Longitudinal Dynamic Web of Determinants – The Future of Injury “Prediction” and “Prevention”

 The deeper I get into research and sports science, the more I think about this one question:

Can we predict and prevent injury?

I know your answer (NO!), but simply viewing injury in sport as a black and white construct is missing the point entirely. Like any challenging athletic endeavor, the true answer likely lies in the grey areas (and there is quite lot when it comes to injury).

(Quick note: For this post, I’m going to draw on my experience as an ACL injury researcher. But really, the concepts I discuss can apply to all other sports injuries.)

I’ve analyzed a ton of injury prediction and prevention literature as an ACL researcher. Here’s what I can conclude…the rates of ACL injury remain incredibly stable over the years.

Figure 1. ACL injury rates over a 20-year period in adolescents (6-18 years old)1
 Figure 2. ACL / MCL injury rates over a 6-year period in NFL athletes2

So, what gives? Why has all the time and effort spent researching and understanding injuries led us no further in being able to, at minimum, reduce their occurrence? To find some answers, we first need to understand the questions we’ve been asking.

Are We Asking the Right Question About Injury Prediction and Prevention?

A scenario you’ve likely come across as a sports scientist: Pre-season camp is starting, and you want to identify which athletes will most likely get injured. What do you do? Typically, we turn to a screening test. For ACL injury, this can come in the form of drop-vertical jumps or jump-cutting assessments. From there, we can analyze variables such as frontal plane knee motion or muscular activity and use this data to drive an intervention in athletes who performed poorly on it. In some athletes, frontal plane knee motion predicts ACL injury.3 In others, it does not.4,5

I do believe these tests can be useful. For one, they at least get us in the ballpark of variables we should be aware of when designing intervention strategies. Additionally, screening tests are usually cost feasible, quick to administer, and can provide a ton of data in a relatively short amount of time.

While well intentioned, the issue in isolated pre-season assessments is that they fail to take into account a very important component: athletes are like hurricanes,6 constantly changing and evolving based upon their interactions with their environment. When you’re conducting a pre-season screening test, you’re only getting one frame of data on an athlete system that will likely never behave in that way again. Further compounding the issue is the use of statistical approaches such as regression to examine how isolated or multiple factors from this single time frame can “predict” athlete behavior many months in advance. This leads us to a linear, isolated risk factor model for injury prediction and prevention (Figure 3).

Figure 3. Isolated risk factor model using pre-season screenings

So instead of asking, “which screening test should I be doing”?, the better question becomes, “how can I predict a hurricane?”

Modeling Injury through a Longitudinal Dynamic Web of Determinants

Hurricane forecasts collect millions of data points each day on variables including wind speed and direction, humidity, temperature, and moisture to inform meteorologists whether a hurricane is upcoming.7 While sports scientists will never come close to this level of data depth in the high-performance setting, there are still principles we can take from hurricane forecasts and apply to injury prediction and prevention.

1) More frequent athlete monitoring

Screening athletes once or twice per year is likely to create a very false sense of the true signal driving high injury risk behavior. Take for example the drop-vertical jump test for ACL injury risk screening. The frontal plane knee motion an athlete demonstrates at pre-season could be drastically different at early-season, mid-season, and / or during post-season. The ultimate goal should be to monitor athletes longitudinally throughout the year to determine trends in athlete behavior. Let’s say our athlete completes drop-vertical jump tests at different frequencies throughout the year. Our frontal plane knee motion may look something like the below figures.

Figure 4. Example frontal plane knee angle tracked with different frequencies during the season

While this example is extremely simplistic, it drives home an important point. Infrequent monitoring of athlete behavior leaves us guessing in terms of risk classification. Longitudinal information can assist the sports scientist (alongside the medical and strength staff) in determining the most appropriate intervention(s) at the most appropriate time(s) to bring athletes back into a lower risk category and keep them on the field.

