Inteligência artificial na vigilância da violência por parceiro íntimo contra mulher a partir de dados administrativos de saúde: revisão de escopo

Leonardo Naves dos Reis; Antônia Mirely Inocêncio da Silva; Igor de Oliveira Reis; Carla Aparecida Arena Ventura

doi:10.1590/SciELOPreprints.16137

##article.authors##

Leonardo Naves dos Reis Universidade do Estado do Amazonas https://orcid.org/0000-0001-6846-7967
- Conceptualization
- Data Curation
- Formal Analysis
- Investigation
- Methodology
- Project Administration
- Supervision
- Writing – Original Draft Preparation
- Writing – Review & Editing
Antônia Mirely Inocêncio da Silva Universidade do Estado do Amazonas https://orcid.org/0009-0004-2581-3814
- Formal Analysis
- Investigation
- Writing – Review & Editing
Igor de Oliveira Reis University of São Paulo https://orcid.org/0000-0002-9834-5538
- Writing – Review & Editing
- Investigation
Carla Aparecida Arena Ventura University of São Paulo https://orcid.org/0000-0003-0379-913X
- Writing – Review & Editing

DOI:

https://doi.org/10.1590/SciELOPreprints.16137

Keywords:

Intimate partner violence, Artificial intelligence, Health surveillance, Administrative data, Health information systems, Scoping review

Abstract

Objective: To map the applications of artificial intelligence and data science in health surveillance of intimate partner violence (IPV) against women using administrative data, in studies published between 2015 and 2026, with a specific analysis of Brazilian production.

Method: Scoping review conducted according to the Joanna Briggs Institute (JBI) methodology and reported following PRISMA-ScR guidelines. The protocol was registered a priori at the Open Science Framework (DOI: 10.17605/OSF.IO/SGEK3). Searches were performed in May 2026 in the following databases: PubMed, Scopus, Web of Science, IEEE Xplore, ACM Digital Library, LILACS, and SciELO. The search strategy was calibrated through a pilot test, with verified retrieval of pre-defined anchor studies. A two-phase screening was conducted by two independent reviewers in blinded mode, with conflict resolution by consensus.

Partial results: 240 records were retrieved, of which 167 remained unique after deduplication. The title-and-abstract screening (Phase 1) resulted in 133 exclusions and 34 records forwarded to full-text reading (Phase 2). Inter-reviewer agreement was almost perfect (Cohen's kappa = 0.93). The distribution of the literature shows a concentration in biomedical databases (PubMed, Scopus, Web of Science), with substantially scarce indexed Latin American production (1 record in LILACS, 2 in SciELO).

Considerations: This is version 1.0 of the preprint, encompassing the registered protocol, the executed search strategy, and the results of Phase 1 screening. Subsequent versions will incorporate the full-text screening results, data extraction, and final synthesis, following the protocol's timeline.