• AI can detect and prevent violence against healthcare staff in real time. • Machine learning predicts aggression using clinical notes and behavior data. • NLP identifies verbal threats and abusive language with high accuracy. • Predictive AI models outperform traditional risk assessment methods in healthcare violence prevention. • AI tools improve staff safety, response times, and overall care quality. The rising incidence of physical, verbal, and psychological assaults on healthcare workers has become a critical occupational and public health concern. Such incidents negatively impact staff well-being, contribute to burnout, and compromise the overall quality of patient care. To explore how artificial intelligence (AI) techniques, including machine learning, deep learning, and natural language processing, can be applied in healthcare settings to detect, prevent, or mitigate workplace assaults against healthcare staff. Additionally, it sought to identify the most effective technical and ethical strategies for implementing AI-based interventions. A systematic review was performed following the PRISMA 2020 guidelines. Searches were conducted between July and August 2025 in electronic databases (PubMed, Scopus, Web of Science, CINAHL, IEEE Xplorer, and Google Scholar) and AI-assisted bibliographic tools (Perplexity, SciSpace, and Elicit). Studies were screened and assessed for methodological quality using the Joanna Briggs Institute critical appraisal tools. Seventeen studies met the inclusion criteria. Evidence indicates that AI can play a substantial role in enhancing staff safety. Approaches include smart sensors for real-time detection of aggressive behaviors, deep learning models analyzing clinical notes to predict violent incidents, advanced natural language processing systems (BERT, GPT, RAG-ECE) for identifying threatening language, and interpretable machine learning algorithms (LightGBM-SHAP-ALE) for pinpointing key risk factors. Predicting and preventing violence in healthcare is a complex challenge requiring multidisciplinary solutions. AI offers promising tools for early detection, accurate prediction, and effective intervention. However, implementation must consider technical limitations, algorithmic bias, privacy concerns, and the necessity of human oversight.
Moreno-Moreno et al. (Wed,) studied this question.