Better to see Chapter 6 Gender, age, and alcohol effects on survival
and
Chapter 7 Older drivers of Traffic Safety (2004)
Below is only words of 1991 book -- the typeset pages may be viewed at no cost by clicking here
Chapter 2. EFFECTS OF SEX AND AGE (From 1991 book Traffic Safety and the Driver)
Most road-user factors important in traffic safety depend strongly on the sex and age of the road-user. US male fatalities outnumber female fatalities by well over a factor of two; there are eight 20-year-old male drivers killed for each 65-year-old driver killed. A focus on the variables sex and age is therefore to be expected. Another reason for the focus on these variables is that they tend to be the only demographic variables available, as Haight [1985] comments. The first question we address is a basic one not confined to traffic, but an important consideration in nearly all studies of fatal injury. This question is, "How does the risk of death from identical physical impacts depend on sex and age?" Although the question relates more to basic human physiology than to traffic, it was answered using a data file collected for traffic safety applications (many examples of its use to address traffic safety questions are described later).
THE FATAL ACCIDENT REPORTING SYSTEM (FARS)
The Fatal Accident Reporting System, which we shall hereafter refer to as FARS, is a computerized data file maintained by the National Highway Traffic Safety Administration, an agency of the US Department of Transportation. The file was set up to document every fatal crash that occurred on a US public road since 1 January 1975. A fatal crash is defined as one in which anyone dies within 30 days of the crash as a result of the crash.
Data on the crash, the people involved, and the vehicles involved are compiled based mainly on information provided by police at the scene of the crash. Because the goal is to include every fatal crash, there is no possibility of post-crash physical examinations of vehicle and site to provide estimates of travel speed, or vehicle change of speed on impact, the most effective indicator of crash severity. The file is largely limited to information readily recorded, such as the number and type of vehicles involved, sex and age of occupants involved, time of day, posted speed limit on the roadway, and the like. More specific vehicular information can be extracted from the Vehicle Identification Number (VIN) and the driver's license number. By the end of 1988 the file had records on over half a million fatal crashes in which over 650 000 people were killed. The number of fatalities necessarily exceeds the number of crashes because there must be at least one fatality for the crash to be included in FARS. A printed publication giving much summary information from the file is issued annually [National Highway Traffic Safety Administration 1989, and prior years].
Even though data in this file
are as reliable as any available, there is still always some uncertainty. There are obviously people, especially
elderly victims, who die from non-trauma causes in hospital within 30 days of a
crash, in which it is not clear whether the death resulted from the crash or
other causes. The total number of
fatalities in the FARS file is less than that estimated by the National Safety
Council [1989] by between 2.9% and 4.5% (average 3.9%) for the years 1975 -
1988. The main contributor to this
difference is that the National Safety Council includes deaths that occur within
one year of the crash; even longer periods would increase the total further,
but by increasingly small amounts.
Although the 3.9% difference corresponds to almost 2000 traffic deaths
per year not being coded in FARS, this is not expected to materially affect the
results of studies described
here, especially as most characteristics of the excluded fatalities are
expected to reasonably match those of the included. Other problems of missing, and possibly inaccurate, data will be
mentioned when specific studies are described.
These few points are mentioned to stress that even fatality data coded
in FARS are far from perfect, even though they are far more reliable than data
for any other level of harm.
RISK OF DEATH FROM THE SAME PHYSICAL IMPACT
Given that a male and female of similar age receive similar physical insults, or impacts, which is more likely to die? Traditional epidemiological studies are unable to answer this question because adequate samples of sufficiently similar cases are not available. Medical treatment naturally focuses more on the injuries than on quantifying the forces that caused them. While it was well known that death and injury risk increase with increasing age [Baker, O'Neill, and Karph 1984; Verbrugge 1982; Waller 1985], the effect was not quantified. By using the method described below to make appropriate inferences from the FARS data, Evans [1988a] addressed this physiological problem using data on over 80 000 fatalities.
Double pair comparison method
The double pair comparison method enables us to make inferences from the FARS data without needing external exposure measures. Two classes of occupants, "subject" occupants and "control" occupant are used. The influence of some characteristic (sex, age, belt-wearing, etc.) of the subject occupant on his or her fatality probability in a crash, other factors being the same, is estimated using the control occupant to standardize conditions -- that is, to estimate exposure. Many different subject and control occupants were used to study the influence of sex and age on fatality risk. For expository convenience the method is described below for one specific case in which the subject occupant is a car driver and the control occupant is a male passenger seated in the right-front seat.
