S-OPTIMA: Surgical Outcome Prediction Utilising Machine Learning

This project aims to develop and implement an innovative digital tool that optimises the management of Intensive Care Unit (ICU) capacity for elective surgical patients across NHS trusts. With ICU beds representing a critically constrained resource, and elective surgical admissions constituting a significant proportion of demand, improved forecasting offers substantial opportunities to enhance patient safety and operational efficiency.

Working within the Digital Patient Safety Laboratory framework, we will develop machine learning models that predict individual patients' probability of requiring ICU admission following elective surgery. The primary model will utilise routinely collected clinical data available in electronic health records, while an enhanced version will incorporate additional pre-assessment consultation information where available.

We will explore several high-impact patient safety opportunities through this approach:
 

• Reducing Day-of-Surgery Cancellations: By providing reliable forecasts of aggregate ICU bed demand, the tool could significantly reduce last-minute cancellations that cause patient distress and operational inefficiency.

• Optimising Resource Allocation: The system will enable more effective planning of ICU staffing and resources, potentially improving care quality for both elective and emergency patients requiring intensive care.

• Enhancing Clinical Decision Support: The tool will systematically identify cases requiring detailed clinical review of post-operative destination planning, allowing targeted intervention where most beneficial.

The system will aggregate individual predictions using probability theory to forecast overall ICU bed demand up to three days before surgery, providing actionable intelligence for bed managers and clinical teams.

This work aligns with NIHR and NHS patient safety priorities, particularly in developing intelligent systems that enhance resource allocation while reducing avoidable harm from healthcare delivery disruptions.

Lead Investigators