CHAPTER 4
Conclusions and Suggested Research

Conclusions

This project examined the prevalence of EOR time as well as the associated CNC risk of looking away from the forward roadway. Both prevalence and risk were assessed for different driving phases and for younger and older teen drivers, as well as those with self-reported ADHD diagnoses. Mixed-effects logistic regression models were employed to calculate ORs, which are accepted estimates of CNC risk.

Overall, the prevalence results indicate that teen drivers may get more comfortable looking away from the forward roadway over time versus during the first 6 months of licensure. However, the ORs are higher for EOR during the first 6 months, suggesting that teen drivers may be learning and getting more judicious when they do look away from the forward roadway during the second 6 months. Additionally, the risk of looking away increases with longer durations of EOR, which is expected; however, the increase in risk by duration is highest for single longest EOR. More detailed conclusions are discussed in the following paragraphs for driving phases, age groups, and teen drivers with ADHD.

Driving Phases

The prevalence prior to SCEs and crash risk of total EOR for all durations was high in the first 6 months and declined in the second 6 months. In the first 6 months, the significance of the OR indicates that crash risk significantly increases with the duration of total EOR. In the second 6 months, the duration of 2 seconds served as a critical threshold at which the crash risk increased significantly for total EOR. A single longest EOR greater than or equal to 2 seconds is associated with nearly a 5-fold increase in crash risk (OR = 4.96), while the total EOR greater than or equal to 2 seconds also increases crash risk but to a lesser extent (OR = 2.91), with both being statistically significant. This result is to be expected as there are important fundamental differences in the two eye-glance metrics. Total EOR time is a calculation of all glances away from the forward roadway in a 6-second window; thus, drivers may have brief glances to the forward roadway and gain some, albeit insufficient, information about the forward roadway environment. However, during the single longest EOR glance, the driver will not gain any information about the forward roadway, as this metric indicates a single glance to a location that does not include the forward roadway.

As compared to the first 6 months, the higher prevalence of the single longest EOR prior to both SCEs and baselines in the second 6 months indicates that teen drivers were more likely to look away from the driving task for at least 1 second as they gained experience over time. Again, this may suggest that they were becoming more judicious about when they chose to look away from the forward roadway for one longer glance.

Age Groups: 16 to 17 Versus 18 to 19

Compared to drivers aged 16 to 17 years, teen drivers aged 18 to 19 years had fewer single prolonged glances greater than or equal to 2 seconds, a higher prevalence of total EOR during baselines, and more accumulated total glances. However, the difference between the two age groups was not statistically significant. Despite this, the crash risk in the 16 to 17 age group was still higher than in the 18 to 19 age group, suggesting that 18- to 19-year-old teens may have learned and become more judicious about when to look away than the 16- to 17-year-old teen drivers.

The significance of ORs for both age groups and both types of EOR indicates that crash risk and the likelihood of SCEs increase with the duration of EOR. For the single longest EOR, crash risk in the 16 to 17 age group surged from 2 fold to nearly 11 fold when a single prolonged glance exceeded 2 seconds. In the 18 to 19 age group, a single prolonged glance exceeding 2 seconds increased crash risk 2.5 fold, but when it exceeded 3 seconds, the risk escalated dramatically to nearly 25 fold, emphasizing the extreme danger of prolonged inattention away from the forward roadway. A higher OR for the single longest EOR, as compared to the total EOR, especially for durations exceeding 2 seconds in the 16 to 17 age group and 3 seconds in the 18 to 19 age group, indicates that a single prolonged glance away from the road is a stronger predictor of crash risk. As previously discussed, this is likely due to the difference between getting no information about the forward roadway (single longest glance off road) and getting some information about the forward roadway (total time EOR).

ADHD Groups

In the 16- to 19-year-old non-ADHD groups, the ORs increased with the duration of EOR, consistent with those observed in the previous analyses. A single prolonged glance exceeding 2 seconds increased crash risk 4 fold, but when it exceeded 3 seconds, the risk escalated dramatically to nearly 27 fold. As compared to teens without ADHD, teens with ADHD had a higher prevalence of total EOR and single longest EOR during baselines, as well as lower ORs for various durations. Although this difference in the prevalence of EOR was not significant, teen drivers with ADHD tend to look away from the road more frequently and may be at an elevated risk for missed hazards. All ORs were not significant, indicating no measurable impacts of EOR durations on crash risk for the ADHD group. There are several limitations to the analysis of ADHD teens. There was a relatively small sample of ADHD teen drivers (N = 22). The ORs that were reported also had fairly wide confidence intervals, which reflect the lack of statistical power in these analyses. Finally, only five participants had a Barkley score above 7, indicating a clinically significant level of inattention or hyperactivity-impulsivity. Possibly, the mixed severity levels of ADHD among all participants may also have contributed to insignificant ORs.

