Development and Effectiveness Verification of an AI Application for Pressure Injury Nursing Based on Clinical Judgment Model

"1.Background and Rationale Pressure injuries, also referred to as pressure ulcers, are a global healthcare issue, particularly severe among elderly and chronically ill patients. The prevalence is rising due to an aging population, prolonged hospital stays, and chronic disease incidence. Conventional prevention and treatment are often based on subjective nursing judgment and personal clinical experience, which limits consistency and efficiency across care settings.

Recent advancements in artificial intelligence (AI) enable integration of evidence-based decision-making tools into clinical workflows. Existing AI solutions for wound care tend to be rule-based and expert-driven, which constrains their adaptability in real clinical environments.

The clinical judgment model offers a systematic decision-making pathway: observing, interpreting, responding, and reflecting. Integrating this model into an AI mobile application for pressure injury nursing can provide consistent, evidence-based, and context-responsive guidance to nurses, improving patient safety and quality of care. This prospective study aims to develop such an AI-based application and evaluate its clinical educational effectiveness among hospital nurses.

• Study Objectives

• Primary objective:

To evaluate the effect of an AI nursing application, based on the clinical judgment model, on nurses' clinical judgment competencies in pressure injury care.

• Secondary objectives:

To measure changes in nurses' knowledge of pressure injury nursing. To assess satisfaction with the AI-based education program. To monitor any change in pressure injury incidence in patients under the care of participating nurses.

• Study Design

This is a prospective, randomized controlled, parallel-group study. Nurses will be allocated to either:

• Experimental arm: AI application-based education
• Control arm: Conventional education without AI support Assessments will be conducted pre-intervention, immediately post-intervention, and at three months post-intervention. Data analysis will compare within- and between-group outcomes.

The development follows the ADDIE instructional design framework:

• Analysis - guideline review and expert interviews
• Design - algorithm development based on the clinical judgment model
• Development - AI mobile application construction and educational content integration
• Implementation - pilot testing in a clinical setting
• Evaluation - randomized controlled trial to assess effectiveness

• Study Phases and Procedures

• Analysis

Review five evidence-based guidelines:

• EPUAP/NPIAP/PPPIA 2021 Prevention & Treatment
• Wound Care 2016 Assessment & Treatment
• NICE 2014 Prevention & Treatment
• KCE 2013 Treatment
• Korean Nurses Association Clinical Practice Guidelines (Pressure Injury Nursing)

Compare recommendations on prevention, staging, diagnosis, risk assessment, and treatment.

Conduct in-depth, semi-structured interviews (~1 hour each) with 15 wound/ostomy continence nurses with ≥3 years' experience. Topics include risk factors, clinical decision criteria, education improvement needs, and expectations for AI.

• Design Develop a pressure injury nursing algorithm aligned to the clinical judgment model (notice → interpret → respond → reflect).

Include stages: prevention, diagnosis/staging, treatment selection, evaluation. Ensure flexibility for varied patient conditions and institutional settings. Validate through expert panel review.

• Development Collaborate with software engineers to build a mobile AI app with: risk prediction, wound image staging, educational resources, and decision-support prompts.

Design an intuitive, efficient UI/UX. Embed visual wound classification tools and context-specific alerts.

• Implementation Pilot the application with experienced wound care nurses. Identify usability or algorithmic issues, revise prior to trial.
• Evaluation Participants: 80 nurses (≥1-year clinical experience), randomized into experimental and control groups (40 each, allowing for dropouts).

Intervention group: 2-hour AI-assisted education plus application practice. Control group: Same duration of lecture-based standard education without AI.

Assessments:

• Clinical judgment (modified Lasater Clinical Judgment Rubric for pressure injuries, score 11-44)
• Knowledge (PZ-PUKT test; 0-39)
• Satisfaction (Likert 1-5, plus qualitative feedback)
• Incidence of pressure injuries in patients under participants' care (if available) Survey time: ~30 minutes. All assessments repeated immediately post-intervention and at 3 months.

Inclusion criteria

• Registered nurse working at a tertiary hospital, directly engaged in patient care
• ≥3 years experience (algorithm development phase) or ≥1 year experience (evaluation phase)
• Written informed consent

Exclusion criteria

• Communication difficulties for interviews (analysis phase)
• Unable to understand education for AI application (evaluation phase)