Validation of a Clinical Decision‐Support Algorithm for Chronic Wound Classification and Treatment: An Expert Consensus

ABSTRACT

Accurate chronic wound classification is essential for appropriate management, yet diagnostic variability persists in routine practice. Transparent, rule-based decision-support tools may improve standardisation but require validation against expert judgement under clearly defined conditions. To evaluate inter-expert agreement, agreement between a rule-based algorithm and an expert-consensus reference standard, diagnostic accuracy as a complementary measure, exploratory comparison with a non-expert nurse, and expert agreement with algorithm-generated therapeutic recommendations. Thirty anonymised standardised clinical cases were classified by the algorithm and one non-expert nurse. Thirty wound-care experts, including 26 nurses, three physicians, and one researcher, were organised into six independent panels of five and classified case subsets, yielding 150 ratings. A consensus reference diagnosis was defined a priori as agreement by at least 3/5 experts. The primary outcome was algorithm–consensus agreement using Cohen's κ. Expert reliability was assessed using Krippendorff's α and Fleiss' κ. Recommendation agreement was dichotomised and analysed exploratorily. Expert agreement was low to moderate (Krippendorff's α = 0.26–0.60), highest for pressure ulcers/injuries and venous leg ulcers, and lowest for mixed or unknown leg ulcers and diabetic foot ulcers. Consensus was reached in 29 of 30 cases. The algorithm achieved 86.2% accuracy (25/29) and substantial agreement (κ = 0.70, 95% CI 0.46–0.94). Nurse accuracy was 72.4% (21/29, p = 0.219). Experts endorsed 85.2% of therapeutic recommendations. The algorithm showed promising agreement under controlled conditions, supporting further prospective validation in larger, balanced real-world datasets.