Predicting wound healing outcomes: a comparative accuracy analysis of AI-driven indices and percent area reduction

Wounds represent a major global health and economic burden, with chronic wounds affecting millions annually and costing medical care providers over $126 billion in the USA alone. Current assessment tools, such as percent area reduction (PAR), are widely used but limited by subjectivity and suboptimal predictive accuracy, particularly for complex wound types. This study aimed to evaluate the performance of an artificial intelligence (AI)-powered Healing Index (HI) in predicting delayed healing compared with PAR, leveraging the growing integration of AI into healthcare to enhance wound assessment and prognostic capabilities.

This retrospective study evaluated the performance of the AI-powered HI in predicting delayed healing for pressure injuries, venous ulcers, diabetic foot ulcers and arterial ulcers. Using a clinically validated dataset of 173 816 wounds collected via a digital wound care solution, the HI model's predictive accuracy was compared with PAR. The HI incorporated objective wound characteristics, such as tissue composition and exudate, to forecast healing trajectories.

By week 3, the HI achieved a balanced accuracy of 65%, surpassing PAR, which reached the same level only in week 4. This earlier prediction enables more timely treatment adjustments, facilitating improved outcomes and reducing healthcare costs.

The AI-powered HI demonstrates significant potential for transforming wound care by providing more accurate, objective and earlier identification of non-healing wounds. Its integration into clinical practice could enhance resource allocation, optimise treatment strategies and reduce the economic burden of chronic wounds. Further validation across diverse healthcare settings is warranted to ensure equitable implementation.