II. TARGET POPULATION AND SAFETY CONCERNS

Advanced air bags have the potential to improve the benefits of air bag systems and to reduce air bag induced fatalities and serious injuries. This chapter estimates the size of the potential target population that would benefit from advanced air bags. Fatalities and injuries are discussed in separate sections.

A. Fatalities

Fatalities reported here were derived from NHTSA's 1997 Fatality Analysis Reporting System (FARS). In 1997, there were a total of 18,136 drivers and right front passengers killed in frontal crashes (see Table II-1) which accounted for about 63 percent of all occupant fatalities. Of the 18,136 fatalities, 14,004 (77 percent) were drivers and 4,132 were right front-seated passengers. The majority (68 percent) of these fatalities were unrestrained occupants ⁽¹⁾.

Table II-2 shows these fatalities disaggregrated by impact speeds and belt use. Note that fatal frontal crashes in FARS are categorized by initial or principal point of impacts (IMPACT1 or IMPACT2). Occupants are considered to be in frontal crashes if their vehicles had an impact force from a 10-2 o'clock direction. Distribution by crash impact speeds was also derived from

Table II-1
1997 Driver and Right Front Passenger Fatalities

Source: NHTSA 1997 Fatality Analysis Reporting System (FARS), 1993-97 Crashworthiness Data System (CDS)

* Frontal crashes are defined as initial or principal impact force from 10-2 o'clock direction.

the 1993 to 1997 CDS. Because of variations in data elements describing crash characteristics, it is not possible to establish a one-to-one association between FARS and CDS; hence frontal crashes are defined somewhat differently for these two databases. Frontal crashes in the NASS

CDS are defined by their principal direction of force (DOF1), their general area of damage (GAD1), and the primary specific horizontal location (SHL1) as either:

or

or

Table II-2
Fatalities In Frontal Impacts By Crash Severity

Source: NHTSA 1993-1997 CDS and 1997 FARS.

Note: Fatalities by crash speeds and belt use were derived from 1993-1997 CDS and adjusted to 1997 FARS level.

The agency has estimated that air bags have saved over 4,547 lives cumulatively from 1987 through August 1, 1999. In calendar year 1997, about 36 percent of the on-road passenger cars and 28 percent of light trucks/vans were equipped with driver side air bags, and 22 percent of passenger cars and 17 percent of light trucks/vans were equipped with passenger side air bags. Air bags saved 842 lives in 1997. If one assumes that all passenger vehicles (cars, utility vehicles, light trucks, and vans) had been equipped with air bags, they would have saved an estimated 3,253 lives annually. In total, there would have been 18,978 (=18,136 + 842) potential fatalities associated with frontal impacts if no vehicles had air bags in 1997. Potential fatalities are defined here as people in frontal crashes that died plus those would have been fatally injured in the absence of air bags.

Table II-3 shows, by several crash impact speed levels, the potential fatalities, lives that would have been saved, and the remaining fatalities if all vehicles in the fleet were equipped with pre-98 air bag systems. Advanced air bags have the potential to reduce the remaining fatalities. Belt use in Table II-3 is the same as found in 1993-1997 CDS at 32 percent.

As shown in Table II-3, an entire fleet of pre-MY 1998 air bags would save about 3,253 lives annually, and thus, are an important source of occupant protection in current passenger vehicles. However, air bags may have adverse effects on occupants who are too close to the air bags when they deploy. Of particular concern are children. As of August 1, 1999, NHTSA's Special Crash Investigation (SCI) Program has identified a total of 164 cases (137 confirmed and 27 still under investigation) of < 25 mph v in which the deployment of an air bag resulted in fatal injuries to an occupant between 1990 and 1998. Of these 164 fatalities, 17 were infants in rear-facing child safety seats (RFCSS), 77 were children aged one to twelve years old, 61 were drivers, and 9 were adult passengers. These cases were then projected to an annual basis under the assumption that all passenger vehicles were equipped with air bags (of the type typically produced during 1987-1997) by multiplying the actual number of incidents by an adjustment factor (f) that adjusts the vehicle fleet to a fleet in which all vehicles have air bags. By assuming that air bag-induced fatalities are proportional to the percentage of the fleet with air bags, the adjustment factor for each year is the ratio of the number of vehicles in operation to the number with air bags, i.e., f=1/r where r is the percentage of the fleet with air bags. The corresponding mathematical formula is:

Table II-3
Estimated Lives Saved and Fatalities Remaining in Frontal Crashes
Assuming the Whole Fleet of Passenger Vehicles Had Air Bags

Source: NHTSA 1993-1997 CDS and 1997 FARS.

