Thursday 7 November | 11:05-12:05

Session format: Presentation

Stream: Safety

Part 1: Using workload capacity indicators to evaluate patient deterioration early warning tools

Many hospitals around the world are using, or seeking to use, rule- and AI- based early warning (EWTs) to recognise clinical deterioration. Traditionally, these tools are compared with tables of data containing sensitivity and precision at different alert thresholds.

However, there is now evidence to suggest that clinician workloads incurred by deterioration alerts are a key factor in determining their effectiveness.

In this session, we use the largest Electronic Medical Record (EMR) patient dataset in Australia, spanning 11 hospitals, to demonstrate a new approach for selecting the best EWT, AI or rule-based, and the most suitable alert thresholds for your hospital, taking into consideration workload capacity.

Anton van der Vegt The University of Queensland; Australia

 

Part 2: Predicting and preventing: protecting hospitalised patients and staff from deterioration events

To date early warning systems have focussed on recognising and responding to deterioration events in our inpatient wards. With digital hospitals and the data they provide, we now have the opportunity to use AI methods to predict and prevent these events rather than recognise and respond to them when it may be too late to change the outcome. Currently we train AI algorithms to predict deterioration based on traditional deterioration end-points, such as unplanned ICU transfer, cardiac arrest or death – but these usually arise well after serious deterioration has occurred. A novel deterioration definition that can identify patient deterioration earlier could augment these traditional deterioration endpoints. Without this, risk prediction AI, and its promise of expediting the delivery of the right information to the right clinician at the right time, will never reach its true potential.

Victoria Campbell The University of Queensland; Australia