Risk assessment of revictimization in cases of domestic violence against women
Resumo
Risk assessment tools for domestic violence against women show limited predictive accuracy when grounded only in structured professional judgment. In Brazil, the National Risk Assessment Form (FONAR) lacks statistically grounded weighting, constraining its effectiveness. This study evaluates FONAR’s predictive capacity and tests whether judicial-data enrichment improves revictimization risk estimation. Administrative and judicial records were integrated through probabilistic linkage, yielding 8,694 cases (1,429 revictimizations) from 10,069 FONAR forms collected between March 2022 and September 2024. Propensity score matching (1:1, nearest-neighbour with caliper on the logit) estimates an effect of protective measures on revictimization of 0.1246 (ATC; 95% CI [0.1089, 0.1412]) and 0.1226 (ATT; 95% CI [0.1064, 0.1399]), consistent with selection of higher-risk cases for treatment. Supervised models were tested under two configurations: FONAR-only and enriched. XGBoost performed best (AUC = 0.6139, recall = 0.8027), outperforming FONAR-only (AUC = 0.5709). Protective-measure indicators and prior judicial proceedings dominate predictions, acting as proxies for case severity rather than causal drivers. The study contributes an empirical assessment of FONAR, evidence that data integration improves risk modelling, and a pipeline to support judicial decision-making.
Referências
Allard, V., Higgs, T., and Slight, M. Psychometric properties of the spousal assault risk assessment from samples of people having perpetrated intimate partner violence. Trauma, Violence, & Abuse 25 (5): 3443–3460, 2024.
Austin, P. C. Optimal caliper widths for propensity-score matching when estimating differences in means and differences in proportions in observational studies. Pharmaceutical Statistics 10 (2): 150–161, 2011.
Berk, R., Heidari, H., Jabbari, S., Kearns, M., and Roth, A. Fairness in Criminal Justice Risk Assessments: The State of the Art. Sociological Methods & Research 50 (1): 3–44, February, 2021.
Caulfield, K. M., Gray, N. S., Edwards, A., and Snowden, R. J. The domestic abuse, stalking and harassment and honour-based violence (dash) risk assessment instrument in predicting deadly or persistent domestic abuse. Forensic Sciences 5 (4): 272–284, 2025.
da Silva, T. D. GESTÃO DE CONFLITOS E DE ALTERNATIVAS PENAIS NO COMBATE À VIOLÊNCIA DOMÉSTICA E FAMILIAR CONTRA A MULHER: O PROGRAMA PROTEÇÃO NA MEDIDA. M.S. thesis, Universidade de Fortaleza (UNIFOR), Fortaleza, Ceará, Brasil, 2021.
Gray, A. L., Mills, J. F., and Forth, A. E. Risk for violent and ipv recidivism among incarcerated men with a history of ipv perpetration: An examination of the predictive validity of the odara, dvrag, and pcl-r. Criminal Justice and Behavior 52 (1): 79–100, 2025.
Grogger, J., Gupta, S., Ivandic, R., and Kirchmaier, T. Comparing conventional and machine-learning approaches to risk assessment in domestic abuse cases. Journal of Empirical Legal Studies 18 (1): 90–130, 2021.
Gupta, A. Xgboost versus random forest. Disponível em: [link]. Acesso em: 31 de dezembro de 2022, 2021.
Júnior, A. P. and Ribeiro, F. Análises de dados de violência doméstica com o uso de aprendizagem de máquina: um mapeamento sistemático. In Anais da VII Escola Regional de Computação do Ceará, Maranhão e Piauí. SBC, Porto Alegre, RS, Brasil, pp. 143–150, 2019.
Lisboa, M., Teixeira, A. L., and Pasinato, W. Formulário de Risk Assessment para o CNVD: um instrumento para o enfrentamento da violência doméstica contra a mulher. [link], Brasil, 2019. info:eurepo/grantAgreement/FCT/5876/147304/PT#UID/SOC/04647/2019.
Lundberg, S. and Lee, S.-I. A unified approach to interpreting model predictions, 2017.
Myhill, A., Hohl, K., and Johnson, K. The ’officer effect’ in risk assessment for domestic abuse: Findings from a mixed methods study in england and wales. European Journal of Criminology 20 (3): 856–877, 2023.
Pham, A. T., Nunes, K. L., Hilton, N. Z., Ennis, L., and Jung, S. The factor structure of intimate partner violence risk. Criminal Justice and Behavior 53 (1): 101–118, 2026.
Rajendran, S. and Sundararajan, P. Predicting criminal recidivism using specialized feature engineering and xgboost. CrimRxiv n/a (n/a): –, 2021.
Rosenbaum, P. R. and Rubin, D. B. The central role of the propensity score in observational studies for causal effects. Biometrika 70 (1): 41–55, 1983.
Ryland, H., Burghart, M., Zhong, S., Fazel, S., and Yu, R. Risk assessment tools used at the policing stage for health and crime outcomes: A systematic review and meta-analysis. Social Science & Medicine vol. 383, pp. 118457, 2025.
Stuart, E. A. Matching methods for causal inference: A review and a guide. Statistical Science 25 (1): 1–21, 2010.
Turner, E., Brown, G., and Medina, J. Predicting domestic abuse (fairly) and police risk assessment. CrimRxiv vol. 31, pp. 145–157, feb 25, 2021. [link].
Turner, E., Medina, J., and Brown, G. Dashing hopes? the predictive accuracy of domestic abuse risk assessment by police. The British Journal of Criminology vol. 59, pp. 1013–1034, 08, 2019.
Viljoen, J. L., Goossens, I., Monjazeb, S., Cochrane, D. M., Vargen, L. M., Jonnson, M. R., Blanchard, A. J. E., Li, S. M. Y., and Jackson, J. R. Are risk assessment tools more accurate than unstructured judgments in predicting violent, any, and sexual offending? A meta-analysis of direct comparison studies. Behavioral Sciences & the Law 43 (1): 75–113, 2025.
Weber, M. A., Schnyder, N., Kirschstein, M. A., Graf, M., Endrass, J., and Rossegger, A. The key role of base rates: systematic review and meta-analysis of the predictive value of four risk assessment instruments. Swiss Medical Weekly 155 (40166): n/a, 2025.
