American Journal of Epidemiology
Copyright © 2004 by the Johns Hopkins Bloomberg School of Public Health
All rights reserved
Vol. 159, No. 6
Printed in U.S.A.
DOI: 10.1093/aje/kwh087
Efficacy of Side Air Bags in Reducing Driver Deaths in Driver-Side Collisions
Elisa R. Braver1,2 and Sergey Y. Kyrychenko1
1
Insurance Institute for Highway Safety, Arlington, VA.
Current address: National Study Center for Trauma and Emergency Medical Services, University of Maryland School of
Medicine, Baltimore, MD.
2
Received for publication October 24, 2003; accepted for publication January 15, 2004.
Side air bags, a relatively new technology designed to protect the head and/or torso in side-impact collisions,
are becoming increasingly common in automobiles. Their efficacy in preventing US driver deaths among cars
struck on the near (driver’s) side was examined using data from the Fatality Analysis Reporting System and the
General Estimates System. Risk ratios for driver death per nearside collision during 1999–2001 were computed
for head/torso and torso-only side air bags in cars from model years 1997–2002, relative to cars without side air
bags. Confounding was addressed by adjusting nearside risk ratios for front- and rear-impact mortality, which is
unaffected by side air bags. Risk ratios were 0.55 (95% confidence interval: 0.43, 0.71) for head/torso air bags
and 0.89 (95% confidence interval: 0.79, 1.01) for torso-only air bags. Risk was reduced when cars with head/
torso air bags were struck by cars/minivans (significant) or pickup trucks/sport utility vehicles (nonsignificant).
Risk was reduced in two-vehicle collisions and among male drivers and drivers aged 16–64 years. Protective
effects associated with torso-only air bags were observed in single-vehicle crashes and among male and 16- to
64-year-old drivers. Head/torso side air bags appear to be very effective in reducing nearside driver deaths,
whereas torso-only air bags appear less protective.
accidents, traffic; air bags; automobiles; mortality; safety; vehicles
Abbreviations: CI, confidence interval; FARS, Fatality Analysis Reporting System; GES, General Estimates System; RR, risk
ratio; SUV, sport utility vehicle; VIN, Vehicle Identification Number.
Side-impact collisions pose higher risks to occupants of
passenger vehicles than do collisions involving impacts to
vehicles’ fronts or rears, because the sides of passenger vehicles have a limited ability to crumple and absorb energy in
collisions (1). Each year, 3.18 million US passenger vehicles
are involved in police-reported crashes in which they are
struck on the side; these crashes result in approximately
9,400 deaths—about 30 percent of all passenger vehicle
occupant deaths (2, 3).
The occupants at highest risk of serious injury, usually
injury to the chest and head, in side-impact collisions are
those seated on the side struck (4–7). An estimated 40–75
percent of passenger vehicle occupant deaths in side-impact
crashes result from head injuries (8). Passenger vehicle
occupants struck in the side by another vehicle are more
likely to sustain serious injuries when the striking vehicle is
a pickup truck or sport utility vehicle (SUV) than when it is
a car (4, 5). Pickups and SUVs cause more fatalities because
of their higher average weight and because their higher
front-end geometry increases the risk of direct head strikes
and other injuries among occupants of struck vehicles (9,
10).
To reduce the risks to occupants of vehicles struck on the
side, side air bags designed to protect the torso (chest and
abdomen) were introduced by Volvo (Volvo Car Corporation, Gothenburg, Sweden) in model year 1995 (11). Torsoprotection side air bags, mounted in doors or seats, are
designed to attenuate and distribute forces on the chest and
abdomen from the intruding vehicle side (12). Side air bags
designed to protect the head were introduced by BMW
(BMW AG, Munich, Germany) in model year 1998, and
more than 90 percent of makes and models offering headprotection side air bags also include torso protection (12).
Head-protection side air bags are designed to keep an occu-
Reprint requests to Dr. Elisa Braver, National Study Center for Trauma and Emergency Medical Services, University of Maryland School of
Medicine, 701 West Pratt Street, Room 526, Baltimore, MD 21201 (e-mail: [email protected]).
556
Am J Epidemiol 2004;159:556–564
Efficacy of Side Air Bags in Reducing Driver Deaths 557
pant’s head from striking interior vehicle structures or being
struck by external objects intruding into the vehicle (13).
Crash tests have suggested that both head-protection side
air bags and torso-only side air bags should be beneficial to
vehicle occupants, yet crash test ratings are not always
improved by the addition of either torso-only or combined
head/torso-protection side air bags (5, 6, 8, 14–16). Whether
side air bags are beneficial in real-world crashes has not been
established by prior research. One German study reported
that injured automobile occupants with torso-only side air
bags had lower proportions of moderate to severe injuries to
the thorax and head; however, the report did not indicate
whether injury incidence was reduced (17). Another study
reported that side air bags failed to reduce injury risk in nearside crashes, but it classified all vehicle models offering
optional side air bags as having side air bags, made no
distinction between side air bags with head protection and
those without head protection, and derived injury severity
data from a source that often misclassifies minor injuries as
serious injuries (18).