Note: Fatalities by crash speeds were derived from 1993-1997 CDS and adjusted to 1997 FARS level.

Table II-4 shows actual and projected fatalities by years. The actual fatalities, except calendar year 1998, were those fatalities caused by pre-MY 1998 air bags. For year 1998, fatalities caused by MY-1998 (redesigned) air bags were also included in the projection because pre-MY 1998 air bags deployed at a greater force and thus would have killed the same occupants if the pre-MY 1998 air bags were installed in the vehicles.

If all passenger vehicles were equipped with air bags, for example in the year 1998 (using the above formula Pd=Ad*f), about 15 (4*1/0.272) infants in RFCSS, 81 (22*1/0.272) children, 11 (3*1/0.272) adult passengers, and 25 (10*1/0.394) drivers would have been killed by air bags (and otherwise probably would not have died if had there been no air bag).

Table II-4
Projected At-Risk Fatalities by Years

1. This row provides information for 1998 vehicle models with redesigned air bags in calender year 1998. These fatalities are included in the 1998 numbers above.

2. The number is derived by assuming 1998 model vehicles accounted for 7.84 percent of the fleet in operation and 87 percent of these new vehicles were equipped with redesigned air bags, i.e., 0.068=0.0784*0.87

In late 1996, the agency started a much broader public awareness program on the potential adverse effect of air bags. In addition, the agency required 1998 new vehicles to have air bag warning labels. Increasing public awareness of the air bag occupant safety issue reduced the air bag risk to rear-facing infants and children. As shown in Table II-4, the number of air bag induced fatalities gradually reduced, especially from 1997 to 1998. To take the effectiveness of the public awareness into account and to reduce year by year fluctuation, this analysis uses the weighted average of 1997 and 1998 projected deaths to estimate an annualized baseline fatal population for the at-risk groups. These projected deaths were weighted by the number of on-road operational vehicles in the fleet. There were about 194,653,000 and 198,401,000 passenger vehicles on the road in 1997 and 1998, respectively. The annualized deaths can be written as following:

Annualized Deaths = (194,653,000*Pd₉₇ + 198,401,000*Pd₉₈)/(194,653,000+198,401,000)

Because more vehicles were on the road in 1998 than in 1997, the annualized projection thus gave a slightly greater weight to 1998 cases. In total, as shown in Table II-5, there would be approximately18 infants in RFCSS, 102 children aged 1-12, 16 adult passengers, and 45 drivers ⁽²⁾ killed by air bags if all vehicles in the fleet were equipped with pre-MY 1998 air bags. For comparison purpose, the projected annual deaths for MY 1998 air bags are presented here. Note, the projected numbers based on the performance of MY 1998 air bags were derived from analysis of limited SCI data. Following is a detailed description of the analysis.

Table II-5
Estimated Annual At-Risk Fatalities With A Full Fleet of Air Bags
by Air Bag Types

*Annual deaths were projected using 1997 and 1998 fatal cases only.

** Based on judgement and analysis of minimal data.

An analysis of Special Crash Investigation (SCI) Fatalities by Model Year and Investigation Date was undertaken to determine how well the redesigned air bags were performing, based on the minimal data available. Table II-6 shows these data, which compare SCI Cases, including those cases not on the official list yet in ( ). This analysis compares what was known to the SCI team one year and nine months after the start of the new model year and compares MY 1996 vehicles to MY 1997 vehicles to MY 1998 vehicles over the same length of time (vehicle months on the road). The results are that there are still fatalities occurring to out-of-position occupants with the redesigned air bags, but fatalities appeared to have been reduced from 14 in MY 96 and 17 in MY 97, to five in MY 98. Table II-7 shows these data compared to Polk registrations (discussed further at length later in this analysis). The average of MY 96 and MY 97 data is a fatality rate of 1.14. Compared to this, the fatality rate for MY 1998 of 0.34, is about 30 percent. The agency does not know how much of this reduction is from redesign, media coverage (putting kids in rear seat), or luck. However, initial data indicate that redesigned air bags are making good progress towards reducing the out-of-position problem.