So, is that it? We just need to track a variable long enough over time to be able to predict and prevent injury? I wish it were that simple! This leads me to my second important point…

2) Viewing injury prediction and prevention as a web of dynamic interactions

In my opinion, a great benefit of isolated risk factor models is that they can provide the sport scientist with multiple perspectives into what ultimately contributes to injury. While excessive frontal plane knee motion is a risk factor for ACL injury,8 other researchers have found that worse neurocognitive performance puts athletes at greater risk.9 Some contribute ACL injury to familial history,10 or perhaps its due to hip strength.11

Taking into account multiple isolated risk factors can help the sports scientist develop what’s known as a dynamic web of determinants.12 Think of this as a web of variables all related to a particular injury, interacting with each other in ways that may or may not be linear. The excessive frontal plane knee motion our athlete demonstrates in weeks 6–8? What if that was from decreased reaction time abilities,13 brought on by a loss of sleep14 due to high stress?15

Athlete behaviors are extremely complex and can elicit a wide range of responses due to small or large changes in how factors interact with each other over time. Taking this into consideration, the sports scientist must consider the updated injury model.

Figure 5. Dynamic web of determinants for ACL injury risk17

While the actual statistical procedures are outside the scope of this post (see the papers from Meeuwisse, 2007,16 Bittencourt, 2016,17 and Stern, 20206 for more specifics), the sports scientist should understand that isolated factor analysis simply cannot account for all of the important grey area information that contributes to the injury event. Therefore, it’s important that the sports scientist examine the most relevant features of a particular injury and attempt to quantify their contributions.

So, how can we put this new model into practice?

Practical Considerations for Injury Prediction and Prevention

In order to make use of the dynamic web of determinants model, sports scientists must be aware of the following factors:

Explain the why

The worst thing a sports scientist can do is roll out a bunch of new assessments at once without explaining their importance to the team and coaching staff. Some athletes may be hesitant at first, as they might think the data will be used against them in terms of playing time and future opportunities. It’s extremely important to continuously discuss with your team how you will leverage this data to keep the athletes on the field at peak performance. This is where our “soft skills” essentially make or break our sports science initiatives. In order for this model to work, everyone within the team (athletes, coaches, sports science, S&C, medical) needs to be on board.

Time and feasibility of assessments

Assessments should be quick to administer and not be a burden to athlete schedules. Bonus points if your assessments can be done in group settings or by position groups.

You will have to tailor your model to meet the needs and demands of your athletes, but ideal settings would allow for weekly assessments. Certain variables, such as stress and sleep quality, can be collected daily via athlete management systems or Excel / Google Sheets. Make use of the wearable technology available, as these can give you accurate information comparable to gold-standard devices (but certainly look for wearables that have been validated against laboratory-based equipment).

Athlete compliance

Athletes are competitive, so assessments should tap into this trait to not only get high quality data, but also high athlete compliance. Make use of weekly leaderboards to display the athletes that scored the highest on each assessment. Do not punish athletes for low compliance. Rather re-iterate the importance of compliance in terms of using the data for their benefit.

Bandwidth for data collection and analysis

At the end of the day, the data we collect needs to be actionable. While it’s important to collect as much information as possible, it’s equally important that we not spend our entire day collecting and analyzing data. In my experiences, coaches and athletes take notice when you are “in the trenches”. Ideally, we can assign a single value to our macro-level metrics of interest and monitor trends over time. Important macro-level variables to consider: workload, recovery, sleep, stress / anxiety, movement biomechanics, strength / power output, neurocognitive performance, nutrition / hydration status, and prior injury history.

Conclusion

Injuries in sports represent the most complex problem for sports scientists. While it is likely impossible that a true injury prevention and prediction solution will ever be possible, there are certainly ways in which we can improve our monitoring techniques to give us the best possible insights into injury. Taking a longitudinal approach and exploring a range of related variables will certainly be helpful for sports scientists.

References

1.         Beck NA, Lawrence JTR, Nordin JD, DeFor TA, Tompkins M. ACL tears in school-aged children and adolescents over 20 years. Pediatrics 2017;139:.

