Traumatic brain injury is a major cause of morbidity and mortality (5,6) and is a major concern in motor sports safety. Head injuries are responsible for up to half of trauma deaths and account for most cases of permanent disability after injury (7). In the United States from 1989 to 1998, an annual average of 53,288 deaths was associated with traumatic brain injury, and 34% of these were secondary to motor vehicle crashes (1).
Significant brain injuries are thought to occur when an impact exceeds the threshold for neuronal injury. Previous studies have evaluated the threshold for a single impact to cause significant brain injury when a force is applied via a direct blow to the head or brain (3,4,9). The mechanism for head injury in motor vehicle crashes, however, is very different to those investigated in these studies.
The investigation of automobile crashes to understand factors of occupant injuries and to develop measures for injury prevention is a well-established practice. Melvin et al. (8) published a paper discussing the addition of an accident data recorder (ADR) as a method to gather data in racing crashes and gave a brief report of crashes to date, but they did not investigate specific injury thresholds. Wright (11) also reported ADR data gathered from Formula 1 cars in 2000, but had insufficient data to draw definite conclusions about correlation with injuries. We are unaware of any studies that yet have established an ability to predict brain injury using car-mounted accelerometers in actual motor vehicle crashes. We are attempting to estimate injury thresholds in Indy Racing League (IRL) car crashes with hopes that in the future we can develop interventions to decrease the likelihood of reaching these thresholds and of sustaining injury. In addition, we can utilize this information to better equip our IRL safety teams as they arrive to the crash scene. As thresholds are established, we can set the accelerometer on the car to trigger a light that signifies to the safety personnel that a certain threshold had been reached. We, therefore, conducted a study to compare the percent of IRL car drivers sustaining brain injury in those crashes where the chassis sustained a peak impact of ≥50 vehicle G forces (G) versus those with a lesser impact.
We analyzed prospectively collected data regarding IRL circuit car crashes. Data were collected from January 1996 to July 2003 and placed in a collision database. We extracted from this crash database information on the vehicle G forces in the crash and information about the driver and injuries sustained. The study was reviewed and approved by the Indiana University investigational review board.
In 1996, the IRL began prospectively collecting data on all racing crashes in the series with the goal of improving safety through the application of crash protection research. An IST Model EDR-3 Environmental Shock and Vibration Recorder (Instrument Sensor Technology, Okemos, MI) was used initially (1996-1999), and Delphi Automotive Systems Accident Data Recorders (ADR2) have captured impact data since and are currently in place. The accident data recorders are mounted in all IRL cars for all events. Data recorded by the ADR was analyzed with PI software and entered into the data base.
An ADR was mounted on the floor of the car's chassis below the driver and as near the center of gravity of the car as possible and registered crash pulse data. The IST Model EDR-3 device is a standalone, three-axis acceleration sensing data recorder that retains data for the 10 most severe events where acceleration exceeds a set threshold (8,11). The currently used ADR2 is a 171 × 188 × 56 mm unit, weighing 1.6 kg. Triggering, logging, and software information is similar to that described by Wright (11) in Formula 1 cars. The ADR2 begins logging on a signal provided by the engine control module named "engine run." This indicates when the engine starts and stops running. When the signal is sent that the engine starts, the ADR2 starts logging and continues logging until the signal goes to the "off" condition. During a crash, that occurs on impact or very soon after. It continues to log for 1 min after the off condition to record any secondary impacts the car might experience. We have crash pulse (accelerometer) data on 374 crashes in this time period. An example of crash pulse data from a crash resulting in a head injury is displayed in Figure 1. Analysis was performed on all crashes in which accelerometer data were recorded. Because not all crashes are sufficiently severe to trigger the recorders, the total number of recordings is always less than the number of crashes. Over the years, few crashes occurred in which the recorders have failed to capture data, either because of human errors or component failures.
Driver injury data were abstracted by IRL nurses at the time of the accident from physician diagnosis (ICD-9CM discharge diagnosis code) both at the track emergency center and any health care facility to which drivers were referred. The nurses also recorded the specifics of each injury, including the specific symptoms and the duration of these symptoms.
We used the head injury aspect of the abbreviated injury scale (AIS) to evaluate brain injury severity (10). (Table 1) The AIS is a simple injury scale with values that represent a compromise between energy absorption by injury and "threat-to-life" criteria. Individual injuries are rated separately in addition to the overall rating that may be assigned.
We defined head injury as listed in the AIS; including anyone who sustained a concussion, as defined by loss of consciousness or altered mental status and anyone with an identifiable anatomic injury on computed tomography (CT), magnetic resonance imaging (MRI), or operative findings. Although this includes injuries such as fractures and severe lacerations, none of the drivers in our analysis sustained these injuries without a brain injury.
