Fracture Incidence in National Collegiate Athletic Association Women’s Sports During 2009–2010 Through 2018–2019

Study Data

During our study period, the NCAA ISP used a convenience sampling scheme coupled with a rolling recruitment model to collect data. We focused on data captured within the following women’s sports (as these represent sports in which data were consistently captured across the study period): basketball, cross-country, field hockey, gymnastics, ice hockey, lacrosse, softball, soccer, swimming and diving, tennis, track and field, and volleyball. Athletic trainers (ATs) at the participating institutions contributed to the data collection, providing exposure and injury data using their clinical electronic medical record systems. Data were extracted from electronic medical record systems using a common data element export standard, and extracted data were deidentified and subjected to a series of automated verification procedures.¹⁴ A reportable injury was defined as an injury that occurred due to participation in an organized intercollegiate practice or competition and necessitated medical attention by a team certified AT or physician, regardless of the incurred time loss.¹⁴ For each reportable injury, the ISP captured crucial details regarding the injury and the circumstances leading to the injury. An exposure was defined as any team-sanctioned athletic activity in which student-athletes participated and were potentially exposed to the risk of injury.¹⁴ Along with injury details, the ISP also captured specifics regarding these exposure events, including the number of participants in each event. We used these details to estimate the at-risk exposure time in terms of AEs, defined as a single athlete participating in one NCAA-sanctioned practice or competition event.^14,17,18

Data Analysis

We analyzed fracture frequencies and distributions by various characteristics such as sport, mechanism of injury (player contact, equipment/apparatus contact, noncontact/overuse, other/unknown), the injured body part, and injury history (new, recurrent, other/unknown). We used a Bayesian framework to estimate rates (per 10 000 AEs) of fractures reported by sport and event type. We briefly describe our conceptual approach for characterizing fracture incidence below; details regarding each step in our analytical process, along with accompanying Stan programs and posterior predictive checks, are included in the Supplemental File.

Using Bayesian Inference in Sports Injury Surveillance

The premise of our approach for estimating incidence density using a Bayesian framework was the conceptualization of injury rates as underlying factors driving the observed injury counts. The goal of Bayesian analysis is to estimate distributions of parameters of interest by combining prior beliefs with observed data through a specified model and subsequently describing the characteristics of these posterior distributions.^19,20 Although delving into the mathematical philosophies underpinning the frequentist (ie, classical) and Bayesian methods is beyond the scope of this manuscript, one primary advantage of the Bayesian approach in this context is the ability to directly comment on the parameters of interest, and particularly their variability.^19,20 For instance, although classical methods allow for the calculation of observed injury rates and corresponding confidence intervals, the Bayesian credible interval (CrI) provides a plausible range of values for the injury rate itself. In contrast, the classical CI is a property of repeated sampling, indicating that if a sampling procedure were repeated n number of times, a certain fraction (ie, commonly 90% or 95%) of the resulting intervals would be expected to contain the true, yet unknown, injury rate. Furthermore, observational studies and injury surveillance often yield instances of sparse data across various variable contrasts or categories. In such cases, classical approaches may render parameters of interest inestimable. The Bayesian framework, however, offers a means to obtain realistic values for those parameters. Ultimately, our objective is not to advocate for the superiority of one approach over the other. Instead, we aim to present a different analytical perspective on injury surveillance data, offering more nuanced inferences than classical approaches.

We used a negative binomial model for calculating injury rates, which accommodated for overdispersion in injury counts. In Bayesian inference, it is typical to assign prior distributions for each parameter included within a modeling framework.^19,20 For the analysis of injury rates using a negative binomial model, we used Gamma (1,500) as the prior for the rate parameter and Beta (5,100) as the prior for the overdispersion parameter. Our analytical models were customized and written in Stan.²¹ We compiled and fit these models using the RStan package in R, running 2000 iterations across 4 chains to ensure convergence and sufficient posterior sampling.^19–21 Models were fit using No-U-Turn sampling, an advanced Markov chain Monte Carlo algorithm and an extension of the Hamiltonian Monte Carlo method.^19–21 Posterior samples of the injury rates and probabilities were extracted for inference, and model diagnostics such as ̂R and an effective sample size were computed to assess the convergence and efficiency of the sampling methods (details on model assessment criteria using these metrics are included in the Supplemental File 1).^19–21

Fracture Distributions by Body Part, Injury Mechanism, and Injury History

During the study period, the NCAA ISP recorded 944 fractures across all women’s sports (Table 1). Fractures were most frequently reported among lower extremity body parts (Table 2). Over one-third of all fractures reported in cross-country (39.3%), gymnastics (32.4%), and track and field (41.1%) were attributed to the foot. Similarly, a large proportion of fractures in cross-country and track and field were attributed to the lower leg (Table 2). The hand and wrist were also commonly fractured body parts, accounting for 55.1% of all reported fractures in field hockey and approximately 44% of all fractures in ice hockey and softball.

