Why have fatality rates among older drivers declined? The relative contributions of changes in survivability and crash involvement

Accident Analysis and Prevention 83 (2015) 67–73

Contents lists available at ScienceDirect

Accident Analysis and Prevention journal homepage: www.elsevier.com/locate/aap

Why have fatality rates among older drivers declined? The relative contributions of changes in survivability and crash involvement Jessica B. Cicchino* Insurance Institute for Highway Safety, 1005 North Glebe Road, Arlington, VA 22201, USA

A R T I C L E I N F O

A B S T R A C T

Article history: Received 27 February 2015 Received in revised form 5 June 2015 Accepted 30 June 2015 Available online xxx

Objectives: This study examined the trend in fatality rates per vehicle miles traveled (VMT) among older drivers relative to middle-aged drivers and quantified the contributions of changes in crash involvement and survivability to this trend. Methods: Using U.S. national databases, changes in driver deaths per crash involvement (marker of death risk when involved in a crash) and crash involvements per VMT (marker of crash risk) from 1995–1998 to 2005–2008 among older drivers aged 70 and over relative to changes among middle-aged drivers aged 35–54 were computed. The contributions of these components to the relative changes in older drivers’ fatality rates per VMT were calculated using the decomposition methodology. Results: Fatality rates per VMT declined more among older drivers than among middle-aged drivers over the study period. Relative to middle-aged drivers, drivers aged 75 and older experienced large declines in crash risk and modest declines in death risk. Relative declines in crash risk accounted for 68–74% of the larger decline in fatalities per VMT among drivers aged 75 and older compared with middle-aged drivers. Drivers aged 70–74 experienced modest relative declines in crash risk and death risk. Declines in death risk among drivers aged 75 and older relative to middle-aged drivers were much larger in side-impact crashes; improvements in crash survivability accounted for nearly half of the relative decline in fatality rates in these crashes. Relative survivability did not change significantly in frontal impacts. Higher death risk was more important than higher crash risk in explaining older drivers’ elevated fatality rates per VMT relative to middle-aged drivers during 1995–1998, and the contribution of heightened death risk was even greater during 2005–2008. Conclusions: Many factors may have reduced crash involvements among drivers 75 and older, including changes in travel patterns, health, and roadway design. In side impacts, side airbags and reduced passenger vehicle incompatibility may have improved survivability for older drivers. Because excess fragility now makes an even larger contribution to older drivers’ elevated fatality rates, future countermeasures that improve survivability can likely provide large benefits. ã 2015 Elsevier Ltd. All rights reserved.

Keywords: Aged Aged, 80 and over Accidents Traffic Trends Older drivers

1. Introduction Driver involvements in fatal crashes per vehicle miles traveled (VMT) begin to rise at age 70 and are strikingly higher than those of middle-aged drivers by age 80 (Cicchino and McCartt, 2014). This is due both to older drivers’ higher likelihood of being involved in a police-reported crash per VMT and to their greater odds of dying when involved in a crash (Cicchino and McCartt, 2014; Li et al., 2003). Age-related declines in cognitive, visual, and physical functioning can heighten an older driver’s crash risk (Anstey et al., 2005; Owsley et al., 1998; Sims et al., 2001; Stutts et al., 1998).

* Fax: +1 703 247 1678. E-mail address: [email protected] (J.B. Cicchino). http://dx.doi.org/10.1016/j.aap.2015.06.012 0001-4575/ ã 2015 Elsevier Ltd. All rights reserved.

Greater susceptibility to injury, or fragility, with age increases the risk that a crash-involved older driver with suffer a fatal injury (Kahane, 2013; Li et al., 2003). Fatalities among drivers aged 70 and older in the United States peaked in 1997 and have since declined, despite large increases in the older population, their licensing rate, and their annual VMT (Cicchino and McCartt, 2014). Similarly, fatal crash involvements per VMT among drivers 70 and older relative to rates among middle-aged drivers fell steeply from 1995 to 2006 to 2008, as did fatal crash involvement rates per licensed driver during 1997– 2012. There are multiple factors that could have contributed to these downward trends. Some potential factors, such as improvements in vehicle crashworthiness or occupant protection that were especially beneficial for older vehicle occupants, would primarily impact crash survivability; others, such as changes in license

