|IV. Other Methodologies Used to Estimate Helmet
By Jonathan Goldstein
Two other methodologies have been employed to analyze helmet effectiveness. Evans and Frick (1986) use a "double pair comparison method" where one of the occupants of a motorcycle is used as a control for the other occupant in motorcycle accidents. Bowman and Schneider (1980) use a computer simulation model of impact responses of motorcycle victims that is extrapolated from the impact responses of primates and dummies. Both of these methodologies suffer from critical flaws and tend to produce unreliable estimates of helmet effectiveness.
1. Evans, L. and M. C. Frick (1986). "Helmet Effectiveness in Preventing Driver and Passenger Fatalities." General Motors Research Publication No. GMR- 5602.
Using the "double pair comparison method (DPCM) where one of the occupants of a motorcycle, either the driver or the passenger, is used as a control for the other occupant in motorcycle accidents involving both a male driver and a male or female passenger in which there is at least one fatality, this study concludes that: (1) helmeted riders are 27% less likely to die compared to non helmeted riders; and (2) the fatality in the driver's seat exceeds that in the passenger's seat by 31%.
While the methodology employed in this paper is superior to previous studies using correlation analysis, the DPCM still fails to control for all of the factors that influence motorcycle fatalities and thus results in bias (unreliable) estimates of helmet effectiveness. In particular, under a realistic set of assumptions, it can be shown that the DPCM produces estimates of helmet effectiveness that are systematically biased upward (overstated).
The authors incorrectly argue that the DPCM controls for all relevant factors by using one occupant as a control for the other occupant. While it is true that many aspects of a crash are the same for both driver and passenger (i.e. crash speed, object struck, road conditions, etc.), other important factors that dramatically influence the probability of death are not controlled for. In particular, the blood alcohol content (BAC) of the passenger and driver -- a major determinant of fatalities (see Goldstein (1986)) -- cannot be assumed to be the same in both occupants nor can the dynamic response to impact of the driver and passenger be assumed to be the same. In this latter situation, the kinetic energy -- potential for bodily damage -imparted to the two occupants can be used up very differently, thus creating dramatic differences in the injuries sustained and the probability of fatality for driver and passenger.
Given that these two very important factors that influence death and injury have not been controlled for, it can be shown that the DPCM approach will erroneously assign the adverse effects of these two factors to the non- use of a helmet. Thus overstating the effectiveness of helmets (a technical appendix that outlines this argument is available).
2. Bowman, B. M. and Schneider, L. W. (1980). "Simulation Analysis of Head/Neck Impact Response for Helmeted and Unhelmeted Motorcyclists." Highway Safety Research Institute, Ann Arbor.
This study estimates the effectiveness of motorcycle helmets through the use of computer simulations of the dynamic response of the head and neck. The study concludes that: (1) helmet use invariably lessens the exposure levels of dynamic responses which have a role in Producing head injury; and (2) helmet use almost always reduces the severity of neck response and for no simulation configuration or condition to greatly increase the likelihood of neck injury. It is further predicted that helmet use significantly reduces the likelihood and severity of both head and neck injuries.
Some interesting findings of this study are the result that helmets can induce neck injuries and the results on the limitations of helmet design (pp. 170-203). In particular, it is found that: (1) a high probability of serious brain injury for a 20 mph head impact with a vertical rigid "truck" surface is equally likely for helmeted and non helmeted riders; (2) little differences between shear forces at the upper neck exist but greater shear forces at the lower neck exist for head impacted helmeted cyclist; (3) at 20 mph head impacts, peak extension torques are double the injury tolerance level for helmeted riders (and even higher for unhelmeted riders); (4) peak flexion torques at the upper neck are significantly greater for head impacted helmeted cyclists; and (5) for chest impacts, neck elongation forces are 30% greater and shear forces are also greater for helmeted riders. Thus helmets are most effective at lower impact velocities (10 mph or less) and helmets can induce neck injuries.
The methodology employed by simulation studies are also subject to criticism. These computer models are based on extrapolations of the dynamic similarities of the post- impact responses between primates and human beings based on impact speeds from 0-7 mph. As is commonly accepted, at least by statisticians, predictions from such models that are made beyond the range of experience (0-7 mph) on which the models are based are highly unreliable. As Goldstein (1986) shows, neck injuries do not occur until 13 mph impacts -- speeds well beyond the range of experience of the simulation models -- implying that such extrapolations are subject to bias and/or have such a large variance around their point estimates that they are of little use for the evaluation of helmet effectiveness.
© Copyright Jonathan P. Goldstein Ph.D. 1986. All Rights Reserved.