The main study outcome was the percent of drivers who sustained a head injury. We compared the likelihood of injury in those sustaining a maximal impact of ≥50 G versus those with a lesser impact. Crashes were classified into either having a maximal chassis impact force of <50 G or a maximal impact force of ≥50 G. Fisher's exact test was used to compare the prevalence of head injures between the two groups. An odds ratio (OR) and an exact 95% confidence interval (CI) were estimated to assess the magnitude of sustaining an injury from a vehicle impact of ≥50 G. In addition, the mean maximal vehicle G force was estimated for those crashes where an injury occurred and for those where none occurred. These means were compared using a Student's t-test. The mean difference and a 95% CI were estimated. The Student's t-test was also used to compare the mean impact G force of an AIS2 or greater head injury with an AIS1 head injury and to compare the mean impact G force of an AIS3 head injury with an AIS1 or AIS2 head injury. Because the cut point of 50 G was based on empirical observation, we performed a logistic regression analysis to assess if a linear relationship between the likelihood of a head injury and the vehicle impact G force exists. From this model, we estimated the G force needed in order to observe head injury rates of 10, 50, and 90% (2). Analyses were performed using the statistical software SAS version 8.2 (SAS Institute, Cary, NC).
In this time period, accelerometer data were collected and analyzed on 374 crashes. The median peak vehicle G recorded at the time of impact was 50 G (range 5-239). Of the 374 crashes, 186 (49.7%) had a recorded impact of <50 G and 188 (50.3%) were ≥50 G. A driver in a crash with a maximal vehicle impact of ≥50 G developed a head injury 16.0% (30/188) of the time versus 1.6% (3/186) of those with a crash of <50 G (P < 0.001). This gives an odds ratio of 11.6 (95% CI: 3.5, 60.1).
The mean peak vehicle G force in those crashes where a head injury occurred was 79.6 (SD 28.5) compared with 50.6 (SD 28.0) in those with no head injury (P < 0.001). This is a difference of 29.0 G (95% CI: 18.9, 39.0). In those 33 crashes where a head injury occurred, no significant differences were seen in the mean peak G force sustained between more severe and less severe head injuries (Fig. 2). The mean chassis impact G for the 23 drivers sustaining an AIS2 or greater head injury was 79.5 (SD 28.5), whereas the mean G for the 10 drivers with a mild head injury was 79.8 (SD 30.0; P = 0.981). For the 7 drivers with an AIS3+ head injury, the mean peak chassis G force was 77.0 (SD 19.6), whereas for the 26 drivers with a less severe head injury, the mean peak G force was 80.3 (SD 30.7; P = 0.734).
A logistic regression model was used to model the probability of a head injury as a linear function of the vehicle impact G force. Results from this analysis yielded parameter estimates of −4.028 (SE 0.453; P < 0.001) for the intercept term and 0.027 (SE 0.006; P < 0.001) for the peak G term. This gives an OR of 1.15 (95% CI: 1.08, 1.21) for each five unit increase in the peak vehicle impact G force. Using this logistic model, we estimated that a crash impact force of 68 G (95% CI: 52, 83) would result in a 10% probability of a head injury. Likewise, a force of 149 G (95% CI: 123, 209) would result in a 50% probability and a force of 230 G (95% CI: 182, 349) would result in a 90% probability of a head injury.
We analyzed impact data on crashes that occurred in the IRL circuit from 1996 to 2003 to investigate peak vehicle G as a predictor of brain injury. Head injury is a cause of significant morbidity and mortality across the world and it is a major concern in motor sports safety. To our knowledge, our study is the first to attempt to correlate chassis crash accelerations and brain injury in motorsports crashes. The accelerometers used in this study measure peak G at the center of gravity of the car's chassis. We are unaware of any such studies that have established an injury threshold criterion from vehicle chassis impact data. Anecdotally, we have noted an apparent increase in injury when the chassis peak G were ≥50 G. We, therefore, evaluated the likelihood of significant head injury based on this number to determine if it could predict the likelihood of brain injury. Our study demonstrated that those drivers who had a crash with peak G forces of ≥50 G were significantly more likely to sustain a head injury than those with a lesser impact. In addition, those crashes resulting in a head injury sustained a mean peak vehicle G force of 29.0 G more than the mean in the group of crashes without a head injury. The regression analysis we performed gives an OR of 1.15 (95% CI: 1.08, 1.21) for each five-unit increase in vehicle G. This further demonstrates that as vehicle G forces sustained in a crash increase, so does the likelihood of sustaining a head injury. This analysis estimated that an injury rate of 10% would be seen in crashes with a vehicle impact of 68 G and this would increase to a head injury rate of 90% for those drivers in cars reaching 230 G in crashes. When considering only those crashes where a head injury occurred, however, no differences were found in the mean peak G force sustained between more severe and less severe head injuries. We anticipated that an increase in vehicle G force would predict an increase in severity of injury just as it predicted the likelihood of sustaining an injury. It is possible that a correlation was not found between vehicle G and severity of injury because of the relatively small number when only evaluating those crashes with injury. The study was sufficiently powered to demonstrate an association of vehicle G and head injury, but when the injuries are broken into different severities, the study may not have identified an association even if one was present.