The purpose of the present study was to evaluate the realworld performance of vehicles equipped with side air bags to
determine whether they reduce mortality among US drivers
in vehicles struck on the left (driver’s) side. Side air bags are
becoming increasingly common: In 2003, more than one
third of all new vehicles sold in the United States had side air
bags, most of which were designed primarily for torso
protection (19). Side air bags are being considered as one
countermeasure to the problem of geometric incompatibility
between passenger cars and SUVs/pickup trucks. Determining whether side air bags work as intended is important
for future decisions concerning vehicle design.
MATERIALS AND METHODS
Mortality during 1999–2001 was studied among drivers of
passenger cars for model years 1997–2002. Pickup trucks
were excluded, because they were rarely equipped with side
air bags during these model years (5). Minivans and SUVs
with side air bags could not be studied, because the numbers
of crashes involving these vehicles in government crash
databases were too low.
The measure studied was risk of death per side-impact
collision among passenger car drivers involved in policereported crashes during 1999–2001. The Fatality Analysis
Reporting System (FARS), a census of fatal collisions on US
public roads in which a death occurs within 30 days of the
crash, was the source for data on driver deaths (20). The
General Estimates System (GES), a national sample of US
police-reported crashes, was the source for data on drivers
involved in police-reported crashes (21). For generation of
national estimates, the 55,000 annual crashes (raw counts)
included in GES are weighted according to their probability
of selection for the stratified national sample.
A vehicle-features database maintained by the Highway
Loss Data Institute (Arlington, Virginia) was used to
compile a list of 1997–2002 model year vehicles with standard or optional side air bags. The list indicates vehicle make
and model, model year, the standard or optional status of
side air bags, and whether the side air bag systems are
Am J Epidemiol 2004;159:556–564
designed to protect the head, the torso, or both. Manufacturers assign each vehicle a 17-digit vehicle identification
number (VIN), the last five digits of which are unique to the
vehicle. To preserve driver privacy, FARS provides only the
first 12 digits, and GES provides only the first 11. These
partial VINs usually indicate whether the vehicles are
equipped with side air bags (standard or optional), but some
vehicles’ VIN codes do not indicate the presence of optional
side air bags. After such vehicles had been identified in the
FARS and GES databases, the National Highway Traffic
Safety Administration sent their full 17-digit VINs to the
manufacturers, who supplied the agency with information
about side air bag presence and type of side air bag. This
information was then transmitted to us by the National
Highway Traffic Safety Administration.
The comparison group of vehicles consisted of model year
1997–2002 cars without side air bags that were involved in
police-reported crashes. These included vehicles for which
side air bags had been an option that was not chosen by
buyers and vehicles for which side air bags were unavailable.
Statistical analysis
Risk ratios and 95 percent confidence intervals were
calculated for nearside death (from FARS) among passenger
car drivers involved in police-reported crashes (from GES)
by side air bag status. Separate risk ratios were calculated for
side air bags that included head protection and side air bags
designed to protect mainly the torso. These ratios were stratified according to crash, vehicle, and driver characteristics.
Analyses were carried out using Microsoft Excel,
SUDAAN, and SAS software (22–24).
Risk ratios were computed as follows. 1) Expected number
of nearside deaths in cars with side air bags = (number of
nearside driver deaths with no side air bags ÷ number of
drivers involved in nearside collisions with no side air
bags) × number of drivers involved in nearside crashes of
cars with side air bags. 2) Crude risk ratio for nearside deaths
in cars with side air bags = observed number of nearside
deaths in cars with side air bags ÷ expected number of nearside deaths.
Because cars with side air bags are more costly, their
drivers are probably of higher socioeconomic status, which
has an inverse relation with motor vehicle-related fatality
risk (25, 26). Thus, drivers with side air bags may differ
systematically from the general population of drivers in
having a lower likelihood of fatal outcomes in collisions in
which they are involved, irrespective of the presence of side
air bags. Speed of travel, especially in single-vehicle
crashes, seat-belt use, percentage of travel on different types
of roads (including urban, rural, and suburban roads), and
occupant compartment safety design are some potentially
confounding factors that may be influenced by socioeconomic status. We examined the mortality experience of
drivers in vehicles with and without side air bags in frontand rear-impact collisions (collisions in which side air bags
should have no effect). To address a wide range of
confounding factors, we then adjusted risk ratios for nearside
collisions for the combined front- and rear-impact mortality
experience among vehicles with side air bags relative to
558 Braver and Kyrychenko
RESULTS
FIGURE 1. Clock points for initial impact positions among passenger cars involved in collisions.
those without side air bags by dividing the crude risk ratios
for nearside collisions by the crude risk ratios for front/rear
collisions. Both crude and adjusted risk ratios are reported so
that the effects of adjustment can be discerned. Formulae for
calculation of adjusted risk ratios and their 95 percent confidence intervals are shown in the Appendix.
