Acute to random workload ratio is ‘as’ associated with injury as acute to actual chronic workload ratio: time to dismiss ACWR and its components
Franco M. Impellizzeri, Stephen Wookcock, Aaron J. Coutts, Maurizio Fanchini, Alan McCall, Andrew D. Vigotsky
Abstract
Aim: The aim of this study is to assess the presence and implications of statistical artefacts created by a commonly used indicator of injury risk in both research and practice: the ratio between acute workload (AL) and chronic workload (CL), named ACWR. Methods: Using previously published data, we generated a contrived ACWR by dividing the AL by fixed and randomly generated CLs, and we compared these results to real data. Results: After reproducing the original analyses with only the ACWR showing effects compatible with higher injury risk (odd ratios, OR: 2.45, 95%CI 1.28 to 4.71), we demonstrated similar findings by dividing AL by the “contrived” fixed and randomly generated CLs: OR=1.95 (1.18 to 3.52) dividing by 1510 (average CL); and OR using random CL= 1.53 (mean), ranging from 1.16 to 2.07. Random ACWR calculated reducing the variance of the original AL further inflated the ORs (mean OR=1.89, from 1.42 to 2.70). ACWR causes artificial reclassification of players compared to AL alone. Finally, neither ACWR nor AL alone confer a predictive advantage to an intercept-only model, even within the training sample (c-statistic = 0.574/0.544 vs. 0.5 in both ACWR/AL and intercept-only models, respectively).Discussion: ACWR is a rescaling of the explanatory variable (AL, numerator), in turn magnifying its effect estimates and decreasing its variance despite conferring no predictive advantage. Other ratio-related transformations (e.g., reducing the variance of the explanatory variable and unjustified reclassifications) further inflate the OR of AL alone with injury risk. These results also disprove the etiological theory behind this ratio and its components. We suggest ACWR be dismissed as a framework and model, and in line with this, injury frameworks, recommendations, and consensus be updated to reflect the lack of predictive value of and statistical artefacts inherent in ACWR models.