Table 1.Fracture Frequencies, Athlete-Exposures (AEs); and Posterior Mean Rates per 10 000 AEs by Sport^a

View Fullscreen

Table 2.Fracture Distributions by Body Part (Data Presented as Observed Frequencies and Proportions)^a

View Fullscreen

Fractures in NCAA women’s sports were most commonly reported as noncontact/overuse injuries (Table 3). Most fractures reported in cross-country (91.1%), swimming and diving (92.3%), and track and field (81.1%) were attributed to such mechanisms. Notably, equipment/apparatus contact accounted for >60% of fractures reported in field hockey and ice hockey and 57.9% of fractures reported in softball (Table 3). Most fractures in women’s sports during the study period were also reported as new injuries (Table 3). The prevalence of recurrent fractures was highest in track and field (17.8%) and gymnastics (17.6%). Among all reported fractures, 17.1% (n = 161) were classified as chronic injuries (72.4% were classified as not chronic).

Table 3.Fracture Distributions by Mechanism of Injury and Injury History (Data Presented as Observed Frequencies and Proportions)^a

View Fullscreen

Fracture Incidence Rates

The 944 fractures captured were recorded over 4 804 395 AEs across the study period. Using our Bayesian approach, we obtained distributions of various injury rates, enabling us to understand their features comprehensively. For the overall fracture rate, the posterior mean rate (obtained using the posterior distribution of the overall rate calculated using both observed data and prior information) was 2.16 per 10 000 AEs (Table 1). The 95% CrI for this rate ranged from 1.39 to 3.44. This interval indicates that there is a 95% probability that the true fracture rate in NCAA women’s sports lies between 1.39 and 3.44 per 10 000 AEs. When stratifying by event type, the posterior mean competition rate was 3.66 per 10 000 AEs, with a 95% CrI from 2.32 to 5.83. For practices, the posterior mean rate was 1.71 per 10 000 AEs, with a 95% CrI from 1.09 to 2.72 (Table 1). The posterior mean rate ratio between competition and practice rates was 2.26, with a 95% CrI ranging from 1.09 to 4.24. This result suggests that we can be 95% confident that the fracture rate during competitions is truly higher than the fracture rate during practices in this population.

The posterior mean overall injury rate was highest in gymnastics (6.29 per 10 000 AEs; 95% CrI = 3.70, 10.31), followed by cross country (4.04 per 10 000 AEs; 95% CrI = 2.42, 6.60), and field hockey (3.38 per 10 000 AEs; 95% CrI = 1.96, 5.86) (Table 1). The posterior mean practice rate was also highest in gymnastics, at 6.17 per 10 000 AEs (95% CrI = 3.70, 10.04). Cross-country (3.96 per 10 000 AEs; 95% CrI = 2.30, 6.74) and basketball (2.60 per 10 000 AEs; 95% CrI = 1.59, 4.27) followed in terms of estimated practice incidence density. In contrast, the posterior mean competition rate was highest in field hockey (8.14 per 10 000 AEs; 95% CrI = 4.48, 14.03), followed by gymnastics (7.42 per 10 000 AEs; 95% CrI = 2.86, 15.22) and soccer (6.88 per 10 000 AEs; 95% CrI = 4.28, 11.20).

Fracture Distributions by Body Part

Our data indicate that fractures predominantly affect the lower extremities in NCAA women’s sports, with a notable prevalence of foot and lower leg fractures across several sports. These findings align with existing literature and are biomechanically plausible given the movement dynamics inherent to the sports examined herein.^15,22 For instance, the high prevalence of foot and lower leg fractures in cross-country is unsurprising, considering the significant stress placed on these areas during long-distance running.²³ We also observed similar results in soccer, which is intuitive given the nature of the sport. Previous studies have identified the lower extremities as the most frequently injured body regions in soccer; however, further investigation is needed to better understand the athlete characteristics, injury-inciting events, and contextual factors contributing to fracture occurrence in this population.¹⁸ With regard to upper extremity fractures, we observed a notable proportion of hand/wrist fractures in field hockey, ice hockey, and volleyball in our study. These body parts are frequently used during gameplay in these sports and may therefore be at risk of inadvertent contact with equipment or opponents.²⁴ Our findings are in alignment with the existing literature on fracture characteristics in these specific sports, and future efforts may be directed toward closely examining the circumstances under which hand/wrist fractures occur in these sports. These fractures may be viable targets for primary prevention strategies, such as the use of protective equipment or the implementation of improved training techniques that emphasize proper hand positioning and impact mitigation strategies.

