QUT researchers have created a powerful new tool that could predict when an athlete is at risk of suffering another injury when returning to play from a previous injury.

It uses data from wearable sensors capturing how the athlete moves in training or game play, along with data about the preceding injury and contextual information to quantify the risk. The study, titled “Next Generation Models for Subsequent Sports Injuries,” is published in Applied Stochastic Models in Business and Industry.

The system was developed by Associate Professor Paul Wu, Distinguished Professor Kerrie Mengersen and Yu Yi Yu from the QUT School of Mathematical Sciences and Center for Data Science, alongside an interdisciplinary team comprising performance health researchers from the Australian Institute of Sport, statisticians from UNSW and informed potential users.

Their hope is that it could help coaches, medical staff and athletes spot danger signs early, avoid costly injuries and keep players performing at their peak.

Approximately 40% to 60% of athletes sustain at least one injury in a given season, with 15% to 40% sustaining a second. From a community perspective, Australians suffered some 3.47 million sports injuries in 2023, with 66,500 needing hospitalization.

“With the rapid rise of wearable and other sensing technologies, the time is ripe for building next generation models to make sense of complex data and patterns, and support anticipative management and prevention of subsequent injuries,” Professor Wu said.

“The idea is to integrate training and competition performance data with injury data to link changes in performance to early warning signs for elevated injury risk.”

The team developed an approach to infer the internal state of the athlete, which was characterized as more or less susceptible to injury. This was linked to injury risk via variables, or features, obtained from a wearable sensor, medical (injury) and contextual data.

Importantly, Professor Wu said, the approach was able to capture changing injury dynamics and susceptibility over the course of a season.

Using data from an AFL club across one season, the model explained injury occurrences correctly 77% of the time with 90% specificity.

“Age emerged as the strongest factor influencing how an athlete might transition from a more susceptible to less susceptible state or vice versa, followed by context (for example, games carry higher risk than training), and the severity of the last injury,” Professor Wu said.

“Self-rated exertion and running speed also proved to be key indicators of injury risk.”

Professor Wu said the model could be particularly valuable in return-to-play situations, where a player is recovering from an injury and wants to minimize the chance of another.

“We can run ‘what-if’ scenarios, such as adjusting training or match loads to see the potential impact on injury risk or estimate an athlete‘s susceptibility right after a game or training session,” he said.

“Our vision is to give athletes, coaches and support staff, whether in elite sport or the community, tools that help them make sense of complex data, to allow them to train and compete at their best while managing the risk of subsequent injury.”