Annals of Biomedical Engineering

, Volume 39, Issue 8, pp 2130–2140 | Cite as

Development of the STAR Evaluation System for Football Helmets: Integrating Player Head Impact Exposure and Risk of Concussion

Article

Abstract

In contrast to the publicly available data on the safety of automobiles, consumers have no analytical mechanism to evaluate the protective performance of football helmets. The objective of this article is to fill this void by introducing a new equation that can be used to evaluate helmet performance by integrating player head impact exposure and risk of concussion. The Summation of Tests for the Analysis of Risk (STAR) equation relates on-field impact exposure to a series of 24 drop tests performed at four impact locations and six impact energy levels. Using 62,974 head acceleration data points collected from football players, the number of impacts experienced for one full season was translated to 24 drop test configurations. A new injury risk function was developed from 32 measured concussions and associated exposure data to assess risk of concussion for each impact. Finally, the data from all 24 drop tests is combined into one number using the STAR formula that incorporates the predicted exposure and injury risk for one player for one full season of practices and games. The new STAR evaluation equation will provide consumers with a meaningful metric to assess the relative performance of football helmets.

Keywords

Concussion Mild traumatic brain injury Acceleration Risk Exposure HITS Impact 

Introduction

Recent research has suggested that there are as many as 3.8 million sports-related concussions each year in the United States, with participation in football resulting in the highest incidence of injury.8,29 Studies showing the potential long-term effects of these injuries have put sports-related concussions under the national spotlight as a primary health concern.16,35,36 Furthermore, there is concern that repetitive sub-concussive head impacts in sports may lead to neurocognitive deficits.18,19,25 While limiting the number of head impacts in sports is an important component of reducing injury incidence, improving head protection is another essential element of injury mitigation.7 This article focuses on a new mechanism to evaluate the protective capabilities of football helmets.

Substantial effort has been invested in researching head acceleration in relation to brain injury.27 Head acceleration is thought to be indicative of the inertial response of the brain, and therefore is used to predict brain injury.27 All head injury safety standards for automobiles and helmets (motorcycle, sports, or bicycle) use measured humanoid head acceleration (or a function of head acceleration) during specified testing conditions to determine whether a product is safe to sell to consumers. While the Federal Motor Vehicle Safety Standards (FMVSS) 201 and 208 govern whether an automobile is safe to sell using pass/fail injury criteria, the New Car Assessment Program (NCAP) provides consumers with a quantitative metric of the relative safety between automobile models.23,28 NCAP is a valuable tool for consumers who are concerned with safety.

In contrast to the publicly available NCAP safety data on automobiles, consumers have no information on the relative impact performance between different helmets; moreover, there is no quantified metric that provides meaningful interpretation of the test results. The National Operating Committee on Standards for Athletic Equipment (NOCSAE) provides a set of voluntary standards that are designed to assess a helmet’s ability to prevent skull fracture. NOCSAE certification involves testing helmets through a series of drop tests, in which every drop test must result in a head form impact response below a specified threshold. The NOCSAE standards have done an excellent job of eliminating skull fractures in helmeted sports, but do not consider concussion or provide data to the public. The lack of publicly available safety performance data for helmets can be partially attributed to the previously limited knowledge of impact-induced brain injury and the challenge of interpreting the data; however, recent studies have provided a more complete understanding of the head acceleration patterns associated with impacts in football as well as possible concussions.

Since 2003, researchers have been instrumenting football players with the Head Impact Telemetry (HIT) System (Simbex, Lebanon, NH) to collect head acceleration data each time a player experiences a head impact.11 While Duma et al.11 were the first to instrument football players with the HIT System, other researchers have adopted this technology to investigate head impacts in football.4,21,40,43 The measurement and analysis of head acceleration data collected from these in-helmet accelerometer arrays have been well-validated and widely accepted.5,11,24,39 By instrumenting and observing a population that is at risk of concussion, researchers have collected data that can be used to provide the foundation for biomechanically characterizing concussion risk. Between all researchers utilizing the HIT System to collect head acceleration data from football players, over 1.5 million head impacts have been recorded to date.

The objective of this study is to develop and introduce the concept of a new evaluation system that can be used to provide quantitative insight into the protective performance of football helmets against concussions. This new evaluation system is designed to integrate the overall helmet impact performance into a singular metric that is derived from the head impact exposure and injury risk based on real-world impact measurements in football players. Head impact exposure is expressed as the number of impacts that are experienced by players in relation to specified drop test configurations. This new system is analogous to NCAP, in that biomechanical data are interpreted for the public in order to provide consumers with a meaningful metric to use when deciding which product to purchase.

Materials and Methods

The STAR Equation

The Summation of Tests for the Analysis of Risk (STAR) equation is presented as a mechanism to compute a singular metric from a total of 24 drop tests that can be used to evaluate the relative performance of football helmets. The STAR equation was developed to relate the head impact exposure that a football player experiences throughout one full season of participation to the injury risk associated with each head impact (Eq. 1). Fundamentally, this equation correlates every head impact that a football player experiences during one season of participation to 24 drop tests (four locations × six drop heights) that can be performed using a NOCSAE-style drop tower and head form.34 The products of impact exposure and injury risk associated with each impact location and drop height are summated to compute a predicted concussion incidence for one player during one season wearing a specific helmet.
$$ STAR = \sum\limits_{L = 1}^{4} {\sum\limits_{H = 1}^{6} {E(L,H) \cdot R(a)} } , $$
(1)
where STAR is the Summation of Tests for the Analysis of Risk, L represents one of four impact locations (front, rear, side, or top), H represents one of six drop heights (60, 48, 36, 24, 12 in., and lowest), E represents head impact exposure as a function of impact location and drop height, R represents injury risk as a function of peak resultant head acceleration, and a represents peak resultant head acceleration resulting from each specific drop height and impact location.

