Background
Older drivers are the fastest growing group of drivers in the United States (US) [
1], both in terms of the number of drivers and the number of annual miles driven. They have a crash rate nearly equal to that of young drivers whose crash rate is the highest among all age groups [
2,
3]. Research over the past several decades has indicated that vision impairment contributes to older drivers’ increased crash risk [
4‐
6]. The vast majority of studies on vision and older driver safety have utilized accident reports from police agencies as the outcome measure for collision involvement. In the Unites States these accident reports are routinely indexed by the state jurisdiction in which they occur, with many states making these reports available to scientists (after appropriate regulatory approval) for the purposes of traffic safety research. While these reports provide a wealth of information about the circumstances of a crash (e.g., driver age, place, weather conditions, vehicles involved), they reveal little to nothing about visual and other mechanisms underlying the occurrence of a crash. An alternative to using accident reports to study risk factors for crash involvement is to conduct naturalistic driving studies [
7,
8]. Naturalistic driving data are generated by participants driving their own vehicles in the course of their everyday life over long observation periods (e.g., one to two years). Their vehicles are unobtrusively equipped with sensors and video cameras, which record vehicle kinematics, global positioning system (GPS) location, presence of near-by objects, driver behavior (e.g., gaze direction, secondary task activity), and the roadway environment. Naturalistic data provide an unprecedented level of objective detail on safety critical events such as crashes including pre-crash information about driver behavior and roadway contextual factors.
It is well established that spatial contrast sensitivity impairment is common among older adults [
9‐
12]. As with most of the vision and driver safety research literature, studies on contrast sensitivity and crash risk have relied on accident reports comparing collision rates for those with contrast sensitivity impairment to those with no or minimal impairment. While some studies report that contrast sensitivity impairment is associated with a recent history of crash involvement [
13,
14] others have reported no association between contrast sensitivity and incident (or future) crash involvement [
15‐
17]. Contrast sensitivity impairment from age-related cataract is associated with a history of higher collision rates [
18] and on-road driving performance problems [
19,
20]. Drivers with contrast sensitivity impairment secondary to Parkinson’s disease experience on-road difficulties [
21‐
24]. Impaired contrast sensitivity regardless of etiology has also been associated with on-road driving problems [
25,
26].
Older driver studies on contrast sensitivity and collision involvement up until now have for the most part assessed contrast sensitivity using a letter chart, specifically the Pelli-Robson chart [
27]. The Pelli-Robson chart measures how much contrast is required to identify letters subtending 2.8° of visual angle; letter size is not varied on the chart, but letter contrast is. The chart is a popular choice in epidemiological and clinical studies because it is brief, easy to administer, and has good reproducibility [
28‐
30]. It also has confirmed construct validity for everyday visual task performance in that reduced contrast sensitivity as measured by the chart is associated with adverse outcomes such as falls [
31], mortality [
32], performance mobility deficits [
33], and slowed reading [
34]. A limitation of the Pelli-Robson chart is that it does not measure a person’s contrast sensitivity as a function of spatial frequency (i.e., target size) [
35].
Prior research using SHRP2 data from older drivers [
36] examined the relationship between contrast sensitivity and crash risk. In SHRP2 contrast sensitivity was assessed by the Optec 6500 P which measures contrast sensitivity for five spatial frequency targets ranging from 1.5 to 18 cycles per degree (cpd) [
37]. Contrast sensitivity results were presented for some but not all spatial frequencies tested, with no explanation provided for why some were omitted. It appears that only statistically significant results were presented.
When a set of thresholds for a range of spatial frequencies is measured as done in SHRP2, a model can be used to fit thresholds to form a spatial contrast sensitivity function (CSF); many CSF models have been put forth [
38,
39]. Some of these models can provide good fits to the raw data with only 4 parameters [
40]. The major advantage of measuring the CSF, rather than contrast sensitivity for a single target size, is that it constitutes a comprehensive summary of visibility for a broad variety of spatial stimuli [
39].
The purpose of this study is to examine the association of photopic and mesopic CSFs with motor vehicle crash involvement by older drivers using naturalistic driving techniques. Photopic vision is mediated by cone photoreceptors only, whereas mesopic vision is mediated by both cone and rod photoreceptors.
