Using artificial intelligence for personal protective equipment guidance for healthcare workers in the COVID-19 pandemic and beyond

Background: Current procedures for effective personal protective equipment (PPE) usage rely on the availability of trained observers or 'buddies' who, during the COVID-19 pandemic, are not always available. The application of artificial intelligence (AI) has the potential to overcome this limitation by assisting in complex task analysis. To date, AI use for PPE protocols has not been studied. In this paper we validate the performance of an AI PPE system in a hospital setting.

Methods: A clinical cohort study of 74 healthcare workers (HCW) at a 144-bed University teaching hospital. Participants were recruited to use the AI system for PPE donning and doffing. Performance was validated by the current gold standard double-buddy system across seven donning and ten doffing steps based on local infection control guidelines.

Results: The AI-PPE platform was 98.9% sensitive on doffing and 85.3% sensitive on donning, when compared to remediated double buddy. On average, buddy correction of PPE was required 3.8 ± 1.5% of the time. The average time taken to don was 240 ± 51.5 seconds and doff was 241 ± 35.3 seconds.

Conclusion: This study demonstrates the ability of an AI model to analyse PPE donning and doffing with real-time feedback for remediation. The AI platform can identify complex multi-task PPE donning and doffing in a single validated system. This AI system can be employed to train, audit, and thereby improve compliance whilst reducing reliance on limited HCW resources. Further studies may permit the development of this educational tool into a medical device with other industry uses for safety.