Injury Severity and Recovery among Soccer Players in the United States: Examining Analytical Methodologies and Key Determinants Open Access
Downloadable ContentDownload PDF
Background: Sports-related injuries present a substantial public health and economic burden. Yet, the short-term and long-term effects of sports-related injuries are not fully understood. This can be attributed to a lack of consensus in the measurement and analysis of sports-related injury severity. While time loss related definitions and evaluations have dominated injury severity assessment, there exist no methodological guidelines for directly evaluating time loss data in sports settings. As a result, there have been very few multivariable examinations of time loss determinants in specific sports settings. So, we aim to compare inferences from distinct approaches for multivariable modeling of time loss, where we consider time loss as a manifestation of both observable covariates, and unobservable injury severity. Moreover, in targeted evaluations of sports-related injury severity, it is also important to consider multiple injury corollaries to best summarize injury severity. That is, while identifying potential determinants of injury severity within specific contexts, it is important to consider multiple context-specific measures of severity. And so, we examine potential determinants of lower-extremity and head/neck injury severity, among youth soccer players in the U.S., using a multivariable perspective, and multiple injury corollaries to capture injury severity. Methods: Data from the National Collegiate Athletic Association- Injury Surveillance System (NCAA-ISS) for the 2004/05-2013/14 years, were used to examine approaches for multivariable modeling of time loss. We examined injury site, injury mechanism, sex, setting, injury history, competition level, and playing surface as determinants of time loss, and used random-effects Poisson regression models, and Weibull Accelerated Failure Time (AFT) Frailty models to obtain ‘severity-adjusted’ estimates. Data from the National Athletic Treatment, Injury and Outcomes Network Surveillance Program (NATION-SP) for the 2011/12-2013/14 years were used to identify potential determinants of youth soccer-related lower-extremity and head/neck injury severity. For lower-extremity injuries, we examined surgeries, season-ending outcomes, and the recovery process, in order to identify potential determinants of injury severity. We evaluated the roles of injury site, sex, setting, and playing surface in jointly determining the odds of a season-ending outcome, and the odds of time loss (given a lower-extremity injury), using multivariable logistic regression models. For lower-extremity injuries resulting in time loss, we evaluated the recovery process using random effects/frailty incorporated multivariable models of time loss. To examine soccer-related head/neck injuries, we evaluated concussion diagnoses, concussion characteristics (specific symptomology and symptom resolution time), and the recovery process. We evaluated the roles of sex and setting in jointly determining the odds of concussion and the odds of time loss, from head/neck injuries, using multivariable logistic regression models. We used Generalized Estimating Equations to assess the impact of injury mechanism, sex, injury history, and setting on determining the odds of specific concussion symptoms, as well as to fit an odds ratio dependence structure for the observed symptoms. We also evaluated the roles of injury mechanism, sex, and setting in determining the odds of longer concussion symptom resolution time, using ordinal logistic regression models. We then evaluated the head/neck injury related recovery process using random effects/frailty incorporated multivariable models of time loss. Results: While evaluating techniques for the multivariable modeling of time loss, we observed generally consistent results between the random effects Poisson regression and the Weibull AFT Frailty models. In the Poisson approach, injury site, injury mechanism and injury history appeared to have the strongest association with time loss due to injury, when examining ‘similarly severe’ injuries. For example, the adjusted time loss due to a knee injury was higher than the time loss due to an ankle/foot injury (referent group), when comparing injuries at ‘similar levels of injury severity’ (Adj. TLR= 1.21, 95% CI: 1.12, 1.31). Similarly, after accounting for latent ‘injury severity,’ injury site, injury mechanism, and injury history had the strongest association with time spent injured in the Weibull AFT Frailty model. From comparison analyses, we also observed that the random effects/frailty based approaches yielded better fitting models than fixed-effects-only approaches. In examining soccer-related lower-extremity injury severity among youth soccer players, we observed that injury site, sex, and setting (or event-type) emerged as significant determinants of season-ending injuries, while the odds of a lower-extremity injury resulting in time loss were significantly higher in young female players compared to young male players (Adj. OR= 1.34, 95% CI= [1.05, 1.71]), and in game settings compared to practice settings (Adj. OR= 2.19, 95% CI= [1.72, 2.80]). From our random effects Poisson regression and Weibull AFT Frailty models, we also observed longer recovery (‘severity-adjusted’ time loss) following knee injuries, compared to ankle/foot injuries (referent group). In examining soccer-related head/neck injury severity, we noted that the odds of concussion diagnosis from head/neck injuries were higher among young female players, compared to their male counterparts (Adj. OR= 1.84, 95% CI= [1.21, 2.79]), and from game-related injuries compared to practice-related injuries (Adj. OR= 2.23, 95% CI= [1.46, 3.41]). From evaluating specific concussion symptomology, we observed higher odds of light sensitivity (Adj. OR= 1.89, 95% CI= [1.01, 3.53]), and drowsiness (Adj. OR= 2.10, 95% CI= [1.08, 4.11]) among young female players compared to young male players. When examining concussion symptom resolution time, we observed lower odds of longer symptom resolution from player-to-player contact resultant concussions (Adj. OR= 0.33, 95% CI= [0.18, 0.59]). From our random effects Poisson regression and Weibull AFT Frailty models, we also observed that sex, and injury history were significant determinants of recovery (‘severity-adjusted’ time loss) following head/neck injuries. Conclusion: ‘Severity-adjusted’ multivariable models of time loss help to isolate covariate effects in determining recovery trajectories. Moving forward, with further evaluation in different player samples, these methods may be used to identify the strongest determinants of recovery from sports-related injuries. In targeted evaluations of sports-related injury severity and recovery, examining multiple context-specific injury corollaries may be considered. Our results indicate that, analyses motivated to identify determinants of severity may take this approach to describe how specific factors contribute to injury severity. Such multi-outcome examinations are important as injury severity is ultimately, a latent construct. Our examinations of surgeries, season-ending outcomes, and recovery following lower-extremity injuries, jointly present a more comprehensive picture of youth soccer-related lower-extremity ‘injury severity,’ than from evaluating time loss alone. Similarly, examining concussion diagnoses, concussion characteristics, and recovery from head/neck injuries help in more comprehensively identifying potential determinants of soccer-related head/neck injury severity. The methods and discussions presented in this work propose additional considerations for investigators aiming to evaluate sports-related injury severity in specific settings. It is up to future investigators to adapt the methods discussed here, to context-specific factors and mechanisms, while identifying potential determinants of sports-related injury severity.