Results
Of the 1019 participants in SHRP2 ≥ 60 years old, 89 were missing both photopic or mesopic contrast sensitivity data, resulting in 930 persons available for analysis. When fitting photopic CSF models, an additional 9 were missing photopic data, and 77 had R2 < 0.5 for the model fit, and thus were deleted, resulting in a total of 844 persons with photopic contrast sensitivity parameters. When fitting mesopic CSF models, an additional 7 were missing mesopic data and 69 had R2 < 0.5 for model fit and thus were deleted, resulting in 854 with mesopic contrast sensitivity parameters.
Approximately ¾ of the sample were in their 60s or 70s with the remaining ≥80 years old (Table
1). There were slightly more men than woman. The vast majority of participants (over 95%) were white of non-Hispanic origin. Table
2 shows the mean and standard deviation of contrast sensitivity parameters for the sample. Consistent with the literature [
53], mesopic AULCSF and peak sensitivity were lower than those parameters for photopic contrast sensitivity.
Table 1
Demographic characteristics of the sample (N = 915)
Age, years |
60–69 | 310 | 33.9 |
70–79 | 396 | 43.3 |
80–90 | 200 | 21.9 |
90–99 | 9 | 1.0 |
Gender |
Men | 491 | 53.7 |
Women | 424 | 46.3 |
Race |
White | 871 | 95.2 |
Black | 15 | 1.6 |
Otherb | 22 | 2.4 |
Unknown | 7 | 0.8 |
Ethnicity |
Hispanic or Latino | 12 | 1.3 |
Not Hispanic or Latino | 872 | 95.3 |
Unknown | 31 | 3.4 |
Table 2
Contrast sensitivity parameters for the sample
Photopic contrast sensitivity parameters (N = 844) |
AULCSF a | 1.70 | 0.37 |
Peak log sensitivity | 1.89 | 0.21 |
Mesopic contrast sensitivity parameters (N = 854) |
AULCSF a | 1.25 | 0.37 |
Peak log sensitivity | 1.75 | 0.23 |
Participants drove a total of 7,417,879 miles with an average of 8107 ± 6967 miles driven over 233 ± 183 h of driving time per person. The most common vehicle driven was a passenger car (68.1%), followed by sport utility vehicle or pickup (17.6%), mini-van (5.8%), crossover (1.9%), and unknown (6.6%). Of the 915 participants, 688 had no crashes, 161 had 1 crash, 44 had 2 crashes, 9 had 3 crashes, 5 had 4 crashes, 2 had 5 crashes, 1 had 6 crashes, 3 had 7 crashes, 1 had 8 crashes, and 1 had 13 crashes. This resulted in 354 total crash events among 227 participants. Of the 354 crashes, 307 were deemed at-fault for the participant driving. The majority of crashes (272 of 354, 77%) and at-fault crashes (237 of 307, 77%) occurred during the daytime.
Table
3 lists the age-adjusted associations between photopic and mesopic contrast sensitivity parameters and crash rate. Photopic AULCSF and peak log contrast sensitivity were not associated with crash rate, whether defined as all crashes or at-fault crashes only (all
p > 0.05). Mesopic AULCSF was associated with an increased crash rate when defined in terms of all crashes; drivers in the lowest AULCSF quartile were 36% more likely to incur crashes per mile driven than were those in the upper three quartiles (RR: 1.36, 95% CI: 1.06–1.72). The association between peak mesopic log sensitivity and crash involvement was slightly weaker than that for AULCSF but statistically significant, with drivers in the lowest log sensitivity quartile 28% more likely to be crash involved (RR: 1.28, 95% CI: 1.01–1.63). Both associations were stronger when considering only at-fault crashes; drivers in the lowest AULCSF and peak log sensitivity quartiles were 50 and 38% more likely, respectively, to be crash involved as compared to those in the upper three quartiles (RR: 1.50, 95% CI: 1.16–1.93; RR: 1.38, 95% CI: 1.07–1.78).
