Predictive Models to Estimate Probabilities of Injuries, Poor Physical Fitness, and Attrition Outcomes in Australian Defense Force Army Recruit Training

Stephen C Allison, Bruce C Cohen, Edward J Zambraski, Mark Jaffrey, Rob Marc Orr

Research output: Book/ReportCommissioned reportResearchpeer-review


The purpose of this investigation was to assess the predictive potential of variables collected during the Australian Defence Force Recruit Training (n=19,769; 7,692 [28-day reservists course]; 12,077 [80-day standard]. The 28-day incurred 17.6% injury rate, 1 stress fracture, 5.2% attrition, 30.0% fitness test failure. The 80-day: 34.3% injury rate, 44 stress fractures, 5.0% attrition, 12.1% fitness test failure. Separate models were derived to predict injuries, attrition, and failure to pass the final physical fitness tests. Areas under the receiver operating characteristic curves (AUCs) for course-specific predictive models were relatively low (ranging from 0.51 to 0.69) consistent with failed to poor predictive accuracy. Course-combined models performed somewhat better, with 2 models having AUCs of 0.70 and 0.78; considered fair predictive accuracy. Although overall predictive accuracy was poor, accuracy was improved in models that included course length (28 vs. 80 day) as a predictor; suggesting the potential for using duration of training as a proxy for physical activity dosage to help predict injury and physical fitness.
Original languageEnglish
Place of PublicationNatick
Publisher US Army Research Institute of Environmental Medicine
Number of pages58
Publication statusPublished - 1 Nov 2015

Cite this