68

J.B. Cicchino / Accident Analysis and Prevention 83 (2015) 67–73

renewal policies or in driving patterns among older drivers, would primarily impact crash involvement; and others, such as better health, could potentially impact both. Cicchino and McCartt (2014) reported that both the odds of dying in a crash and involvement rates per licensed driver in crashes of various severities declined more among older than middle-aged drivers during 1997–2008, but because different methods were used to examine these components, it is unclear which made a larger contribution. This would be helpful to know so that the possible mechanisms underlying the change could be better understood. Such understanding could be used to develop countermeasures to further reduce fatalities among older drivers. The decomposition methodology, which separates fatality rates per unit traveled into the component parts that contribute to them, has been used to quantify the relative contributions of crash involvement, crash survivability, and VMT to older drivers’ elevated fatality and serious injury rates compared with middleaged drivers (Dellinger et al., 2002; Li et al., 2003; Meuleners et al., 2006). Research using the decomposition methodology that examined crashes that occurred in the mid-1990s found that increased fragility was more important than increased crash risk in explaining older drivers’ elevated fatality rates relative to those among middle-aged drivers (Li et al., 2003). The current study extends previous research by using this method over a second dimension—time—to estimate the relative contributions of changes in crash survivability and crash involvement to the recent decline in fatality rates among older drivers. Because crash involvement rates and survivability differ for older drivers in crashes resulting in side and frontal impacts (Kahane, 2013; Li et al., 2003), changes were also examined for crashes involving these vehicle impacts. 2. Methods Decomposition was used to disaggregate changes in older driver death rates per VMT from 1995–1998 to 2005–2008 into changes in two components: driver deaths per police-reported crash involvement (marker of death risk when involved in a crash), and crash involvements per VMT (marker of crash risk). Analyses did not extend beyond 2008 because more recent VMT data are not available. This excluded several years of the U.S. recession when relative declines in older driver fatal crash involvement rates slowed. Changes among older drivers aged 70 and older were assessed relative to changes among middle-aged drivers aged 35– 54. Passenger vehicle driver fatalities during 1995–1998 and 2005– 2008 were obtained from the National Highway Traffic Safety Administration’s (NHTSA) Fatality Analysis Reporting System (FARS), a census of motor vehicle crashes on public roads that resulted in a fatality within 30 days. Crash involvements of passenger vehicle drivers during these years were obtained from NHTSA’s General Estimates System (GES), a nationally representative probability sample of police-reported crashes that can be weighted to produce national estimates. Standard errors for GES data were obtained from tables provided by the National Highway Traffic Safety Administration (2014). Data on passenger vehicle miles traveled were obtained from the household travel surveys that are conducted periodically by the Federal Highway Administration (FHWA), with the most recent administrations providing estimates of 1 year of travel during April 1995–March 1996, April 2001–2002, and January 2008–December 2008 (Federal Highway Administration, 2009). Information is collected on travel from all members of a sample of U.S. households, and can be weighted to estimate total VMT for 1 year by driver age. For simplicity, the 1995–1996 and 2001–2002 surveys are treated as if they were conducted in 1995 and 2001,

respectively, which were the calendar years that corresponded with the majority of the months of travel estimated by those surveys. Annual national VMT counts from FHWA’s Highway Performance Monitoring System (HPMS) (Federal Highway Administration, 2014) were used to interpolate VMT by driver age from the household travel surveys to the study years as described below. The national VMT estimates from the HPMS are based largely on data reported by the states but do not account for VMT by driver age. Data from the national household surveys and the HPMS were used to estimate total VMT by driver age during the study periods of 1995–1998 and 2005–2008. Using national VMT estimates from the HPMS, the changes in VMT from 1995 to 1996, 1996 to 1997, and 1997 to 1998 were calculated. The percentage by which these changes differed from the average annual change in VMT during 1995–2001 was determined. The average annual changes in VMT from the household travel surveys by driver age during 1995– 2001 were also calculated by dividing the total change for each age group by 6. Average annual changes by driver age from the household travel surveys were adjusted for 1996, 1997, and 1998 by applying the percentage by which the change in VMT from the HPMS for that year differed from the average annual change during 1995–2001. For example, the increase in national VMT from HPMS from 1995 to 1996 was 1% smaller than the average annual increase from HPMS during 1995–2001. Thus, for each age group, the average annual increase in VMT for 1996 from the household travel surveys was adjusted downwards by 1%. To estimate VMT by driver age for 1996, the adjusted average annual change in VMT by driver age from the household travel survey for that year was added to the estimated mileage for that age group from the 1995 survey. Adjusted average annual changes for 1997 and 1998 by driver age were added to the previous year’s VMT estimate to determine VMT for those years. The same process was used to estimate total VMT by driver age during 2005–2008, using VMT estimates from the HPMS during 2001–2008 and from the household travel surveys during 2001 and 2008. Standard errors for the VMT data were computed separately for each of the three household travel surveys for each year of VMT estimates they were used to interpolate. As explained above, estimates of VMT from the 1995 and 2001 surveys were used to interpolate VMT values during 1995–1998, and estimates from the 2001 and 2008 surveys were used to interpolate VMT values during 2005–2008. Standard errors from the 2001 and 2008 surveys were computed using the replicate weights provided by FHWA, and standard errors from the 1995 survey were computed using a SAS macro (Schmoyer, n.d.). The higher standard error resulting from the weights from either survey involved in interpolation was used when computing 95% confidence intervals. Rate ratios (RRs) of driver deaths per 100 million VMT, driver deaths per 1000 crash involvements, and crash involvements per 100 million VMT during 1995–1998 and 2005–2008 were computed for older drivers age groups relative to middle-aged drivers. Ratios of the RR for 2005–2008 to the RR for 1995–1998 were then computed to represent changes in rates over time among older drivers relative to the change among middle-aged drivers. For instance, the RR ratio for the relative change in driver fatalities per VMT was computed using Eq. (1), where D = driver fatalities, T = VMT, o = older drivers, m = middle-aged drivers, 1 = time 1 (1995–1998), and 2 = time 2 (2005–2008). RR ratios less than 1 with 95% confidence intervals that exclude 1.0 indicated that the RR for the older driver group decreased from 1995–1998 to 2005– 2008 relative to the middle-aged driver group significantly at the 0.05 level (See Appendix A for formulae for 95% confidence intervals).

