Writers of Pro Football Prospectus 2008

18 Oct 2017

The Truth Behind Rising Injury Rates

by Zachary O. Binney

In 2015 we published an analysis that found the number of reported NFL injuries has been rising substantially since about 2007. The word "reported" is important in that sentence. Our database, as with any database reliant on public information, contains only injuries that appear on an injury report.

This means an observed increase could reflect a few different things:

  • A true increase in the underlying NFL injury rate. That is, more players really are getting hurt.
  • Reporting bias: better reporting of injuries that occur by teams. We might expect this to lead to more minor injuries being reported, but to have little impact on more severe injuries.
  • Diagnostic bias: better identification (and subsequent reporting) of injuries that were previously occurring but had been unrecognized. We might expect to see this with concussions and other head injuries, especially.

To separate out these effects we make one large but defensible assumption: injuries that caused missed games have been consistently reported regardless of the year. There is a lot of room for teams to manipulate the injury report for minor injuries, but if you're out of the game, you're out of the game.

We took the injuries in our database and split them by severity (minor, no games missed; moderate, one or two games missed; and severe, three-plus games missed) and type (concussion/head injury vs. other injuries). If the increase is due primarily to diagnostic bias, we would expect to see large increases in concussions and much more moderate increases in other injuries. If the increase is due to reporting bias, we would expect to see large increases in the number of minor injuries reported, while more severe injuries remained relatively flat. If there is a true increase in the underlying NFL injury rate, we would expect all groups to exhibit an increase.

While in previous articles we have reported injury rates per 1,000 athlete-exposures (a player participating in one practice or game), we're going to just provide counts here instead. This is because during the regular season the number of athlete-exposures is expected to have remained roughly constant from 2007 to 2015; whenever someone gets hurt, someone else immediately takes their reps or roster spot. Only injuries occurring in the regular season are counted, because preseason roster sizes have changed over time.

Overall Changes in Injury Rate

This is an update of our previous analysis that includes the 2015 season; 2016 is not comparable with prior seasons because the "Probable" designation was eliminated from the game status portion of the injury reports. In Figure 1, we can see a pretty clear increase in the injury rate over time, particularly from 2007 it 2012 before it plateaus from 2012 to 2015. However, as noted above, this doesn't allow us to tease out a true increase from reporting and diagnostic bias.

Injury Rates Stratified by Severity and Type

Figure 2 depicts what we would expect if the increases are primarily, but not entirely, due to diagnostic and reporting bias rather than a true increase in the NFL injury rate.

Head Injuries: The number of head injuries increased regardless of severity. This is consistent with both better identification of head injuries and more careful treatment of those injuries that are identified. We could hypothesize that 1) better identification causes an overall increase in the number of concussions that are reported, and 2) more careful treatment causes increases in the higher severity groups since what would have been a "minor" concussion in earlier years becomes a "moderate" or "severe" concussion in later years.

Non-Head Injuries: The largest increase by far is for minor injuries, which increased steadily by about 50 percent from 2007 to 2015. Moderate and severe injuries rose steadily from 2007 to 2011 before dropping and then rising again, but the overall difference between 2007 and 2015 is about 10 percent. Taken together, these suggest that much of the observed increase is due to better reporting of minor injuries.

But how much is due to reporting and how much to true changes? To estimate that, we can do some back-of-the-envelope math. First, we'll assume (generously) that there has been a true increase of 10 percent in the NFL's underlying injury rate. Overall, there has been about a 34 percent increase in non-head injury rates, so 10/34 = about 30 percent could be due to a true increase and 70 percent due to better reporting.

The proportion of the increase due to better reporting is likely well above 70 percent for head injuries, but it's tricky to calculate without a group that has steady reporting over time.

Conclusions

A majority (at least 70 percent) of the increase in reported NFL injuries appears to be driven by increased identification and reporting of concussions and minor non-head injuries rather than an actual change in NFL injury rates. True injury rates likely increased about 10 percent from 2007 to 2015. That's not negligible, but it's much lower than the 35 percent or so that just looking at the overall change might suggest.