Two sets of fatal crashes are selected. The first set consists of crashes involving cars each containing a female driver and a male passenger, at least one of whom is killed. From the numbers of female driver and male passenger fatalities, a female driver to male passenger fatality ratio is calculated. From a second set of crashes involving cars containing male drivers and male passengers, a male driver to male passenger ratio is similarly calculated. Under assumptions discussed in Evans [1986], dividing the first fatality ratio by the second gives the probability that a female driver is killed compared to the corresponding probability that a male driver is killed, averaged over the distribution of crashes that occur in actual traffic; thus we obtain the risk to the female driver compared to the risk to the male driver averaged over a wide range of impact severities.
The control occupant, the male
passenger, does not enter directly into the result. Because of this key feature of the method, many separate
estimates can be calculated by choosing various control occupants. This helps avoid confounding influences due
to interactions between subject and control occupant; the basic assumptions of
the method require that, given crashes of identical severity, the probability
of a passenger death should not depend (in the present example) on the sex of
the driver. This assumption would be
violated if, for example, the same physical insult was more likely to kill a
passenger travelling with a male driver than one travelling with a female
driver. Departures from this assumption
could arise if, for example, passengers travelling with male drivers tend to be
older than those travelling
with female drivers. The potentially
biasing influences of such confounding interactions can be removed by disaggregating
the passenger into age categories so that male and female drivers are examined
when accompanied by passengers of similar age.
Consequently, the control occupant will be disaggregated into as many
categories as the data allow. It is
further assumed that the information coded in the FARS data is correct; the
potentially biasing effects of missing data are discussed in detail for a
number of specific cases by Evans [1988b].
Let us give the specific example of comparing unbelted car driver fatality risk for females aged 33-37 to that for males in the same age interval (call them 35-year-old drivers). For the control occupant we choose unbelted male right-front passengers aged 16-24, hereafter referred to as 20-year-old male passengers. The 1975-1983 FARS data used by Evans [1988a] show that, in the set of cars used in the first of the two comparisons, 43 female drivers aged 35 were killed while travelling with 20-year-old male passengers, while 23 male passengers aged 20 were killed while travelling with female drivers aged 35. From these values we compute the 35-year-old female driver to 20-year-old male passenger fatality ratio
r1 = 43/23 = 1.870 . Eqn 2-1
That is, for this case, 1.870 female drivers were killed per male passenger killed.
The corresponding information from the set of crashes used for the second comparison, in which the driver is male rather than female, gives a 35-year-old male driver to 20-year-old male passenger ratio
r2 = 206/128 = 1.609 . Eqn 2-2
The female to male fatality risk for 35-year-old car drivers (using 20-year-old male passengers as the control occupant) is given by
R = r1/r2 = 1.162 . Eqn 2-3
Note that any asymmetries between the safety of the driver and passenger seating positions operate in one direction in the first comparison (Eqn 2-1), but in the opposite direction in the second comparison (Eqn 2-2), so that when the Eqn 2-3 ratio is computed any such effects cancel. The driver risk far exceeds the passenger risk regardless of the sex of the driver in the specific case illustrated because of the age difference between subject and control occupants, as discussed later. Based on the raw data above, the standard error in R can be calculated as described in Evans [1986; 1988c], and for this case is 0.332. The relatively large uncertainty flows mainly from the smallest of the four frequency counts in the calculation, namely, the 23 male passengers, aged 16-24, killed travelling with 35-year-old female drivers. Inferences about relative fatality risk in crashes were made by Partyka [1984] and Kahane [1986] using the same ratios explored formally in the double pair comparison method [Evans 1986a].
The above estimate of fatality risk to 35-year-old female drivers compared to 35-year-old male drivers was based on one set of control occupants -- males aged 16-24. In Evans [1988a] the same calculation is applied for 8 classes of control occupants (male and female passengers, each in four age categories). Each calculation provides an independent estimate of the fatality risk to 35-year-old female drivers compared to that to 35-year-old male drivers, and procedures are described in Evans [1986; 1988c] to determine a weighted average and associated standard error.
