First, I think we need to talk about why a prevention/prediction model is difficult and what has to be accounted for if we wish to eventually build one. While injury prevention is our long term goal, we have too many variables to control for at this point. Here are a few points highlighted by Kitman Labs, which I will elaborate on, that speak directly into this topic:
- We cannot really know that injury prediction works – it relies on an event to actually happen AND for the predictor to demonstrate that they knew about it beforehand.
- If we are going to try to “predict” injury beforehand, how often were we right vs. how often were we wrong…false positives, so to speak. If we predict an injury will happen, did we predict this 10x and get it right 1x? Not good odds, right?
- What “types” of injuries happen? Contact vs. non-contact? Soft tissue or skeletal injuries? Do we account for spine injuries or upper extremity injuries or only focus on lower extremity injuries at this time? Is an injury categorized strictly by “time loss”? If so, what counts as time loss? If not, how do you categorize it?
- The prior point leads to this one: We have to store injury data accurately and regularly. Athletic trainers seem to have a system for this…PTs do not, really. So, before moving forward with a prediction model, we need to know if the old data is still relevant? Maybe it is, but only in that prior injuries often increase the risk of future injuries – so, we would know who has been hurt in the past to maybe factor that into our equation(s)?
- Sample size – needs to be large enough to be statistically significant and showcase an accurate ability to “predict” without large false negatives
THESE are the things that it will take to build a prediction model. We have to know which key markers of what is tested actually allows us to adequately address the needs of an uninjured population and reduce injury? Even saying that is going to be tough without upfront data & robust tracking (those that we predicted and were right, those that we predicted and were wrong, those that we didn’t predict but were injured anyway, those that we did not predict and did not get injured). Then, see how our prediction model stacks up. MOSTLY, I think we need to temper our claims for injury prevention and injury prediction and hopefully be able to get by with words like injury risk.
Risk is more associated with things like high blood pressure, smoking, being overweight. These things place people at a higher risk of conditions like heart disease, strokes, cancer. So, to even start with the categorization of risk, we need to try to determine what is a correlation and what is causation. Not to get scientific or overly dorky, but that matters with making claims on how we operate in the rehab/fitness/sports world. Are our findings just linked in some way or does one cause the other? For example, maybe those who eat bacon for breakfast score 5 points higher on an IQ test. Can anyone say that eating bacon makes you smarter? Maybe you are smarter and so you eat bacon…this is just a silly example (because, why not bacon), but it illustrates that one does not CAUSE the other. They are just linked. Correlated.
So, let’s forget the bold claims about prevention…that word muddies the good work being accomplished in sports medicine. Instead, we should focus more on injury risk. Which things are simply correlated and which things cause consequences to be more likely to happen? For that, we need to determine the questions we need to ask (and attempt to answer) to address any risk. The top questions, in my opinion, are:
- What makes a given athlete more susceptible to a given injury?
- How do we best assess this?
- What factor(s) can be modified to minimize that susceptibility/risk?
These are questions for another post – stay tuned to dig into this further…