Road traffic crashes result in the deaths of approximately 1.19 million people around the world each year and leave between 20 and 50 million people with non-fatal injuries.
— National Road Safety Action Grants Program
The Challenge
Road traffic accidents continue to be a major contributor to injuries and fatalities globally, raising significant public health and safety concerns. In Australia, we witness approximately 1,200 road-related deaths annually, along with around 40,000 severe injuries. These accidents impose a considerable social and economic burden, with an estimated cost of $30 billion to our economy each year (National Road Strategy 2021-30). Risky road behaviour (e.g., speeding, distracted driving, driving under the influence of alcohol or drugs) has been identified as a major cause of traffic accidents. While awareness of the risks linked to individual driver behaviour is increasing, effectively addressing this issue, and targeting safety campaigns, demands a thorough grasp of human behaviour's complexity.
The broad brush demographics, typically provided as national statistics in road accidents, provide a highly aggregated story of road trauma across Australia. These measures are effective in tackling road safety and aid in implementing policy at scale. However, once the broad implementation of policy and penalties reaches a plateau of effectiveness, a deeper understanding of social influences may be required to continue reducing the national road toll.
The Approach
Our research project is analysing various aspects of road accidents, including their types and severity, along with social well being indicators and social cohesion, to develop the whole-of-society comprehensive approach to the identification and prevention of risky road behaviour. The aim is to identify small cohorts of at risk road users, enabling niche target strategies that will lead to a reduction in road injuries and fatalities.
Understanding these niche populations of road users will require modern, cutting-edge data analysis and modelling. We will adopt an approach of statistics and machine learning in this research that will be extensible to other analyses in the future. Our primary research question is:
How do community characteristics, such as socio-economic disadvantage, residential mobility, and ethnic heterogeneity, influence the frequency and severity of road accidents in Australia?
Our initial investigations are focused on Queensland, using open data, such as the road crash dataset, the Australian census and the Queensland crime incidents, to build a picture of communities and the road safety issues they face.
This project is being conducted jointly with Dr Zarina Vakhitova (Monash University). This National Road Safety Action Grant was funded by the Australian Government.
The Solution
To take advantage of multiple open data sources, PAG have constructed a data lake to store information collected from various APIs that maybe useful in determining the relationship between socio-economic status, crime and traffic incidence. Essentially, the more variability we can explain through the models, the less chance there is of missing key factors that can identify smaller populations who have different risks in terms of road safety, than the general population. This can be achieved, by modelling at a more granular level.
The data lake will provide the most current and granular level data to facilitate building statistical and machine learning models to identify the relationships between community structure and road traffic accidents. The data will be constantly updated, from live sources, which will also allow us to create a real-time dashboard, with both modelled and summary statistics. This dashboard will provide continuous updates of the latest trends and insight into driver behaviour.
The Impact
The resulting data modelling, combining socio-economic and well being with road incident data, will provide a more complete profile of driver behaviour than standard approaches of considering demographics in isolation. It will also avoid the highly detailed approach of studying individual behaviour over time, which risks being unnecessarily invasive of privacy and may provide driver profiles that are hard to convert into policy.
In this research, we will relate temporal changes in drug-related crime and other social change, at small geographical levels, to forecast if this trend is likely to continue increasing, plateau or decrease. The modelling outcomes will also identify regions where, for example, drug-related activity (or other criminal offences) is associated with poor traffic outcomes.
In addition to risky road behaviour, our modelling approach will address regional and remote road safety by examining regional variations in social structure and types and severity of road accidents. Our research is in line with the National Road Safety Strategy 2021–30, which prioritises improved road safety outcomes for Aboriginal and Torres Strait Islander people, due to these communities bearing a higher burden of road trauma. We will use an indigenous-specific Socio-Economic Indexes for Areas (SEIFA) indicator derived from the Australian Census data in our modelling to examine road incidents concerning Aboriginal and Torres Strait Islander communities.
The project will use data from Queensland to serve as an example of the analysis possible at the national level.