The methods presented in this article detail the derivation of the head impact exposure and injury risk components of the STAR equation. Next, as an example of how the equation should be used, the STAR value of a hypothetical helmet is computed and its interpretation is discussed.

Head Impact Exposure

For the purpose of the STAR equation, head impact exposure is defined as the number of impacts that one football player will experience through one complete season of participation. Crisco et al.6 quantified impact exposure in detail for three collegiate football teams over the duration of a single season; and reported that players can experience over 1400 impacts over the course of one season. These data were utilized for this study because their analysis investigated exposure on a per player basis. As expected, this study reported that the number of head impacts per player per season increased with the number of games and practices in which they participated.6 The 90th percentile head impact exposure per player per season was chosen in order to account for a full season of participation at a conservatively high level. Given these data, the STAR equation uses 1000 head impacts as the total exposure that a collegiate player participating in one full season would experience.

Exposure per player per season was further investigated by impact location by utilizing 62,974 head impacts that were recorded between 2009 and 2010 with the HIT System at Virginia Tech.11,12 These impacts were analyzed to determine the head impact exposure on an impact location basis.5 Each recorded head impact was categorized into one of four general impact locations based on the computed azimuth and elevation of the impact vector (Fig. 1). Any impact with an elevation >65° was categorized as an impact to the top of the helmet. Impacts with elevations <65° and azimuths between 45° and −45° were categorized as impacts to the front of the helmet. Impacts with elevations <65° and azimuths between 135° and −135° were categorized as impacts to the rear of the helmet. All other impacts were categorized as impacts to the side of the helmet, given the symmetry of the human head about the sagittal plane. The number of impacts to each generalized location was normalized by the total number of impacts to determine the percentage of impacts to each location. The percentage of impacts to each location was multiplied by the number of impacts that each player experiences throughout a season of full participation to determine the number of impacts to each location per player per season. Using this normalization approach allows for the location exposure to be scaled up or down based on the targeted number of impacts throughout a season.
Figure 1

Definition of parameters used to group impact locations. Azimuth and elevation values for the impact vector were computed for each impact based on the head acceleration data measured

Exposure per player per season for each impact location was further investigated by impact severity. The severity distributions were determined by using 51,008 head impacts that were recorded between 2006 and 2010 with the HIT System from players wearing Riddell VSR4 helmets (Riddell Inc, Elyria, OH).11,12 Only one helmet type was used so that potential variations in helmet response between types would be removed for this part of the analysis. This analysis is geared toward quantifying impact exposure as impact energy input to the helmet, rather than resulting head acceleration from impact as a function of helmet response. These impacts were categorized into the aforementioned impact locations and analyzed to determine head impact severity distributions. A Weibull distribution was fit to the peak linear head acceleration data for each impact location. The Weibull probability density function (pdf) takes the form of Eq. (2). The pdf represents the probability of a player experiencing any given head acceleration for one impact. The area under the entire pdf curve represents the probability of an impact resulting in any head acceleration, which is 100%. Moreover, the pdf can be integrated over specific bounds to determine the probability of an impact between the upper and lower bounds. The quality of the distribution fit was assessed by comparing an empirical cumulative distribution of the raw data to the computed Weibull cumulative density function (cdf) (Eq. 3). The Weibull pdf for each impact location was multiplied by the number of impacts to each impact location so that the area under each curve represented the number of impacts to each location.
$$ w_{pdf} = \frac{{\alpha (x - \theta )^{\alpha - 1} }}{{\beta^{\alpha } }}e^{{ - \left( {\frac{x - \theta }{\beta }} \right)^{\alpha } }} $$
(2)
$$ w_{cdf} = 1 - e^{{ - \left( {\frac{x - \theta }{\beta }} \right)^{\alpha } }} , $$
(3)
where wpdf is the Weibull probability density function, wcdf is the Weibull cumulative density function, α is the shape parameter, β is the scale parameter, θ is the location parameter, and x is the peak resultant head acceleration.
Next, it was important to decouple the specific helmet used to collect the head acceleration data from the overall STAR analysis. In order to do this, the head acceleration distributions were transformed to impact energy distributions, which are then independent of the specific helmet response. This was accomplished by defining the relationship between impact energy and resulting head acceleration for the specific helmet used for data collection. A total of three Large Riddell VSR4 helmets were tested on a medium NOCSAE head form using a NOCSAE-style drop tower. Drop heights ranged from 6 to 66 in. and were incremented by 6 in. For each of the three VSR4 helmets, these 11 drop heights were tested at the front, rear, side, and top impact locations that are defined by NOCSAE; resulting in a total of 132 tests. All equipment and instrumentation were calibrated to NOCSAE specification. Acceleration data were sampled at 20,000 Hz and filtered to NOCSAE specification (SAE J211 CFC 1000) for each test. Peak resultant head acceleration and Severity Index were computed for each test.15 A second order polynomial regression analysis was performed using a least squares technique to determine the relationship between drop height and the average VSR4 peak head acceleration for each impact location (Eq. 4).
$$ H = p_{1} a^{2} + p_{2} a + p_{3} , $$
(4)
where p1, p2, and p3 are regression coefficients, H is drop height, and a is peak resultant head acceleration.
Utilizing the polynomial regression models for each impact location, head acceleration distributions were transformed to drop height distributions. Once the distributions were representative of the impact energy to the helmet, not the head acceleration resulting from the response of the helmet to impact, the number of impacts associated with varying severities was determined for each impact location. Drop heights of 12, 24, 36, 48, and 60 in. were selected because they encompass the impact energies associated with a wide range of impacts seen on the field. The Weibull pdf for each location was integrated over the bounds defined in Table 1 to determine the number of impacts experienced by one player participating in one season for each drop height. Each number of impacts was rounded to the nearest integer, with the exception that any number between 0.0 and 1.0 was rounded up to 1.0. The “lowest” category was used to describe impacts with head accelerations <19 g. These impacts were separated from the analysis as they are not considered to be relevant given that the lowest reported concussion in the literature is 42 g.12,21 Moreover, the drop test height for this value would be only a few inches and that would not be a reasonable or necessary test to perform.
Table 1

