You asked that we identify for a recent time period the number of people who were killed in traffic accidents in Connecticut for which one of these four negative driving behaviors was a relevant factor—driving under the influence of alcohol, speeding, failure to wear seat belts, or failure to wear a motorcycle helmet. You also asked for our assessment of how many of these fatalities might not have occurred had these four negative factors not occurred.

SUMMARY

We must note at the outset of this response that an assessment of this sort is highly speculative. To estimate the number of potentially avoidable fatalities that might be attributable to the elimination of the four fatality-producing behaviors, several broad assumptions had to be made and an accounting had to be made for ways in which these factors overlap. As noted in the analysis, the assumption was made that in the fatalities attributed to alcohol and excessive speed no other factors would have caused the accidents. The potentially avoidable fatalities attributable to alcohol were based on victims who were associated with a driver who had a known blood alcohol content (BAC) of . 10% or more as shown in the National Highway traffic Safety Administration’s (NHTSA) Fatality Analysis Reporting System (FARS). This was the per se level that determined the offense of driving while under the influence in 2001. Assumptions regarding seat belt and motorcycle helmet use effectiveness in preventing fatal injuries are based on NHTSA studies estimating such effectiveness.

Data from FARS for 2001 shows a total of 318 fatalities including 213 drivers, 67 passengers, and 38 others (33 pedestrians, three bicyclists, and two others). Our analysis based on the assumptions noted above suggests that 213 of the 318 fatalities were potentially avoidable had no drivers operated at BACs of . 10% or more or driven at speeds too fast for conditions or above posted speed limits, all victims had worn seat belts, and all fatally injured motorcyclists worn helmets.

Table 1 breaks out the potentially avoidable fatalities that we were able to identify in 2001 attributable to each of the negative behaviors in which you were interested.

Table 1: Estimated 2001 Fatalities Possibly Avoidable Due to Four Selected Driving or Riding Behaviors

ALCOHOL RELATED FATALITIES

DOT statistics indicate that 144 of the total of motor vehicle related fatalities occurring in 2001 were “alcohol related. ” Under federal and state statistical schemes an accident or fatality is considered alcohol related if a driver or a non-motorist involved in the accident has a measurable or estimated blood alcohol concentration (BAC) of . 01% or more or there is an indication in the police report that the driver or non-motorist had or may have consumed alcohol. (In 2001, the BAC level that determined the per se offense of intoxication was . 10%. It was lowered to . 08% in 2002. )

It is important to note two things about the use of the term “alcohol related. ” First, there is not necessarily an automatic or direct causal link between the alcohol involvement determination and a determination of responsibility for the accident. In other words, it does not necessarily mean that in all cases (1) the person who had consumed alcohol caused the accident or (2) if the person who consumed alcohol caused the accident that the alcohol consumption contributed to the accident. Second, there may be no evidence other than the police officer’s belief as indicated in the police report that a participant in the accident had consumed an amount of alcohol at some point prior to the accident. Thus, some alcohol-involved accidents could, and frequently do, involve a driver who has consumed no alcohol and strikes a pedestrian who had been drinking or a driver who had consumed a small amount of alcohol prior to the accident and who was not intoxicated, but was noted by the investigating officer as “had consumed alcohol” even though he was not cited as responsible for the accident.

NHTSA notes in FARS that alcohol test result data from this database must be “interpreted with caution. ” It states that alcohol test results obtained from the database are the actual state-reported data. However, this data differs from NHTSA’s published estimates of alcohol involvement because it employs a statistical model to impute the presence of alcohol in cases where there may be no direct evidence of alcohol in the record, but certain factors relating to the accident, such as the age and sex of the driver and the time of the accident, fit a profile of other accidents that are known to have involved alcohol. NHTSA assigns a blood alcohol level to the driver based on this statistical model. For 2001, this statistical imputation formula inflates the percentage of fatalities in the Connecticut data that were deemed alcohol related from 45. 3% to 50. 6%. In effect, in its statistical compilations for Connecticut, NHTSA has attributed an additional 14 fatalities as alcohol-related using this model for which there is no actual BAC evidence. For this analysis,

we have relied on the state-reported data (144 fatalities) based on actual BAC test results rather than the statistically-enhanced NHTSA data (158 fatalities).