Using FARS, nearside collisions were defined as crashes
with initial impacts at the 8 o’clock, 9 o’clock, and 10
o’clock points (the left side of the occupant compartment),
and front and rear collisions were defined as crashes with
initial impact points at 12 o’clock and 6 o’clock (see figure
1). The clock points 1, 5, 7, and 11 were excluded from the
group of front and rear collisions because of the small possibility that side air bags could deploy after impacts in these
positions. The GES coding system is different from that of
FARS; initial points of impact on the near side were those
labeled “left side” in the GES database, and front and rear
collisions were those labeled “front” and “back.” As in
FARS, impacts to the right and left corners of the vehicle in
the front and back were excluded from the definition of
front/rear collisions in GES.
In estimated police-reported crashes involving drivers
struck on the left (driver’s) side during 1999–2001, a total of
35 nearside driver deaths occurred in passenger cars with
side air bags designed to protect the head, and 105 deaths
occurred in cars with torso-only side air bags (table 1).
Approximately 99 percent of cars with head-protection side
air bags included in the study also had torso protection.
Figure 2 shows rates of passenger-car driver death per
100,000 drivers in crashes occurring during 1999–2001 for
impact points on the near (driver’s) side and in the front/rear,
according to whether side air bags were optional or standard
equipment and according to the type of side air bag. The
lowest nearside driver death rates were observed among cars
with head-protection side air bags, and the next lowest nearside driver death rates were observed in cars with torso-only
side air bags. Cars with no side air bags available had the
highest rates of driver death for both nearside and front/rear
impact points. For both nearside and front/rear impact
points, driver death rates in cars where optional air bags were
absent were much lower than in cars that did not have side
air bags offered as standard or optional equipment, indicating the presence of factors confounding the relation
between side air bags and driver fatality risk.
Side air bags with head-protection systems
Risk of driver death in a nearside impact was significantly
reduced in passenger cars that had side air bags with head
protection (adjusted risk ratio (RR) = 0.55, 95 percent confidence interval (CI): 0.43, 0.71) (table 2). In cars with headprotection air bags, significant reductions in risk were
observed among male drivers, drivers aged 16–64 years, and
cars with four doors. Reductions in risk among female
drivers were similar to those of male drivers; reductions
among drivers aged 65 years or more were similar to those of
drivers aged 16–64 years. The low numbers of deaths among
females and drivers aged 65 years or more led to wide confidence intervals that included 1.0.
Significant protective effects were also observed among
the smallest and largest cars equipped with head-protection
TABLE 1. Numbers of passenger-car driver deaths and involvements in police-reported collisions, by side air bag
status and initial impact point,* for model years 1997–2002, Fatality Analysis Reporting System and General
Estimates System, 1999–2001
No. of drivers involved in crashes
No. of driver deaths
Side air bag status
Near side
Near side
Side air bags with head protection
Torso-only side air bags
No side air bags
Front/rear
Weighted†
Front/rear
Raw
Weighted†
Raw
35
108
22,289
180
93,622
739
105
208
45,640
390
199,275
1,588
1,800
3,044
522,935
4,439
2,185,876
17,623
* The near side (driver’s side) is defined as clock positions 8, 9, and 10 in the Fatality Analysis Reporting System and the
left side in the General Estimates System. Front/rear impacts are defined as impacts in clock positions 12 and 6 in the
Fatality Analysis Reporting System and the front and back in the General Estimates System.
† National estimates were derived by weighting raw counts of drivers in the sample according to the probability of
selection of crashes for the stratified national sample.
Am J Epidemiol 2004;159:556–564
Efficacy of Side Air Bags in Reducing Driver Deaths 559
FIGURE 2. Rates of passenger-car driver death per 100,000 drivers involved in police-reported collisions, by side air bag status and initial
impact point, United States, 1999–2001. Bars, 95% confidence interval.
side air bags. The adjusted risk ratio showed no reduction in
risk for midsize cars, though the crude risk ratio suggested
comparable decreases in risk by car size. Among drivers of
cars with head-protection side air bags, there was a significant decrease in nearside fatality risk in two-vehicle sideimpact crashes and a nonsignificant decrease in risk in
single-vehicle nearside crashes. Head-protection side air
bags afforded a large degree of protection in two-vehicle
crashes in which the other vehicle (the “crash partner”)
striking the car on the driver’s side was a passenger car/
minivan (adjusted RR = 0.26, 95 percent CI: 0.11, 0.64).
Reduced risk of driver death, albeit nonsignificant, was also
observed when cars’ crash partners were pickups or SUVs.