J.B. Cicchino / Accident Analysis and Prevention 83 (2015) 67–73




Do2 =To2 Dm2 =Tm2

Driver fatalities per VMT RR ratio ¼


Do1 =T o1 Dm1 =T m1

69





(1)

The RR ratio of driver fatalities per VMT from 2005–2008 to 1995–1998 can also be expressed as the product of the RR ratio of driver deaths per crash involvement from 2005–2008 to 1995– 1998 and the RR ratio of crash involvements per VMT from 2005– 2008 to 1995–2008, as illustrated in Eq. (2), where C = driver crash involvements.

3, 4, 8, 9, or 10 o’clock position, and frontal-impact crashes were defined as crashes with an initial impact at the 11, 12, or 1 o’clock position. Impact points are characterized differently in GES. Sideimpact crashes in GES were defined as crashes in which the subject driver’s vehicle sustained an initial impact to the left or right side, and frontal-impact crashes included impacts to the front, front left corner, and front right corner. Imputed data were used in GES for the initial point of impact variable. Few (<2%) values were imputed. Age was not imputed because an excessive number of drivers were coded as age 90 using the imputed age in 1997, which suggests errors in the data. Age was



 Do2 =Co2 D =C  m2 m2 Do1 =C o1 Driver fatalities per VMT RR ratio ¼ Dm1 =C m1 RR ratio 1 : Driver deaths per crash





 C o2 =T o2 C =T  m2 m2  C o1 =T o1 C m1 =T m1 RR ratio 2 : Driver crash involvements

involvement ðdeath riskÞ

The contribution decomposes the effects of relative changes in driver deaths per crash (death risk, RR ratio 1) and in crash involvements per VMT (crash risk, RR ratio 2) to relative changes in the driver fatality rate per VMT, using Eq. (3). jlnðRR ratioi Þj Contributioni ¼ P2  100% i¼1 jlnðRR ratioi Þj

(3)

A contribution of more than 50% indicates that the change of the component, either death risk or crash risk, had the greater role in the decline in older drivers’ fatality rate per VMT relative to the change among middle-aged drivers. RR ratios of driver fatality rates per VMT, driver death rates per crash involvement, and crash involvements per VMT, as well as the contributions of the latter two rates to the driver fatality rate per VMT, were also computed for crashes involving side and frontal impacts. Total VMT by time period and age group were used as denominators in rates per VMT in these calculations; more specific information on mileage exposure was not available. In FARS, side-impact crashes were defined as crashes with an initial impact to the older or middle-aged driver’s vehicle at the 2,

(2)

per VMT ðcrash riskÞ

unknown among 6% of drivers in GES during the earlier and later study periods. Eq. (3) was also used to compute the contributions of death risk and crash risk to older drivers’ excessive death rate per VMT relative to that of middle-aged drivers during 1995–1998 and 2005–2008 separately, with rate ratios for deaths per crash and crashes per VMT for a single time period used in the equation instead of RR ratios. Here, contributions of more than 50% indicate that a component has the greatest influence on older drivers’ death rate per VMT, compared with the rate among middle-aged drivers. 3. Results Driver deaths per VMT, driver deaths per crash, and crash involvements per VMT increased with driver age during both time periods (Table 1,Fig. 1). The numbers of driver deaths and driver involvements in police-reported crashes decreased between the study periods for all age groups, whereas the amount of driving increased for all age groups. The increases in the amount of driving were much larger among drivers aged 75–79 (40%) and 80 and older (86%) than among drivers aged 70–74 (7%) or middle-aged drivers (15%). The mileage-based rates of driver deaths and driver

Table 1 Driver deaths per 100 million vehicle-miles of travel (VMT), driver deaths per 1000 crash involvements (death risk), and driver crash involvements per 100 million VMT (crash risk), by driver age and period, 1995–1998 and 2005–2008. Driver age and Passenger vehicle period driver deaths

Drivers in policereported crashes

VMT (in millions)a

Driver deaths per 100 million VMT

Driver deaths per 1000 crash involvements (death risk)

35–54 1995–1998 2005–2008

24,871 23,991

14,040,289 12,533,828

3,647,171 4,223,362

0.68 0.57

1.77 1.91

385 297

70–74 1995–1998 2005–2008

3792 2921

968,269 761,463

214,528 230,160

1.77 1.27

3.92 3.84

451 331

75–79 1995–1998 2005–2008

3622 2896

733,852 612,625

113,402 158,283

3.19 1.83

4.94 4.73

647 387

80+ 1995–1998 2005–2008

5049 4766

643,780 639,620

61,113 113,892

8.26 4.18

7.84 7.45

1053 562

a

Interpolated between household travel survey study years and adjusted for annual changes in national VMT.