In 2001, the FARS data show that a total of 88 drivers involved in fatal accidents in Connecticut had BAC test results exceeding the legal limit for operating a motor vehicle of . 10% or more. Table 2 shows the BAC distribution for these 88 drivers as indicated in the FARS data.

Table 2: Known BAC Test Result Distribution of Drivers Involved in Fatal Motor Vehicle Accidents in Connecticut in 2001

Fatality Avoidance Assessment

Our analysis of the FARS data files involving drivers who were operating above the legal BAC limit suggests that 52 fatalities were potentially avoidable had there been no incidents of a driver operating with a BAC of . 10% or more—the BAC that determined intoxication in Connecticut in 2001. This estimate is based on the assumption that the participants’ BAC levels were in fact the primary reason why the accident occurred, that the intoxicated driver was actually at-fault for the accident, and that no other factor could have led to the result. This is a significant assumption, in that it is probably not true in all cases, but is necessary for the sake of this analysis.

The figure of 52 potentially avoidable fatalities includes 33 drivers known to have had BACs indicating intoxication, 13 passengers, four pedestrians, and two drivers of other vehicles involved in the accident. Several pedestrians and one bicyclist who had BACs above what would

be allowed if they had been driving a motor vehicle were not considered as potentially avoidable since there was no evidence that the driver of the vehicle had consumed alcohol.

The 52 fatalities represent situations where the 33 drivers involved were operating with an illegal BAC level and there was no evidence of excessive speed. As noted below, the group of drivers and the associated fatalities where there were indications of both driving with illegal BACs and excessive speed were treated separately from the groups that exhibited only illegal BACs or excessive speed.

SPEEDING

In 2001, the FARS data showed driving too fast for conditions or in excess of the posted speed limit as a related factor in 111 fatalities (80 drivers, 31 passengers), approximately 35% of the total fatalities. Once again, based on the assumption that this was the primary reason for the accident, all of these fatalities would be potentially avoidable if the negative driving behavior was eliminated. However, in many cases, the speeding behavior occurred in conjunction with consumption of alcohol at levels indicating intoxication. Thus, a significant portion of these fatalities have already been counted in the alcohol-related avoidable fatalities. Our analysis of the individual accident files indicates that 55 of the fatalities (43 drivers, 12 passengers) fall into this overlapping alcohol/speed category and have been separated out as shown in the following section. Therefore, the number of potentially avoidable fatalities we attributed solely to excessive speed was 60 (40 drivers, 20 passengers).

FATALITIES POTENTIALLY SUBJECT TO COUNTING AS EITHER SPEED OR ALCOHOL RELATED

The FARS data shows that some of the fatalities involved factors that related to both intoxication and excessive speed. Since these could be counted in either category, we have treated them separately in the analysis. The potentially avoidable fatalities (based on the same assumptions noted above) falling in this group total 53—40 drivers, 12 vehicle passengers, and one driver of another vehicle involved in the accident.

USE OF SEAT BELTS

Fatality Overview

Of the 318 motor vehicle related fatalities reported in FARS in 2001, 280 were vehicle occupants (213 drivers, 67 passengers). After eliminating motorcyclists and victims whose restraint use could not be determined, the FARS database identifies a total of 241 fatally injured people (183 drivers, 58 passengers) for whom restraint use could be determined. This includes children who should be in child safety seats rather than seat belts. The FARS data shows that 67 of the fatally injured drivers and 22 of the fatally injured passengers were in restraints (one child was killed in an improperly used child safety seat). The remaining 152 fatally injured people (116 drivers, 36 passengers) were not shown to have been using restraints. However, 41 of the 152 were riding in vehicles in which air bags deployed. Of the 111 fatalities that remained (81 drivers, 30 passengers), in six cases air bags were available in the vehicle, but it was not known if they deployed. For this analysis, we have assumed that in these six unknown cases, air bags were not a factor. Of the 30 unrestrained passengers, 15 were in the front seat, 14 were in the second seat, and one was in some other part of the passenger compartment or cargo area of the vehicle.