Distribution of Fractures by Mechanism of Injury

Data examined in this study indicate that fractures in NCAA women’s sports predominantly occur due to noncontact and overuse mechanisms, which aligns with the existing literature on bone injuries in female athletes.²⁵ Such fractures were prevalently reported among cross-country and track and field athletes. Existing research suggests that female runners may be at increased risk of overuse and stress fractures due to low energy availability, defined as insufficient caloric intake relative to energy expenditure.²⁶ Low energy availability can lead to disruptions in menstrual cycles and subsequently decreased bone mineral density, thereby increasing fracture susceptibility.^27,28 Previous research also suggests that stress fractures from repetitive load bearing, as in cross-country running, may predispose athletes to chronic bone density issues and heightened risk for future fractures.²⁹ The observed prevalence of noncontact and overuse fractures in our study underscores the potential role of diet and nutrition in injury risk assessment.^30,31 Future longitudinal studies on bone health markers are necessary among female runners in particular to identify potential relationships between dietary patterns and bone health, and to determine appropriate intervention points for improving bone health outcomes.

In contrast with the above-mentioned pattern in fracture mechanisms, fractures in soccer and basketball were most commonly attributed to player contact mechanisms. Indeed, we noted that approximately 41% of all fractures in soccer and basketball were attributed to player contact mechanisms. The prevalence of player contact–resultant fractures in soccer and basketball may be considered reflective of the high-contact nature of these sports, in which players frequently engage in physical confrontations.^18,32–34 Previous research among soccer athletes in particular suggests that tackling and sudden directional changes may be notable contributors to contact injuries in this population.³⁵ Although player contact is also a well-recognized aspect of basketball gameplay, further investigation is needed to identify the specific inciting events that contribute to contact-related injuries in this sport.^34,36,37 Understanding the interaction between frequently fractured body parts and fracture mechanisms is also essential in this regard, as the dynamics between anatomical site and injury mechanism can shape long-term bone health and recovery trajectories in female athletes. Prior studies indicate that fractures caused by high-impact mechanisms, such as player contact, may increase susceptibility to joint degeneration and early-onset osteoarthritis.³⁸ Therefore, future researchers should seek to clarify the mechanisms driving site-specific fractures in female athletes, with a particular emphasis on the potential interplay between injury location and mechanism on long-term bone health.

Equipment and apparatus contact emerged as a notable mechanism of fracture injury in field hockey, ice hockey, and softball. This may be unsurprising given the nature of these sports, in which athletes are vulnerable to abrupt equipment contact during gameplay.^39–41 As noted above, these data suggest that the implementation of enhanced protective equipment and equipment-related safety measures may be effective in reducing the fracture burden in these sports. Future researchers should focus on elucidating the specific inciting mechanisms of injury involved in these fractures to inform advancements in protective equipment design and use that may aid in prevention efforts.

Recurrent Fractures and Chronic Fractures

The prevalence of fracture recurrence varied by sport in our sample, with recurrent fractures most notably reported in gymnastics and track and field. We note that in the ISP, injury history is documented based on whether the AT classifies the injury as new or as a recurrence from the current or previous academic year. Fracture recurrence can have significant short-term and long-term implications. Recurrent fractures can lead to chronic pain, reduced mobility, and an increased risk of osteoporosis, all of which further elevate the likelihood of additional fractures and long-term disability.³ Moreover, recurrent fractures may indicate underlying bone health issues with serious long-term implications, such as reduced bone mineral density.³ We also noted that approximately 17% of all reported fractures in our study were reported as chronic injuries. This is noteworthy and further underscores the complexity of bone health in female athletes, as chronic fractures may develop gradually due to cumulative stress and inadequate bone remodeling. Existing evidence indicates that energy availability, hormonal imbalances, and low bone mineral density can impair bone remodeling and increase the risk of chronic fractures over time.^42–45 Although these physiological characteristics are challenging to elucidate using injury surveillance data, they should be specifically targeted in future, prospective studies. Such studies are necessary to determine how bone health indicators can serve as early markers for recurrent fracture risk and to understand how bone health progresses through the aging process in female athletes and active women.⁴⁶

