Can your running data predict an injury before it happens?

By Canadian Running

· 1 min. read

Can your running data predict an injury before it happens?
Source: Canadian Running

Researchers studied whether running data from smartwatches and training apps could predict injuries before they occur. A machine-learning analysis of 643 runners training for the 2022 New York City Marathon found that training metrics alone couldn't reliably forecast injuries in otherwise healthy runners. The findings suggest current tracking technology has significant limitations for injury prevention.

Key points

  • Nearly half of marathon trainees reported at least one injury during the study period. Out of 643 runners followed over 16 weeks, 307 experienced injuries that forced them to modify their training.
  • The first predictive model succeeded mainly by identifying runners already experiencing symptoms. The model's strongest indicator was whether a runner had already reported injury or pain the previous week, making it less useful for true prediction.
  • A stricter second model showed dramatically reduced accuracy when tested on healthy runners. Researchers created this version using only weeks when runners weren't already injured, which more closely mimicked real-world prediction needs.
  • Training data alone wasn't sufficient to reliably forecast injuries in advance. The model incorporated running history, previous injuries, fatigue, illness, weekly distance, and training load changes but still failed to accurately predict future injury status.
  • Injury risk relationships were too complex for simple universal thresholds. The study found that "the relationship between training patterns and injury risk was complicated, rather than pointing to a simple mileage threshold that worked for everyone."
  • Researchers identified limitations that future studies could address to improve predictions. Self-reported injuries and imprecise injury timing made analysis difficult, and experts noted that "larger datasets and more precise information about when injuries begin could help future models improve."

Summarized by MalakyeAI

Links Cited on this Page