Table 3
Age-adjusted associations between impaired contrast sensitivity and rate of motor vehicle crash involvement for all crashes and for at-fault crashes
Photopic contrast sensitivity parameters |
AULCSFa | 0.80 (0.61–1.05) | 0.102 | 0.77 (0.57–1.03) | 0.077 |
Peak log sensitivity | 0.80 (0.61–1.04) | 0.091 | 0.77 (0.58–1.03) | 0.072 |
Mesopic contrast sensitivity parameters |
AULCSFa | 1.36 (1.06–1.72) | 0.015 | 1.50 (1.16–1.93) | 0.002 |
Peak log sensitivity | 1.28 (1.01–1.63) | 0.042 | 1.38 (1.07–1.78) | 0.014 |
Discussion
In a prospective population-based study on older drivers in the US, impaired photopic contrast sensitivity was not associated with an increased rate for incident motor vehicle crash involvement. Our results agree with two previous prospective, population-based studies on older drivers, also performed in the US [
16,
17]. It is the case that other studies have found positive associations between photopic contrast sensitivity deficits and crash involvement but the outcome variable in these studies was a history of crashes, not future crashes [
13,
14]. The limitation with using study designs where the outcome is crashes retrospective to enrollment is that one does not know whether, for example, the photopic contrast sensitivity impairment existed before the crash. These studies taken together provide little support for photopic contrast sensitivity testing at the population level for the purposes of reducing the burden of motor vehicle crashes.
It is useful to consider why photopic contrast sensitivity deficit is unrelated to future crash risk. First, studies have shown that contrast sensitivity impairment is related to self-reported driving difficulty [
54] as well as self-regulation [
55]. Driver’s with photopic contrast sensitivity deficits are more likely to avoid challenging driving situations [
55,
56], reduce driving exposure (i.e., mileage) [
55], and are at greater risk for driving cessation [
57,
58]. Second, one of the most common causes of contrast sensitivity loss is age-related cataract. However, most older adults in the US undergo cataract surgery and intraocular lens implantation after cataracts start hampering their visual daily activities. Cataract surgery reduces motor vehicle crash risk by 50% [
59]. Thus, while contrast sensitivity loss has a negative impact on driving habits, its potential negative impact on crash risk appears to be mitigated by changes in driving habits and the wide-spread availability of cataract surgery.
Although the association between photopic contrast sensitivity and crash involvement was null in our study, mesopic contrast sensitivity impairment was associated with incident crashes and at-fault crashes. For example, drivers with impaired mesopic sensitivity as defined by AULCSF were 50% more likely to incur an at-fault crash as compared to those who were unimpaired. That mesopic contrast sensitivity impairment elevates collision risk in older drivers is a novel finding for the literature. Previous studies suggest that mesopic vision deficits contribute to driver safety and performance problems in studies of samples of drivers that have a wide range of ages, but not specifically focused on older drivers. Lachenmayr et al. [
60] reported that drivers with worse mesopic vision were more likely to be involved in night-time collisions, however 2/3 of those studied were not older adults, but were ≤ 60 years old. A study on bus and truck drivers found that those with “reduced twilight vision” were more frequently involved in collisions [
61]. Older drivers with mesopic vision acuity impairment exhibited worse night-time driving performance on a closed driving course [
62]. Black et al. found that correcting astigmatism in young drivers with toric contact lenses improved on-road driving performance which was also linked to better mesopic vision [
63].
A strength of this study is the use of incident crashes as an outcome measure based on naturalistic driving recordings. The study was population-based and included a large sample of older drivers. This is the first study to utilize a contrast sensitivity function model and the parameters it generates to examine associations between older driver safety and contrast sensitivity. A strength of this approach is that it comprehensively considers the entire spatial envelope of visibility, not each spatial target’s piecemeal role in collision risk (the approach used in a previous report using SHRP2 data [
36]). In that report [
36], the association between collision involvement and contrast sensitivity was evaluated separately at each of 4 spatial frequencies under photopic and mesopic viewing conditions; that is, they did not fit a CSF model to the data before analysis. A limitation of our study is an insufficient sample size and number of crash events in order to study contrast sensitivity and crash involvement stratified by day- versus night-time crashes.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.