Skip to main content
SearchLoginLogin or Signup

The Limits of Predicting Near Lethal and Lethal Family and Intimate Partner Violence

Published onJul 09, 2024
The Limits of Predicting Near Lethal and Lethal Family and Intimate Partner Violence


Accurate assessment of the risk of fatal family or intimate partner violence (FIPV), particularly of intimate partner homicide (IPH), has long been a clinical and research goal.  Although several risk assessments have been developed to assist in this task, limited validation methodologies in past research mean that it remains unclear whether fatal FIPV can be accurately predicted. This study aimed to address this research gap using a prospective methodology. Prediction of lethal and near lethal FIPV was tested in a full population cohort (n = 38,454) of unique perpetrator and victim dyads recorded by police over 7 months between September 2019 and March 2020 in the Australian state of Victoria. The predictive validity of an actuarial risk assessment instrument (ARAI) previously validated for general FIPV risk and of 12 widely recognized risk factors for male-to-female IPH were assessed over a 12-month follow-up period. Neither the ARAI, individual IPH risk factors, nor a model developed from the IPH risk factors, accurately predicted which dyads subsequently came to police attention for lethal or near lethal FIPV. All methods misclassified many thousands of dyads as high risk who did not have subsequently reported lethal/near lethal FIPV. These findings were attributable to both the low base rate of lethal/near lethal FIPV during the follow-up period (a universal phenomenon), and the commonality of widely recognized risk markers for male-to-female IPH in FIPV cases reported to police. These results indicate that when evaluated using a prospective design, existing assessments of FIPV homicide risk are likely to overestimate the risk of homicide in a very substantial number of cases. This suggests that a public health approach to the prevention of FIPV homicide may be of greater utility than trying to accurately identify cases at high risk of a low base rate event such as homicide.

No comments here
Why not start the discussion?