Effect of sex on fatality risk
Applying the above procedure to car drivers in five-year age cells generates the results plotted in the top left graph of Fig. 2-1. The larger standard errors at older ages arise because of fewer fatalities to older drivers. The discussion has so far focused exclusively on drivers as subjects. However, the subject can be any vehicle occupant. The top right plot in Fig. 2-1 is for right-front passengers as subjects, using drivers as control occupants; in this case we have data for ages below the minimum for driving. The other plots in Fig. 2-1 show corresponding data for other subject occupants for which substantial numbers of male and female data are available. The rear-car-seat estimates use drivers as control subjects. Fatality risk to female and male motorcycle drivers cannot be compared because of insufficient female fatalities; motorcycle drivers (male) are used as control subjects for the motorcycle passenger comparison.
------------------------------------------------------------------------------
Fig. 2-1 about here
------------------------------------------------------------------------------
Subjects in the eight
categories in Fig. 2-1 are killed by a wide range of impact mechanisms. For example, car occupant fatalities usually
are associated with impact upon the vehicle interior, while motorcyclist
fatalities with impact upon objects not related to the vehicle. The absence or presence of steering wheels,
safety belts, helmets, cushioning effects of motorcyclist drivers in front,
etc. all affect the details of the injury insult. Given these differences, the extent to which the eight plots in
Fig. 2-1 show similar features for ages at which there are data in common is
notable. For example, for all eight
cases, at average age 30, the female
fatality risk exceeds the male fatality risk.
The eight values range from a low of 15% (that is, R=1.15) for car
drivers to a high of 49% for unhelmeted motorcycle passengers. The weighted average of the eight values is
(31 + 6)%. There is no indication
that any of the eight individual estimates departs in a statistically
significant way from this average value, supporting the interpretation that the
same physical insult is 31% more likely to kill a 30-year-old female than a
30-year-old male.
The features in common in the eight plots in Fig. 2-1 suggest that the effect displayed is due essentially to differences in basic susceptibility to fatal trauma as a function of sex, with the specific nature of the traffic crash being of secondary importance. That is, we would anticipate that fatality risk would similarly depend on sex for other types of potentially fatal physical insults, such as severe falls or blows from objects (including vehicles - the present method cannot be applied to investigate pedestrian fatality risk).
Fig. 2-2 presents a synthesis of all the information in Fig. 2-1. The point plotted at each age is the weighted average and standard error for all the points at the same age. The number of points contributing to the average varies between 2 and 8. From about age 15 to age 45, the same physical insult is approximately 25% more likely to kill a female than a male of the same age. For ages less than about 5, the risk is higher for males than for females. At ages greater than 60 there is a suggestion that the fatality risk may again become higher for males, although the uncertainty is too great to justify any definitive conclusion.
------------------------------------------------------------------------------
Fig. 2-2 about here
------------------------------------------------------------------------------
Effect of age on fatality risk
Age effects are more difficult to determine than the sex effects because of extremely strong interactions between driver age and passenger age -- drivers of a given age tend to travel with passengers of similar age, but of opposite sex. Although both tendencies facilitated the sex analysis, they make the age analysis more difficult because the cases which provide the crucial information are those less common ones in which the occupants differ in age. Further, the likelihood of finding subject and control occupants of dissimilar age decreases steeply with the magnitude of the age discrepancy. Of even greater concern is the possibility of strong unknown confounding factors. For example, crashes by cars containing two 20-year olds might differ in so many ways from crashes by cars containing a 20-year old and a 70-year old that important violations of the assumptions on which the method is based could occur. To reduce the possibility of such potentially confounding effects, the following approach was adopted, which will be explained in terms of male car drivers.
First, fatality risk for
25-year-old male drivers was determined relative to fatality risk for
20-year-old males. Then fatality risk
for 30-year-old males was determined in the same way relative to 25-year-old
males (not 20-year-olds). The ratio of
fatality risk for 30-year olds relative to 20-year olds was then obtained as
the product of the two ratios, and the standard error as a combination of the
errors for both. The risk for
35-year-old male drivers relative to 20-year-old male drivers was then obtained
by multiplying this by the 35 to 30 year ratio, and so on for all ages above
(and below) 20 . In this way fatality
risks for all the ages relative to age 20 are obtained without individual
comparisons involving large differences in age. As the error at any age reflects contributions from each step
away from the
reference, it will necessarily increase as we move further from the reference
age. Applying this process to the data
for each of the individual subject-occupant categories generates the 10 plots
shown in Fig. 2-3; the two plots more than in Fig. 2-1 are motorcycle
drivers. A figure corresponding to Fig.
2-3, but for females, is given in Evans [1988a].
