Older People Avoidable Attendances: Identifying Avoidable Hospital Attendances in Older People

DC0037

Understanding the trajectory of older patients in emergency and urgent care systems, including their care pathways and outcomes, is crucial, especially for those with frailty and long-term conditions. This knowledge can help predict necessary hospital attendances and follow-up admissions. However, deriving attendance patterns analytically is challenging.
With large structured datasets like CUREd+, data-driven methods can be applied. Machine learning (ML), a subset of AI, can model historical patient behaviors by learning from past data and adapting decision parameters accordingly. A trained ML model could help determine whether a patient’s attendance is necessary based on their individual characteristics.
However, ML models are complex and often function as “opaque boxes,” making their decision-making process difficult to interpret. Users typically see only a set of parameters and performance results, which can introduce uncertainty. To address this, explainable ML (XML) techniques are needed to enhance transparency and trust. XML methods help identify key variables influencing the model’s predictions and explain how they interact to produce outcomes. By integrating XML, we aim to provide interpretable insights, ensuring that AI-driven decisions are understandable and reliable for clinical use.

9 September 2025

In progress

This study aims to improve the efficiency and effectiveness of emergency care systems in England, particularly for older patients with frailty and long-term conditions. The direct benefits to patients and the healthcare system will include:

Better Decision-Making for Hospital Attendance – By developing machine learning (ML) models that are driven by historical knowledges within the massive data, we can determine whether a patient’s attendance at emergency care services is necessary. This will help reduce unnecessary hospital visits, minimizing patient stress and improving resource allocation.

Optimized Patient Pathways – Understanding the trajectories of older patients in emergency care settings will enable healthcare providers to predict and streamline care pathways, leading to more targeted interventions and improved patient outcomes.

Enhanced Trust in AI-Driven Decisions – By incorporating explainable AI (XAI) methods, the study will provide insights into the key factors driving model decisions. This will support clinicians in interpreting and trusting AI recommendations, ultimately leading to safer and more transparent healthcare practices.

Policy and Practice Changes – The findings will inform policymakers and NHS decision-makers about how ML can optimize emergency care services. This could lead to new guidelines for hospital admissions, resource distribution, and patient management strategies.

No

Yes

Yorkshire and Humber Secure Data Environment