Identifying individual-based injury patterns in multi-trauma road users by using an association rule mining method

In many road crashes the human body is exposed to high forces, commonly resulting in multiple injuries. This study of linked road crash data aimed to identify co-occurring injuries in multiple injured road users by using a novel application of a data mining technique commonly used in Market Basket Analysis. We expected that some injuries are statistically associated with each other and form Individual-Based Injury Patterns (IBIPs) and further that specific road users are associated with certain IBIPs. First, a new injury taxonomy was developed through a four-step process to allow the use of injury data recorded from either of the two major dictionaries used to document anatomical injury. Then data from the Swedish Traffic Accident Data Acquisition, which includes crash circumstances from the police and injury information from hospitals, was analysed for the years 2011 to 2017. The injury data was analysed using the Apriori algorithm to identify statistical association between injuries (IBIP). Each IBIP were then used as the outcome variable in logistic regression modelling to identify associations between specific road user types and IBIPs. A total of 48,544 individuals were included in the analysis of which 36,480 (75.1%) had a single injury category recorded and 12,064 (24.9%) were considered multiply injured. The data mining analysis identified 77 IBIPs in the multiply injured sample and 16 of these were associated with only one road user type. IBIPs and their relation to road user type are one step on the journey towards developing a tool to better understand and quantify injury severity and thereby improve the evidence-base supporting prioritisation of road safety countermeasures.