Driver crash involvements per 100 million VMT (crash risk)

70

J.B. Cicchino / Accident Analysis and Prevention 83 (2015) 67–73

9 8

(a) 1995-98

7

2005-08

6 5 4 3 2 1 0

9 8

35-54

70-74

75-79

80+

70-74

75-79

80+

(b) 1995-98

7

2005-08

6 5 4 3 2 1 0

1200 1000

35-54

(c)

crash involvements declined between the study periods for all driver age groups, especially the two oldest age groups. In contrast, the rate of driver deaths per 1000 crash involvements declined for the older driver groups but increased for middle-aged drivers. Older drivers’ elevated fatality rates per VMT during 1995– 1998 were attributed mainly to their heightened death risk, and this continued during 2005–2008 (Table 2). Greater death risk relative to middle-aged drivers contributed to an even larger proportion of older drivers’ elevated fatality rates per VMT in this latter period, accounting for 86%, 77%, and 68% of the increased rates among drivers aged 70–74, 75–79, and 80 and older, respectively. In both study periods the contribution of death risk to older drivers’ elevated fatality rates per VMT declined with age. The declines from 1995–1998 to 2005–2008 in the rate of driver fatalities per VMT among all older age groups were larger than the decline among middle-aged drivers (Table 3). Crash risk among drivers aged 75–79 and 80 and older declined relative to middleaged drivers by 22% and 31%, respectively. Declines in death risk relative to middle-aged drivers were not as large as the relative declines in crash risk among drivers aged 75 and older, falling 11% among drivers aged 75–79 and 12% among drivers aged 80 and older; the latter decline was significant. Among drivers aged 75– 79 and 80 and older, declines in crash risk accounted for 68% and 74%, respectively, of the relative decline in fatality rates per VMT. Relative declines in death risk (9%) and crash risk (5%) were more modest among drivers aged 70–74 (Table 3). Neither reached statistical significance. For these drivers, declines in death risk accounted for two-thirds of the decline in their fatality rate per VMT. 3.1. Vehicle impact point

1995-98 2005-08

800 600 400 200 0

35-54

70-74

75-79

80+

Fig. 1. Driver deaths per 100 million VMT (a), driver deaths per 1000 crash involvements (b), and crash involvements per 100 million VMT (c), by driver age, 1995–1998 and 2005–2008.

Relative declines in driver fatality rates per VMT were decomposed for crashes involving frontal and side impacts to the vehicles of drivers aged 75 and older, who experienced the largest declines among older drivers in overall driver fatality rates per VMT. Deaths in these crashes comprised more than 90% of the fatalities among drivers 75 and older during 1995–1998 and 2005– 2008, and these crashes accounted for more than 80% of their crashes (Table 4). Rate ratios of driver deaths per VMT, driver deaths per crash, and crash involvements per VMT relative to middle-aged drivers were higher in side-impact than in frontalimpact crashes. During 1995–1998, drivers aged 75 and older were nearly 13 times as likely as middle-aged drivers to die in a sideimpact collision per VMT, and during 2005–2008 the relative rate

Table 2 Rate ratio (RR) of driver deaths per VMT relative to middle-aged drivers during 1995–1998 and 2005–2008 by driver age, and contributions of death risk and crash risk to excess fatality rates. Driver age and period

Driver deaths per 100 million VMT

Driver deaths per 1000 crash involvements (death risk) Driver crash involvements per 100 million VMT (crash risk)

RR relative to middle-aged drivers

RR relative to middle-aged drivers

Contribution% (95% CI)

RR relative to middle-aged drivers

Contribution% (95% CI)

70–74 1995–1998 2005–2008

2.59 2.23

2.21 2.00

83 (76–91) 86 (79–94)

1.17 1.11

17 (15–19) 14 (12–15)

75–79 1995–1998 2005–2008

4.68 3.22

2.79 2.47

66 (61–73) 77 (71–84)

1.68 1.30

34 (30–38) 23 (21–25)

80+ 1995–1998 2005–2008

12.12 7.37

4.43 3.89

60 (55–65) 68 (62–74)

2.71 1.89

40 (35–46) 32 (29–35)

