[The appropriateness of pressure injury treatment plans generated by large language models: a multicentre pilot...

. The appropriateness of pressure injury treatment plans generated by large language models: a multicentre pilot study.

Pressure injuries (PIs) are a major clinical and organizational challenge associated with high healthcare costs and complex management, variability in dressing selection and adherence to guidelines. Large language models (LLMs) may assist nurses in PI management by generating evidence-based treatment plans.

To evaluate the appropriateness of PI treatment plans generated by an artificial intelligence (AI) system based on a LLM compared with international guideline recommendations.

A multicenter cross-sectional study was conducted between July and October 2025. Nurses working in different healthcare settings submitted real-world PI cases through an online questionnaire, reporting wound characteristics, comorbidities, and clinical images when available. Cases were processed using GPT-4o configured with a structured prompt based on international wound care guidelines. Generated treatment plans underwent internal validation by the research team and external evaluation through participant feedback.

Twenty-six clinical cases were submitted by 20 nurses. The AI system generated treatment plans fully consistent with international recommendations. In cases including wound photographs, it provided more detailed recommendations. Five of the seven nurses who provided feedback, considered the treatment appropriate, while two reported educational benefits and no safety concerns.

LLMs may represent a promising and educational support in wound care, potentially improving therapeutic appropriateness and standardization of clinical practice. Studies on real clinical outcomes are required.