Predicting self-reported injury status among runners training for the New York City Marathon.
Mark A Fontana, Jamie S Egbert, Brett G Toresdahl
PMID 41906622WHAT IT FOUND
For runners already modifying training due to injury, models predicted next week's injury status well.
For runners not yet modifying training, prediction was poor. Current pain and injury status were the strongest predictors. Training load metrics like ACWR showed weaker associations.
Key findings
01The model performed well for runners regardless of prior injury status (AUROC 87.3%, AUPRC 52.1%) but poorly for those not already modifying training (AUROC 67.3%, AUPRC 7.7%).
02Prior week pain and injury status were the top predictors in the general model, while acute:chronic workload ratio (ACWR) was a top predictor in the model for uninjured runners.
03Injury status tended to persist week to week, with most runners recovering rather than worsening, highlighting the importance of early recognition.
STILL TO COME
How it was doneWhat they foundWhat it means for PTsWhat it means for OTs
Read the rest of this summary
You get three full summaries a month, free, and we do not ask for a card. Search, the TL;DRs and your library stay unlimited either way.
What it does not show
Low predictive power for runners not already modifying training, limiting clinical utility for primary prevention in asymptomatic runners. Sample was self-selected (Strava users, NYC Marathon registrants) and may not represent the general running population. Injury status and training logs were self-reported, introducing potential recall bias. ACWR has documented statistical limitations, including artifacts and irrelevance during tapering. The study did not predict acute traumatic injuries, focusing only on overuse injuries.
Declared interests
The authors have no competing interests to report. Dr. Fontana acknowledges support from Schmidt Sciences, LLC.
The easy way to misread this
Do not interpret the high accuracy of the first model as evidence that these tools can predict injury in healthy runners. The model performed well largely because it could predict that runners already injured would remain injured. For runners not yet modifying training, the predictive power was very low (AUPRC 7.7%), meaning the model cannot reliably identify who will become injured next.
Summarised by AI from the full paper, without a clinician reviewing it. Check it against the source before it changes what you do. Read it on PubMed →