J.B. Cicchino / Accident Analysis and Prevention 83 (2015) 67–73

71

Table 3 Ratio of the rate ratio (RR) of driver deaths per VMT for older drivers compared with middle-aged drivers during 2005–2008, relative to that during 1995–1998, and contributions of death risk and crash risk. Driver age

70–74 75–79 80+

Driver deaths per 100 million VMT

Driver deaths per 1000 crash involvements (death risk) Driver crash involvements per 100 million VMT (crash risk)

Ratio of RR during 2005–2008 relative to 1995–1998 (95% CI)

Ratio of RR during 2005–2008 relative to 1995–1998 (95% CI)

Contribution% (95% CI)

Ratio of RR during 2005–2008 relative to Contribution% 1995–1998 (95% CI)) (95% CI)

0.86 (0.81–0.92) 0.69 (0.64–0.74) 0.61 (0.55–0.67)

0.91 (0.80–1.02) 0.89 (0.78–1.00) 0.88 (0.78–1.00)

66 (58–75) 32 (28–36) 26 (23–29)

0.95 (0.83–1.09) 0.78 (0.67–0.90) 0.69 (0.59–0.81)

34 (30–39) 68 (59–78) 74 (63–87)

Table 4 Fatalities and crash involvements among drivers aged 75 and older in side-impact and frontal-impact crashes during 1995–1998 and 2005–2008, and rate ratios (RR) of driver deaths per 100 million VMT, driver deaths per 1000 crash involvements (death risk), and driver crash involvements per 100 million VMT (crash risk) relative to middle-aged drivers. Impact point and year

Older driver fatalities

Older driver crash involvements

RR relative to middle-aged drivers Driver deaths per 100 million VMT

Driver deaths per 1000 crash involvements (death risk)

Driver crash involvements per 100 million VMT (crash risk)

Side 1995–1998 2005–2008

3630 2886

576,087 446,249

12.94 8.19

4.83 3.95

2.68 2.07

Frontal 1995–1998 2005–2008

4412 4169

605,931 587,656

6.33 4.53

2.92 2.77

2.17 1.64

declined to about 8 times as high. Drivers aged 75 and older were about 6 times as likely as middle-aged drivers to die in a frontalimpact collision per VMT during 1995–1998, and about 5 times as likely during 2005–2008. Drivers aged 75 and older experienced similarly large declines of 23% and 25% in side- and frontal-impact crash risk, respectively, relative to middle-aged drivers (Table 5). However, relative declines in death risk were not equivalent. In side impacts older drivers’ relative death risk per crash fell significantly by 18%, whereas in frontal impacts their relative death risk fell nonsignificantly by 5%. Declines in death risk accounted for 44% of the drop in older drivers’ mileage-based death rate in side impacts, compared with only 15% in frontal impacts. 4. Discussion Excess death risk made a larger contribution than crash risk to older drivers’ elevated fatality rates per VMT relative to middleaged drivers during 1995–1998 and 2005–2008. This finding is consistent with prior research (Li et al., 2003) and likely reflects older vehicle occupants’ greater vulnerability to injury and lesser ability to recover from injuries sustained from a given crash impact compared with younger occupants (Evans, 2001; Kahane, 2013). Among drivers aged 70–74, crash risk was not greatly elevated relative to middle-aged drivers during the earlier or later portions of the study period. Relative improvements in crash survivability

accounted for the majority of the larger decrease in fatality rates per VMT among this age group compared with middle-aged drivers Drivers aged 75–79 and 80 and older were 2–3 times as likely to crash per VMT as middle-aged drivers during 1995–1998, and this gap narrowed considerably by 2005–2008. Crash risk for these drivers relative to middle-aged drivers decreased by 22–31%, which accounted for more than two-thirds of their decrease in fatalities per VMT relative to middle-aged drivers. In part because of this substantial decline in relative crash risk among drivers 75 and older, excess death risk made an even greater contribution to older drivers’ elevated fatality rates per VMT relative to middleaged drivers during 2005–2008 than during 1995–1998. Many factors potentially contributed to the large drop in crash risk among drivers aged 75 and older. Low-mileage drivers tend to have higher crash rates per VMT, possibly because they tend to drive a larger proportion of miles on local roads with more conflict points or because they have chosen to drive less due to impairments that increase their crash risk (Langford et al., 2006). Average annual VMT per driver increased from 1995– 2006 to 2008 by 60% for drivers aged 75–59 and by 51% for drivers aged 80 and older (Cicchino and McCartt, 2014), which suggests that the percentage of low-mileage drivers may have declined. Decreases in functional disabilities and visual impairments among the older population may have contributed to this trend and also reduced crash involvement propensity (Hung et al., 2011; Lee et al., 2009).