2.         Injury Data Since 2015. NFL.com.

3.         Hewett TE, Myer GD, Ford KR, Heidt RS, Colosimo AJ, McLean SG, van den Bogert AJ, Paterno MV, Succop P. Biomechanical measures of neuromuscular control and valgus loading of the knee predict anterior cruciate ligament injury risk in female athletes: a prospective study. Am J Sports Med 2005;33:492–501.

4.         Krosshaug T, Steffen K, Kristianslund E, Nilstad A, Mok KM, Myklebust G, Andersen TE, Holme I, Engebretsen L, Bahr R. The Vertical Drop Jump Is a Poor Screening Test for ACL Injuries in Female Elite Soccer and Handball Players: A Prospective Cohort Study of 710 Athletes. Am J Sports Med 2016;44:874–883.

5.         Nilstad A, Petushek E, Mok KM, Bahr R, Krosshaug T. Kiss goodbye to the ‘kissing knees’: no association between frontal plane inward knee motion and risk of future non-contact ACL injury in elite female athletes. Sports Biomech 2021;1–15.

6.         Stern BD, Hegedus EJ, Lai YC. Injury prediction as a non-linear system. Phys Ther Sport 2020;41:43–48.

7.         Magnusson L, Bidlot JR, Bonavita M, Brown AR, Browne PA, Chiara GD, Dahoui M, Lang STK, McNally T, Mogensen KS, Pappenberger F, Prates F, Rabier F, Richardson DS, Vitart F, Malardel S. ECMWF Activities for Improved Hurricane Forecasts. Bulletin of the American Meteorological Society 2019;100:445–458.

8.         Della Villa F, Buckthorpe M, Grassi A, Nabiuzzi A, Tosarelli F, Zaffagnini S, Della Villa S. Systematic video analysis of ACL injuries in professional male football (soccer): injury mechanisms, situational patterns and biomechanics study on 134 consecutive cases. Br J Sports Med 2020;54:1423–1432.

9.         Swanik, Covassin T, Stearne DJ, Schatz P. The relationship between neurocognitive function and noncontact anterior cruciate ligament injuries. Am J Sports Med 2007;35:943–948.

10.       Myer GD, Heidt RS, Waits C, Finck S, Stanfield D, Posthumus M, Hewett TE. Sex comparison of familial predisposition to anterior cruciate ligament injury. Knee Surg Sports Traumatol Arthrosc 2014;22:387–391.

11.       Khayambashi K, Ghoddosi N, Straub RK, Powers CM. Hip Muscle Strength Predicts Noncontact Anterior Cruciate Ligament Injury in Male and Female Athletes: A Prospective Study. Am J Sports Med 2016;44:355–361.

12.       Philippe P, Mansi O. Nonlinearity in the epidemiology of complex health and disease processes. Theor Med Bioeth 1998;19:591–607.

13.       Herman, Barth JT. Drop-jump landing varies with baseline neurocognition: implications for anterior cruciate ligament injury risk and prevention. Am J Sports Med 2016;44:2347–2353.

14.       Benjaminse A, Webster KE, Kimp A, Meijer M, Gokeler A. Revised Approach to the Role of Fatigue in Anterior Cruciate Ligament Injury Prevention: A Systematic Review with Meta-Analyses. Sports Med 2019;49:565–586.

15.       Almojali AI, Almalki SA, Alothman AS, Masuadi EM, Alaqeel MK. The prevalence and association of stress with sleep quality among medical students. Journal of Epidemiology and Global Health 2017;7:169–174.

16.       Meeuwisse WH, Tyreman H, Hagel B, Emery C. A dynamic model of etiology in sport injury: the recursive nature of risk and causation. Clin J Sport Med 2007;17:215–219.

17.       Bittencourt NFN, Meeuwisse WH, Mendonça LD, Nettel-Aguirre A, Ocarino JM, Fonseca ST. Complex systems approach for sports injuries: moving from risk factor identification to injury pattern recognition—narrative review and new concept. Br J Sports Med 2016;50:1309–1314.

Leave a Reply