Machine Learning Model for Readmission Prediction of Patients With Heart Failure Based on Electronic Health Records:...

Area of Science:

• Health Informatics
• Clinical Decision Support Systems
• Machine Learning in Healthcare

Background:

• Heart failure (HF) care incurs significant healthcare costs, largely due to high 30-day readmission rates.
• Uncertainty in discharge timing decisions for HF patients leads to prolonged hospital stays, increased costs, and potential negative impacts on patient outcomes.
• Machine learning (ML) predictive models are being developed to mitigate this uncertainty by identifying high-risk patients for readmission.

Purpose of the Study:

• To investigate how a clinical decision support (CDS) tool influences healthcare professionals' decision-making processes regarding HF patient discharge.
• To analyze the usability and practical implementation aspects of the CDS tool by capturing practitioners' experiences with its outputs.
• To assess the impact of the CDS tool on decision consistency, quality, and work efficiency.

Main Methods:

• A quasi-experimental design with a randomized crossover assessment involving 12 physicians and nurses using 20 HF patient scenarios.
• The test group received patient data alongside ML-based CDS tool outputs predicting readmission risk, while the control group received only patient data.
• Data collection included interviews and observations, focusing on decision consistency, quality, efficiency, usability, and confidence in the CDS tool.

Main Results:

• The study is currently in the recruitment and scenario selection phase (September 2023 - March 2024).
• Ethical approval has been secured from the Swedish ethical review authority.
• The project is funded by the Knowledge Foundation as part of the Center for Applied Intelligent Systems Research Health research profile.

Conclusions:

• This study protocol is designed to inform future formative evaluation studies.
• It will contribute to the development and testing of ML models in collaboration with clinical professionals.
• The findings will provide insights into the effective integration of CDS tools into routine clinical workflows for HF patient management.