Posted by: Zach Binney on 18 Oct 2017

9 comments, Last at 24 Oct 2017, 8:05am by jtr

Comments

1
by theslothook :: Wed, 10/18/2017 - 4:30pm

I guess it would require a deeper time series dive to see if 10 percent increase in injuries is a lot or a little. With no real experience with the data, I can only suspect the noise per year is pretty large.

2
by Zach Binney :: Wed, 10/18/2017 - 5:48pm

You can get an idea of the expected noise by using the confidence intervals. The relatively sustained rise over time increases my confidence the trends are real.

3
by ChrisLong :: Thu, 10/19/2017 - 7:57am

Did you test for statistical significance with these data? Looking at the CIs and trends, I would suspect the 10% increase you're seeing and proposing as the "real" increase may not actually be a statistically significant increase over time. Furthermore, where are you drawing these confidence intervals from?

4
by Zach Binney :: Thu, 10/19/2017 - 8:55am

Chris,

Thanks for your comment! Since these are (large) count data, the CIs are approximate Poisson CIs - I simply add or subtract 1.96 x the square root of the counts.

As for statistical significance, I try to avoid null hypothesis significance testing (that is, I don't draw conclusions from whether p is < 0.05 or > 0.05 for any given comparison) because of its many problems documented elsewhere. I prefer to present the data and its uncertainty and let people draw their own conclusions. If you look at the moderate/severe non-head injury data and see no reliable evidence of change, that's fine! I see a fairly consistent, sustained rise that led me to conclude (generously, I admit) that there's been very roughly a 10% rise in the underlying injury rate.

7
by ChrisLong :: Thu, 10/19/2017 - 8:18pm

Yeah I don't like the use of alpha values either, I prefer model selection approaches. Thanks for compiling all the data!

5
by jtr :: Thu, 10/19/2017 - 10:33am

The severe injuries don't seem to have any real trend beyond noise to them. Head injury trends are almost certainly due to an improved effort to diagnose concussions. I would speculate that the increase in non-head minor injuries is just due to reporting; high-def TV's make it easier for fans to notice that a given player has missed some snaps, and social media and the 24-hour news cycle means that more info comes out about those injuries.

The upward trend in moderate injuries is pretty subtle. It's possible that an increase there is caused by a decrease in painkiller usage by teams. If teams/players are less willing to dose up on toradol to play through injuries, that would cause moderately injured players to miss a game or two when in the past they might have leaned on drugs to play though it.

6
by Zach Binney :: Thu, 10/19/2017 - 11:15am

Totally agree on head injuries. I hope I got at that in the article. I also agree about non-head minor injuries being an issue of reporting; I definitely got at that in the article. And I'm happy to agree to disagree on the moderate/severe injury trends. Some will say that's dangerously subjective, but I argue that's realistically how science proceeds and we should recognize and embrace it!

Your painkiller idea is a really interesting one, but I think it's broader than that. One might hope that players and team medical and training staffs are becoming more aware of the long-term performance degradations from injuries and are more often taking the appropriate (longer) time to recover. Whether that's true or not I don't know, but it could be an alternative explanation for any trends that are there.

8
by The Ninjalectual :: Sun, 10/22/2017 - 4:26am

I wonder if teams aren't using MORE painkillers than in the past. I think expecting about 2/3 of the players on any given NFL field to be playing on pain pills is reasonable today. I have no idea about past eras.

9
by jtr :: Tue, 10/24/2017 - 8:05am

The give-out-handfuls-of-opiates-on-the-team-plane approach got a lot of media and DEA attention a few years ago. At the very least, the NFL has made sure that teams are being more discreet than that now. And just anecdotally, it seems like players are more aware than ever of the dangers of casually taking so many opioids. I suppose the team trainers are probably the only people in the NFL that truly know how painkiller use is trending.