Results
Figure 1A,B, respectively, shows the number of players whose leg injuries were severe enough that the injured player had to miss at least one regular season game, and the mean number of games the injured players missed for each of the last 10 NFL seasons (2010–2019). There was no noticeable trend—increase or decrease—in the number of players missing at least one game of the regular season owing to injury to the leg—the slope (m=2.879±4.852) of the linear fit did not significantly differ from zero (t(8)= 0.593, p=0.569). In terms of the number of games the injured player missed, the slope (m=0.091±0.066) of the linear fit was not significant as well (t(8)= 1.366, p=0.209).
Figure 1Number of most serious leg (ankle/knee) injuries and resultant games missed (years 2010–2019). (A) shows a bar graph of the total number of players who missed at least one game because of a leg injury (ordinate) as a function of NFL season (abscissa; 2010–2019). Asterisks indicate rule changes adopted before the corresponding season to specifically protect players from leg injuries. (B) shows a bar graph of the number of weeks of game time that the injured player missed on average (ordinate) as a function of NFL season (abscissa; 2010–2019). NFL, National Football League.
The number of players who missed game time owing to injuries to the back (spine) did not vary much (slope—m=0.109± 0.420) over the last decade, as confirmed by the lack of statistical significance of the trend (t(8) = 0.260, p=0.802; figure 2A). The linear trend in game time missed (figure 2B; m=−0.034± 0.086) because of injury to the back was not significant either (t(8) = –0.399, p=0.700).
Figure 2Number of most serious back (spine/back/skeleton) injuries and resultant games missed (years 2010–2019). (A) shows a bar graph of the total number of players who missed at least one game because of injury to the back (ordinate) as a function of NFL season (abscissa). There are no asterisks, as there have not been any rule changes adopted in the last decade to specifically protect players from back injuries. (B) shows a bar graph of the number of weeks of game time that the injured player missed on average (ordinate) as a function of NFL season (abscissa; 2010–2019). NFL, National Football League.
In contrast, the number of players missing at least one game of the regular season owing to injury to the arm showed a small but significantly increasing trend across the last 10 NFL seasons: the slope (m=2.279±0.868) of the linear fit to the data in figure 3A was significantly greater than zero (t(8)= 2.626, p=0.030, two-tailed). The slope (m=0.129±0.076) of the average number of weeks of game time missed over seasons 2010–2019 was not significantly different from zero, however (t(8) = 1.699, p=0.128).
Figure 3Number of most serious arm (shoulder/arm) injuries and resultant games missed (years 2010–2019). (A) shows a bar graph of the total number of players who missed at least one game because of an arm injury (ordinate) as a function of NFL season (abscissa). Asterisks indicate rule changes adopted before the corresponding season to specifically protect players from injuries to the arm. (B) shows a bar graph of the number of weeks of game time that the injured player missed on average (ordinate) as a function of NFL season (abscissa; 2010–2019). NFL, National Football League.
The number of players who missed game time owing to head injuries showed no appreciable trend in either direction (slope—m=1.491± 1.954) over the last 10 NFL seasons (t(8) = 0.763, p=0.467; figure 4A). On the other hand, there was a strong linearly increasing trend in game time missed because of head injury (figure 4B): the slope was positive (m=0.138±0.037) and significant (t(8) = 3.779, p=0.0055; two-tailed).
Figure 4Number of most serious head injuries and resultant games missed (years 2010–2019). (A) shows a bar graph of the total number of players who missed at least one game because of a head injury (ordinate) as a function of NFL season (abscissa). Asterisks indicate rule changes adopted before the corresponding season to specifically protect players from injuries to the head. (B) shows a bar graph of the number of weeks of game time that the injured player missed on average (ordinate) as a function of NFL season (abscissa; 2010–2019). NFL, National Football League.
In a companion analysis, we compared two models—a linear model versus a non-linear family of step functions—over the data to see if the data were better fitted by a line, which is reflective of a gradual trend over the last decade, or by a sharp, abrupt change, reflective of some underlying change in the rules or improvement in protective gear or in the game itself. In brief, the non-linear step function was not a substantially better fit of data on the number of players who suffered leg, back, arm or head injuries (see online supplementary figures 5A-8A and online supplementary material for details). The step function was not a significantly better fit of games missed due to injuries to the leg (see online supplementary figure 5B and online supplementary material for details) and back (online supplementary figure 6B) and trended towards significance for injuries to the arm (p=0.086; online supplementary figure 7B) and head (p=0.071; online supplementary figure 8B).