------------------------------------------------------------------------------
Fig. 2-3 about here
------------------------------------------------------------------------------
Comments previously made regarding Fig. 2-1 apply to Fig. 2-3 (and to the corresponding data in Evans [1988a] for females). The plots are interpreted as showing the age dependence of basic physiological response to physical impact, with the specific details of the physical insult being of less central importance. There is every reason to expect that these same relationships apply to physical insults unrelated to occupant injuries; for example to people struck by vehicles, or to injuries unrelated to traffic. Additional supportive evidence for this interpretation is the finding in Evans [1988a] that age effects are relatively similar for car passengers involved in different types of crashes.
The age effects are summarized in Fig. 2-4, in which each point is the average extracted from the 10 different male occupants and 8 different female occupants. The form of both figures at young ages explains part of the increase in the number of occupant fatalities per capita with declining age for young children reported by Baker [1979]. The relations in Fig. 2-4 can be expressed analytically as
Rmales(A) = exp 0.0231 (A - 20) = 0.630 exp ( 0.0231 A) Eqn 2-4
and
Rfemales(A) = 1.3 exp 0.0197 (A - 20) = 0.877 exp ( 0.0197 A) Eqn 2-5
for A ³ 20, where A is age in years and R is probability that a given impact will prove fatal relative to the probability that the same impact will kill a 20-year-old male. Once age exceeds about 20, fatality risk grows at an approximately uniform rate of (2.3 + 0.2)% per year for males and (2.0 + 0.2)% per year for females. At age 70 the risk is about three times what it is at age 20. Dividing Eqn 2-5 by Eqn 2-4 produces the broad effect, but not the details, of the Fig. 2-2 direct comparison of female to male risk.
------------------------------------------------------------------------------
Fig. 2-4 about here
------------------------------------------------------------------------------
The data used to derive the age and sex effects were included irrespective of alcohol involvement. Because alcohol use increases the risk of death from the same impact (Chapter 7), and its use is more associated with males and the young, it is likely that the results reported above underestimate the degree to which females are more likely than males to be killed by the same physical impact, and the extent to which risk increases with age.
VARIOUS DRIVER FATALITY RATES
For the remainder of this
chapter driver means a driver of any motorized vehicle, including a motorcycle,
truck, bus, etc. This choice insures a
simple categorization of all traffic fatalities as either drivers or
non-drivers; pedalcyclists are considered non-drivers. The dependence of driver fatalities on age
and sex examined below, which is based on Evans [1988c], thus reflects choice
of vehicle, how it is used and what the consequences of a
crash are, given that one occurs, all factors which are themselves strongly
influenced by age and sex. (Some information parallel to that presented here,
but for car drivers only, is given in Evans [1987]).
Fig. 2-5 shows the number of driver fatalities versus age and sex for FARS 1981-1985. The pattern is very stable from year to year - the plots (with the ordinate scale to 1000 rather than 5000) for each of the individual FARS years look essentially the same as Fig. 2-5, except for an increase in random variation, or noise. The steep decline with age (the value for 65-year-old males is 0.123 times what it is for 20-year-old males) is in part due to the fact that there are more younger than older people in the US population.
------------------------------------------------------------------------------
Fig. 2-5 about here
------------------------------------------------------------------------------
Fig. 2-6 shows the data in Fig. 2-5 normalized for population using data from the Bureau of the Census [1987] giving estimates of the resident population on July 1 by age (in 1 year increments) and sex. The point plotted in Fig. 2-6 at (say) age 65 is the number of 65-year-old drivers killed from 1981-1985 divided by the sum of the number of 65-year olds in each of these years. This plot is therefore an average, weighted by population, of graphs for individual years, all of which look similar (with more noise) to Fig. 2-6. Fig. 2-6 shows driver deaths per capita increase with age for males over about age 65.
------------------------------------------------------------------------------
Fig. 2-6 about here
------------------------------------------------------------------------------
Fig. 2-7 shows driver
fatalities per licensed driver, as given by the Federal Highway Administration [1983]. The data plotted here are for 1983 to ensure compatibility with the
estimates of distance of travel introduced for
the next graph. The general pattern in
Fig. 2-7 differs from that in Fig. 2-6 only insofar as the fraction of the
population holding driving licenses varies with age; in particular, older
females are less likely than those in mid life (in 1983) to have driver
licenses; for example about 70 percent of 70-year-old females have licenses,
compared to over 90 percent for similarly aged males or 30-year-old females
[Federal Highway Administration 1983].
Thus, in contrast to Fig. 2-6, the rates in Fig. 2-7 increase with age
for males and females older than about 65.
