A simple metric for predicting repeated intimate partner violence harm based on the level of harm of the index offence (… as long as a non-linear statistic is applied)

Abstract

PurposeA recent body of evidence investigated repeated intimate partner violence (IPV) using crime harm indices (the severity of victimisation), instead of crime counts (the number of additional victimisation incidents). Yet, the predictive utility of harm scores in IPV remains unclear – except that high-harm IPV is not usually followed by any additional IPV incidents. The authors take cases of repeat IPV from North Zealand Police, Denmark, to predict subsequent IPV harm and counts based on the level of harm of the first reported IPV offence.Design/methodology/approachUsing the Danish crime harm index (CHI) to estimate harm levels, non-linear regression models are applied (due to the non-linear nature of the data) to show that the CHI level of the index offence validly predicts gains in future CHI but does not predict IPV counts.FindingsThe findings suggest that whilst high-harm IPV is a rare event and repeat high-harm IPV even rarer, when they do occur, escalation in harm is likely to occur.Practical implicationsA simple metric of harm of the first reported IPV offence can validly predict future harm – however, scholars should apply more fitting analytical techniques than crude descriptive statistics, which fail to take into account the non-linear distribution of police records.Originality/valueThis is the first study to show the value of predicting future harm based on prior harm in IPV.

Overcoming unreported violence using place‐based ambulance data: The case for mapping hotspots based on health data for crime prevention initiatives

“…Police‐recorded data constitute the predominant method for analyzing crime, an approach limited by the willingness of members of the public to disclose their victimization (see Clark et al, 2022; McKee et al, 2022) or the wiliness of the police to focus more intently on victims (Lay et al, 2023). Consequently, supplementary or complementary datasets are necessary (see discussion in Ariel, 2023; and Loewenstein et al, 2023). With the increased availability and utilization of health data, police forces can develop interagency data sharing in order to comprehend where and when violence occurs in their communities (Sutherland et al, 2021).…”

Using machine learning to forecast domestic homicide via police data and super learning

“…Indeed, as far back as 1928, practitioners have been using tools to forecast events like recidivism, yet not all of these tools were based on data; many were based on intuition, and as a result, they were susceptible to poor performance. Statistical tools were later developed 41 , yet these tools were limited to the confines of the generalized linear models they inherited from, and thus, their predictive accuracy suffered 42 . When viewed in this light, machine learning tools are simply the next iteration of a century-long tradition of using tools to forecast criminal-justice-related events.…”