Clinical Applications of Artificial Intelligence in Occupational Health: A Systematic Literature Review (December 2024)-Clinical Applications of Artificial Intelligence in Occupational Health: A Systematic Literature Review

Clinical Applications of Artificial Intelligence in Occupational Health: A Systematic Literature Review

Pdf Summary

The study by Zaira S. Chaudhry, MD, MPH, and Avishek Choudhury, PhD, systematically reviews the clinical applications of artificial intelligence (AI) in occupational health. Searching databases like PubMed and IEEE Xplore, they identified 27 relevant articles published between 2014 and 2024. These articles were categorized into four groups based on their focus: health risk assessment, return to work and disability duration, injury severity, and injury management.<br /><br />Key findings reveal the predominant use of AI for health risk assessments, such as predicting noise-induced hearing loss and employee susceptibility to musculoskeletal disorders, lung diseases, and metabolic syndromes. Within health risk assessment, AI showed significant potential in enhancing occupational health surveillance, predicting illnesses, and offering interventions to mitigate risks.<br /><br />The review highlights diverse AI algorithms implemented in the studies, including neural networks, support vector machines, and random forests, with model accuracy ranging from 0.60 to 0.99. Importantly, most studies showed moderate quality using the APPRAISE-AI tool. However, a significant concern is the lack of explainability in many AI models, which could hinder their clinical adoption in occupational health due to the need for models that provide clear, evidence-based outputs.<br /><br />Studies also explored AI predictive models for facilitating return to work, managing disability, and determining injury severity, all crucial in informing decision-making within workers’ compensation contexts. Other studies assessed AI in injury management and rehabilitation, though results suggested that current AI models do not always outperform human judgment.<br /><br />The authors assert that, although AI shows immense promise in occupational health, future research should focus on developing explainable, clinically validated AI models within diverse real-world settings. They recommend integrating clinician expertise into AI model development to ensure practical applicability in occupational health settings, as well as improving data practices to address potential biases and privacy concerns.

Keywords

artificial intelligence

occupational health

health risk assessment

injury management

predictive models

neural networks

explainability

clinical applications

worker compensation

data privacy