Artificial intelligence and occupational health and safety: a systematic review

Abstract Objectives The objective of this study is to The present systematic review sought to evaluate the application of artificial intelligence (AI) in the domain of occupational health and safety (OHS), with a particular focus on practical implementations, benefits, and the ethical, social, and operational challenges involved. The review also provides insights into current developments and future perspectives regarding AI integration in occupational contexts . Study design A systematic review of experimental and observational studies has been conducted. Methods The review was registered in PROSPERO (CRD42024568795) and conducted in accordance with the PRISMA 2020 guidelines. A comprehensive search was conducted in PubMed, Scopus, and Google Scholar. The studies included in the review were peer-reviewed, written in English, and focused on the application of artificial intelligence in the field of occupational health and safety. The selection of studies and extraction of data was conducted by independent reviewers, employing the TROSH-IA instrument to assess the quality of the studies. The inter-reviewer reliability was found to be substantial (Cohen's κ = 0.87). The data were synthesised into visual summaries and comparative tables . Results Following a comprehensive review of 540 records, 43 studies were deemed to meet the inclusion criteria. The potential applications of AI encompass a range of domains, including risk prediction, PPE detection, ergonomic monitoring, and hazard prevention. The models employed in this study encompassed Random Forest, CNN, LSTM, and YOLO. The majority of studies documented a high predictive accuracy (F1 > 0.80). The quality of the research varied, with a focus on the industrial and construction sectors. Conclusions Recent advancements in artificial intelligence have demonstrated significant potential in enhancing workplace safety through the implementation of predictive analytics and automation. Nevertheless, challenges pertaining to data quality, ethical concerns, and a paucity of standardisation persist. It is imperative that future research considers the entire AI lifecycle, from design to implementation, in order to ensure transparency, fairness and sector-wide applicability.