ACT Policing will narrow the focus of the tool it uses to assess the risk of repeat family violence in a bid to improve its accuracy, following a review that found the odds of the tool predicting such incidents were only "marginally better than chance".
An Australian Institute of Criminology review of the Family Violence Risk Assessment Tool described it as "not a strong predictor" of repeat domestic violence.
The tool was also found to be less effective than most other Australian and international tools, though many of those have been in place longer and have been adjusted following initial evaluations.
Since March 2017, ACT Policing officers have been required to use the Family Violence Risk Assessment Tool when responding to domestic violence incidents.
The tool produces a score reflecting the likelihood of repeat family violence, based on police officers' responses to 37 different items that include the offender's criminal history, the history of violence in a relationship, and any mental health issues.
Perpetrators who score between 0 and 13 are determined to be low-risk in terms of potential reoffending, while those who score between 14 and 27 are classified as medium-risk. Offenders who score 28 or above are considered high-risk. Police officers can revise ratings based on their own analysis of the information provided.
"ACT Policing mandates that officers prioritise cases based on their risk rating, and proactively implement immediate measures to ensure the safety of victims in higher risk cases," the Institute of Criminology review said.
As part of their study, which was commissioned by ACT Policing, the institute's principal research analyst Dr Christopher Dowling and research manager Anthony Morgan analysed 350 family violence incidents in which police used the tool between March and December 2017.
They then considered any repeat family violence offences reported within six months of the initial report.
They found that ACT Policing's tool was "not a strong predictor" of repeat family violence.
In the incidents examined, repeat domestic violence occurred in only 19 per cent of cases where perpetrators were rated high-risk.
"False positives, where cases are incorrectly categorised as high-risk, have significant resource implications for police and other agencies," the criminologists said in their review.
In 40 per cent of the cases where there was further domestic violence, the offenders had been classified as low-risk.
The criminologists said false negatives - cases classified as low-risk where repeat domestic violence did occur - meant a victim may not be protected from further harm.
While police officers in the ACT were able to revise risk ratings based on their professional judgment, the review said "it became apparent through this wave of data collection that this was not occurring".
The results of the review showed some factors like alcohol or drug use and psychological health issues to not be significant predictors of short-term repeat domestic violence.
Dr Dowling and Mr Morgan discovered that refining the Family Violence Risk Assessment Tool and using just 10 of its most important items, rather than all 37, produced "markedly more accurate" predictions that were comparable to those of the better performing domestic violence risk assessment tools used by other police forces.
An ACT Policing spokeswoman said since the review, the Family Violence Risk Assessment Tool had been amended based on the findings.
"The revised version is expected to be rolled out across ACT Policing this year," she said.
"The tool now consists of 10 individual assessment items which have [been] shown to improve the predictive nature of the tool.
"The new tool will be re-evaluated again in 12 months."
The 10 assessment items include whether the offender has assaulted the victim during the most recent incident or in the past, and whether the level of violence in the relationship has escalated.
Some of the other important considerations covered by the revised tool include whether there is a pregnancy or baby in the relationship, whether the offender and victim have recently separated, and whether the offender was experiencing financial problems.