Table 5 Ratio of the rate ratio (RR) of driver deaths per VMT for drivers aged 75 and older compared with middle-aged drivers during 2005–2008, relative to that during 1995–1998, and contributions of death risk and crash risk, in side-impact and frontal-impact crashes. Impact point

Side Frontal

Driver deaths per 100 million VMT

Driver deaths per 1000 crash involvements (death risk)

Driver crash involvements per 100 million VMT (crash risk)

Ratio of RR relative to middle-aged drivers (95% CI)

Ratio of RR relative to middle-aged drivers (95% CI)

Contribution% (95% CI)

Ratio of relative to middle-aged drivers (95% CI)

Contribution% (95% CI)

0.63 (0.60–0.67) 0.72 (0.67–0.76)

0.82 (0.72–0.93) 0.95 (0.84–1.08)

44 (39–50) 15 (14–17)

0.77 (0.67–0.89) 0.75 (0.65–0.87)

56 (49–65) 85 (74–97)

72

J.B. Cicchino / Accident Analysis and Prevention 83 (2015) 67–73

Roadway improvements that were beneficial to older drivers could have also played a role. FHWA began issuing guidelines in 1998 on roadway design practices that may be especially helpful for older drivers (Staplin et al., 1998), including increasing conspicuity of signals, signs, and roadway markings and modifying road geometry at intersections to improve visibility. By 2007, 24 states reported incorporating at least half of the FHWA recommendations into their state design guides (Government Accountability Office, 2007). Cities in Michigan reported greater reductions in intersection crashes among older than middle-aged drivers after implementing some of these improvements (Bagdade, 2004). Drivers aged 75 and older made larger relative gains in survivability in side impacts compared with frontal impacts. Some vehicle improvements since the mid-to-late 1990s have been especially beneficial for protecting older occupants in side impacts. Side airbags have been more effective in preventing fatalities among front-seat occupants aged 70–96 than among those aged 13–49, while frontal airbags have been equally effective for both age groups (Kahane, 2013). Relative to middle-aged drivers, a larger percentage of older drivers travel in cars than in SUVs, pickup trucks, and minivans (Cicchino and McCartt, 2014), which are larger and heavier than cars, on average, and thus provide better protection in a crash. During the earlier study period in the current research, occupants of cars were at particularly high risk of injury and death when struck in the side by heavier vehicles, but this gap narrowed over time with improved vehicle design (Teoh and Nolan, 2012). Limitations of the current research should be noted. A vehicle occupant’s death risk when involved in a crash depends on the severity and type of impact experienced in the crash, the strength of occupant protection features in mitigating the effects of the crash impact, the occupant’s vulnerability to injury from a given impact, and the occupant’s ability to recover from injuries. It is unknown to what degree these factors influenced the changes in death risk seen in this study. Death risk increased slightly for middle-aged drivers from the earlier to the later study period, despite improvements in occupant protection during this time that are associated with lower fatality risk, such as the increased number of vehicles with frontal and side airbags and with structures that have improved crashworthiness (Highway Loss Data Institute, 2012, 2013). This suggests that factors additional to changes in occupant protection affected shifts in death risk during this time. Trends in crash severity were not examined because candidate variables for examining crash severity in GES, such as speed limit, are likely not reliable indicators of crash severity (Farmer, 2003). GES does not include minor crashes not reported to police. Older drivers’ crashes may be more often reported because they are more likely to be injured in a crash of a given severity. Reporting criteria may have changed in areas sampled by GES over time. Because GES and the household travel surveys are samples, estimates are subject to sampling error. Estimates for both are based on fewer older than middle-aged drivers. The household travel surveys do not include information on VMT by road type, and changes in travel on different road types (e.g., through intersections, on highways) would affect crash involvement rates in the crash types that were examined. Passenger vehicle VMT by driver age was estimated for years in which the household travel surveys were not conducted, with adjustments made based on changes in total national VMT from HPMS. HPMS VMT counts include mileage from vehicles other than passenger vehicles and from commercial travel, and annual changes in personal passenger vehicle travel may have differed. Household travel survey collects travel data from a nationally representative sample of households, whereas the HPMS travel estimates are based largely on traffic counts submitted by states and do not reflect a rigorous sampling strategy. However, results

