Re-establishing Confidence in Confidence Intervals: An Evaluation of Recent Practices in Sport Injury Epidemiology

Several studies investigating injury burden have used "standard" formulae for injury rates when calculating their 95% confidence intervals. However, this may have led to artificially narrow confidence intervals because the authors' calculations did not account for violations of important underlying assumptions such as (1) clustering (e.g. when some athletes have multiple injuries), and (2) other assumptions unrelated to clustering. Although previous authors have recommended appropriate methods such as bootstrapping to solve these challenges, there is little guidance for sport medicine researchers on how to implement bootstrapping given the complexity of data in our field. The purposes of this article are (1) to illustrate when the "standard" formulae for injury rate confidence intervals can be used in sport medicine research, (2) why the standard formulae for injury rate confidence intervals are inappropriate when estimating injury burden and (3) provide more detailed instructions on how to use bootstrapping for confidence intervals in the context of any sport medicine study that includes repeated measures.