ASsuring Safe artificial Intelligence in ambulance Service 999 Triaging (ASSIST)
Improving the chances of surviving an out-of-hospital cardiac arrest by using AI to support ambulance service call centre staff.
The challenge
People who suffer out-of-hospital cardiac arrest have very high mortality. The number of deaths could be reduced if cardiopulmonary resuscitation could be given within three to five minutes of the onset of the cardiac arrest. Every minute of delay reduces the chances of survival by 10%. It is crucial, therefore, that ambulance service call centre staff recognise out-of-hospital cardiac arrest so that they can dispatch an ambulance quickly and provide instructions over the telephone to bystanders. However, the evidence suggests that at least 25% of out-of-hospital cardiac arrests are not recognised. The introduction of an artificial intelligence (AI) solution could improve recognition and reduce premature mortality.
The research
The project team will adapt an existing Corti AI platform, which has been piloted in Copenhagen, for use within the Welsh Ambulance Service (WAST). The assurance activities will contribute to the development of a real-world Body of Knowledge for assurance cases of AI in critical sectors.
The progress
The team has completed their work on understanding and specifying the operating environment for the Corti AI system and determining safety assurance requirements at the clinical system level. For this, they interviewed healthcare staff and analysed the data from these interviews to help define the operational design domain and clinical system-level safety assurance requirements.
The development of a self-contained safety case argument for the use of the Corti AI system within the WAST NHS Trust context has started. The team has developed an understanding of the data acquisition and management element of this work and will then move on to technology development and adaptation and testing and performance evaluation.
The final area of work for the project is about embedding the system in the real world. This entails engaging stakeholders from regulatory and standardisation bodies and from ambulance services nationally. This is progressing well and the team’s work has been presented at the HSJ Patient Safety Congress and the WAST Digital Leadership Group. Additional opportunities to disseminate the work are planned for early 2022.
Papers and presentations
-
NHS R&D Forum Conference, May 2023: Trust building as a foundation for assurance of healthcare AI: Experiences of the Assuring safe artificial intelligence in critical ambulance service response (ASSIST)
-
Rees, N., Holding, K., and Sujan, M. "Information governance as a socio-technical process in the development of trustworthy healthcare AI" in Frontiers in Computer Science, March 2023.
- EURO and UK NAVIGATOR (2020) International Academies of Emergency Dispatch peer-reviewed oral conference presentation: Assuring safe AI in ambulance response.
- ASSIST included in an oral presentation at REASON (2021) UKRI Trustworthy Autonomous Systems Node in Resilience, online workshop with project partners: Pre-hospital and ambulance services experiences and opportunities with autonomous systems.