Risk reductions were observed regardless of whether the
crash partners that were passenger vehicles weighed less
than 3,800 pounds (<1,724 kg) (significant) or 3,800 pounds
or more (≥1,724 kg) (nonsignificant). A large percentage of
driver deaths in cars with head-protection side air bags
occurred in two-vehicle collisions in which the crash partner
was a large commercial vehicle, and these side air bags did
not increase protection in such crashes.
vehicle crashes. Decreased risk, though not statistically
significant, appeared to be present among cars of all sizes
and among both two-door and four-door cars.
Male drivers and drivers aged 16–64 years had significant
reductions in nearside mortality risk. Adjusted risk ratios for
female drivers and drivers aged 65 years or more showed no
benefits from such air bags; however, the ability to detect
decreases in risk was less than optimal for females and
drivers aged 65 years or more. Additional analyses stratified
by age and gender (not shown in table) suggested that
females aged 65 years or more had an increase in nearside
mortality associated with torso-only side air bags (adjusted
RR = 2.16, 95 percent CI: 1.08, 4.33).
On the basis of adjusted risk ratios, no protective effects of
torso-only side air bags were observed in two-vehicle
crashes. Neither did adjusted risk ratios indicate benefits
when crash partners were cars/minivans or pickups/SUVs in
two-vehicle side-impact crashes. In addition, the adjusted
risk ratio suggested increased driver fatality risk when crash
partners were passenger vehicles that weighed 3,800 pounds
or more.
Torso-only side air bags
Head-protection versus torso-only side air bags
For passenger cars, torso-only side air bags were associated with a nearly significant reduction in driver death rates
in nearside crashes (adjusted RR = 0.89, 95 percent CI: 0.79,
1.01) (table 2). Greater risk reductions occurred in single-
Head-protection side air bags were significantly more
effective than torso-only side air bags in reducing nearside
driver fatality risk overall and for various categories of vehicles and crash types (table 2). Direct comparisons are valid,
Am J Epidemiol 2004;159:556–564
560 Braver and Kyrychenko
TABLE 2. Risk ratios for passenger-car driver deaths in US nearside* collisions occurring
during 1999–2001, by type of side air bag and by crash, vehicle, and driver characteristics,
for model years 1997–2002
No. of deaths
Variable
Observed Expected
Risk ratio
Crude
RR†
Adjusted
RR‡
95% CI† for
adjusted RR
Side air bags with head protection
Total no. of cars
35
77
0.46
0.55
0.43, 0.71
Single-vehicle
15
20
0.76
0.76
0.35, 1.65
Two-vehicle
18
49
0.37
0.47
0.35, 0.64
Car/minivan
3
18
0.16
0.26
0.11, 0.64
Sport utility vehicle/pickup truck
5
17
0.29
0.47
0.20, 1.12
Large truck/bus
8
3
2.40
1.93
0.66, 5.69
Other/unknown
2
Crash type
Crash partner vehicle type (twovehicle crashes)
Passenger vehicle crash partner
weight
<3,800 pounds (<1,724 kg)
6
22
0.28
0.37
0.21, 0.66
≥3,800 pounds (≥1,724 kg)
4
15
0.26
0.45
0.19, 1.06
Car size
Mini/small
6
12
0.52
0.51
0.27, 0.95
Midsize
12
22
0.55
1.04
0.66, 1.64
Large/very large
17
44
0.39
0.41
0.30, 0.57
No. of doors
2
2
7
0.30
0.38
0.10, 1.53
≥4
33
69
0.48
0.55
0.42, 0.73
Male
25
49
0.51
0.56
0.42, 0.74
Female
10
30
0.33
0.67
0.41, 1.08
26
53
0.49
0.56
0.43, 0.75
9
27
0.33
0.59
0.27, 1.26
Driver gender
Driver age (years)
16–64
≥65
Table continues
because the same control group (vehicles without side air
bags) and the same adjustment methods were used to evaluate the two types of side air bag systems.
DISCUSSION
Side air bags designed to protect the head appear to be very
effective in reducing mortality in nearside collisions among
drivers of passenger cars. Air bags with head protection
appeared to provide benefits for both male and female
drivers and for both younger and older drivers, though the
reductions observed among females and drivers aged 65
years or more were not statistically significant.
Because very few cars in our study (<1 percent) offered
head protection without torso protection, it was not possible
to identify the separate contributions of head and torso
protection in vehicles with head-protection side air bags.
However, the estimated risk reduction in nearside collisions
for drivers of cars with torso-only side air bag systems was
substantially lower, suggesting that the incremental benefit
of head protection is considerable.