------------------------------------------------------------------------------
Fig. 2-7 about here
------------------------------------------------------------------------------
Fig. 2-8 shows, on a logarithmic scale, the number of driver fatalities per unit distance of travel, which is estimated by dividing driver fatalities by the product of the number of licensed drivers and the average distance of travel per driver as determined in the Nationwide Personal Transportation Study data [US Department of Transportation 1985] for 1983. Fig. 2-8 shows further elevation for older and younger ages above the average because older and younger drivers travel less than average drivers do. For example, the average annual distance of travel for male drivers aged 65-69, aged 35-39, and aged 16 is, respectively, 14 500 km, 31 400 km, and 2 200 km.
------------------------------------------------------------------------------
Fig. 2-8 about here
------------------------------------------------------------------------------
INVOLVEMENT RATES IN SEVERE CRASHES
Increases with age like those
in Figs 2-7 and 2-8 have contributed to concerns regarding the driving
performance of older drivers. Such
concerns
find additional support in research showing various changes in mental and
sensory functions as humans age [Charness 1985; Reff and Schneider 1982;
Welford 1981]. Declines with age have
been found for such driving tasks as reading signs at night [Sivak, Olson, and
Pastalan 1981], perceiving and reacting to roadway hazards [Olson and Sivak
1986], and general driver performance [Yanik 1985]. Ranney and Pulling [1989] find reaction times for skills related
to vehicle control increase with age.
Involvement rates in fatal crashes do not correctly reflect such
changes, because of the strong influence of age on fatality risk when a crash
occurs. The number of drivers of given
age and sex killed should be considered to be the product of two factors:
1. The number of involvements in very serious crashes
2. The probability that involvement proves fatal.
The first factor reflects influences due to all use and behavioral factors, such as amount and type of driving, driver capabilities, type of vehicle driven, time of day, degree of intoxication, and driving risks. The second factor can be influenced also by such behavioral factors as safety belt wearing and alcohol consumption. Apart from such considerations, which will be discussed in Chapters 7 and 9, the probability that a given crash results in death is essentially physiological rather than behavioral in nature, and for the present purposes can be adequately approximated by Eqns 2-4 and 2-5. When driver age is 16 to 20, we assume R = 1 for males and R = 1.3 for females; that is, the fatality risk from the same severity crash is the same as for a 20-year-old driver of the same sex.
Fatality rates focus on the
outcome, not the severity of the crash that led to the death. Here we examine involvement rates in crashes
of similar
severity by considering crashes in a severity range greater than or equal to
that sufficient to kill 80-year-old male drivers, for which case R has a value
of 4.0 (Eqn 2-4). Consider a set of
crashes in which N fatalities occur to 80-year-old males. If these crashes were repeated keeping all
factors the same except the drivers, then we would expect 0.25N fatalities for
20-year-old male drivers and 0.325N fatalities for 20-year-old female drivers
(Eqn 2-5). In order to obtain the same
number of fatalities, 4.0 times as many crashes by 20-year-old drivers, and 3.1
times as many crashes by 20-year-old female drivers are required. In this way we can use the observed numbers
of fatalities to infer involvement rates in crashes in the severity range
sufficient to kill 80-year-old male drivers.
Fig. 2-9 shows the number of involvements in crashes in the same severity range (that necessary to kill 80-year-old males) per licensed driver versus age and sex. In contrast to Fig. 2-7, there is now no longer any noticeable upward trend at older ages. The upward trend in Fig. 2-7 was caused by the increase in fatality risk from the same impact with increasing age, and not by an increase in involvements with increasing age.
------------------------------------------------------------------------------
Fig. 2-9 about here
------------------------------------------------------------------------------
Severe crash involvements per unit distance of travel (Fig. 2-10) increase with increasing driver age for ages above about 60. However, the increase is smaller than in Fig. 2-8; even at the oldest age plotted, the rates for males and females are still less than those for male drivers under 30.