The data are not robust enough to have any confidence about how well redesigned air bags are working for the four individual categories of out-of-position occupants (rear facing infants, forward facing children, adult passengers, and drivers). However, the potential difference is significant enough that the agency will perform a sensitivity analysis, assuming redesigned air bags reduce the potential target population to 30 percent of its estimated total based on pre-MY 1998 models. For Table II-5, a distribution for the at-risk groups is provided based on the roughly one-third fatality rate. With no infant fatalities in rear facing child safety seats, the estimate of 10 is based on engineering judgment comparing the "aggressiveness" of pre-MY 1998 air bags to MY-1998 air bags, in general.

Table II-6
Special Crash Investigation Cases

* Cases under investigation, but not on official list yet.

Table II-7*
SCI Cases per Million Registered Vehicles

* See endnote at the end of this chapter.

Table II-8 presents the same statistics from the analysis described above but for fatalities that occurred over the last 21 months (10/01/97 to 07/01/99). The comparison indicates fatalities are a little lower over the last 21 month period. The MYs 96-97 averaged 15.5 fatalities (see Table II-7) over the first 21 months and 13 fatalities over the last 21 months. The data still show that redesigned air bags perform better for out-of-position occupants. However, statistics for MY 1998 redesigned air bags were based on limited data and might not be a reliable indication of any trend.

Table II-8
SCI Cases per Million Registered Vehicles

Another analysis to assess the impact of the MY 1998 redesigned air bags on baseline population estimation is to examine the 1998 FARS. A First Cut Analysis of FARS, for MY 1998 redesigned vehicles over the first 12 months of 1998, compared to MY 1997 vehicles over the 12 months of 1997, and MY 1996 vehicles over the 12 months of 1996, was undertaken. The question we were trying to answer is whether the frontal fatality rate increased with the decrease in power in redesigned air bags in MY 1998 vehicles. Testing results showed no difference for belted occupants and a slight difference for unbelted occupants. Most vehicles met the 30 mph unbelted test anyway. We would expect that no difference could be found without substantially more data. No statistically significant difference was found.

The percent of fatalities that were frontal are:

The most relevant comparison might be the 39.6 % for MY 1998 vehicles to all the data available for MY 1997 vehicles (in calendar year 1997 and the 6-month file of 1998) of 40.3%. This results in a risk ratio of 0.972, or a 2.8 percent reduction in frontal fatalities. This is not a statistically significant difference.

These pre-MY 1998 air bags, would save 3,253 lives annually, however, 181 occupants would be killed by the air bags. Thus, the net estimated lives saved would be 3,072 (3,253 - 181). Table II-9 summarizes these estimates in detail. It is important to note that the projections were based on all identified (confirmed and unconfirmed) cases. However, there are 5 unconfirmed cases in 1997 and 14 unconfirmed cases in 1998, therefore, the projected annualized at-risk population could be smaller. Equally important is the fact that all the estimates are based on the assumption that, in the future years, there are no changes in occupant demographics, driver/passenger behavior, belt use, child restraint use, or the percent of children sitting in the front seat. As public education programs are more successful in creating better awareness of occupant safety issues, and as auto manufacturers voluntarily phase in improved air bags, the potential negative safety impacts of air bags would be further reduced.

Table II-9
Estimated Full Fleet Impacts of Pre-MY 1998 Air Bags on Fatalities

* Potentially there are benefits from air bags for correctly positioned children in high severity impacts. Sled test data do show a reduction in injury measures for correctly positioned child dummies with air bags compared to belted child dummies in 30 mph impacts. This does not appear to be the case for infants in rear facing child safety seats. All RFCSS tests have indicated an increased probability of head injury with air bags. Statistical analyses have shown negative effectiveness of air bags for children. This implies that the negative impacts of air bags for children at low speeds are overwhelming the benefits, if any, for children at high speeds. It is impossible to prove that an air bag saved a life in a particular high speed crash, since about 50 percent of unbelted occupants survive (with injuries) in crashes with a change in velocity (delta V) of 30 to 40 mph. Until there are enough data available to do a statistical analysis of the effectiveness of air bags for children at different speeds, the agency cannot estimate the benefits of air bags for children under the age of 12.

B. Injuries

The injury population assessment uses two data sources: the 1993-1997 CDS and the 1997 General Estimates System (GES)⁽³⁾. GES is the main database used by the agency to produce national statistics on nonfatal crashes in the U.S. However, GES is a sample taken directly from police-reported crashes and does not provide in-depth investigations of injury profiles and crash configurations as does CDS. This analysis uses GES to estimate the size of injury populations and CDS to describe crash characteristics such as MAIS injury severity and delta v for crash severity.