Fracture Rates

Our findings indicate that fracture incidence in NCAA women’s sports is nearly twice as high in competitions compared with practices, aligning with other sports medicine research that indicates a higher injury incidence in competition settings.²² This can be attributed to the inherently dynamic nature of competition gameplay and the relatively uncontrolled environments characteristic of competitive events compared with practices. Prior authors have discussed how the unpredictability of competitive environments increases the likelihood of fractures due to sudden, high-impact events.⁴⁷ As such, it becomes vital to further examine the sports in which competition fracture incidence densities are highest. Our results suggest that competition fracture incidence is highest in field hockey, followed by gymnastics and then soccer. Coupled with the fracture characteristics described herein, these findings offer important insights for targeted evaluation and intervention. Specifically, we observed higher prevalences of contact than noncontact fractures in field hockey and soccer. This suggests that in field hockey and soccer in particular, collision and impact events during play might be significant contributors to fracture risk.⁴⁸ Emphasizing the importance of proper technique, player/body awareness, and the use of adequate protective equipment could potentially reduce the incidence of contact-related fractures in these sports.⁴⁹ Furthermore, a closer examination of the injury-inciting events, particularly within the competition setting, could help identify if policy or gameplay considerations could reduce fracture incidence in these sports. There exist examples in youth sport settings in which effective gameplay modifications have been shown to positively affect injury incidence.⁵⁰ Although not directly applicable to the present setting, this prior research may offer blueprints for similar evaluations in collegiate field hockey and women’s soccer.

Our findings also indicate that overall fracture incidence among NCAA women’s sports is highest in gymnastics and cross-country. The nature of gymnastics, involving high-impact landings and complex routines, may inherently contribute to fracture risk among these athletes.⁵¹ Similarly, cross-country athletes may be prone to overuse injuries and stress fractures due to repetitive high-mileage training on uneven terrain.⁵² Previous authors have also noted a high fracture incidence in athletes of these sports, reinforcing the need for increased focus on these populations.^15,22 Considerable attention has been directed toward energy availability and nutrition in runners, particularly in conversations around their risk of stress fractures. For instance, research has shown that inadequate calcium and vitamin D intake significantly increases the risk of stress fractures in runners.^53,54 Additionally, biomechanical factors such as running gait and footwear have been identified as contributing factors to fracture risk in runners.⁵⁵ Coupled with the existing literature, our findings support the need for developing sport-specific risk assessment and injury prevention programs that address the unique risks associated with gymnastics and cross-country.

Limitations

Findings presented herein provide a critical overview of fracture epidemiology in NCAA women’s sports; however, they are limited by a number of factors. The NCAA ISP uses a convenience sampling scheme with a rolling recruitment model. As a result, participation varies by sport and year, potentially limiting the external validity of these findings. That said, NCAA ISP recruitment strategies have evolved significantly over time, leading to a substantial improvement in participation throughout the study period (reflecting, for instance, support and communication from the NCAA Sport Science Institute). Regarding the estimates presented, we acknowledge the ongoing debate within sports injury surveillance surrounding the expression of at-risk exposure time in terms of AEs. Although the use of AEs offers an efficient reporting solution, it may not represent the most precise measure of exposure time, potentially compromising the precision of injury incidence estimates. Additionally, we note that the ATs responsible for reporting data are not provided with study-specific diagnostic criteria or variable guidelines when documenting injuries. This is particularly relevant to the reporting of injury history and chronicity, as the ISP relies on AT clinical expertise and familiarity with an athlete’s medical background to document these elements. As such, although NCAA ISP methods are designed to standardize reporting practices, the potential remains for variability among ATs in reporting, which could result in nondifferential misclassification of the variables examined. Furthermore, as noted above, we used a distinctive analytical approach in this study. It is recognized that Bayesian methods inherently involve subjective decision points, especially concerning prior specifications. Our decisions were made with plausible and conservative considerations in mind, and were supported by prior sensitivity checks. Nevertheless, we have transparently disclosed our decisions within the programs shared in the supplemental file. We acknowledge that future investigators may adjust these parameterizations in different applications of these methods using surveillance data.