Using machine learning to predict volunteer acceptance rates.

Out of hospital cardiac arrest (OHCA) is a leading cause of death, responsible for 350,000-700,000 deaths annually in Europe. For every one-minute delay in treatment, the likelihood of survival decreases by 7-10%. Ambulances often do not arrive on the scene quick enough to save the patient.

In the Netherlands, an innovative app, HartslagNu, alerts registered volunteers when an OHCA occurs nearby. Nearby volunteers receive the alert, have the option to accept or reject it and receive the optimal route either directly to the victim -to start CPR- or first to an AED, if they accept the alert. There are 24,000 registered publicly available AEDs and 245,000 citizen rescuers with real time location information in the Netherlands.

Strategic location of AEDs (facilities) with respect to both OHCA victims (demand) and available volunteers (supply) is vital to decrease response time. However, there is very little known how volunteers behave, under what conditions they would accept an alert, and whether this acceptance rate has relation with location and time of the alert, volunteer’s age, gender and location and other factors that are available in the data.

The aim of this project is to predict the (positive) response rate of volunteers given the available volunteer and alert data. Modelling this acceptance behaviour will enable us to make more realistic assumptions on availability and number of volunteers in the AED deployment problem.