Integral bounds used to determine the number of impacts associated with each drop height for each impact location

Drop height

Integral bounds

Lower

Upper

Lowest

0 g

19 g

12 in.

19 g

18 in.

24 in.

18 in.

30 in.

36 in.

30 in.

42 in.

48 in.

42 in.

54 in.

60 in.

60 in.

Infinity

Note that the lowest bounds include all impacts below 19 g (median impact). This drop height is not evaluated because these impacts are not associated with injury

Injury Risk

The sub-concussive head acceleration data set utilized in the risk analysis consisted of 62,974 head impacts that were recorded between 2009 and 2010 with the HIT System at Virginia Tech. Concussive head acceleration data were compiled from three separate studies using identical data collection protocols with the HIT System to create a data set of 32 clinically diagnosed concussions.3,12,21 These data sets are capable of accurately defining the distributions of head accelerations associated with sub-concussive and concussive impacts. Sub-concussive impacts were fit to a Weibull distribution in the form of Eq. (2). Concussive impacts were found to be normally distributed, and therefore were fit to a normal probability density function (Eq. 5).
$$ n_{pdf} = \frac{1}{{\sigma \sqrt {2\pi } }}e^{{\frac{{ - (x - \mu )^{2} }}{{2\sigma^{2} }}}} , $$
(5)
where npdf is the normal probability density function, μ is the mean, and σ is the standard deviation.

Published injury incidence rates for game participation were used to determine the proper weighting between sub-concussive and concussive head acceleration distributions. For collegiate athletes, there are 5.56 concussions per 1000 athletic exposures, where an athletic exposure is defined as one athlete participating in at least one play of one game or practice.1 For athletes in the National Football League (NFL), an injury incidence rate of 0.41 concussions per game was considered for professional athletes.37 To relate the number of concussions to the number of sub-concussive impacts for both the collegiate and NFL groups, it was assumed that the median player experiences 16.3 impacts per game.6 For collegiate athletes, 5.56 concussions per 1000 games played with 16.3 impacts per game per player can be expressed as an injury incidence rate of 0.341 concussions per 1000 impacts. For NFL athletes, 0.41 concussions per game with 88 players participating in each game and 16.3 impacts per game per player can be expressed as 0.286 concussions per 1000 impacts. It is important to note that current research suggests that as many as 53% of concussions go unreported.30 This underreporting rate was applied to both calculated injury incidence rates, resulting in 0.726 concussions per 1000 impacts for collegiate athletes and 0.609 concussions per 1000 impacts for NFL athletes.

Next, estimated injury incidence rates were used to combine the sub-concussive and concussive head acceleration distributions in order to have the proper sub-concussive to concussive impact ratio. The weighting of each distribution based on injury incidence rates allows for an unbiased risk analysis. A logistic regression analysis based on the weighted sub-concussive and concussive head acceleration distributions was used to express risk as a function of head acceleration for both the collegiate and NFL groups (Eq. 6). The regression coefficients were determined using a generalized linear model technique.
$$ R(a) = \frac{1}{{1 + e^{ - (\alpha + \beta x)} }}, $$
(6)
where R(a) is the probability of injury, α and β are regression coefficients.

STAR Value Assessment

For illustrative purposes, a STAR value was calculated using hypothetical head acceleration data. Example peak resultant head acceleration values were generated for the 12, 24, 36, 48, and 60 in. drop tests for each impact location. Equation (6) was used to calculate probability of concussion for each testing configuration based on collegiate injury incidence rates. Risk of concussion was multiplied by the number of impacts that a player experiences for each testing configuration and summed to calculate an overall concussion incidence (Eq. 1).

Results

Head Impact Exposure

Overall, impacts to the front of the helmet occurred most frequently, and were followed by impacts to the rear, top, and side of the helmet (Fig. 2). Using these percentages, the number of impacts to each impact location for a single player participating in a complete season was computed based on the assumption that a total of 1000 head impact are experienced. This transformation gives that for a single season, a player will experience 347 impacts to the front of the helmet, 319 impacts to the rear of the helmet, 171 impacts to the top of the helmet, and 163 impacts to the sides of the helmet.
Figure 2

Percentage of impacts associated with each impact location (left). Assuming 1000 impacts per player per season, the number of impacts to each location was determined (right)

Impact severity distributions in terms of peak resultant head acceleration were determined for each impact location. Each distribution was fit to a Weibull pdf (Eq. 2) and Table 2 displays the computed parameters for each impact location. As an example, Fig. 3 displays the Weibull pdf for impacts to the front of the helmet. A comparison of the Weibull cdf to the empirical cdf is also shown in Fig. 3. The quality of fit was consistently good for all impact locations.
Table 2

Weibull probability density function parameters for impact severity distributions separated by each impact location

Impact location

α

β

θ

Front

15.1228

0.8897

10.0

Rear

16.5724

0.9333

10.0

Side

13.3692

0.8837

10.0

Top

21.8923

0.8899

10.0

Figure 3

Weibull probability density function fitted to head accelerations resulting from impacts to the front of the helmet (left). Quality of fit can be investigated through comparison of the Weibull cumulative density function to an empirical cumulative density function of the same data (right)