------------------------------------------------------------------------------
Fig. 2-10 about here
------------------------------------------------------------------------------
THREAT TO OTHER ROAD USERS
All the above focused on how the age and sex of a driver influence the threat to the driver's own life. In many ways (for example, legal) this risk is presumed to be largely under the control of the driver. Here we address how the risk a driver poses to other road users depends on the driver's age and sex. This question raises a host of different legal and moral issues which are relevant to discussion of driver licensing policy, in particular licensing test procedures that may make it more difficult for the elderly to obtain licenses. We investigate the threat to other road users by examining the number of crashes in which pedestrians are killed as a function of the age and sex of drivers (of any type of motorized vehicle) involved in the crashes. Attention is confined to single vehicle crashes because when more than one vehicle is involved it is not always possible to determine from the FARS data which vehicle struck the pedestrian. In addition, involvement in multiple vehicle crashes poses threats to drivers different from those of single vehicle crashes in which pedestrians are killed; the drivers of cars in single-vehicle pedestrian-fatality crashes are themselves usually not seriously injured. No assumption is made regarding responsibility in pedestrian fatality crashes; the FARS data show about one third of fatally injured pedestrians have blood alcohol concentrations in excess of 0.1 percent by volume, the legal limit for intoxication in most US states.
Figs 2-11 through 2-14 show the
variables for crashes involving pedestrian fatalities corresponding to those
for driver fatalities in Figs 2-5 through 2-8.
The similarity between each corresponding set of curves reflects the
extent to which pedestrian fatalities are proportional to driver fatalities,
the basis of the pedestrian fatality exposure approach to be discussed in
Chapter 4. The only curve that suggests
any increase in threat to other road
users as drivers age is Fig. 2-14, which shows pedestrian fatality crashes per
unit distance of travel. Here the
increase is small, and applies only at ages above about 70; it is quite
overshadowed by the much greater values associated with young drivers of either
sex.
------------------------------------------------------------------------------
Figs 2-11 through 2-14 all about here
------------------------------------------------------------------------------
HOW SERIOUS IS THE OLDER-DRIVER PROBLEM?
The ten figures (Figs 2-5 through 2-14) show relations between driver age and a variety of measures of risk of involvement in severe and fatal traffic crashes for male and female drivers. For all 20 comparisons of values at age 65 to age 20 for the same sex, the value at age 65 was less than (in most cases substantially less than) the value at age 20 (numerical values are given in Evans [1988c]). For example, 65-year-old male drivers were involved in 88 percent fewer single vehicle crashes in which pedestrians died than were 20-year-old male drivers; when normalized to the same distance of driving, the older driver's rate becomes 72 percent less.
When 65-year-old drivers are compared to 40-year-old drivers, the only variable which is greater for the older male drivers is driver fatalities per unit distance of travel; here the rate for 65-year-olds exceeds that for 40-year-olds by 33 percent. For female drivers, three variables are larger at age 65 than at age 40, namely number of drivers killed per licensed driver (5 percent larger), driver fatalities per unit distance of travel (77 percent larger), and driver involvements in severe crashes per unit distance of travel (13 percent larger).
Thus, for some measures 65-year-old drivers are more at risk than 40-year-old drivers, though not more at risk than 20-year-old drivers. In all the cases where the risk at age 65 exceeds that at age 40, the increased risk is borne by the driver; in no case studied did the 65-year-old driver pose a greater threat to pedestrians than did the 40-year-old driver. The source data for driver licenses and distances of travel place all drivers 70 and older in the same category, which precluded any detailed examination at ages beyond the 60's. It would be desirable if organizations tabulating data such as the number of licensed drivers would avoid such broad aggregation when substantial quantities of data are available at older ages. The fatality and population data, which are available in one year increments to age 85, indicate a declining threat to other users (Fig. 2-12) through age 85 for both sexes on a per capita basis. The upward trend in the corresponding graph for male driver fatalities (Fig. 2-6) is largely explained by the greater likelihood of fatality in a crash.
The graphs which best reflect the behavioral aspects of driving, namely, driver involvements in crashes in the same high severity range per unit distance of travel, and crashes in which pedestrians are killed per unit distance of travel (Figs 2-10 and 2-14) show remarkably similar features. Drivers from about age 30 to 60 have the lowest involvement rates. As age decreases below 30, rates increase at an increasing rate. For ages greater than about 60, rates increase somewhat, but much less rapidly than as one approaches the younger ages in the graphs. Male rates are consistently higher than those for females.
Much larger than any
proportionate increase in driver risk with increasing age is the decline in
distance of driving. For example, male
drivers 70 and over drive, on average, 9 300 km/year, compared to 31 000
km/year for 35 to 39 year-old drivers; the corresponding values for female
drivers are 4 300
km/year and 12 600 km/year, respectively.
The problem of aging may thus be more one of reduced mobility than of
reduced safety. As mental and sensory
abilities decline, the dominant response is less driving, especially under
conditions of elevated risk, rather than a net increase in risk from driving;
as people age, the threat they pose to other road users declines.
