AI-Assisted Skin Assessment for Pressure Injury Prevention in Critical Care Nurses

A Randomized Controlled Trial on the Application of Artificial Intelligence (AI) in Skin Assessment for Pressure Injury Prevention and Staging by Critical Care Nurses

The goal of this clinical trial is to learn whether an artificial intelligence (AI)-assisted skin assessment tool can improve the accuracy of pressure-injury staging in critical-care nurses. The study also aims to understand whether the AI tool increases nurses' knowledge and confidence in performing skin assessments. The main questions it aims to answer are:

Does AI-assisted assessment improve the accuracy of pressure-injury staging compared with standard visual assessment?

Does the use of AI improve nurses' knowledge and confidence related to skin assessment and pressure-injury staging?

Researchers will compare nurses who use an AI-assisted mobile application with nurses who perform standard manual assessments to see whether the AI tool improves staging accuracy and supports early identification of pressure injuries.

Participants will:

Complete brief questionnaires about their knowledge and confidence before and after training

Perform skin assessments on their assigned ICU patients using either standard methods or the AI tool.

Have their assessments compared with those of a blinded wound-care specialist, who will determine the most accurate staging

Pressure injuries remain a significant and largely preventable complication among critically ill patients, with ICU populations at particularly high risk due to immobility, hemodynamic instability, and complex medical needs. At KFSHRC-Jeddah, more than half of all hospital-acquired pressure injuries reported in 2024 occurred in critical-care settings, underscoring ongoing challenges in early detection and consistent staging. Although the organization follows evidence-based practices and uses tools such as the Braden Scale and NPIAP staging guidelines, variability in nurses' knowledge, skill, and confidence continues to influence prevention quality and accuracy of assessment.

Traditional skin assessment relies primarily on visual inspection and clinical judgement, which can lead to inconsistent interpretation of early tissue changes, particularly in darker skin tones, deep tissue injuries, and moisture-associated skin damage. These limitations highlight the need for innovative approaches that support more consistent and objective staging.

Artificial intelligence (AI)-assisted image recognition has emerged as a potentially valuable adjunct to standard nursing assessment. By analyzing skin characteristics such as color, texture, and contour, AI tools may assist nurses in identifying early-stage changes and provide decision support aligned with NPIAP criteria. Integrating AI into routine practice has the potential to enhance early detection, improve staging accuracy, and reduce practice variation.

This randomized controlled trial evaluates the use of an AI-assisted mobile application compared with standard manual skin assessment performed by critical-care nurses. The intervention uses an image-recognition tool that analyzes standardized photographs of high-risk skin areas and provides staging recommendations based on NPIAP definitions. Nurses in the control group will continue performing traditional visual and palpation-based assessments according to existing hospital protocols.

All participating nurses will receive pre-intervention education on pressure injury prevention, comprehensive skin assessment, and NPIAP staging to establish a consistent baseline. The intervention group will undergo additional training on standardized image capture to ensure appropriate lighting, distance, and positioning. A blinded wound-care specialist will independently review all assessments and images; this external review serves as the reference standard for evaluating accuracy and inter-rater reliability.

In addition to examining staging accuracy, the study will assess changes in nurses' knowledge and confidence before and after the intervention using validated instruments. It will also explore the feasibility and acceptability of integrating AI into ICU workflows. The findings are expected to inform how AI technology can support nursing practice, enhance clinical decision-making, and help reduce the incidence of hospital-acquired pressure injuries in critical-care environments.

Inclusion criteria

• Nurses working within the organisation for at least 6 months
• Nurses involved in direct patient care for over 50% of their work time.
• Skin assessments and staging for patients at risk for developing pressure injuries (Using the Braden Scoring system).
• Adult Patients (18 years and older)
• Patients who are currently admitted to the ICU and are receiving critical care treatment.
• No current severe skin conditions patients without active severe dermatological conditions (e.g., large open wounds, severe rashes) that would interfere with the AI-based skin assessment process.

Exclusion criteria

• Nurses working within the organization for less than 6 months
• Nurses involved in direct patient care for less than 50% of their work time
• End-of-Life Care or Terminal Illness- patients receiving end-of-life care or those with a terminal diagnosis, where the prevention of pressure injuries may not be a priority and where participation in the study may not align with their care goals.
• Severe or active dermatological conditions- patients with active skin conditions such as severe rashes, burns, or other dermatological issues that could interfere with accurate skin assessments by AI or confound the study results.
• Recent Skin Grafts or Advanced Wound Care- patients who have recently undergone skin grafts or those receiving complex wound care treatments that are outside the scope of typical pressure injury prevention practices.
• Inability to Maintain Required Positioning for Skin Assessment- patients who are physically unable to remain in the necessary position for the skin assessments, either due to severe mobility restrictions or critical medical conditions.