This month, Greg was finally released from prison. In fact, many people in his county were released from prison this month, some of them under parole supervision. Unfortunately, some of the released individuals will commit at least one crime in the coming several years. Many people relapse into criminal behaviour because the factors that initially facilitated their crimes often persist. Additionally, it can be quite challenging to reintegrate into society after prison.

Although three years later some former prisoners will reoffend and end up in prison again, the majority will desist from crime. Policymakers and practitioners would like to know in advance who among released prisoners is likely to reoffend, so they can keep a closer eye on them and provide additional support if necessary, such as access to substance abuse treatment programmes or vocational training.

But how can they do it, and how can they do it effectively?

The main way to predict potential future reoffending is through the use of some form of risk assessment, such as standardised risk assessment instruments. These instruments evaluate various risk factors and protective factors that comprise an individual’s characteristics and circumstances.

You can watch the video accompanying this text on YouTube ->

Imagine that we implemented a risk assessment instrument and measured the reoffending risk of Greg, using a risk score that ranges from 0 to 100. He scored somewhere around 30. Then we measured the risk in 10 more individuals, who each received different scores, ranging from 5 to 90.

We let these people go about their lives and checked on them after 3 years.

In general, yes. However, AUC, although an important metric, only presents half of the picture. To decide who we consider at high risk of reoffending, we need to choose a cutoff risk score. Individuals scoring at or above the chosen score would be considered high risk, while those scoring lower would be considered low risk.

In the hypothetical example on the right, if we set the cutoff score at around 60, we would correctly identify some reoffenders and desisters but also misidentify a few. We would get 7 correct and 4 wrong. If we set a lower cutoff score, we would get 6 correct and 5 wrong. Overall, in this hypothetical case, we would be more likely to be correct using the higher threshold. However, in real-life applications, the choice of the threshold depends on the intended application of risk assessment results and other societal factors external to risk assessment.

Overall, such properties of the risk assessment instrument as sensitivity and specificity depend on the choice of the cutoff. Practitioners and policymakers should pay closer attention to them and not focus too much on AUC, as most commonly used risk assessment tools have comparable AUC values.

If we want to prioritise the prevention of crime and provide intensive supervision to most potential reoffenders, we might consider a lower cutoff. Reoffenders would have a higher weight in our decision-making. On the other hand, if we want to prioritise less intrusive supervision, such as in a cohort of individuals who committed less dangerous crimes, and facilitate independence and reintegration, we might choose a higher threshold. This way, desisters will have a higher weight in our decision-making.