The above discussion has focused on how various measures depend on average chronological age. Not only do various measures of driver performance decline with age, but variability amongst individuals also increases [Ranney and Pulling 1990], underlying the importance of not judging an individual's driving ability on the basis of chronological age. As individuals age, risk of death increases rapidly, so the fraction of the total risk that is due to motor vehicle crashes declines (Fig. 14-2 and related discussion).
PEDESTRIAN INVOLVEMENTS IN FATAL AND SEVERE CRASHES
Above we examined the age and sex of drivers involved in crashes in which pedestrians were killed. We now examine the age and sex of the pedestrians involved without regard to the characteristics of the involved drivers, using 1981-1985 FARS data [Evans 1987].
Fig. 2-15 shows the distribution of pedestrian fatalities by pedestrian age and sex. In order to make the graph easier to follow, the male data have been joined by lines, and the peak values are identified in the caption. The 1981-1985 FARS data show 596 six-year-old child fatalities (an average of 73 six-year-old boys killed per year and 46 six-year-old girls).
------------------------------------------------------------------------------
Fig. 2-15 about here
------------------------------------------------------------------------------
The same data normalized by population are shown in Fig. 2-16. In Chapter 6 (Fig. 6-5) we comment further on the large systematic difference dependent on sex, showing higher male than female pedestrian fatality rates for all ages, including the first year of life.
------------------------------------------------------------------------------
Fig. 2-16 about here
------------------------------------------------------------------------------
Part of the large increase in pedestrian fatalities per capita with increasing older ages in Fig. 2-16 is due to the greater likelihood that the older person is killed in a crash which a younger one would survive. In order estimate the risk of involvement in a severe crash, as distinct from the outcome, we again use the relationships between risk of death from the same impact and sex and age given in Eqns 2-4 and 2-5. Fig. 2-17 shows the number of pedestrian involvements in crashes in the severity range equal to or greater than that necessary to kill an 80-year-old male pedestrian. Like the driver fatality data, the pedestrian fatality data show peaks at about age 20 for males and females. The increasing involvement in severe pedestrian crashes with increasing age at ages above about 60 is probably reflecting decreasing perceptual and agility skills, and also perhaps increased pedestrian exposure related to driving less.
------------------------------------------------------------------------------
Fig. 2-17 about here
------------------------------------------------------------------------------
OVERINVOLVEMENT OF YOUNG ROAD USERS
A feature common to all ten
figures (Figs 2-5 through 2-14) showing driver rates versus sex and age is the
dramatically higher values for drivers in
their late teens and early twenties, especially male drivers. Similarly, young males pedestrians had
higher involvement rates in severe crashes than persons of either sex at any
other age for which data are available (up to 85 years) (Fig. 2-17). The only rate for which young males did not
have the highest value was pedestrian fatalities per capita (Fig. 2-16), a rate
which increases rapidly with age largely because of increased likelihood of
death from crashes which younger pedestrians would survive.
The overinvolvement of young, and male, road users is one of the largest and most consistently observed phenomena in traffic throughout the world. It is so robust and repeatable that it is almost like a law of nature. Its magnitude suggests that it must involve much more than a mere lack of driving (or road-crossing) experience. This question is of such central importance in traffic safety that various additional aspects of it will be discussed in many of the later chapters.
CONCLUSIONS
From about age 15 to age 45,
the same physical insult is approximately 25% more likely to kill a female than
a male of the same age. For ages less
than about 5, the risk is higher for males than for females. These results were obtained applying double
pair comparison (a method which allows inferences without external measures of
exposure) to 80 000 fatalities coded in the FARS data. Because effects were similar for different
occupants (unbelted car drivers, helmeted motorcycle passengers, etc.), the
results were interpreted to apply to physical insults in general, and not just
to those sustained in traffic crashes.
Applying the same approach to investigate age effects shows that for age
greater than about 20, fatality risk grows at an approximately
uniform rate of (2.3 + 0.2)% per year for males and (2.0 + 0.2)%
per year for females; at age 70 the risk is about three times what it is at 20.
The number of fatalities suffered by a group of road users does not measure the number of crashes in which the group is involved because the outcome (the fatality) depends on a behavioral and a physiological factor -- the involvement in the crash and the probability that the impacts sustained in the crash lead to death. The dependence of the behavioral factor was inferred from the observed number of fatalities using the relationships between probability of death from the same impact and sex and age. Ten measures of driver involvement in crashes (fatalities per capita, involvements in severe crashes per unit distance of travel, involvements in crashes in which pedestrians were killed, etc.) consistently showed much higher rates for drivers in their late teens and early twenties, and higher values for male drivers. Male pedestrians of these same ages were overinvolved in pedestrian crashes, indicating consistently the disproportionate contribution of this group of road users to traffic crashes.