A total of 132 drop tests were performed investigating average head acceleration as a function of drop height and impact location for three Riddell VSR4 helmets. Table 3 displays the regression coefficients and R2 values for the relationship between head acceleration and drop height for these tests (Eq. 4). Figure 4 displays how the number of impacts for each drop height was determined from the Weibull distribution for the front impact location, while Table 4 displays the number of impacts for each drop height for all impact locations.
Table 3

The second order polynomial regression coefficients and R2 values for drop height and head acceleration relationships (Eq. 4)

 

p1

p2

p3

R2

Front

0.009

1.428

15.010

0.999

Rear

−0.026

3.826

5.042

0.992

Side

−0.023

3.241

12.975

0.998

Top

−0.015

2.870

11.403

0.997

Figure 4

Number of impacts associated with each drop height for the front impact location, with integral bounds displayed for each drop height

Table 4

Number of impacts per player per season associated with each impact location and drop height used in the STAR evaluation methodology, which is representative of the 90th percentile player

Drop height

Front

Rear

Side

Top

Impacts <19 g

164

139

81

63

12 in.

138

165

75

85

24 in.

31

11

4

14

36 in.

10

2

1

5

48 in.

3

1

1

2

60 in.

1

1

1

2

Total

347

319

163

171

Note: impacts <19 g are not considered in the STAR testing protocol

Injury Risk

The 62,974 sub-concussive impacts had an average head acceleration of 26 ± 20 g (median of 19 g), while concussive impacts had an average head acceleration of 105 ± 27 g (median of 103 g). Figure 5 displays the pdfs for sub-concussive and concussive impacts. Injury risk curves based on the collegiate and NFL injury rates are shown in Fig. 6 with corresponding parameter values for Eq. (6) shown in Table 5.
Figure 5

Weibull probability density function for all sub-concussive impacts and normal probability density function for all concussive impacts (left). Comparison of distribution fits for sub-concussive and concussive data to empirical data using cumulative density functions (right)

Figure 6

Injury risk curves based on the collegiate and NFL injury rates using resultant peak linear head acceleration

Table 5

Logistic regression parameters for the college and NFL injury incidence rate-based risk curves (Eq. 6)

 

α

β

Collegiate

−9.805

0.051

Professional

−9.928

0.497

STAR Value Assessment

Table 6 displays the hypothetical head accelerations associated with a hypothetical helmet for each drop height and impact location configuration that result in a STAR value of 1.501. This value can be interpreted on average as: if one player wore the hypothetical helmet for one season of 1000 head impacts in practices and games, then the player would theoretically sustain approximately 1.5 concussions.
Table 6

STAR value assessment of a hypothetical helmet that resulted in the listed head accelerations for each drop height and impact location, and exposure per season is representative of 1000 impacts for the 90th percentile player

Impact location

Drop height

Peak G

Risk of injury

Exposure per season

Incidence per season

Front

Impacts <19 g

0.0000

164

0.00

Front

12 in.

50

0.0007

138

0.10

Front

24 in.

70

0.0019

31

0.06

Front

36 in.

90

0.0053

10

0.04

Front

48 in.

120

0.0239

3

0.07

Front

60 in.

150

0.1011

1

0.10

Side

Impacts <19 g

0.0000

81

0.00

Side

12 in.

55

0.0009

75

0.07

Side

24 in.

90

0.0053

4

0.02

Side

36 in.

105

0.0113

1

0.01

Side

48 in.

125

0.0306

1

0.03

Side

60 in.

150

0.1011

1

0.10

Rear

Impacts <19 g

0.0000

139

0.00

Rear

12 in.

65

0.0015

165

0.25

Rear

24 in.

90

0.0053

11

0.06

Rear

36 in.

120

0.0239

2

0.05

Rear

48 in.

135

0.0499

1

0.05

Rear

60 in.

155

0.1267

1

0.13

Top

Impacts <19 g

0.0000

63

0.00

Top

12 in.

50

0.0007

85

0.06

Top

24 in.

80

0.0032

14

0.04

Top

36 in.

100

0.0088

5

0.04

Top

48 in.

120

0.0239

2

0.05

Top

60 in.

145

0.0809

2

0.16

   

STAR value:

1.501

Discussion

It is important to note that no helmet will ever be perfect and that there will always be a risk of head injury in any sporting activity, regardless of the effectiveness of the protective equipment. There are many variables that affect the risk of concussion in sports, with player history and genetic differences likely dominating the variation. Having noted this, the primary purpose of this study was to introduce the STAR equation and methodology, which can be used to evaluate football helmet performance. The STAR evaluation system combines true head impact exposure with injury risk to predict concussion incidence for a specific helmet throughout the course of a football season. Both the impact exposure and injury risk components of the STAR evaluation system are based on real-world data collected from human athletes; and as a result, they reflect the impacts that the average player actually experiences on the field. The methods used to incorporate head impact exposure and injury risk are parametric analyses that determine complete distributions, and are similar to previous work using a more limited data set.14

The STAR evaluation system is not intended to replace or criticize the role of NOCSAE in ensuring the safety of athletic equipment. In fact, our data illustrate that the NOCSAE-style drop test and head form configuration do an excellent job of replicating the type of impacts football players experience on the field. Figure 7 displays a comparison of typical NOCSAE head form impact response to football head impact acceleration corridors for the front and side impact locations. This figure illustrates that the NOCSAE head form accurately models head acceleration pulse shape and duration for impacts in football. For this reason, the NOCSAE-style drop tests are utilized as the core of the STAR system.
Figure 7

