This project, initiated in 2024 and carried out in partnership with the Paraná Secretariat of Public Security (SESP-PR) and the Inter-American Development Bank (IDB), developed a predictive model to assess the risk of revictimization among women in domestic violence cases in Paraná. The tool was designed to classify cases according to the likelihood of subsequent victimization, providing insights to support the targeting of public policies aimed at preventing revictimization and femicide.

Following the development and technical validation of the algorithm, the project will move into a new phase focused on piloting its implementation, funded by the Crime and Violence Initiative (CVI) at J-PAL. Scheduled to begin in 2026, this phase will assess the adoption of the algorithm in operational settings, its use by public officials, and its effects on resource allocation and protocols for assisting victims of domestic violence.

The results showed that the tool can make follow-up visit allocation more efficient by directing available capacity toward victims with a higher likelihood of experiencing another episode of violence. In a simulation involving the 100 cases classified as highest risk, the model identified more than twice as many victims who were subsequently revictimized compared with a random selection. The findings indicate that data already generated by public security institutions can support more objective decision-making and improve the targeting of protection policies, particularly in contexts where the number of cases exceeds available service capacity.

The development of the model also involved integrating more than 18 million police report records with other databases relevant to the context of domestic violence. This data linkage helped identify gaps and inconsistencies in administrative records and generated recommendations for improving the completion of police reports, some of which have already been adopted by the Paraná Secretariat of Public Security.