A Case-Based Reasoning Approach to Post-injury Training Recommendations for Marathon Runners

Recreational running is a popular way for people to exercise, but it attracts a high rate of injury. The widespread adoption of wearable sensors has led to a large volume of real-world data about how people train and recover. While such data has been used for performance prediction and injury risk assessment, little attention has been paid to how recreational runners return to training after an injury. We consider this novel application by recommending a suitable training workload when they return to training, based on how similar runners have returned to training in the past. Our results indicate that, contrary to the conventional wisdom, a conservative return to training may not always be the best option for a runner. Higher initial workloads are associated with improved race-day performance, but without materially increasing the risk of future injuries, compared to lower training workloads.