Keele researchers develop improved model for predicting injuries in footballers

Research undertaken by a group of physiotherapists, statisticians, bioengineers and computer scientists at Keele University will help to improve identification of risk factors for significant injuries in football and advance the use of statistical models in sports injury research.

The research, published in the British Medical Open Sports and Exercise Medicine Journal, could have an impact on the ability of sports performance and medical teams to identify the most important factors for injury.

The researchers looked at the injury data from 24 male football players from a single English football team’s 2015/16 season, and found that injury prediction models in football can be improved by combining modern statistical methods with existing traditional methods. In the first stage of the research, modern statistical methods were used to select the most important predictors for significant injury.

A model was then developed using the previously identified predictors in combination with existing traditional methods (zero-inflated Poisson). By combining these methods, statistical models for predicting injury in football can improve their efficiency and accuracy.

Using these new methods, the researchers were successfully able to improve the real-life application of injury prediction models using the data from these players.

This model would also be appropriate for use in a real-world football club environment and provide medical teams with an objective tool for helping to guide their decision making about whether a footballer should train or play before each session.

Lead author Dr Fraser Philp said: “Statistical models haven’t been widely adopted for use in football clubs. This is usually because they require large sample sizes (bigger than what is available in a single football team) and expensive or complex measurements. Sports and exercise medicine practitioners are also reluctant to use models if they make their job harder or they are unable to explain the choices made by the model.

“Our new research has shown that the models can streamline the information that sports practitioners need to collect and provide a rationale for why these factors are important. We hope to test this model in other football clubs and ultimately help inform decision making in sports and exercise medicine teams, reducing the injury burden on footballers.”


Top 10 CUG 1st Guardian Athena SWAN