Hi everybody. Thank you for tuning in. My name is Nick Traer and today I'm excited to share research about prognosticating head injury with early impact test results. So to get started, I just want to talk a little bit about the impact test for those of you who aren't familiar with it. This testing battery is used across health care, education, and sports to assess head injuries and follow them over time. Basically you'll take an assessment sometime before your injury, normally at the beginning of your sports season, then immediately post-injury and compare those results. And the reason we do this is to detect deficits that may not be captured in clinical symptoms. So for example, some of the more complex aspects of a concussion like verbal memory, visual memory, processing speed, and reaction time. And this is what the test looks like. Those of you who haven't seen it before, it's a number of different cognitive tests. And you'll get these results at the end and you can see down here that you'll have these composite scores that we just touched on, as well as a total symptom score. And together, the way we interpret these results is that two or more significant deviations from your baseline test from before injury is associated with concussion. So two of the five composite scores, if they deviate significantly outside the 80% confidence interval, then that has an 82% sensitivity and 89% specificity of detecting a concussion. And so we want to take these results and go one step further and try to predict recovery. So severity index is the metric that we created. We'll call it SI throughout this research. And basically it is the sum of significant deviations from baseline. So in other words, if you're outside of that 80% confidence interval, any composite scores that fall out of that 80% confidence interval, we included and summated those so that we could have one cohesive metric to gauge severity of concussion. And ultimately, we want to, oh, well, and so if any of you want to calculate severity index for yourself here, just as a little more detail, but basically, as I was saying before, it's just the sum of significant deviations from baseline. And so what we want to do is see if severity index could predict concussion at follow-up. So when patients come for their second post-injury assessment, so there's post-injury one, where they're diagnosed with a concussion, then post-injury two, their first follow-up appointment, are they still concussed at that time based on the criteria that we used before, the two significant deviations from baseline? And what does their symptom burden look like at follow-up? Can we predict that using the severity of the first test? And finally, how will recovery change over time based on different levels of severity index? And ultimately, we want to compare severity index to other prognosticators as well, and we'll get into that later in the talk. So to do this, we collected a large data pool from impact applications who administer the test. And they gave us 25,000 tests or 26,000 tests from two testing centers in Florida and Colorado that have large populations. And ultimately, only 7,000 tests were analyzed because, you know, most of these tests by the nature of the workflow of an impact test are baselines. So if all these students are coming back and getting new assessments every year for baselines, you know, there's a lot of repeat testing. And so that represented around 2,300 unique head injuries amongst almost 2,200 unique students. And to get a little breakdown of those students, the majority were male, but still a good chunk female in there. And, you know, here's their age distribution on this graph, and ultimately, the number of concussions in each group. And if you want to get an idea of the distribution of sports across the different, you know, the different populations, we have a certain predilection towards football just because they're most commonly tested for fear of head injury. But there's a lot of sports included, and we tried to have a wide analysis. So ultimately, our outcomes, you know, going back to the questions that we touched on before, our concussion at follow-up, the symptoms at follow-up, and the overall time to no concussion or the recovery. And we phrase it that way, so we can do a bit of a, you know, time to event analysis that we'll touch on in a few slides. So the first metric is, as you'd expect, the higher severity concussions are associated with, you know, higher rates of concussion at follow-up as well. So the odds of still being concussed at follow-up increased by 12.2% with each unit increase in severity index. And to show that graphically, if you look here on the right, clearly the bin of students who fell into the 12 plus, you know, the higher severity index here on the right, this group were more likely to have concussions at first follow-up as compared to the group with the smallest severity. And then if you look at symptoms as well, it's also predictive in all of the symptom clusters that we analyze. So migraines, sleep disturbances, cognitive disturbances, and neuropsychiatric disturbances. And all of the neurocognitive domains that are reported by IMPACT all had worse outcomes as well. And to look at those graphically, first, we have the neurocognitive domains, the composite scores that IMPACT gives you. And you can see that they just decrease steadily as you get more severe, higher severity index concussions. And that is true for all of the metrics and reaction time, you would prefer to be lower. So that's again, a negative outcome. And we'll see the same more symptoms in each domain cluster with each with increasing severity. And just so you know, these are all disturbances. So cognitive disturbance is not an increase in cognition. And then, you know, given those results that we just showed it, it makes sense that severity index is associated with longer recovery. So we looked out, we've been to patients by the severity index, as we have been doing in the past groups, and follow their recovery, how many had recovered it each day throughout a six week period. And you can see that at any point in the curve, the 12 plus group is there are more likely to be concussed than any of the other groups. And so for example, here, if we look at 20 days, you know, there's still a good percent around 40% still concussed for the 12 plus group, whereas only, you know, around 10%, or even less are concussed in the zero to four group. I think that this is really kind of the where our research applies clinically, if someone's coming to you to you with a concussion and ask, you know, how long they can expect to be concussed, saying that half of all people with concussions like yours appear in the more severe group are still concussed at 10 days, versus only 25% for the zero to four group starts to allow us to prognosticate a little bit on what we can expect recovery to look like. And this is a pretty good model in predicting our area under the curve is 0.74. And that's controlling for sex, age, number of previous concussion, days between tests and location, and only severity index and age were significant predictors in that model. And ultimately, this is an improvement on past models. So first of all, including these neurocognitive domains, in addition to symptoms has already been shown in the literature to improve the predictive value. And then specifically, the real study that established that and the benchmark that we're comparing to is from Lau et al. in 2012. And they looked at symptoms as well as how many significant deviations you had total. So that's to say, of those five composite scores, did you have two or did you have four significant deviations? And so it's just kind of a lump sum. But the problem with that value is that it doesn't, you know, differentiate for the degree of deviation. So for example, three standard deviations from your baseline verbal memory score would be treated the same as 10 standard deviations, which obviously the 10 standard deviations is a very severe head injury. And so we treat that differently in analysis. And so the takeaways from that is really that, you know, SI is an intuitive metric, it's basically just the amount of deficit that you're experiencing from your baseline values. And again, that accounts for larger deviations in even fewer domains. And ultimately, that goes on to predict longer lasting concussions and a greater symptom burden, as you would expect. So in the future, we really just want to ensure that we would like to try to use SI to interpret head injuries that do not meet the threshold for concussion. So you'll recall that in order to be diagnosed with a concussion, you need two of the five composite scores to deviate significantly from baseline. If only one deviates, even if it's a major deviation, that doesn't meet the criteria for concussion based on previous analysis. But you can imagine that if someone has a massive drop in a single neurocognitive domain, that there's still some concerns that the head injury may have serious impact. And so for those patients, SI is just as applicable as in other models, but it'd be good to specifically target those patients, have a larger sample where we could analyze specifically patients who are kind of sub-concussed head injuries and see how good SI is at predicting their outcomes as well. And then ultimately, we might improve this model slightly by weighing some of the domains differently. In the past, Lau et al showed that weighing certain neurocognitive domains over other ones allowed them to achieve higher levels of prediction. And so by weighting ours differently as well, we might be able to fine tune it a little bit. And finally, obviously, we want to expand this to the adult population, include head injuries that aren't just related to sports. So that's just about it. And thank you very much for tuning in today and hope you have a wonderful week.

Impact Test

head injuries

clinical symptoms

Severity Index

concussion outcomes