Information

Approximately 17,000 people in the Netherlands suffer from out-of-hospital cardiac arrest (OHCA) each year. Many die or have poor neurologic outcomes because Emergency Medical Services (EMS) are alerted too late or not at all (unwitnessed cardiac arrest). We want to develop a technical solution to automatically and reliably detect OHCA using smartwatch sensor data and trigger an immediate emergency response to alert EMS systems. 

On this page we will explain in detail how we plan to achieve this goal. We have divided the project into 6 workpackages (WP). Below you will find a short description of each workpackage.

Schematic overview of the project

WP 1 User Research

In the first workpackage we focus on user research. We want to understand the needs, preferences and motivations of the target groups and involve users in the development process. We will do this by conducting surveys, interviews and focus groups. The inital focus will be on potential users who will use the technology to protect themselves (primary users). We are particularly interested in motivations why people would or would not use the technology, but also in identifying what features the technology should have from the user's perspective. We will consider diversity related subjects into account such as sex and socio-economic status. 

Second, we will also focus on the health professionals who respond to the OHCA alert (secondary users), such as ambulance personnel and citizen rescuers, and their opinions and attitudes towards this technology.  In this way, we hope to identify and address barriers to the technology in the early stages of development.

WP2 Sensor Data Collection

Data collection is an integral part and the foundation of the HEART-SAFE project because we will use it to develop/train the algorithm. We have designed a series of studies to collect sensor data from cardiac arrests under a variety of conditions. We will collect data from artificially induced cardiac arrests, patients with OHCA presenting to the emergency department, collect data from artificially induced cardiac arrests,  experimantally mimic cardiac arrests using tourniquete or blood pressure cuff, monitor patients at high risk of cardiac arrest and monitor patients in a hospice setting.

WP3 Open-source cardiac arrest detection algorithm

We aim to develop a pattern recognition algorithm that runs on smartwatches from multiple manufacturers and uses
trained artificial intelligence models to detect cardiac arrest, using the data collected in WP2. This package will consist of fundamental research, proof of principle tests, identifying patterns in the raw sensor data,  applying machine learning techniques for training and validation of the algorithm. Exploring all available sensors to detect the absense of circulation, indicating a cardiac arrest, but mainly focusing on the photoplethysmography (PPG) sensor. The goal is to aquire a high diagnostic accuracy with a sensitivity of at least 99% and a specificity of 99.9%. However, we are willing to initially accept a lower accuracy as we work towards optimizing the algorithm with data collected during field testing (WP5).

WP4 Engine block engineering

The focus of WP 4 is to provide a fully functional open access interface to integrate the algorithm into the "chain of survival". One of the consortium partners is provides its dHealthAI data platform for data processing and signal storage. The algorithm developed in WP3 will be an important part of this platform. An interface will be created between the dHealthAI environment and the fully operational HartslagNu citizen rescuer alert system as well as the alert systems of the EMS dispatch centers. It will thus be integrated into the "chain of survival". In addition, the HEART-SAFE appplication will provide an audible alarm in the event of a cardiac arrest to alert bystanders.

WP5 Field testing

The fifth WP is the field testing of the product created through all the previous WPs. The aim is to generate prehospital sensor data to test and further refine the algorithm; to evaluate the technology in the environment in which it will be used after implementation. In collaboration with the  EMS systems in the Netherlands, we will evaluate the accuracy of the algorithm and refine the technology. The target for this field test is 10.000-25.000 smartwatch users at risk of cardiac arrest. We will focus on the technical aspects (e.g. connectivity, accuracy, artefacts etc) but also on 'human factors' (e.g. psychological impact, user experience, behavioral patterns). Primary and secondary users of the technology will be surveyed in order to identify weak links in the chain.

WP6 Simulation Studies

The final work package will use simulation models to identify the target groups for which the proposed technology is most effective. It will also refine the parameters from the initial alert of the cardiac arrest to the arrival of the EMS. This refinement is essential to optimize the entire chain. Furthermore, the data collected in the previous workpackages will be used to estimate the impact on survival, quality of life after cardiac arrest and economic impact.