As drivers age they pose ever decreasing threats to other road users, as indicated by the number of pedestrian fatality crashes in which they are involved. One reason why they pose a reduced threat to others is a reduction in driving with increasing age. For every one of a larger number of measures of crash involvement examined, 65-year-old drivers had lower rates than 20-year-old drivers. Even though by some measures 65-year-old drivers had higher rates than 40-year-old drivers, the problem of the aging driver may be more one of reduced mobility than of reduced safety.
REFERENCES (CHAPTER 2)
Baker, S.P. Motor vehicle occupant deaths in young children. Pediatrics 64:860-861; 1979.
Baker, S.P.; O'Neill, B.; Karph, R.S. The injury fact book. Lexington, MA: Lexington Books; 1984.
Bureau of the Census. Estimates of the population of the United States, by age, sex and race: 1980 to 1985. Series P-25, no. 985. Washington, DC: US Department of Commerce, Washington, Government Printing Office; 1987.
Charness, N., editor. Ageing and human performance. New York, NY: John Wiley; 1985.
Evans, L. Double pair comparison -- a new method to determine how occupant characteristics affect fatality risk in traffic crashes. Accident Analysis and Prevention 18:217-227; 1986.
Evans, L. Young driver involvement in severe car crashes. Alcohol, Drugs, and Driving 3(3-4):63-78; 1987.
Evans, L. Risk of fatality from physical trauma versus sex and age. Journal of Trauma 28:368-378; 1988a.
Evans, L. Examination of some possible biases in double pair comparison estimates of safety belt effectiveness. Accident Analysis and Prevention 20:215-218; 1988b.
Evans, L. Older driver involvement in fatal and severe traffic crashes. Journal of Gerontology: Social Sciences 43:S186-S193; 1988c.
Federal Highway Administration. Highway statistics, 1983. Washington, DC: US Department of Transportation, Government Printing Office; 1983.
Haight, F.A. The place of safety research in transportation research. Transportation Research 19A:373-376; 1985.
Kahane, C. J. An evaluation of child passenger safety: the effectiveness and benefits of safety seats. Washington, DC: National Highway Traffic Safety Administration, report DOT HS-806 890; February 1986.
National Highway Traffic Safety Administration. Fatal Accident Reporting System 1988. Document DOT HS 807 507. Washington, DC; December 1989.
National Safety Council. Accident facts. Chicago, IL; 1989 edition (issued annually).
Olson, P.L.; Sivak, M. Perception-response time to unexpected roadway hazards. Human Factors 28:91-96; 1986.
Partyka, S.C. Restraint use and fatality risk for infants and toddlers. Washington, DC: National Highway Traffic Safety Administration; May 1984.
Ranney, T.A.; Pulling, N.H. Relation of individual differences in information-processing ability to driving performance. Human Factors Society, Proceedings of the 33rd Annual Meeting, Denver, CO, p. 965-969; 16-20 October 1989.
Ranney, T.A.; Pulling, N.H. Performance differences on driving and laboratory tasks between drivers of different ages. Paper 890270, presented to the 69th Annual Meeting of the Transportation Research Board, Washington, DC; 7-11 January 1990.
Reff, M.E.; Schneider, E.L., editors. Biological markers of ageing. Washington, DC: US Department of Health and Human Services, NIH publication 82-2221; April 1982.
Sivak, M.; Olson, P.L.; Pastalan, L.A. Effect of driver's age on nighttime legibility of highway signs. Human Factors 23:59-64; 1981.
US Department of Transportation. Survey data tabulations; 1983-1984. Nationwide Personal Transportation Study. Report DOT-P36-85-1. Washington, DC; November 1985.
Verbrugge, L.M. Sex differentials in health. Public Health Reports 97:417-432; 1982.
Waller, J.A. Injury control -- a guide to the causes and prevention of trauma. Lexington, MA: Lexington Books; 1985.
Welford, A.T. Signal, noise, performance, and age. Human Factors 23:97-109; 1981.
Yanik, A.J. What accident data reveal about elderly drivers. SAE paper 851688. Warrendale, PA: Society of Automotive Engineers; 1985.