Multivariate modelling of subjective and objective monitoring data improve the detection of non-contact injury risk in elite Australian footballers

Marcus J. Colby, Brian Dawson, Peter Peeling, Jarryd Heasman, Brent Rogalski, Michael K. Drew, Jordan Stares, Hassane Zouhal, Leanne Lester

    Research output: Contribution to journalArticlepeer-review

    47 Citations (Scopus)


    Objectives To assess the association between workload, subjective wellness, musculoskeletal screening measures and non-contact injury risk in elite Australian footballers. Design Prospective cohort study. Methods Across 4 seasons in 70 players from one club, cumulative weekly workloads (acute; 1 week, chronic; 2-, 3-, 4-week) and acute:chronic workload ratio's (ACWR: 1-week load/average 4-weekly load) for session-Rating of Perceived Exertion (sRPE) and GPS-derived distance and sprint distance were calculated. Wellness, screening and non-contact injury data were also documented. Univariate and multivariate regression models determined injury incidence rate ratios (IRR) while accounting for interaction/moderating effects. Receiver operating characteristics determined model predictive accuracy (area under curve: AUC). Results Very low cumulative chronic (2-, 3-, 4- week) workloads were associated with the greatest injury risk (univariate IRR = 1.71–2.16, 95% CI = 1.10–4.52) in the subsequent week. In multivariate analysis, the interaction between a low chronic load and a very high distance (adj-IRR = 2.60, 95% CI = 1.07–6.34) or low sRPE ACWR (adj-IRR = 2.52, 95% CI = 1.01–6.29) was associated with increased injury risk. Subjectively reporting “yes” (vs. “no”) for old lower limb pain and heavy non-football activity in the previous 7 days (multivariate adj-IRR = 2.01–2.25, 95% CI = 1.02–4.95) and playing experience (>9 years) (multivariate adj-IRR = 2.05, 95% CI = 1.03–4.06) was also associated with increased injury risk, but screening data were not. Predictive capacity of multivariate models was significantly better than univariate (AUCmultivariate = 0.70, 95% CI 0.64–0.75; AUCunivariate range = 0.51–0.60). Conclusions Chronic load is an important moderating factor in the workload–injury relationship. Low chronic loads coupled with low or very high ACWR are associated with increased injury risk.

    Original languageEnglish
    Pages (from-to)1068-1074
    Number of pages7
    JournalJournal of Science and Medicine in Sport
    Issue number12
    Publication statusPublished - 1 Dec 2017


    Dive into the research topics of 'Multivariate modelling of subjective and objective monitoring data improve the detection of non-contact injury risk in elite Australian footballers'. Together they form a unique fingerprint.

    Cite this