Background The acute: chronic workload ratio (ACWR) is a widely adopted training load aggregation tool to manage injury risk in sport. Recently, methodological concerns have been raised regarding this approach.
Objective To establish best-fit calculation methods for the ACWR when assessing injury risk, and to assess reproducibility of methods between professional Rugby Union teams playing in the same league.
Design Observational cohort study.
Setting Thirteen professional rugby clubs over two seasons.
Patients (or Participants) During two seasons, 433 and 569 players were recruited, meaning 1002 player seasons from 696 unique players.
Interventions (or Assessment of Risk Factors) Calculation methods: rolling averages(RA) versus exponentially-weighted moving averages (EWMA), coupled versus uncoupled, acute time windows (3–9 day), and chronic time windows (14–35 day).
Main Outcome Measurements Akaike Information Criterion (AIC) and Area Under the Curve (AUC) of model fit to injury risk.
Results 129,448 training load values were collected and aggregated into ACWR values to assess and compare their model fit to 1718 recorded injuries. In the 13 clubs there were 8 different ‘best fit’ ACWR calculations according to AIC score, with 3-day acute loads, 14-day chronic loads, EWMA and coupled approaches being the most common ‘best fit’ models. When the data was pooled, an EWMA, coupled 3:14 day ACWR provided the best fit; this finding was supported using AUC. Irrespective of the averaging or coupling method used, there was very little support for the commonly cited 7:28 day ACWR values.
Conclusions The commonly described 7:28 day average ACWR value may not be the most appropriate in a rugby setting, with 3:14 day EWMA coupled ACWRs providing better model fits. In addition, the best-fitting ACWR is highly variable across a somewhat homogenous set of clubs. Therefore, teams wishing to use ACWRs should model their own data to identify the version that is most appropriate for their setting, while the limitations of this metric should be understood when interpreting the data produced.