AI-enabled digital wound monitoring after cardiac surgery: a randomised controlled feasibility, safety, and...

Surgical site infection after cardiac surgery is a common cause of morbidity and unplanned healthcare use, with most infections developing after hospital discharge. Remote wound monitoring using smartphone technology and artificial intelligence (AI) may support earlier identification of complications.

To evaluate the feasibility, acceptability, and safety of an AI-enabled digital wound monitoring platform plus usual care (Isla-AI) compared with usual care (UC) alone.

This multi-centre, two-arm randomised controlled feasibility trial was conducted at two U.K. hospitals between August 2024 and January 2025. Adults undergoing cardiac surgery were randomised to receive Isla-AI or UC. The study was not powered to assess effectiveness.

120 patients were randomised and participated (Isla-AI N = 62; UC N = 58). Feasibility targets were exceeded: 60% of eligible patients approached consented, 95% of Isla-AI participants submitted at least one image, and 92% completed the study. Ninety-eight percent of images were suitable for clinical assessment. Clinician agreement with AI priority flags was 87%. AI prioritisation performance was slightly better for patients with darker skin tones. More than half of participants required assistance to capture or submit wound images. Patient- and staff-acceptability of AI was largely favourable. Adverse and serious adverse event rates were similar across both groups. The proportion of patients accessing National Health Service (NHS) resources for wound-related problems and antibiotics was lower in the Isla-AI group.

These findings support progression to a large, definitive multi-centre effectiveness trial, with further attention to equity, usability, and workflow integration.

IRAS 338141; local project UHDB/2022/024. ISRCTN16900119.

GOV: NCT06475703. Date registered: 20/06/2024.