Large decreases in nearside fatality risk associated with
head-protection side air bags were observed when crash partners were cars/minivans or SUVs/pickup trucks. This
suggests that such air bags may be addressing some of the
problems of incompatibility when passenger cars are struck
on the side by vehicles with higher ride heights. These
systems were not effective when cars were struck on the side
by commercial vehicles with gross vehicle weight ratings in
excess of 10,000 pounds (>4,536 kg) (primarily large trucks
and buses). The most likely explanation is that head-protection air bags and other restraints cannot overcome extreme
Am J Epidemiol 2004;159:556–564
Efficacy of Side Air Bags in Reducing Driver Deaths 561
TABLE 2. Continued
No. of deaths
Variable
Risk ratio
Crude
RR
Adjusted
RR
95% CI
for adjusted RR
157
0.67
0.89
0.79, 1.01
Observed Expected
Torso-only side air bags
Total no. of cars
105
Crash type
Single-vehicle
26
38
0.69
0.62
0.40, 0.96
Two-vehicle
70
101
0.70
1.11
0.94, 1.31
Car/minivan
19
29
0.66
1.06
0.79, 1.41
Sport utility vehicle/pickup truck
28
49
0.57
1.10
0.79, 1.53
Large truck/bus
16
20
0.78
1.02
0.55, 1.88
Other/unknown
7
Crash partner vehicle type (twovehicle crashes)
Passenger vehicle crash partner
weight
<3,800 pounds (<1,724 kg)
28
42
0.67
1.02
0.82, 1.29
≥3,800 pounds (≥1,724 kg)
22
31
0.70
1.73
1.12, 2.68
Mini/small
22
42
0.52
0.79
0.58, 1.09
Midsize
37
60
0.62
0.89
0.70, 1.12
Large/very large
46
54
0.85
0.89
0.70, 1.14
Car size
No. of doors
2
18
29
0.62
0.86
0.57, 1.28
≥4
87
126
0.69
0.89
0.78, 1.02
Male
58
90
0.64
0.79
0.64, 0.99
Female
47
67
0.70
1.21
0.98, 1.50
16–64
59
104
0.57
0.79
0.67, 0.93
≥65
46
76
0.61
0.98
0.65, 1.46
Driver gender
Driver age (years)
* The near side (driver’s side) was defined as clock points 8, 9, and 10 (see figure 1).
† RR, risk ratio; CI, confidence interval.
‡ Adjusted on the basis of the risk ratio for driver death in initial impacts that were front or rear.
weight mismatches, which often result in massive collapse
of the occupant compartment in lighter vehicles.
The benefits of torso-only air bags appear less substantial,
with the exception of single-vehicle nearside crashes. Moreover, torso-only air bags provided less uniform protection:
Men and drivers younger than 65 years had clear risk reductions, but this was not true for female drivers or older drivers.
Some findings hint that torso-only side air bags could be
placing female drivers, particularly those aged 65 years or
more, at increased risk in nearside crashes; however, more
data are needed before such a conclusion can be drawn.
There also was an indication that torso-only side air bags
were ineffective in two-vehicle crashes.
Based on the patterns of death rates, using front/rear collisions to adjust nearside risk ratios provides a rational method
for controlling some confounding factors associated with
side air bags. Figure 2 indicated that the magnitude of the
Am J Epidemiol 2004;159:556–564
confounding for nearside driver death rates was similar for
front/rear death rates. This method addressed both known
and unknown factors because it was not confined to specific
known confounders. However, it may not have been appropriate for all variables and may have resulted in incorrect
upward or downward adjustment of the crude risk ratios. If
automobile manufacturers who installed side air bags simultaneously introduced innovations that also improved frontal
impact protection in vehicles with side air bags, then
adjusting the crude nearside risk ratios for front/rear
mortality may have led to underestimates of the true effectiveness of side air bags. When interpreting the findings of
this study, both crude and adjusted risk ratios should be
considered. The adjusted risk ratios differed greatly from the
crude risk ratios for some variables, particularly for torsoonly side air bags, for which crude risk ratios suggested
decreases of approximately 30 percent in nearside driver
562 Braver and Kyrychenko
fatality risk. Such differences occurred among midsize cars
with head-protection side air bags, and among cars with
torso-only side air bags they were present for two-vehicle
crashes, crash partner passenger vehicles that weighed 3,800
pounds or more, female drivers (especially older females),
and older drivers. For example, the aforementioned increase
in risk among older female drivers differed greatly from the
corresponding crude risk ratio, which estimated a 15 percent
decrease in risk. The differences between crude and adjusted
risk ratios reflected varying front/rear mortality experience
for these variables. Perhaps the true effect of side air bags on
fatality risk for struck-side drivers lies somewhere between
the crude and adjusted risk ratios in these cases.
Other potential methods of adjusting for confounding were
not reasonable. Using cars within the same cost class without
side air bags as a reference group was not possible, because
most cars in the high-end cost classes are equipped with side
air bags. Using a comparison group composed solely of vehicles without side air bags for which the option of air bags
was available was not feasible because of inadequate
numbers in this group. Multivariate regression analyses
would have been desirable, but such analyses could not be
performed because of insufficient numbers, insufficient data
on potentially confounding variables in FARS and GES, and
the difficulties of combining two very different databases.
