Implementing a Domestic Violence Revictimization Risk Algorithm: A Pilot Study

Domestic violence is a widespread and persistent challenge in Brazil, where nearly 40.7% of women aged 16 and older have experienced intimate partner violence. Revictimization is alarmingly common: 70% of victims report suffering multiple episodes (FBSP and Datafolha, 2025). Despite the presence of second-response programs, limited financial and human resources often prevent timely follow-up, especially for those at greatest risk. In Parana, where the state police operate a dedicated domestic violence response service, existing institutional partnerships offer a unique opportunity to pilot a risk-based targeting strategy. This study will test the feasibility and operational value of integrating a predictive risk algorithm into the workflow of specialized police units. 

The algorithm, scheduled to be completed in 2025, is trained on historical police records and generates a high-risk alert when case and victim characteristics indicate a heightened likelihood of revictimization. The tool will assist police officers in identifying high-risk victims, reducing the likelihood that these victims do not receive assistance from the police. Led by researchers from Leme, who have previously conducted successful experimental studies in public security with the Parana State Department of Public Safety, the pilot will be implemented in partnership with the state's police force, which has formally endorsed the intervention. 
The pilot will allow refinement of the implementation protocols before scaling to a full RCT. Planning activities to prepare for implementation in 2026 are already fully funded. In parallel, we are actively pursuing additional resources from the Parana State Department of Public Safety to support the full RCT.