were very similar when analyses were conducted using VMT interpolated from the national household survey data without adjusting for changes in total national VMT from HPMS. Analyses did not include years beyond 2008 because VMT data by driver age from the household travel surveys were not available. The FHWA plans to begin collecting data for the next household travel survey in late 2015 (Federal Highway Administration, 2015); when these data are available, analyses can be extended to more recent years. In sum, large gains have been made in reducing crashes among older drivers. Crashes could be further reduced by directing countermeasures at the types of errors older drivers most often make, such as looking at but not seeing oncoming vehicles at intersections (Cicchino and McCartt, 2015). Changing intersections so that left turns across traffic are eliminated by constructing roundabouts or diverging diamond interchanges would largely eliminate opportunities to make this error. Vehicle-to-vehicle and vehicle-to-infrastructure communications that alert drivers to oncoming traffic at intersections could also be helpful. However, as the gap in crash risk between older drivers and middle-aged drivers narrows, it is likely that even greater gains can be realized by increasing older occupants’ crash survivability. NHTSA has proposed creating a rating system to help older drivers choose vehicles that would potentially be safer for them (Office of the Federal Register, 2013), but generally, vehicles that are the safest for older drivers are also the safest for drivers of other ages (Insurance Institute for Highway Safety, 2013). Improvements to vehicles that increase their crashworthiness and better access to emergency medical services can improve survivability for all vehicle occupants. Acknowledgements The author thanks Chuck Farmer for assistance with statistical methods and Anne McCartt for helpful comments that improved this paper. This work was supported by the Insurance Institute for Highway Safety. Appendix A. The equations for calculating 95% confidence intervals for the contributions of fragility and crash involvement to older drivers’ excess fatality rate per VMT during a single period were as follows: 95% confidence intervals for contribution of deaths per crash during single period (Table 2) = ( sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi) VarðC o Þ VarðC m Þ þ Exp lnðcontributionÞ  1:96 C 2o C 2m where C = crashes, o = older drivers, and m = middle-aged drivers. 95% confidence intervals for contribution of crashes per VMT during single period (Table 2) = ( sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi) VarðC o Þ VarðC m Þ VarðT o Þ VarðT m Þ þ þ þ Exp lnðcontributionÞ  1:96 T 2o T 2m C 2o C 2m where T = VMT. The equations for calculating 95% confidence intervals for changes in fatality rates per VMT, fragility, and crash involvement rates per VMT among older drivers relative to middle-aged drivers from 1995–1998 to 2005–2008 were as follows: 95% confidence intervals for RR ratios of deaths per VMT of 2005–2008 to 1995–1998 (Tables 3 and 5) = ( sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi) VarðT o1 Þ VarðT m1 Þ VarðT o2 Þ VarðT m2 Þ þ þ þ Exp lnðRR RatioÞ  1:96 T 2o1 T2m1 T2o2 T2m2 where 1 = time 1 (1995–1998) and 2 = time 2 (2005–2008).

J.B. Cicchino / Accident Analysis and Prevention 83 (2015) 67–73

95% confidence intervals for RR ratios of deaths per crash of 2005–2008 to 1995–1998 (Tables 3 and 5) =

(

73

Hung, W.W., Ross, J.S., Boockvar, K.S., Siu, A.L., 2011. Recent trends in chronic disease, impairment and disability among older adults in the Unites States. BMC Geriatr.

sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi) VarðC o1 Þ VarðC m1 Þ VarðC o2 Þ VarðC m2 Þ þ þ þ C 2o1 C 2m1 C 2o2 C 2m2

Exp lnðRR RatioÞ  1:96

95% confidence intervals for RR ratios of crashes per VMT of 2005–2008 to 1995–1998 (Tables 3 and 5) =

(

11 doi:http://dx.doi.org/10.1186/1471-2318-11-47. Insurance Institute for Highway Safety, 2013. Comment to the National Highway

sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi) VarðC o1 Þ VarðC m1 Þ VarðC o2 Þ VarðC m2 Þ VarðT o1 Þ VarðT m1 Þ VarðT o2 Þ VarðT m2 Þ þ þ þ þ þ þ þ T 2o1 T 2m1 T 2o2 T 2m2 C 2o1 C 2m1 C 2o2 C 2m2

Exp lnðRR RatioÞ  1:96

95% confidence intervals for contributions of changes in deaths per crash and crash involvements per VMT used the same equations as above, substituting the contribution for the RR ratio. References Anstey, K.J., Wood, J., Lord, S., Walker, J.G., 2005. Cognitive, sensory and physical factors enabling driving safety in older adults. Clin. Psychol. Rev. 25, 45–65. doi: http://dx.doi.org/10.1016/j.cpr.2004.07.008. Bagdade, J.S., 2004. Low cost intersection improvements reduce crashes for senior drivers. ITE Annual Meeting and Exhibit Compendium of Technical Papers (CDROM). Institute of Transportation Engineers, Washington, DC. Cicchino, J.B., McCartt, A.T., 2014. Trends in older driver crash involvement rates and fragility in the United States: an update. Accid. Anal. Prev. 72, 44–54. doi:http:// dx.doi.org/10.1016/j.aap.2014.06.011. Cicchino, J.B., McCartt, A.T., 2015. Critical older driver errors in a national sample of serious U.S. crashes. Accid. Anal. Prev. 80, 211–219. doi:http://dx.doi.org/ 10.1016/j.aap.2015.04.015. Dellinger, A.M., Langlois, J.A., Li, G., 2002. Fatal crashes among older drivers: decomposition of rates into contributing factors. Am. J. Epidemiol. 155, 234–241. doi:http://dx.doi.org/10.1093/aje/155.3.234. Evans, L., 2001. Age and fatality risk from similar severity impacts. J. Traffic Med. 29, 10–19. Farmer, C.M., 2003. Reliability of police-reported information for determining crash and injury severity. Traffic Inj. Prev. 4, 38–44. doi:http://dx.doi.org/10.1080/ 15389580309855. Federal Highway Administration, 2009. National Household Travel Survey and National Personal Transportation Survey. U.S. Department of Transportation, Washington, DC http://nhts.ornl.gov/download.shtml (accessed 27.02.15.). Federal Highway Administration, 2014. Highway Statistics Series. U.S. Department of Transportation, Washington, DC http://www.fhwa.dot.gov/ policyinformation/statistics.cfm (accessed 27.02.15.). Federal Highway Administration, 2015. FHWA NHTS Newsletter: National Household Travel Survey, January 2015. U.S. Department of Transportation, Washington, DC http://nhts.ornl.gov/TRBNewsletter.pdf (accessed 26.05.15.). Government Accountability Office, 2007. Older Driver Safety: Knowledge Sharing Should Help States Prepare for Increase in Older Driver Population (Report no. GAO-07-413). Government Accountability Office, Washington, DC. Highway Loss Data Institute, 2012. Predicted Availability of Safety Features on Registered Vehicles. Loss Bulletin 28(26). Highway Loss Data Institute, Arlington, VA. Highway Loss Data Institute, 2013. IIHS Crashworthiness Evaluation Programs and the U.S. Vehicle Fleet. Loss Bulletin 30(7). Highway Loss Data Institute, Arlington, VA.

