: 

The sample comprised 2,670 IPV cases reported in 2017 from Portland, Oregon. Data included offender and victim characteristics, criminal history, and incident-specific details.
 

Analysis Used:

The study employed logistic regression, cross-validation methods, and bivariate/multivariate analyses to identify significant predictors of IPV recidivism and create a simplified risk assessment tool. The predictive accuracy was tested using receiver operating characteristic (ROC) analyses (AUC values).

Outcome

Key Findings:

• Primary Predictors: Offender's recent and past criminal history, including domestic and non-domestic offenses, were strong predictors of recidivism. Victim characteristics (e.g., age, criminal history) had predictive value but added limited benefit to existing data.
• Supplemental Data: Supplemental victim-reported data (e.g., risk ratings, suspect unemployment, or substance abuse) did not significantly improve predictive accuracy beyond NIBRS data.
• Risk Tool Performance: A 10-item actuarial tool based on NIBRS data predicted IPV recidivism with an AUC of 0.72, demonstrating comparable or superior performance to existing tools.
  
   

Implications or Recommendations: 

1. Law enforcement agencies can prioritize investigations using an automated NIBRS-based risk tool, reducing the need for resource-intensive data collection and enhancing decision-making efficiency.
2. The findings highlight potential racial disparities in predictive accuracy, requiring further research to ensure equitable application across demographic groups.
3. Agencies should evaluate the cost-benefit of supplemental data collection and consider relying solely on existing records for IPV risk assessment.
4. Broader adoption of standardized data (e.g., NIBRS) could facilitate scalable, automated risk assessment tools across jurisdictions.

This study provides actionable insights into optimizing IPV investigation prioritization and improving law enforcement resource allocation.