Comparison of typical normalized NOCSAE head form impact response to football head impact acceleration corridors for the front (left) and side (right) impact locations. All acceleration curves were normalized to their peak values

For the STAR evaluation system, a single collegiate player participating in every game and practice throughout a football season was assumed to experience 1000 head impacts. The head impact exposure per player per season used in this study is consistent with other published studies. Guskiewicz et al.21 reported that the average collegiate football player experiences 950 impacts per season. In addition, Schnebel et al.43 found the average collegiate player to experience 1353 impacts per season. However, it should be noted that these numbers may not be directly applicable to high school athletes. Broglio et al.4 measured an average of 565 impacts per player in high school athletes. Differences among these studies can be attributed to the number of athletic exposures each player was exposed to, as well as the relative playing time per event. In any case, the exposure number is a scalable constant and can be adjusted for any level (Table 7).
Table 7

Exposure per season for each drop configuration is presented for 50th, 75th, and 90th percentile impact exposures

Impact location

Drop height

Exposure per season

50th percentile

75th percentile

90th percentile

Front

Impacts <19 g

71

110

164

Front

12 in.

61

94

138

Front

24 in.

14

21

31

Front

36 in.

4

7

10

Front

48 in.

1

2

3

Front

60 in.

1

1

1

Side

Impacts <19 g

33

53

81

Side

12 in.

33

51

75

Side

24 in.

2

3

4

Side

36 in.

1

1

1

Side

48 in.

1

1

1

Side

60 in.

1

1

1

Rear

Impacts <19 g

59

93

139

Rear

12 in.

73

112

165

Rear

24 in.

5

8

11

Rear

36 in.

1

1

2

Rear

48 in.

1

1

1

Rear

60 in.

1

1

1

Top

Impacts <19 g

28

44

63

Top

12 in.

37

58

85

Top

24 in.

6

9

14

Top

36 in.

2

3

5

Top

48 in.

1

1

2

Top

60 in.

1

1

2

 

Total impacts:

438

677

1000

Furthermore, the head impact exposure per player per impact location is nearly identical to other published studies. Table 8 compares the impact location weightings used in the STAR evaluation system to that of another study that utilized the same data collection methodologies.31 Moreover, these results are consistent with a study of three collegiate teams.6 It is important to note that exposure on a player position basis was not considered although it has been shown the impact frequency and magnitude vary by position.6,31 Since close to every player on the field is instrumented, the exposure values reported in this study represent an average exposure for all player positions. Moreover, most players play multiple positions and no position-specific helmets currently exist.
Table 8

Comparison of the distribution of impact locations based on data collected from instrumented collegiate football players

Impact location

Percentage of all impacts

VT data

Mihalik et al.31

Front

34.7

35.9

Rear

31.9

30.9

Side

16.3

14.4

Top

17.1

18.8

In order to determine the proper weighting between sub-concussive and concussive head acceleration distributions, published injury incidence rates were considered. Most injury incidence rates are reported as the number of concussions per 1000 athletic exposures, where an athletic exposure is defined as one athlete participating in at least one play of one game or practice. Table 9 is a summary of published injury incidences rates for concussion in football.1,9,20 Booher et al.1 reported higher injury incidence rates because that study included injuries that did not result in loss of playing time, where Guskiewicz et al.20 and Dick et al.9 only considered injuries associated with loss of playing time. In addition, Pellman et al.37 reported an injury incidence rate of 0.41 concussions per game in the NFL. To side with conservatism, only game data were considered and injury incidence rates of 5.56 concussions per 1000 athletic exposures and 0.41 concussions per game were analyzed in this study. Furthermore, an underreporting rate was applied to the injury incidence rates, as previous studies have suggested that underreporting is a prevalent issue with the diagnosis of concussion.2,13,40,46
Table 9

Published injury incidence rates expressed as concussions per 1000 athletic exposures in collegiate football1,9,20

Study

Concussions/1000 A–E

Game

Practice

Total

Booher et al.

5.56

0.25

0.74

Guskiewicz et al.

3.81

0.47

0.81

Dick et al.

2.34

0.21

0.37

Injury risk curves were generated from a logistic regression analysis of properly weighted sub-concussive and concussive data distributions. Logistic risk curves were deemed appropriate due to the censored nature of the data, as the exact head accelerations associated with the onset of injury cannot be identified. While there are advocates of non-parametric approaches,10 such approaches are only representative of the data that are experimentally recorded.26 Moreover, Kent and Funk26 have shown that parametric injury risk functions typically fall within the 95% confidence intervals of non-parametric injury risk functions for any given data set. They noted the largest discrepancy between the two approaches at the risk function tails (around 0 and 100% risk) where experimental data was limited. The real-world data that the injury risk curves are based on are unique in that the distributions are well defined throughout the entire continuum of head accelerations experienced by football players. This likely provides a better representation of actual risk around the tails of the risk curves than most biomechanical experiments used to characterize risk. The parametric approach utilized in this study was chosen so that the risk curve could be representative of the entire population, rather than only the data collected in the experiment.