Next, we explored if new rules that the NFL enacts every season have a short-term impact on player safety. The NFL tweaks several rules annually prior to the start of the new NFL season later that year in order to make the game safer and more exciting. Online supplementary material highlights the key rule changes that the NFL has adopted each year to make the game safer for the players as well as part of body—leg, back, arm, head or some combination thereof (conveniently highlighted in red text)—that was putatively impacted by each such rule change. Arrowheads at the top of online supplementary figures 1-4 show the number of rules that were adopted prior to each NFL season, with each rule represented by a single arrowhead. For example, right before the 2013 season, three rules (indicated by three arrowheads) were changed, which putatively protected from injuries to the head. We analysed whether the adoption of new rules prior to a given season was correlated with change in the number of injuries and/or change in the number of weeks the player had to miss game time because of injury in the given season as compared with the season prior. The number of rules adopted to protect players from injuries to the leg (indicated by counting the number of arrowheads for a given NFL season in online supplementary figure 1A) and change in the number of players who suffered a major leg injury (that caused them to miss at least 1 week of the regular season) were not significantly negatively correlated (r=−0.413, p=0.269; online supplementary figure 9A). Corresponding correlations for back (r=−0.272, p=0.479; online supplementary figure 10A), arm (r=−0.100, p=0.798; online supplementary figure 11A) and head (r=−0.319, p=0.402; online supplementary figure 12A) injuries were not statistically distinguishable from chance either. We conducted a similar set of analyses for the relationship between number of rules changes made and year-to-year change in the number of games the injured player had to miss—to the leg (r=0.099, p=0.799; online supplementary figure 9B), back (r=0.052, p=0.894; online supplementary figure 10B), arm (r=0.081, p=0.837; online supplementary figure 11B) and head (r=−0.157, p=0.687; online supplementary figure 12B); none of the correlations was statistically significant.
Rules that are meant to enhance player safety overall may be too general, and their impact too diffuse to be able to pinpoint any tangible benefit to player safety. A complementary way of investigating the short-term impact of changes in NFL rules is to focus on rules targeted to protect specific parts of the body. For example, a rule that prohibits helmet-on-helmet contact protects players from concussions. Such specific, targeted rule changes are provided by asterisks at the top of figures 1–4, for example, an asterisk at the top of figure 4 (head injuries) above the 2018 season indicates that a rule change(s) was adopted prior to the 2018 season that was tailored specifically to protect players from concussion. We conducted an analysis of the relationship between specific rule changes adopted to protect specific body parts and change in the two measures investigated here. The results were negative in regards to change in number of injured players: seasons in which new rules were adopted to protect specifically the leg (t(7) = 0.901, p=0.398), arms (t(7) = 1.5434, p=0.167) or head (t(7) = −0.220, p=0.832) did not yield statistically fewer injuries as compared with seasons in which no new rules were adopted to target specifically said body parts (we did not identify any rules adopted to protect specifically the back; therefore, a similar analysis on injuries to the back was not possible). The results were somewhat different for the second measure. Whereas seasons in which rules were adopted to protect specifically the leg (t(7) = 0.118, p=0.910) were not statistically related to change in the number of games the injured player had to miss, seasons that saw the adoption of rules to protect specifically the arms (t(7) = 4.194, p=0.004) and head (t(7) = 3.125, p=0.017) did show a significant decrease in the number of games the injured player had to miss in a particular season as compared with the previous season; the result also shows that seasons where no targeted rules were adopted saw corresponding increase in the number of games that players with injuries to the arm and head missed.
Finally, we address if the two measures provide similar or different information. The measures we have adopted—the number of players who missed game time because of injury and the number of weeks of game time missed—were intended to provide two related but nonetheless different perspectives into injuries and their effect or injury management. One way of understanding how related the two measures are is to calculate their correlation for each major body part investigated. The analysis revealed a statistically significant correlation between the two measures in regard to injuries to the leg (r=0.729; 95% CI (0.184 to 0.931), p=0.017); however, the correlations between the two measures in regards to injuries to the back (r=0.315; 95% CI (−0.392 to 0.788), p=0.375), arm (r=0.466; 95% CI (−0.231 to 0.847), p=0.174) and head (r=0.062; 95% CI (−0.590 to 0.864), p=0.854) were not significant.