• Three pedestrian injury risk prediction models were developed in this study
• A mass crash data model was validated using in-depth crash data
• This model could theoretically be used in a pedestrian AACN system
• A refined model could potentially improve pedestrian collision injury outcomes
• Such a model would need to be widely deployed in an AACN system to be effective
Advanced Automated Crash Notification (AACN) systems can inform emergency services of a serious road crash with minimal delay, giving the precise location of the crash and transmitting key information from the vehicle’s event data recorder, including: the crashed vehicle’s delta-V (vehicle change in velocity resulting from the crash), occupant seatbelt use, airbag deployment, and travelling speed. This information can be used to determine the likelihood of serious injury within the crashed vehicle using a suitable injury prediction algorithm. The purpose of this study was to examine two pedestrian crash data sets to develop pedestrian injury risk models using logistic regression analysis. Vehicle speed was used as the predictor variable and injury outcome was the response variable. The crash data used was from the in-depth crash database collected by the Centre for Automotive Safety Research (CASR) and from the South Australian Traffic Accident Reporting System (TARS) mass crash database. Three injury prediction models were developed and a discussion of the data and models are presented. Ultimately, the TARS data injury prediction model was selected as the most suitable injury prediction model, and this model was validated with the CASR in-depth data using receiver operator characteristic analysis. Suitability of the final model for use in a pedestrian AACN system was assessed using an injury threshold analysis. By accepting an injury underestimate rate of 10%, the minimum threshold for injury (for an AACN system activation) is 23%, which occurs at a vehicle speed 23 km/h; the corresponding injury over estimation rate was 84%.