A study warns that 80% of prisoners involved in violent incidents had no risk assessment

BarcelonaOf the 7,302 prisoners who were involved in a violent incident between 2017 and 2022 in Catalan prisons, 5,785 did not have a valid risk assessment at that time, either because it had not yet been carried out or because its validity, which is six months, had expired. This represents 79.22% of prisoners, according to the study. Institutional violence in Catalonia's prisons, The algorithm is based on all the incidents recorded by prison workers in Catalonia in the computerised penitentiary system and includes, among others, assaults, self-harm and disturbances of order. According to ARA, the Justice Department is working on an update to the RisCanvi prediction algorithm – a test to assess the risk of each inmate reoffending – to make it more accurate and with the intention of having it active this year.

"We see that there is a mismatch in the prediction," explains to ARA the professor of psychology at the University of Barcelona (UB) Antonio Andrés-Pueyo, one of the creators of RisCanvi and author of the study together with David Férez-Mangas. When comparing the classifications that had been made of the prisoners with their incident histories in the centre, they found that "many inmates who had been classified as low risk had committed violent acts, and some who were considered high risk had not demonstrated it", although in this second group Andrés-Pueyo points out that it cannot be confirmed that violence has prevented it from happening.

Given this mismatch in some predictions, the two academics propose to the Generalitat some improvements to have more accuracy in the deliberations of the RisCanvi system. The researchers emphasise that the configuration of what is considered a risk factor should be improved, and not focus only on modifying the cut-off point from where it is considered that there is a risk. As for the dates, it must be taken into account that the data on which the study is based are prior to the most serious violent incident experienced in Catalan prisons, not yet a year ago, when an inmate murdered a cook at the Mas d'Enric prison.

The Justice Department has explained to the ARA that in 2023 it commissioned an external audit of the RisCanvi algorithm, which concluded that "it has a robust predictive capacity, comparable to international instruments that work along the same lines." Government sources assure that the level of success of RisCanvi is between 75% and 80%, although they add that "it is a very precise and sensitive tool, which can sometimes give false negatives" due to this sensitivity. However, they indicate that it is constantly under review, and at the moment the department "is working on an update for a new calibration" of the algorithm with the expectation that it will be operational this year.

Two levels of evaluation

As for the volume of evaluations carried out with RisCanvi, the data provided by the department correspond to the month of December 2024, that is, at the end of the year in which the crime in Mas d'Enric marked the Catalan penitentiary system. Government data indicate that, of the 8,488 inmates who were in the penitentiary system at that time, 6,805 (80.1%) had undergone this evaluation. Of these, 12.9% had undergone a less exhaustive version of RisCanvi, called screening and which takes into account ten risk factors. The full version of RisCanvi, which by the end of the year had been carried out by 67.57% of prisoners in Catalonia according to data from the department, is more exhaustive and takes into account 43 different risk factors.

The study by Andrés-Pueyo and Férez-Mangas focuses on the complete RisCanvi evaluations. In fact, one of the reasons that Justice points out when looking for the reasons why almost 80% of prisoners involved in violent incidents did not have a risk classification is that they had not been given a complete evaluation. In the same sense, an educator from the penitentiary system – who wants to remain anonymous – explains to ARA that it is usual to do a screening During the first month of an inmate in prison, but if no risk element is detected and the inmate does not cause any incident, the full assessment can be postponed.

This is even more frequent among pretrial prisoners, taking into account that they may be released shortly after, adds the worker himself. He also indicates that, sometimes, a full assessment requires time to gather all the information to be entered into the system. For example, if the center does not have previous psychiatric information on the prisoner, obtaining a diagnosis may take time.

Another reason that Justice believes may explain why almost 6,000 prisoners involved in violent incidents between 2017 and 2022 did not have a risk classification is that in some cases the assessment had not yet been done because the inmate had just entered or was on remand. In addition, they indicate that the assessment is closed when a prisoner is released.