Technology Acceptance of a Machine Learning Algorithm Predicting Delirium in a Clinical Setting: a Mixed-Methods...

Area of Science:

• Medical Informatics
• Artificial Intelligence in Healthcare
• Clinical Decision Support Systems

Background:

• Early identification of patients at risk of delirium is critical for timely intervention.
• The integration of machine learning (ML) models into clinical practice faces challenges in user acceptance.
• Assessing user acceptance is key to the successful implementation of AI-driven healthcare tools.

Purpose of the Study:

• To evaluate the user acceptance of an implemented ML-based application for predicting inpatient delirium risk.
• To understand healthcare professionals' opinions and concerns regarding the use of this AI tool.
• To assess the application's perceived ease of use, usefulness, system use, and output quality.

Main Methods:

• A mixed-methods design was employed, combining questionnaires and expert group meetings.
• The Technology Acceptance Model (TAM) framework guided the evaluation.
• Data were collected from 47 nurses and physicians using the application and through four expert group discussions.

Main Results:

• The ML application for delirium prediction was rated positively for overall usefulness by healthcare professionals.
• Users found the application's visualization and information understandable, easy to use, and appreciated the added delirium management support.
• The application did not increase workload, but its actual system use was low during the pilot phase.

Conclusions:

• The study provides valuable insights into the user acceptance of ML-based decision support for delirium management.
• High user acceptance is crucial for the successful integration of AI tools to improve healthcare quality and safety.
• Future efforts should focus on predicting actionable events and ensuring strong user adoption of computerized decision support systems.