The analysis used in this study is unique because it weights sub-concussive and concussive acceleration distributions based on calculated injury incidence rates so that the injury risk curves are unbiased. Both the collegiate and NFL injury incidence rates produced very similar risk curves. Figure 8 compares the injury risk curves determined in this study to that of previously published studies. Pellman et al.38 created injury risk curves based on reconstructed concussive impacts from NFL football games. Hybrid III crash test dummies were used to reconstruct 31 impacts based on game video. The risk curves from that experiment are limited because they are biased to concussive impacts. Their analysis was based on 25 concussive and 33 sub-concussive data points, and therefore overestimates risk of concussion. Interestingly, the average linear acceleration associated with concussion reported by Pellman et al.38 was 98 ± 27 g, which is nearly identical to the average concussive linear acceleration collected from human volunteers in this study (105 ± 27 g). Although acceleration magnitude is accurately captured using the Hybrid III, it is likely that acceleration duration is less accurate due to the stiffness of the Hybrid III neck.22 Funk et al.14 quantified risk using unique statistical analysis of data collected from human volunteers. However, that risk curve was based on only four concussive data points, and was thought to underestimate risk. For head accelerations that are associated with <10% risk, the risk curve from Funk et al.14 closely matches the risk curves generated from this study. While the head accelerations associated with concussion are well defined in this study, injury risk associated with these head accelerations is low (<10%) because impacts of these severities occur frequently in collegiate football without concussion. Since true head impact exposure was paired with the concussive impact distribution, these risk curves are thought to be accurate.
Figure 8

Comparison of injury risk curves generated in this study to that of previously published research. The cumulative density function of the normally distributed concussive data set (concussive cdf) is displayed to illustrate the head accelerations associated with only concussive impacts

The injury risk curves utilized in the STAR equation currently only incorporate linear acceleration. It should be noted that these linear accelerations are associated with impact durations of 8–12 ms. Although rotational acceleration is not considered in this analysis, work is underway investigating human tolerance to rotational kinematics.40 Furthermore, rotational acceleration cannot be applied to this testing methodology until protocols for testing rotational kinematics are developed. It has also been suggested that a composite injury metric consisting of many biomechanical parameters may have the best predictive capabilities for concussion.17 The objective of these biomechanical parameters is to predict the tissue-level response of the brain to impact. While researchers are currently investigating the physiologic response of the brain tissue to mechanical insult, linking the tissue-level response of the brain to the kinematics of the skull remains a challenge.32,33 Finite element models will likely be an important factor in understanding this relationship.45 Once a better understanding of these relationships is gathered, more complex methods can be used to predict brain injury in the future.

The STAR equation is determined through relating impact exposure to drop test performance for a specific helmet. A helmet that has lower accelerations associated with each drop test will therefore predict a lower incidence of concussions. Traditionally, it has been challenging to account for all the scenarios in sports that can cause injury in a laboratory testing.41,42,44 This is a result of not knowing the true exposure of impact, as well as the vast amount of possible scenarios in sports. The STAR evaluation system generalizes all possible head impacts in football into 24 NOCSAE-style drop tests, which consists of four impact locations and six impact severities. This system can be a valuable tool in educating consumers on helmet performance, much like the current NCAP rating system aids consumers intending to purchase automobiles. The STAR system is conceptual and its ability to differentiate between helmet designs will be addressed in subsequent work. In the future, the underlying principles of the STAR evaluation system can be applied to other levels of football, as well as other sports, once the exposure data are known.

Notes

Acknowledgments

The authors gratefully acknowledge our sponsors for this research, including the National Highway Traffic Safety Administration and National Institutes of Health (National Institute for Child Health and Human Development) (Contract No. R01HD048638).

