Let us stop throwing out the baby with the bathwater: towards better analysis of longitudinal injury data

Sports injury prevention is a priority area in BJSM 1 and recent commentaries have stressed the need to consider relationships between sports injuries in longitudinal data sets.2 ,3 Players can sustain none, one, or more than one injury over a season of follow-up. Subsequent injuries are statistically related to prior injuries because they occur in the same person. This is true even when the two injuries are clinically distinct.2 ,4 ,5 It is always important to collect, analyse and report data on subsequent injuries in injury incidence studies. Figure 1 shows a hypothetical cohort of five players followed over one season of 10-week duration. This example assumes that all players are injury-free at the start of the season and addresses acute onset injuries rather than those due to repetitive microtrauma. The figure contains considerable information on each player: how long they were followed up; the number of injuries they sustained; how long before they sustained their first injury; and how long after any injury it took before they sustained their next one. Open circles indicate points at which players were injured, and stars indicate when follow-up was censored (eg, player 2 was only followed up for 7-weeks for non-injury reasons). A player does not accrue time at risk when he/she is unable to participate. For example, all of …