The most important limitation of this study is that side air
bags, particularly those incorporating head-protection
systems, are relatively new, so there have been few crashes
and deaths involving cars with side air bags. Investigation of
the effects of side air bags according to variables of interest
was impeded by low statistical power. Age probably
confounded the relation between vehicle size and driver
mortality, because a higher proportion of older drivers, who
are more vulnerable to fatal injury, drive large cars. In addition, potential differences in efficacy could not be examined
among different types of air bags in the same general category, including first-generation side air bags versus those
developed after initial experience and curtain versus seatmounted head-protection air bags (27).
Another limitation stems from missing data in the GES
database. This study used only vehicles with valid VINs for
GES denominators (drivers involved in crashes). While the
FARS database (deaths in numerators) has nearly 100
percent valid VINs, only about 70–75 percent of vehicles in
the GES have valid VINs. Some states do not record VINs,
so their driver deaths were present in the numerators but not
the denominators. This led to underestimation of denominators for death rates; however, this could have biased the
results only if both side air bag status and initial impact
points influenced the likelihood of vehicles’ having valid
VINs, which seems improbable.
Like front air bags, side air bags could pose a risk of injury
to some vehicle occupants, particularly children who are too
close to the air bags; however, there have been few serious
injuries attributed to side air bags thus far (28–30). Automobile manufacturers conduct voluntary tests to ensure that
inflating side air bags do not exceed agreed-upon levels of
injury measures among test dummies representing small
females and children positioned in close proximity to the air
bags (31, 32).
Side air bags with both head and torso protection appear to
work as intended in terms of decreasing the risk of driver
death in nearside collisions, the most dangerous type of
impact for motor vehicle occupants. Current findings
suggest that such air bags may be an effective method for
reducing but not eliminating the adverse effects of incompatibility in ride height between colliding vehicles. In spite of
study limitations and inconsistencies, the evidence for major
reductions in risk from head-protection side air bags is
robust. Torso-only air bags appear to confer lesser benefits
on drivers. The numbers of vehicles equipped with side air
bags in this study were relatively low, so follow-up of these
findings is necessary, as are studies examining the mortality
and morbidity of passengers, particularly children, in vehicles with side air bags. Further research is also needed to
determine whether side air bags benefit occupants of SUVs,
minivans, and pickup trucks.
ACKNOWLEDGMENTS
This research was supported by the Insurance Institute for
Highway Safety (Arlington, Virginia).
The authors gratefully acknowledge the automobile manufacturers (BMW, Honda, Mazda, Mitsubishi, Nissan,
Toyota, and Volvo) who assisted in identifying the presence
of side air bags in vehicles and Barry Eisemann and John
Donaldson of the National Highway Traffic Safety Administration, who facilitated the exchange of information with the
manufacturers. The authors also thank staff members of the
Highway Loss Data Institute and the Insurance Institute for
Highway Safety for compiling the list of vehicles with side
air bags, including Deidre McPherson, Drew Knoblauch,
Matthew Moore, and Richard Firestone. Susan A. Ferguson
and Adrian K. Lund of the Insurance Institute for Highway
Safety provided helpful comments on the statistical methods
and on early drafts of the manuscript. The authors appreciate
the assistance with data processing provided by Maria B.
Penny of the Insurance Institute for Highway Safety and
Cherian Varghese of Rainbow Technology, Inc. (Olney,
Maryland).
REFERENCES
1. Bédard M, Guyatt GH, Stones MJ, et al. The independent contribution of driver, crash, and vehicle characteristics to driver
fatalities. Accid Anal Prev 2002;34:717–27.
2. Insurance Institute for Highway Safety. Fatality facts: general.
Arlington, VA: Insurance Institute for Highway Safety, 2002.
(World Wide Web URL: http://www.highwaysafety.org/
safety_facts/fatality_facts/general.pdf).
3. National Highway Traffic Safety Administration, US Department of Transportation. Traffic safety facts 2001. (Report no.
DOT HS-809-484). Washington, DC: National Highway Traffic Safety Administration, 2002. (World Wide Web URL:
http://www-nrd.nhtsa.dot.gov/pdf/nrd-30/NCSA/TSFAnn/
TSF2001.pdf).
4. Farmer CM, Braver ER, Mitter EL. Two-vehicle side impact
crashes: the relationship of vehicle and crash characteristics to
Am J Epidemiol 2004;159:556–564
Efficacy of Side Air Bags in Reducing Driver Deaths 563
injury severity. Accid Anal Prev 1997;29:399–406.
5. Samaha RR, Elliott DS. NHTSA side impact research: motivation for updated test procedures. (Paper no. 492). In: Proceedings of the 18th International Technical Conference on the
Enhanced Safety of Vehicles, Nagoya, Japan, May 19–22,
2003. (Report no. DOT HS-809-543 (CD-ROM)). Washington,
DC: National Highway Traffic Safety Administration, 2003.
