Investigation of pressure injuries with visual ChatGPT integration: a descriptive cross-sectional study

Abstract

AimThis study aimed to assess the performance of Visual ChatGPT in staging pressure injuries using real patient images, compare it to manual staging by expert nurses, and evaluate its applicability as a supportive tool in wound care management.DesignThis study used a descriptive and comparative cross-sectional design.MethodsThe study analysed 155 patient pressure injury images from a hospital database, staged by expert nurses and Visual ChatGPT using the National Pressure Injury Advisory Panel guidelines. Visual ChatGPT's performance was tested in two scenarios: with images only and with images plus wound characteristics. Diagnostic performance was evaluated, including sensitivity, specificity, accuracy, and inter-rater agreement (Kappa).ResultsExpert nurses demonstrated superior accuracy and specificity across most pressure injury stages. Visual ChatGPT performed comparably in early-stage pressure injuries, especially when wound characteristics were included, but struggled with unstageable and deep-tissue pressure injuries.ConclusionVisual ChatGPT shows potential as an artificial intelligence tool for pressure injury staging and wound management in nursing. However, improvements are necessary for complex cases, ensuring that artificial intelligence complements clinical judgement.Implications for Profession and/or Patient CareVisual ChatGPT can serve as an innovative artificial intelligence tool in clinical settings, assisting less experienced nurses and those in areas with limited wound care specialists in staging and managing pressure injuries.Reporting MethodThe STROBE checklist was followed for reporting cross-sectional studies in line with the relevant EQUATOR guidelines.Patient ContributionNo patient or public contribution.