CDS contains data on all passenger vehicle crashes where at least one passenger vehicle was towed, while GES is a sample of all police-reported crashes not limited to passenger vehicle tow-away crashes. Therefore, injury counts derived from CDS were adjusted only to the GES CDS-equivalent level. As with FARS, this adjustment cannot establish a one-to-one association between GES CDS-equivalent crashes and CDS crashes. CDS equivalent frontal crashes in GES are defined by Hotdeck imputed initial point of impact (IMPACT_H) from 10, 11, 12, 1, or 2 o'clock.

In 1997, there were 291,859 driver and right front passenger MAIS 2-5 and 1,678,192 MAIS 1 non-fatal injuries associated with frontal crashes. MAIS 1-5 ⁽⁴⁾ injuries reported in Table II-10 and Table II-11 were adjusted to 1997 GES CDS-equivalent injury levels.

Table II-10
1997 Drivers and Right Front Passengers With MAIS 2-5 Injuries

Source: NHTSA 1997 National Automotive Sampling System - General Estimated System (NASS-GES) and - 1993-1997 Crashworthiness Data System (CDS)

Note: MAIS 2-5 injuries were derived from 1993-1997 CDS and adjusted to 1997 GES-CDS equivalent level.

Table II-11
1997 Drivers and Right Front Passengers With MAIS 1 Injuries

Source: NHTSA 1997 National Automotive Sampling System - General Estimated System (NASS-GES) and - 1993-1997 Crashworthiness Data System (CDS)

Note: MAIS 1 injuries were derived from 1993-1997 CDS and adjusted to 1997 GES-CDS equivalent level.

Air bags proved to be 10 percent⁽⁵⁾ (not statistically significant) effective in reducing MAIS 2-5 injuries. Annually, air bags would reduce about 30,211 MAIS 2-5 injuries. Table II-12 shows three types of MAIS 2-5 injury estimates in frontal crashes by person role (driver, passenger) and crash impact speeds. These estimates are: number of MAIS 2-5 injuries with no air bags (potential MAIS 2-5 injuries), injuries reduced, and number of remaining MAIS 2-5 injuries if the whole fleet had air bags. There would be a total of 271,892 MAIS 2-5 injuries remaining annually if all vehicles had pre-MY 1998 air bags. Advanced air bags would have the potential to further reduce these remaining injuries. Note that the distribution of MAIS 2-5 injuries by person role, crash impact speeds (delta v), and restrained use were derived from 1993-1997 CDS statistics. Of the total shown, MAIS 2 injuries were 63.8 percent, MAIS 3 were 28.2 percent, MAIS 4 were 6.1 percent, and MAIS 5 were 1.9 percent. Belt use in Table II-12 is the same level found in 1993-1997 CDS at 57 percent.

A table like Table II-12 was not derived for MAIS 1 injuries since the agency believed the effectiveness of air bags for AIS 1 injuries is minimal. Many occupants have a red face from "bag slap" which is considered an AIS 1 injury. Thus, the effectiveness of reducing overall AIS 1 injuries with pre-MY 1998 air bag is believed to be minimal.

Table II-12
Estimated MAIS 2-5 Injuries Remaining
Assuming the Whole Fleet of Passenger Vehicles Had Air Bags

Source: NHTSA 1993-1997 NASS CDS and 1997 GES.

Note: Injuries by crash speeds were derived from 1993-1997 CDS and adjusted to 1997 GES CDS-equivalent level.

In addition to air bag induced fatalities, SCI also identified some cases where children and adults who sit too close to air bags were seriously injured when air bags deployed in low speed impacts. But these SCI cases are by no means comprehensive, and thus might underestimate air bag induced serious injuries if used as the basis to project annual at-risk serious injuries (MAIS 3-5). Instead, the at-risk fatalities were used as the basis. For each MAIS 3-5 injury level, the estimate of annualized at-risk fatalities is multiplied by the ratio (adjustment factors) of injuries to fatalities. The adjustment factors and the ratio of air bag induced injuries to fatalities were derived from 1993-1998 CDS nonweighted cases. The 1998 CDS data were used here to include more air bag induced cases. Because 1998 CDS does not have the appropriate national weights available yet, the nonweighted cases were used to derive the ratios of injuries to fatalities. The analysis for final rule will be updated using weighted cases. Table II-13 shows that annually a total of 9 RFCSS, 195 children, 14 adult passengers, and 37 drivers were seriously injured by air bags.