References

  1. 1.
    Booher, M. A., J. Wisniewski, B. W. Smith, and A. Sigurdsson. Comparison of reporting systems to determine concussion incidence in NCAA Division I collegiate football. Clin. J. Sport Med. 13:93–95, 2003.PubMedCrossRefGoogle Scholar
  2. 2.
    Broglio, S. P., M. S. Ferrara, S. G. Piland, R. B. Anderson, and A. Collie. Concussion history is not a predictor of computerised neurocognitive performance. Br. J. Sports Med. 40:802–805, 2006; discussion 5.PubMedCrossRefGoogle Scholar
  3. 3.
    Broglio, S. P., B. Schnebel, J. J. Sosnoff, S. Shin, X. Fend, X. He, and J. Zimmerman. Biomechanical properties of concussions in high school football. Med. Sci. Sports Exerc. 42:2064–2071, 2010.PubMedCrossRefGoogle Scholar
  4. 4.
    Broglio, S. P., J. J. Sosnoff, S. Shin, X. He, C. Alcaraz, and J. Zimmerman. Head impacts during high school football: a biomechanical assessment. J. Athl. Train. 44:342–349, 2009.PubMedCrossRefGoogle Scholar
  5. 5.
    Crisco, J. J., J. J. Chu, and R. M. Greenwald. An algorithm for estimating acceleration magnitude and impact location using multiple nonorthogonal single-axis accelerometers. J. Biomech. Eng. 126:849–854, 2004.PubMedCrossRefGoogle Scholar
  6. 6.
    Crisco, J. J., R. Fiore, J. G. Beckwith, J. J. Chu, P. G. Brolinson, S. Duma, T. W. McAllister, A. C. Duhaime, and R. M. Greenwald. Frequency and location of head impact exposures in individual collegiate football players. J. Athl. Train. 45:549–559, 2010.PubMedCrossRefGoogle Scholar
  7. 7.
    Crisco, J. J., and R. M. Greenwald. Let’s get the head further out of the game: a proposal for reducing brain injuries in helmeted contact sports. Curr. Sports Med. Rep. 10:7–9, 2011.PubMedGoogle Scholar
  8. 8.
    Daneshvar, D. H., C. J. Nowinski, A. C. McKee, and R. C. Cantu. The epidemiology of sport-related concussion. Clin. Sports Med. 30:1–17, 2011.PubMedCrossRefGoogle Scholar
  9. 9.
    Dick, R., M. S. Ferrara, J. Agel, R. Courson, S. W. Marshall, M. J. Hanley, and F. Reifsteck. Descriptive epidemiology of collegiate men’s football injuries: national collegiate athletic association injury surveillance system, 1988–1989 through 2003–2004. J. Athl. Train. 42:221–233, 2007.PubMedGoogle Scholar
  10. 10.
    Domenico, L. D., and G. Nusholtz. Comparison of Parametric and Non-Parametric Methods for Determining Injury Risk. SAE Technical Paper Series, SAE 2003-01-1362, 2003.Google Scholar
  11. 11.
    Duma, S. M., S. J. Manoogian, W. R. Bussone, P. G. Brolinson, M. W. Goforth, J. J. Donnenwerth, R. M. Greenwald, J. J. Chu, and J. J. Crisco. Analysis of real-time head accelerations in collegiate football players. Clin. J. Sport Med. 15:3–8, 2005.PubMedCrossRefGoogle Scholar
  12. 12.
    Duma, S. M., and S. Rowson. Every newton hertz: a macro to micro approach to investigating brain injury. Conf. Proc. IEEE Eng. Med. Biol. Soc. 1:1123–1126, 2009.Google Scholar
  13. 13.
    Duma, S. M., and S. Rowson. Past, present, and future of head injury research. Exerc. Sport Sci. Rev. 39:2–3, 2011.PubMedCrossRefGoogle Scholar
  14. 14.
    Funk, J. R., S. M. Duma, S. J. Manoogian, and S. Rowson. Biomechanical risk estimates for mild traumatic brain injury. Annu. Proc. Assoc. Adv. Automot. Med. 51:343–361, 2007.PubMedGoogle Scholar
  15. 15.
    Gadd, C. W. Use of a weighted-impulse criterion for estimating injury hazard. In: Proceedings of the 10th Stapp Car Crash Conference, SAE 660793, 1966.Google Scholar
  16. 16.
    Gavett, B. E., R. A. Stern, and A. C. McKee. Chronic traumatic encephalopathy: a potential late effect of sport-related concussive and subconcussive head trauma. Clin. Sports Med. 30:179–188, 2011.PubMedCrossRefGoogle Scholar
  17. 17.
    Greenwald, R. M., J. T. Gwin, J. J. Chu, and J. J. Crisco. Head impact severity measures for evaluating mild traumatic brain injury risk exposure. Neurosurgery 62:789–798, 2008; discussion 98.PubMedCrossRefGoogle Scholar
  18. 18.
    Guskiewicz, K. M., S. W. Marshall, J. Bailes, M. McCrea, R. C. Cantu, C. Randolph, and B. D. Jordan. Association between recurrent concussion and late-life cognitive impairment in retired professional football players. Neurosurgery 57:719–726, 2005; discussion 26.PubMedCrossRefGoogle Scholar
  19. 19.
    Guskiewicz, K. M., S. W. Marshall, J. Bailes, M. McCrea, H. P. Harding, Jr., A. Matthews, J. R. Mihalik, and R. C. Cantu. Recurrent concussion and risk of depression in retired professional football players. Med. Sci. Sports Exerc. 39:903–909, 2007.PubMedCrossRefGoogle Scholar
  20. 20.
    Guskiewicz, K. M., M. McCrea, S. W. Marshall, R. C. Cantu, C. Randolph, W. Barr, J. A. Onate, and J. P. Kelly. Cumulative effects associated with recurrent concussion in collegiate football players: the NCAA concussion study. JAMA 290:2549–2555, 2003.PubMedCrossRefGoogle Scholar
  21. 21.
    Guskiewicz, K. M., J. P. Mihalik, V. Shankar, S. W. Marshall, D. H. Crowell, S. M. Oliaro, M. F. Ciocca, and D. N. Hooker. Measurement of head impacts in collegiate football players: relationship between head impact biomechanics and acute clinical outcome after concussion. Neurosurgery 61:1244–1253, 2007.PubMedCrossRefGoogle Scholar
  22. 22.
    Gwin, J. T., J. J. Chu, S. G. Diamond, P. D. Halstead, J. J. Crisco, and R. M. Greenwald. An investigation of the NOCSAE linear impactor test method based on in vivo measures of head impact acceleration in American football. J. Biomech. Eng. 132:011006, 2010.PubMedCrossRefGoogle Scholar
  23. 23.
    Hackney, J. R., and C. J. Kahane. The New Car Assessment Program: Five Star Rating System and Vehicle Safety Performance Characteristics. SAE Technical Paper Series, SAE 851245, 1995.Google Scholar
  24. 24.
