The Effect of Missing Item Data on the Relative Predictive Accuracy of Intimate Partner Violence Risk Assessment...

Area of Science:

• Forensic Psychology
• Criminology
• Risk Assessment

Background:

• Correctional risk assessment tools are crucial for predicting recidivism.
• Missing data can potentially bias risk assessment outcomes.
• Previous research suggests some tools are robust to missing data.

Purpose of the Study:

• To evaluate the impact of varying missing data percentages on the predictive accuracy of the ODARA and SARA-V2.
• To compare the effectiveness of different data imputation methods (summation, proration, multiple imputation) under missing data conditions.
• To extend previous findings on risk assessment tools and missing data to intimate partner violence samples.

Main Methods:

• A high-risk intimate partner violence (IPV) sample (N=267) was used.
• Simulated missing data at 1%-50% levels were applied to the ODARA and SARA-V2.
• Relative and absolute predictive accuracy were assessed across different data handling methods.

Main Results:

• Missing data and imputation methods did not significantly affect the relative predictive accuracy of the ODARA and SARA-V2.
• Summation methods consistently underestimated absolute risk scores.
• Proration and multiple imputation yielded similar relative predictive validity.

Conclusions:

• The ODARA and SARA-V2 demonstrate robustness to missing data in terms of relative risk prediction.
• Proration or multiple imputation are recommended for managing missing items in IPV risk assessment research.
• Proration is suggested as a more practical approach than summed totals for IPV risk assessment in clinical practice.