Table II-13
Projected Annualized At-Risk MAIS 3-5 Injuries

Source: SCI cases August 1, 1999. 1993-1998 CDS

Note: Baseline is the projected annualized at-risk fatalities.

The following Tables II-14 to II-16 show fatalities and different MAIS injury levels by crash severity (delta v) in frontal crashes. Broken out further, Table II-14 presents these statistics for all front-outboard occupants, belted and unbelted combined; Table II-15 presents statistics for unbelted front-outboard occupants; while Table II-16 presents statistics for belted front-outboard occupants. These tables serve as additional background information to make a necessary adjustment of the overall target population and to analyze benefits for these tests. Figure II-1 graphically depicts the cumulative percentages of fatalities and MAIS 2-5 injuries by crash impact speeds (delta v). Figure II-2 shows the percentage distribution by principal direction of force.

Table II-14
Front-Outboard Occupant Fatalities and MAIS 2-5 Injuries
in Frontal Crashes by Crash Impact Speeds

Source:1993-1997 CDS.

Table II-15
Front-Outboard Occupant Unbelted Fatalities and MAIS 2-5 Injuries
in Frontal Crashes by Crash Impact Speeds

Source:1993-1997 CDS.

Note: Fatalities and MAIS 2-5 injuries with unknown crash impact speeds were excluded in the analysis.

Table II-16
Front-Outboard Occupant Belted Fatalities and MAIS 2-5 Injuries
in Frontal Crashes by Crash Impact Speeds

Source:1993-1997 CDS.

Note: Fatalities and MAIS 2-5 injuries with unknown crash impact speeds were excluded in the analysis.

Endnote for Table II-7,

A small analysis was performed to examine the potential impact of the increase in belt use between 1996 and 1998 on the apparent change in fatality rates for redesigned MY 1998 air bag vehicles. Results from the National Occupant Protection Use Survey (NOPUS) show that the observed average overall safety belt use in front seat outboard passenger cars and light trucks was 61.3 percent in 1996 and 68.9 percent in 1998. Thus, safety belt use increased during the period. The estimates use the average effectiveness of safety belts for passenger cars and light trucks of about 51 percent (45 percent for passenger cars and 60 percent for light trucks). It is estimated that instead of the 14 fatalities that occurred over the first 21 months of the MY 1996 vehicles shown in Table II-7 with 61.3 percent belt use, that there would have been 13 fatalities if belt use had been 68.9 percent [14/(1-0.51*0.613) = 20.37; 20.37*0.51*0.689 = 7.16; 20.37-7.16 = 13]. The SCI fatalities per million registered MY 1996 vehicles using 13 fatalities instead of 14 would be 0.99.

The same calculation for 1997, assuming average belt use in 1997 midway between 1996 and 1998 of 65.1 percent resulted in no change in the number of SCI fatalities of 17 and the fatality rate remaining at 1.20 for MY 1997 vehicles. Thus the average SCI fatality rate for MY's 1996/97 would be 1.10 [(0.99 + 1.20)/2]. Comparing this to MY 1998 rate of 0.34 results in 30.9 percent (0.34/1.10) rather than the previous estimate of 29.8 percent (0.34/1.14), still roughly 30 percent as used in the analysis for Table II-5.

^1.. The restraint use distribution was based on the 1993 to 1997 National Automotive Sampling System (NASS) Crashworthiness Data System (CDS), so that this table would be consistent with Table II-2. Table II-2 provides a distribution of fatalities by delta v. Delta v is only available in NASS-CDS.

^2. The figures in the table are slightly different from the estimates in the "Preliminary Economic Assessment, FMVSS No.208, Advanced Air Bags" August 1998, NHTSA, because this analysis uses 1997 FARS and 1993-1997 CDS crash data. Also, this analysis used a different projection approach to estimate the annual at-risk population. Finally, the number killed by air bags was projected using later data from the Special Crash Investigation program, up to August 1, 1999.

^3. General Estimates System Coding Manual 1997.

^4. Maximum Abbreviated Injury Scale, 1-Minor Injury, 2-Moderate Injury, 3-Serious Injury, 4-Severe Injury, 5-Critical Injury. Only one injury with the most severity is counted per occupants.

^5. The Fourth Report to Congress, Effectiveness of Occupant Protection Systems and Their Use, May 1999.