Preliminary Derivation of Test Item Clusters for Predicting Injuries, Poor Physical Performance, and Overall Attrition in Basic Combat Training
Abstract
Analytic methods including test item clusters (TICs) employed in medical diagnostic testing have potential for estimating probabilities of negative military training outcomes in individual trainees. Baseline attributes and performance scores that discriminate between groups experiencing negative vs. positive training outcomes were combined to maximize predictive power and accuracy. Predictive models were derived from 15 baseline variables using existing data (518 men and 416 women Basic Combat Training (BCT) trainees) to predict: Army Physical Fitness Test (APFT) failure at week 7; overuse injury during BCT; and failure to complete BCT with peers. The models included from one to four predictors per TIC. Large shifts in pre-test to post-test probability (as high as from 16% pre-test probability to 90% post-test probability) were observed with TICs to predict APFT failure for both men and women, and to predict overuse injuries in men. Smaller probability shifts were seen with the single tests identified to predict BCT attrition for both men and women. No useful model for predicting overuse injuries in women was derived from the methods and data employed in this study. This study suggests good potential for these analytic methods to derive useful combinations of prognostic variable for predicting negative outcomes in BCT.
Document Details
- Document Type
- Technical Report
- Publication Date
- Dec 01, 2006
- Accession Number
- ADA460374
Entities
People
- Joseph J Knapik
- Marilyn A. Sharp
- Stephen C. Allison
Organizations
- United States Army Research Institute of Environmental Medicine