Population estimation in injury surveillance using a quasi-bootstrapping approach with convenience sample data

The National Collegiate Athletic Association Injury Surveillance Program (NCAA ISP) is among the longest-standing sport injury surveillance systems in the world. The NCAA ISP relies on a convenience sample of reporting institutions to estimate population-level injury metrics by scaling reported counts by an inverse school-reporting fraction and a fixed under-reporting correction. We propose a new method for improving population estimates of injury incidence obtained in the NCAA ISP, through a quasi-bootstrapping algorithm that capitalises on the historically archived NCAA ISP data.

We employed a phased approach including both a simulation study and a practical application, grounded in the operational methods of the NCAA ISP.

Across a range of simulated conditions, the proposed method approximated true population values with near-zero bias, produced more stable estimates than the current method and generated valid uncertainty intervals for both injuries and exposures. We also noted that when applied to real-world data, the proposed method produced plausible and credible estimates of injury and exposure counts.

By incorporating real-world variability from archived data, this approach improves point and interval estimation even under conditions of low or biased reporting.

The proposed method demonstrates strong potential as a scalable and statistically robust alternative to current population estimation practices in injury surveillance systems that rely on convenience samples.