    Hanlon, E., and C. Bir. Validation of a wireless head acceleration measurement system for use in soccer play. J. Appl. Biomech. 26:424–431, 2010.PubMedGoogle Scholar
  25. 25.
    Janda, D. H., C. A. Bir, and A. L. Cheney. An evaluation of the cumulative concussive effect of soccer heading in the youth population. Inj. Control Saf. Promot. 9:25–31, 2002.PubMedCrossRefGoogle Scholar
  26. 26.
    Kent, R. W., and J. R. Funk. Data Censoring and Parametric Distribution Assignment in the Development of Injury Risk Functions from Biomechanical Data. SAE Technical Paper Series, SAE 2004-01-0317, 2004.Google Scholar
  27. 27.
    King, A. I., K. H. Yang, L. Zhang, W. Hardy, and D. C. Viano. Is head injury caused by linear or angular acceleration? In: Proceedings of the International Research Conference on the Biomechanics of Impact (IRCOBI), 2003.Google Scholar
  28. 28.
    Laituri, T. R., S. Henry, K. Sullivan, and M. Nutt. Considerations of “combined probability of injury” in the next-generation USA frontal NCAP. Traffic Inj. Prev. 11:371–381, 2010.PubMedCrossRefGoogle Scholar
  29. 29.
    Langlois, J. A., W. Rutland-Brown, and M. M. Wald. The epidemiology and impact of traumatic brain injury: a brief overview. J. Head Trauma Rehabil. 21:375–378, 2006.PubMedCrossRefGoogle Scholar
  30. 30.
    McCrea, M., T. Hammeke, G. Olsen, P. Leo, and K. Guskiewicz. Unreported concussion in high school football players: implications for prevention. Clin. J. Sport Med. 14:13–17, 2004.PubMedCrossRefGoogle Scholar
  31. 31.
    Mihalik, J. P., D. R. Bell, S. W. Marshall, and K. M. Guskiewicz. Measurement of head impacts in collegiate football players: an investigation of positional and event-type differences. Neurosurgery 61:1229–1235, 2007; discussion 35.PubMedCrossRefGoogle Scholar
  32. 32.
    Morrison, B., III, D. F. Meaney, and T. K. McIntosh. Mechanical characterization of an in vitro device designed to quantitatively injure living brain tissue. Ann. Biomed. Eng. 26:381–390, 1998.PubMedCrossRefGoogle Scholar
  33. 33.
    Morrison, B., III, K. E. Saatman, D. F. Meaney, and T. K. McIntosh. In vitro central nervous system models of mechanically induced trauma: a review. J. Neurotrauma 15:911–928, 1998.PubMedCrossRefGoogle Scholar
  34. 34.
    NOCSAE. Standard test method and equipment used in evaluating the performance characteristics of protective headgear/equipment. National Operating Committee on Standards for Athletic Equipment: NOCSAE DOC (ND)001-08m10, 2009.Google Scholar
  35. 35.
    Omalu, B. I., S. T. DeKosky, R. L. Hamilton, R. L. Minster, M. I. Kamboh, A. M. Shakir, and C. H. Wecht. Chronic traumatic encephalopathy in a national football league player: part II. Neurosurgery 59:1086–1092, 2006; discussion 92–93.PubMedGoogle Scholar
  36. 36.
    Omalu, B. I., S. T. DeKosky, R. L. Minster, M. I. Kamboh, R. L. Hamilton, and C. H. Wecht. Chronic traumatic encephalopathy in a national football league player. Neurosurgery 57:128–134, 2005; discussion 34.PubMedCrossRefGoogle Scholar
  37. 37.
    Pellman, E. J., J. W. Powell, D. C. Viano, I. R. Casson, A. M. Tucker, H. Feuer, M. Lovell, J. F. Waeckerle, and D. W. Robertson. Concussion in professional football: epidemiological features of game injuries and review of the literature–part 3. Neurosurgery 54:81–94, 2004; discussion 6.PubMedCrossRefGoogle Scholar
  38. 38.
    Pellman, E. J., D. C. Viano, A. M. Tucker, I. R. Casson, and J. F. Waeckerle. Concussion in professional football: reconstruction of game impacts and injuries. Neurosurgery 53:799–812, 2003; discussion 4.PubMedGoogle Scholar
  39. 39.
    Rowson, S., J. G. Beckwith, J. J. Chu, D. S. Leonard, R. M. Greenwald, and S. M. Duma. A six degree of freedom head acceleration measurement device for use in football. J. Appl. Biomech. 27:8–14, 2011.PubMedGoogle Scholar
  40. 40.
    Rowson, S., G. Brolinson, M. Goforth, D. Dietter, and S. M. Duma. Linear and angular head acceleration measurements in collegiate football. J. Biomech. Eng. 131:061016, 2009.PubMedCrossRefGoogle Scholar
  41. 41.
    Rowson, S., C. McNally, and S. M. Duma. Can footwear affect achilles tendon loading? Clin. J. Sport Med. 20:344–349, 2010.PubMedCrossRefGoogle Scholar
  42. 42.
    Rowson, S., D. E. McNeely, P. G. Brolinson, and S. M. Duma. Biomechanical analysis of football neck collars. Clin. J. Sport Med. 18:316–321, 2008.PubMedCrossRefGoogle Scholar
  43. 43.
    Schnebel, B., J. T. Gwin, S. Anderson, and R. Gatlin. In vivo study of head impacts in football: a comparison of national collegiate athletic association division I versus high school impacts. Neurosurgery 60:490–495, 2007; discussion 5–6.PubMedCrossRefGoogle Scholar
  44. 44.
    Shain, K. S., M. L. Madigan, S. Rowson, J. Bisplinghoff, and S. M. Duma. Analysis of the ability of catcher’s masks to attenuate head accelerations on impact with a baseball. Clin. J. Sport Med. 20:422–427, 2010.PubMedCrossRefGoogle Scholar
  45. 45.
    Takhounts, E. G., S. A. Ridella, V. Hasija, R. E. Tannous, J. Q. Campbell, D. Malone, K. Danelson, J. Stitzel, S. Rowson, and S. Duma. Investigation of traumatic brain injuries using the next generation of simulated injury monitor (simon) finite element head model. Stapp Car Crash J. 52:1–31, 2008.PubMedGoogle Scholar
  46. 46.
    Williamson, I. J., and D. Goodman. Converging evidence for the under-reporting of concussions in youth ice hockey. Br. J. Sports Med. 40:128–132, 2006; discussion 32.PubMedCrossRefGoogle Scholar

Copyright information

© Biomedical Engineering Society 2011

Authors and Affiliations

  1. 1.Virginia Tech – Wake Forest University, School of Biomedical Engineering and SciencesBlacksburgUSA

Personalised recommendations