(World Wide Web URL: http://www-nrd.nhtsa.dot.gov/pdf/
nrd-01/esv/esv18/CD/Files/18ESV-000492.pdf).
6. Insurance Institute for Highway Safety. Side impact: how 12
small SUVs fared. (IIHS status report, vol 38, no. 7). Arlington,
VA: Insurance Institute for Highway Safety, 2003. (World
Wide Web URL: http://www.highwaysafety.org/srpdfs/
sr3807.pdf).
7. Lestina DC, Gloyns PF, Rattenbury SJ. Fatally injured occupants in side impact crashes. (Paper no. S5-W-30). In: Proceedings of the 13th International Technical Conference on
Experimental Safety Vehicles, Paris, France, November 4–7,
1991. (Report no. DOT HS-807-990). Washington, DC:
National Highway Traffic Safety Administration, 1991:701–7.
8. Insurance Institute for Highway Safety. Side impacts with fatal
head injuries. (IIHS status report, vol 36, no. 1). Arlington, VA:
Insurance Institute for Highway Safety, 2001. (World Wide
Web URL: http://www.highwaysafety.org/srpdfs/sr3601.pdf).
9. Lund AK, Chapline JF. Potential strategies for improving crash
compatibility in the US vehicle fleet. (SAE technical paper
series 1999-01-0066). In: Vehicle aggressivity and compatibility in automotive crashes. (Publication no. SP-1442). Warrendale, PA: Society of Automotive Engineers, 1999:33–44.
10. Nolan JM, Powell MR, Preuss CA, et al. Factors contributing to
front-side compatibility: a comparison of crash test results.
(SAE technical paper series 99SC02). In: 43rd Stapp Car Crash
Conference proceedings. (Publication no. P-350). Warrendale,
PA: Society of Automotive Engineers, 1999:13–24.
11. vptuning.com. American crash tests prove the efficiency of the
Volvo S80 inflatable curtain. (December 14, 2000). Winchester, MA: vptuning.com, 2000. (World Wide Web URL:
http://www.vvspy.com/news/0012/001214.php3).
12. Autoliv, Inc. Products: side-impact airbags. Stockholm, Sweden: Autoliv, Inc, 2003. (World Wide Web URL: http://www.
autoliv.com).
13. AutoWorld.com. BMW’s head protection system. Bath, OH:
Vehicle Information Services, Inc, 2001. (World Wide Web
URL: http://www.autoworld.com/news/BMW/SafeCar.htm).
14. European New Car Assessment Programme. New Renault
Mégane first small family car to achieve 5 stars. (Press release,
November 26, 2002). Brussels, Belgium: Federation Internationale de l’Automobile, 2002. (World Wide Web URL: http://
www.EuroNCAP.com/content/media/press_releases/
phase_11plus.html).
15. New Car Assessment Program, National Highway Traffic
Safety Administration, US Department of Transportation. New
Car Assessment Program—2002 light passenger cars. Washington, DC: National Highway Traffic Safety Administration,
2002. (World Wide Web URL: http://www.nhtsa.dot.gov/
NCAP/Cars/2002SubC.html).
16. Stoffer H. Side airbags don’t assure 5-star rating by NHTSA.
Automotive News 2002;Dec 9:20.
17. Bakker J, Herrmann R. Methodology for assessing the efficiency of car safety systems based on German accident databases. In: Proceedings of the Innovative Occupant and Partner
Crash Protection Conference, Berlin, Germany, September 6–
7, 2001. Düsseldorf, Germany: VDI-Gesellschaft Fahrzeug und
Verkehrstechnik, 2001.
Am J Epidemiol 2004;159:556–564
18. McGwin G, Metzger J, Porterfield JR, et al. Association
between side air bags and risk of injury in motor vehicle collisions with near-side impact. J Trauma 2003;55:430–6.
19. Stoffer H. Safe with side air bags? It’s not that simple…
Automotive News 2002;Dec 9:1.
20. National Highway Traffic Safety Administration, US Department of Transportation. Fatality Analysis Reporting System
(1999–2001). Washington, DC: National Highway Traffic
Safety Administration, 2000–2002.
21. National Highway Traffic Safety Administration, US Department of Transportation. General Estimates System (1999–
2001). Washington, DC: National Highway Traffic Safety
Administration, 2000–2002.
22. Microsoft Corporation. Microsoft Excel 2000. Redmond, WA:
Microsoft Corporation, 2000.
23. RTI International, Inc. SUDAAN, version 8. Research Triangle
Park, NC: RTI International, Inc, 2001.
24. SAS Institute, Inc. SAS, version 8.2. Cary, NC: SAS Institute,
Inc, 2001.
