Grant: Project Grant

Each year, there are 40,000 treatable out-of-hospital sudden cardiac arrests (SCA) in Canada. Bystander cardiopulmonary resuscitation (CPR) can double SCA survival but needs to be started within minutes. In most communities, less than 50% of victims receive CPR before the ambulance arrives. Currently, less than 10% of SCA victims will leave the hospital alive. 9-1-1 telecommunicator-assisted CPR instructions to 9-1-1 callers (previously trained in CPR or not) can greatly improve bystander CPR and survival rates for out-of-hospital SCA. Unfortunately, many SCA victims gasp for air in the first few minutes. This can fool the 9-1-1 callers and telecommunicators into thinking the victim is still alive, and results in delayed or no CPR instructions to callers. We successfully built and developed an artificial intelligence architecture which integrated existing foundation Large Language Models in a single site. We are now ready to pilot test the feasibility of implementing the AI architecture in five 9-1-1 communication centres. We will also determine the required elements that will ensure the future success of a future larger national trial following this success of this pilot randomized controlled trial. The resulting system, when integrated within 9-1-1 telecommunicators' workflow, will listen to 9-1-1 conversations and help improve the ability and speed with which 9-1-1 telecommunicators can recognize SCA, initiate timely CPR instructions, and improve survival for out-of-hospital SCA victims. When implemented across Canada, this could save 2,000 lives each year.