Table 1.

Description of six data cleaning methods analyzed

Method Number	Description
Method 1	All HAEs without cleaning, regardless of classification or timing.
Method 2	Time-windowed HAEs limited to active game play, excluding periods before the start of the game, after the end of the game, and during breaks (e.g., halftime, time outs).
Method 3	HAEs classified as true positives by prevent biometrics’ proprietary classification algorithm without time-windowing.
Method 4	Time-windowed HAEs classified as true positives by Prevent Biometrics’ proprietary classification algorithm.
Method 5	Video verification, where a trained research team member performed a guided video review of all time-windowed HAEs recorded by the IMGs in Sportscode (Hudl, Lincoln, NE) and coded HAEs as true or false positive based on observable contact synchronized to the recorded game footage. Only iMG recorded HAEs were included (visually identifiable HAEs that did not trigger iMG recordings were not included). HAEs recorded by iMGs that occurred outside of the camera view (n = 90) and could not be verified were excluded from analyses.
Method 6	Time-windowed, HAEs classified as true positives by both the Prevent Biometrics’ proprietary classification algorithm and confirmed through video verification.