25. Braver E. Race, Hispanic origin, and socioeconomic status in
relation to motor vehicle occupant death rates and risk factors
among adults. Accid Anal Prev 2003;35:295–309.
26. Cubbin C, LeClere FB, Smith GS. Socioeconomic status and
the occurrence of fatal and nonfatal injury in the United States.
Am J Public Health 2000;90:70–7.
27. Sugimoto T. Honda side impact protection study. Presented at
the Society of Automotive Engineers government/industry
meeting, Washington, DC, May 12–14, 2003. Washington, DC:
National Highway Traffic Safety Administration, 2003.
28. Chidester A. Side impact inflatable restraint systems: real
world experience of side impact air bags in the special crash
investigations. Presented at the Side Airbag Out-of-Position
Technical Working Group public meeting, October 25, 2001.
Washington, DC: National Highway Traffic Safety Administration, 2001. (World Wide Web URL: http://www-nrd.nhtsa.
dot.gov/pdf/nrd-01/NRDmtgs/2001/1001chip.pdf).
29. Kirk A, Morris A. Side air bag deployments in the UK—initial
case reviews. (Paper no. 100). In: Proceedings of the 18th International Technical Conference on the Enhanced Safety of Vehicles, Nagoya, Japan, May 19–22, 2003. (Report no. DOT HS809-543 (CD-ROM)). Washington, DC: National Highway
Traffic Safety Administration, 2003. (World Wide Web URL:
http://www-nrd.nhtsa.dot.gov/pdf/nrd-01/esv/esv18/CD/Files/
18ESV-000100.pdf).
30. National Center for Statistics and Analysis, National Highway
Traffic Safety Administration, US Department of Transportation. NHTSA NCSA side air bag case summary. (October 1,
2003). Washington, DC: National Highway Traffic Safety
Administration, 2003. (World Wide Web URL: http://
www-nrd.nhtsa.dot.gov/pdf/nrd-30/NCSA/SCI/3Q_2003/
SABCST.pdf).
31. Prasad AK, Samaha RR, Louden AE. Evaluation of injury risk
from side impact air bags. (Paper no. 331). In: Proceedings of
the 17th International Technical Conference on the Enhanced
Safety of Vehicles, Amsterdam, the Netherlands, June 4–7,
2001. (Report no. DOT HS-809-220 (CD-ROM)). Washington,
DC: National Highway Traffic Safety Administration, 2001.
(World Wide Web URL: http://www-nrd.nhtsa.dot.gov/pdf/
nrd-11/17ESV_331.pdf).
32. Side Airbag Out-of-Position Injury Technical Working Group.
Recommended procedures for evaluating occupant injury risk
from deploying side airbags. Arlington, VA: Insurance Institute for
Highway Safety, 2000. (World Wide Web URL: http://www.
highwaysafety.org/vehicle_ratings/twg_final_procedures.pdf).
564 Braver and Kyrychenko
APPENDIX
Formulae for Calculation of Adjusted Risk Ratios and 95 Percent Confidence Intervals
Adjustment method
1) Expected number of deaths in front/rear impacts in cars with side air bags = (number of front/rear driver deaths with no
side air bags ÷ number of drivers involved in front/rear collisions with no side air bags) × number of drivers involved in front/
rear crashes of cars with side air bags.
2) Crude risk ratio for front/rear deaths in cars with side air bags = observed number of front/rear deaths in cars with side air
bags ÷ expected number of front/rear deaths.
3) Adjusted risk ratio for nearside deaths in cars with side air bags = crude risk ratio for nearside deaths ÷ crude risk ratio for
front/rear deaths.
Adjusted risk ratio and 95 percent confidence interval
1) Adjusted risk ratio for nearside deaths = D × C, where
D = (number of nearside driver deaths with side air bags × number of front/rear driver deaths with no side air bags) ÷ (number
of nearside driver deaths with no side air bags × number of front/rear driver deaths with side air bags)
and
C = (number of drivers in nearside crashes with no side air bags × number of drivers in front/rear crashes with side air bags) ÷
(number of drivers in nearside crashes with side air bags × number of drivers in front/rear crashes with no side air bags).
2) 95 percent confidence interval for the adjusted risk ratio for nearside deaths:
Lower confidence limit = D × C × exp–1.96(standard error(ln C + ln D)).
Upper confidence limit = D × C × exp1.96(standard error(ln C + ln D)).
Standard error(ln C) is obtained from SUDAAN using the Taylor series linearization method combined with variance estimation formulae specific to the General Estimates System sample design.
Standard error ( ln D ) =
1
1- + ---1- + ---1- + ----,
---d1 d 2 d 3 d 4
where d1 = number of nearside driver deaths with side air bags, d2 = number of nearside driver deaths with no side air bags,
d3 = number of front/rear driver deaths with side air bags, and d4 = number of front/rear driver deaths with no side air bags.
Am J Epidemiol 2004;159:556–564