Fatality Avoidance Assessment

This is the most complex of the four areas for which to determine potentially avoidable fatalities. NHTSA has done several studies over the years that have examined FARS records to estimate occupant protection system effectiveness. Generally speaking, restraint system effectiveness can depend on several factors, such as the presence of air bags, the principal point of impact, the age of the victim, vehicle weight, and if the most harmful event in the accident was a rollover. Restraints are most effective in purely frontal crashes (12 o’clock position on the vehicle), relatively effective in partial head-on crashes (10 o’clock to 2 o’clock position), and somewhat less effective for other points of impact (side and rear-end crashes). Vehicle rollover also serves to reduce restraint effectiveness significantly.

The NHTSA analysis states that belted drivers in cars with air bags have a 21% fatality reduction in purely frontal crashes relative to belted drivers in cars without air bags (and 11% in all types of crashes). Unbelted drivers in cars with air bags have a 36% fatality reduction in purely frontal crashes relative to unbelted drivers in cars without air bags (and 14% in all types of crashes). Belts alone are estimated to

reduce fatality risk by 45% (Fifth/Sixth Report to Congress—Effectiveness of Occupant Protection Systems and Their Use, DOT HS 809-442, November 2001).

To estimate the number of potentially avoidable fatalities based on lack of restraint use we first had to account for the drivers with known BACs of . 10% or more already accounted for in the alcohol-related analysis. The FARS data indicates that 135 fatally injured drivers were apparently not using occupant restraints. However, 57 of them had known BACs of . 10% or more. Since they have already been accounted for in the alcohol-related analysis, they were discounted here to avoid double counting. This leaves 78 fatally injured drivers (1) who had no known alcohol in their systems (47), (2) who had BACs of . 01-. 09% (10), or (3) for whom a test was not administered, was administered but the results were not known, or it was unknown if a test was performed (21).

Of the 78 drivers, The FARS data show that for 58 no air bag was available or there was no evidence of deployment. For the other 20 drivers, the data indicated deployment of an air bag. To estimate the number of potentially avoidable driver fatalities from this group of 78 drivers we applied the NHTSA 45% seat belt effectiveness estimate to the 58 drivers without air bag deployment and a fatality reduction of 6%--the difference between the 45% effectiveness for seat belts alone and the 51% effectiveness NHTSA estimates for air bags plus lap-shoulder belts—to the 20 drivers for whom air bags deployed. This yields 29 potentially avoidable fatalities attributable to lack of restraint use (27 with no air bag deployment and 2 when air bag deployment occurred).

Of 33 passengers not already accounted for in the alcohol analysis, six were front seat passengers for whom air bags deployed and the other 25 were either front seat passengers without air bag deployment or were seated elsewhere in the vehicle. Applying the same factors used above, we estimated that 13 of them might have been potentially avoidable fatalities had restraints been used (12 not covered by air bags and one for whom the air bag deployed.

USE OF MOTORCYCLE HELMETS

In 2001, there were 46 motorcyclist fatalities (44 operators, 2 passengers). DOT data show that 12 of the 46 people (26. 1%) who died were wearing helmets (26. 1%) and the other 34 were not wearing helmets (73. 9%). NHTSA has published several studies in which it has attempted to estimate the effectiveness of helmets in preventing fatal injuries to motorcyclists. These studies use FARS data to compare crashes involving motorcycles with two occupants, at least one of whom was killed. NHTSA recently published an update of its previous effectiveness estimates in which in concludes that helmets are 37% effective in preventing fatal injuries to motorcycle riders (Motorcycle Helmet Effectiveness Revisited, DOT HS 809-715, March 2004).

Alcohol use and, in particular, excessive speed were significantly associated with the motorcycle fatalities. Of the 34 motorcycle fatalities involving unhelmeted victims, 13 were associated with excessive speed, 4 with operator BAC levels of . 10% or more, and 5 with both excessive speed and BAC levels of . 10% or more. Therefore, these 22 fatalities are already accounted for in the totals for the three analytical categories dealing with those specific factors and, to avoid double counting, they have been excluded from this section of the analysis. Applying NHTSA’s 37% helmet effectiveness to the remaining 14 unhelmeted fatalities that did not involve either excessive speed or illegal BAC levels yields 6 potentially avoidable fatalities had all 14 of them been wearing helmets.

JF: nf