Complex Systems and Running Injuries: Part 2 - The Bigger Picture

In Part 1 we introduced the problem (running injuries are multi-factorial) and an adjunct to the solution (complex systems modeling). Part two is about how all of this works and why it's relevant to the future of figuring out the cause and prevention of running injuries. To start, lets review the process of what needs to happen for effective injury prevention to occur: 

What's worth mentioning is that assessing efficacy (it works in a laboratory) vs. effectiveness (it works in real-life) are two very different things, and therefore represent different stages. For the sake of our topic, TRIPP stage 2 is the "let's figure out the cause of running injuries" and TRIPP stage 3 is the "let's figure out ways to prevent running injuries." 

Complex systems as they relate to identifying mechanism of injury (TRIPP stage 2)

So most of what we know about running-related injuries right now is from TRIPP stage 1 injury surveillance data (mostly on incidence in different running populations) and TRIPP stage 2 risk factor identification studies. The main issue that most risk factor studies have is that they're black box in nature; we know the input and output, but we don't really know how it works inside the box.

WHAT'S IN THE BOX?

If all you have are black box studies, you're left with a lot of variables that are kind of associated with each other but you generally don't have a clear way of explaining it. The design of the study is important as well; observational studies can't indicate causal factors, only associations. Another issue is that if there are as many variables as we think there are that play a role in injury development, we need HUGE sample sizes in randomized controlled trials or prospective studies to be reasonably confident in their statistical power. Most of the time, it just isn't feasible to get a sample size of 10,000 runners. What's the alternative, then?

Agent-Based Modelling (ABM)

This is where the robots come in. Agent-based models (ABM) are computer-based simulations that take real data and apply it on an individual-basis (the "agent") to examine the effects of interacting variables. ABM simulations address the difficulties of traditional sports science research; they can be performed with large sample sizes, multiple times, and results can be compared under different conditions. This is an ideal complementary test for a hypothesis that is either logistically impossible or unethical.

Is this... running?

Systems Dynamics (SD) Modelling

If ABM modelling is most applicable identifying potential risk factors on the individual-level, SD modeling is useful for analyzing how populations interact. SD modelling is a computational method with focus on causal effects, making it useful in determining contributing factors to injury. Hulme and colleagues use a great example about studying the unintended effects of an intervention on a population-level in running: a new shoe becomes available touting injury reduction properties and results in runners increasing their training load too much too soon, resulting in more injuries. But, I mean, what shoe would ever claim to reduce injury risk...?

Systems dynamics causal loop diagram from Hulme et al. 2018

Complex systems as they relate to developing preventative measures (TRIPP stage 3)

There's a lot of parallels between studying running injuries and incident/risk management in ergonomics science. In 2017, Hulme and colleagues proposed the Systems Theoretic Accident Mapping and Processes (STAMP) model as applied to the Australian distance running system. It's pretty comprehensive. Like "I can't actually post an image of it because the file is too big" comprehensive. What it notably brings to the table are the differentiation of "upstream" and "downstream" factors. Upstream factors are on a macro-level (say a sporting association policy) whereas downstream factors are on a micro-level to the incident itself (training load, for one). All of the factors much further upstream from the actual incident (injury) itself are therefore related in some capacity and many haven't been studied yet. The model includes five levels from top-down:

Level 1: Parliament & Legislatures

I'm going to go ahead and assume that most of us wouldn't initially blame the government for running injuries. That said, laws, rules, and regulations have a direct impact on and receive feedback from the next level downstream:

Level 2: Government Agencies, Sporting Associations, Funding, & Research Organizations

This is a big tier that includes government research funding bodies, national sporting associations, state legislation, and insurance companies. This has direct effects on every level below spanning education, funding, research priorities, and standards of practice.

Level 3: General Service & Healthcare Providers

This group contains running clubs, healthcare services like physical therapy, product retailers, and media. Policies, performance measures, scientific findings, and education then influence:

Level 4: Running Management, Supervision, & Injury Prevention

The sports medicine staff, including coaches, physicians, rehab staff, and sports scientists. The instruction, education, and treatment directly effects the runner and is influenced by their feedback in the form of performance and injury reporting.

Level 5: Runner & The Running Process

This one should look familiar actually:

Surely this is running

This is basically the Bertelsen model described in Part 1 of this blog post. It's the "here's what running injuries are" on an individual level without taking into account all of the other upstream factors.

Yes, this is fairly abstract and it's impossible to know how much of each level is involved on an individual case basis. This is the big picture perspective of running-related injuries, though. Establishing context is a necessary step in injury prevention development. While the complex systems approach doesn't give us the concrete answer to "why?", it explains how the parts interact with the whole. It's impossible to see this model and tell the runner in front of you that their injury is because of [insert single causal factor here], or that they could have avoided injury by not doing that.

The relationships demonstrated in the STAMP model lay the groundwork for future prevention programs to be tested further upstream from the individual runner. When we think of classic injury prevention programs, they mostly just involve the runner and the sports medicine staff (levels 4 and 5). We're pretty sure training load plays a role in the development of injury on the individual level, but we also know most (if not all) runners have the choice to run. If behavior change is therefore the next target, what's the most effective method of achieving it? Education from a coach? Education from the media? Is education even the best way to achieve behavior change? Another issue, of course, is that traditional methods of research like RCTs are no longer possible when we try to examine the effects of government or sporting association policy on injury incidence. ABM and SD modelling are then useful on an individual- and population-level to simulate the effects of different interventions and examine how other variables in a system may react. 

Pictured: a Westworld Season 3 screenwriter

If you take a few key points away from this post, they should be:

1) Complex doesn't always mean complicated, but this stuff is complicated.
2) There is much more at play than just training load in the development of running-related injuries.
3) We need to look further upstream to develop meaningful injury prevention programs.
4) I'm more sure that we exist in a simulation now than I was before I started studying complex systems.

If you've made it this far, thank you for reading! You might be in the simulation too...


Jason Tuori, PT, DPT, CSCS 



References:

1. Finch C. A new framework for research leading to sports injury prevention. J Sci Med Sport. 2006;9(1-2):3-9.
2. 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(21):1309-1314.
3. Bertelsen ML, Hulme A, Petersen J, et al. A framework for the etiology of running-related injuries. Scand J Med Sci Sports. 2017;27(11):1170-1180.
4. Hulme A, Salmon PM, Nielsen RO, Read GJM, Finch CF. Closing Pandora's Box: adapting a systems ergonomics methodology for better understanding the ecological complexity underpinning the development and prevention of running-related injury, Theor Issues Ergon Sci. 2017; 18(4): 338-359
5. Hulme A, Salmon PM, Nielsen RO, Read GJM, Finch CF. From control to causation: Validating a 'complex systems model' of running-related injury development and prevention. Appl Ergon. 2017;65:345-354.
6. Hulme A, Mclean S, Salmon PM, Thompson J, Lane BR, Nielsen RO. Computational methods to model complex systems in sports injury research: agent-based modelling (ABM) and systems dynamics (SD) modelling. Br J Sports Med. 2019;53(24):1507-1510.
7. Hulme A, Thompson J, Nielsen RO, Read GJM, Salmon PM. Towards a complex systems approach in sports injury research: simulating running-related injury development with agent-based modelling. Br J Sports Med. 2019;53(9):560-569.