There are several limitations to our study. We present limited clinical data in that we only evaluate diagnosis at the time of injury and no follow-up data, including neuropsychometric testing. Although we understand that our choosing 50 G based on gestalt of level that injuries occurred is a weakness of our study, we did so in an attempt to establish an injury threshold that would need further prospective validation. This study does not answer the question of whether the G force itself caused the head injuries but instead only evaluates the maximal G force measured as a predictor of brain injury. G forces experienced are a composite of the force, the direction of the acceleration, and the time interval during which the acceleration occurs. We did not analyze each crash for direction of impact or the time interval. The study only evaluates the crash based on maximal G forces measured. As we evaluate the correlation of G forces delivered to the head in the future using earpiece accelerometers, we will correlate the car measurements with the earpiece and evaluate the measured G force and time. In addition, we will document the direction of acceleration in an attempt to identify important determinants of injury. We also cannot comment on which drivers endured contact with a portion of the passenger compartment and which sustained only G-force deceleration without impact contact. It is likely that this has a great influence on the likelihood of injury. Our study only evaluates which drivers (in a similar environment and restraint system) sustained injury at various G forces and does not evaluate the exact cause of the injury at these G forces (contact or the G forces themselves). Also, the accelerometers were mounted on the car's chassis and it is possible that a variety of safety equipment, different chassis, and different types of crashes will significantly alter the forces translating from the center of gravity of the car to the driver's head. And finally, we only have data on 374 of 647 crashes and it is possible that this incomplete data could falsely represent all crashes. We are reporting, however, all crashes in which ADR data was recorded.
We plan future studies with accelerometers mounted in earpieces to determine correlation with chassis accelerometers in an attempt to better measure the forces delivered to the driver's head. We also plan to incorporate other measures such as the head injury criterion (HIC) and the severity index coefficient (SIC) in future investigations with the ear and car accelerometers.
In conclusion, our study demonstrates that IRL car crashes with peak vehicle G ≥ 50 are significantly more likely to develop traumatic brain injuries than those with lesser impact. Future investigations to decrease G forces may decrease head injuries in the IRL series and even highway crashes.
This article was previously submitted as a Poster Presentation to the Society for Academic Emergency Medicine Annual Meeting, May 17, 2004, Orlando, Florida.
1. Adekoya, N., D. J. Thurman, D. D. White, et al. Surveillance for traumatic brain injury deaths-United States, 1989-1998. 2002, Atlanta: Center for Disease Control and Prevention, pp. 1-20.
2. Agresti, A. An Introduction to Categorical Data Analysis
. New York: John Wiley & Sons, 1996, pp. 104-105.
3. Gurdjian, E. S., H. R. Lissner, F. R. Latimer, et al. Quantitative determination of acceleration and intracranial pressure in experimental head injury. Neurology
4. Gurdjian, E. S., V. L. Roberts, and L. M. Thomas. Tolerance curves of acceleration and intracranial pressure and protective index in experimental head injury. Journal of Trauma-Injury Infection & Critical Care
5. Jennett, B. Epidemiology of head injury. J. Neurol. Neurosurg. Psychiatry
6. Jennett, B., and R. MacMillan. Epidemiology of head injury. Br. Med. J. (Clin. Res. Ed.).
7. KRAUS, J. F. Epidemiology of head injury. In: Head Injury
, 3rd ed., P. R. Cooper (Ed.). Baltimore: Williams & Wilkins, 1993, 1-25.
8. Melvin, J. W., K. J. Baron, W. C. Little, et al. Investigation of Indy Car Crashes Using Impact Recorders.
SAE Motorsports Engineering Conference and Exposition. Dearborn, Michigan, 1996. Warrendale, PA: Society of Automotive Engineers Inc., 1996, pp. 127-143.
9. Naunheim, R. S., J. Standeven, C. Richter, et al. Comparison of impact data in hockey, football, and soccer. Journal of Trauma-Injury Infection & Critical Care
10. States, J. D. The abbreviated and the comprehensive research injury scales. In: 13th STAPP Car Crash Conference, Boston, 1969. Warrendale, PA: Society of Automotive Engineers Inc., 1969, pp. 282-294.
11. Wright, P. G. The analysis of accident data recorder (ADR) data in Formula 1. In: SAE Motorsports Engineering Conference & Exposition, Dearborn, Michigan, 2000. Warrendale, PA: Society of Automotive Engineers Inc. Technical paper 2000-01-3551 in P-361.