Traffic Safety Administration Concerning the New Car Assessment Program. U.S. Department of Transportation, Washington, DC Docket No. NHTSA-2012-0180, July 3, 2013. Kahane, C.J., 2013. Injury Vulnerability and Effectiveness of Occupant Protection Technologies for Older Occupants and Women. Report No. DOT-HS-811-766. National Highway Traffic Safety Administration, Washington, DC. Langford, J., Methorst, R., Hakamies-Blomqvist, L., 2006. Older drivers do not have a high crash risk—a replication of low mileage bias. Accid. Anal. Prev. 38, 574–578. doi:http://dx.doi.org/10.1016/j.aap.2005.12.002. Lee, D.J., Arheart, K.L., Lam, B.L., Zheng, D., Christ, S.L., McCollister, K.E., Davila, E.P., Caban-Martinez, A.J., 2009. Trends in reported visual impairment in United States adults. Opthalmic Epidemiol. 16, 42–49. doi:http://dx.doi.org/10.1080/ 08658080262443. Li, G., Braver, E.R., Chen, L., 2003. Fragility versus crash involvement as determinants of high death rates per vehicle-mile of travel among older drivers. Accid. Anal. Prev. 227–235. doi:http://dx.doi.org/10.1016/S0001-4575(01)107-5. Meuleners, L.B., Harding, A., Lee, A.H., Legge, M., 2006. Fragility and crash overrepresentation among older drivers in Western Australia. Accid. Anal. Prev. 38, 1000–1010. doi:http://dx.doi.org/10.1016/j.aap.2006.04.005. National Highway Traffic Safety Administration, 2014. National Automotive Sampling System (NASS) General Estimates System (GES) Analytical User’s Manual 1988–2012 (Report no. DOT HS-811-853). U.S. Department of Transportation, Washington, DC. Office of the Federal Register, 2013. National Highway Traffic Safety Administration, Docket No. NHTSA-2012-0180, New Car Assessment Program (NCAP). Federal Register, 78(66. National Archives and Records Administration, Washington, DC, pp. 20597–20604. Owsley, C., Ball, K., McGwin, G., Sloane, M.E., Roenker, D.I., White, M.F., Overley, E.T., 1998. Visual processing impairment and risk of motor vehicle crash among older adults. J. Am. Med. Assoc. 279, 1083–1088. doi:http://dx.doi.org/10.1001/ jama.279.14.1083. Schmoyer, R., n.d. Standard Error Calculations for NPTS Estimates, U.S. Department of Transportation, Washington, DC. http://nhts.ornl.gov/1995/download/se.pdf (accessed 27.02.15.). Sims, R.V., McGwin Jr., G., Pulley, L., Roseman, J.M., 2001. Mobility impairments in crash-involved older drivers. J. Aging Health 13, 430–438. doi:http://dx.doi.org/ 10.1177/089826430101300306. Staplin, L., Lococo, K., Byington, S., 1998. Older Driver Highway Design Handbook (Report no. FHWA-RD-97-135). Federal Highway Administration, Washington, DC. Stutts, J.C., Stewart, J.R., Martell, C.M., 1998. Cognitive test performance and crash risk in an older driver population. Accid. Anal. Prev. 30, 337–346. doi:http://dx. doi.org/10.1016/S0001-4575(97)108-5. Teoh, E.R., Nolan, J.M., 2012. Is passenger vehicle incompatibility still a problem? Traffic Inj. Prev. 13, 585–591. doi:http://dx.doi.org/10.1080/ 15389588.2012.6762.