General Conclusions

Teen drivers aged 18 to 19 years had a lower CNC risk than teen drivers aged 16 to 17 years. Additionally, teens aged 16 to 17 years had lower CNC risk during Phase 2 as compared to Phase 1. Specifically, teen drivers had a slightly higher prevalence of total EOR and a lower prevalence of single prolonged glances in Phase 1 prior to SCEs than in Phase 2. This pattern indicates that newly licensed teens are more prone to frequent, brief distractions; as they gain driving experience, they may glance away less often but for slightly longer durations. The higher crash risk associated with a single prolonged glance, as compared to the total EOR, resulted in lower prevalence, regardless of the age and driving experience. The longer the time of EOR, the higher the crash risk. Therefore, the comparison of EOR across driving phases and age groups aligns with findings from previous studies (Simons-Morton et al., 2014; Curry et al., 2017; Curry et al., 2019). Specifically, teen drivers

have a higher CNC risk and are more likely to look away from the road in the first 6 months than in the second 6 months after licensure. Interestingly, teens aged 18 to 19 years are more likely to have more single longest EOR greater than or equal to 1 second and total EOR greater than or equal to 1, 2, and 3 seconds as compared to teens aged 16 to 17 years.

Most of the research using the simulation method has indicated that teens with ADHD have difficulty maintaining their attention and are more likely to take long glances away from the road (Kingery et al., 2015; Curry et al., 2022). The results of this study showed that teen drivers with ADHD had lower prevalence of EOR and lower ORs across four duration thresholds as compared to those without ADHD. This raises a challenging question: Do teens with ADHD take fewer glances away from the road and have lower crash risk than teens without ADHD? Obviously, that is not the case. Instead, it is possible that other factors, such as compensatory strategies, metrics used in this study, or limitations in the data sample, contributed to these unexpected results.

One possible reason for the unexpected results is that the two EOR metrics may not fully capture the inattention traits of teen drivers with ADHD. Deficits in fixation control and saccade suppression (Munoz et al., 2003) might communicate differences in glance frequency rather than glance duration from teens without ADHD. In simulated driving, Michaelis et al. (2012) found that adult drivers with ADHD did not exhibit more saccadic eye movements and shorter fixation duration as compared to those without ADHD. Similarly, Groom et al. (2015) reported no significant differences in fixation duration between two such groups. The prevalence of total EOR, single longest EOR, and fixation duration may not essentially reflect the ADHD-related deficits.

Previous research has utilized variability of lane position, speed, and brake time to assess ADHD-related impairments. For example, Narad et al. (2013) documented that teens with ADHD have significant risk for adverse driving performance outcomes. However, Stavrinos et al. (2015) did not observe the same driving performance impairments of ADHD. That is, driving performance metrics have produced conflicting estimates of ADHD crash risk. When the number of long glances away from the roadway (> 2 s), lane position variability, and speed were considered as adverse driving outcomes for teen drivers with ADHD, an increase in driving experience only reduced the crash risk related to the number of long off-road glances (> 2 s) but not to any other adverse driving. Kingery et al. (2015) employed the number of extended glances away from the roadway exceeding 2 seconds to measure visual inattention that was implicated as a mechanism for ADHD- and texting-related deficits in lane position variability. Both teen drivers with ADHD and those texting during driving would divert their eyes from the roadway for a greater percentage of time, and with a greater number of extended glances, than those without ADHD under either no-distraction conditions or hands-free phone conversation conditions.

Another possible reason for the unexpected results is the smaller sample size of the ADHD group as compared to the non-ADHD group in the SHRP 2 NDS. A total of 32 teen drivers with ADHD were identified based on medical records. However, only 22 participants with ADHD had valid EOR data. The random variations in small sample sizes can heavily impact the prevalence and OR estimates, increasing the likelihood of wide confidence intervals and reduced statistical power. This can lead to challenges in detecting significant differences between ADHD and non-ADHD drivers, even if meaningful patterns exist.

Suggested Research

Future research on the impact of EOR and teen driversʼ CNC risk could delve further into the contextual aspects of teens and their choices for looking away from the forward roadway. This could include greater understanding of the roadway environment (e.g., high or low traffic density, proximity to intersections versus straight road segments, type of roadway, etc.), as well as types of secondary tasks that require greater EOR duration. Additionally, glance metrics such

as the proportion of time that teens are looking forward versus looking away, and the conditions under which such proportions change, may also add insight into those scenarios that increase risk for teen drivers.

Future research should add more ADHD data to better investigate the relationship between crash risk and ADHD. Expanding the dataset could provide deeper insights into the factors contributing to the elevated crash risk among teen drivers with ADHD and inform targeted interventions. Larger, more balanced samples could help improve the reliability of the prevalence and risk estimates, providing a clear understanding of the relationship between ADHD and visual attention patterns during driving. The difference in the proportion of total EOR between the SHRP 2 NDS and the SPDS NDS introduced new opportunities for comparative analysis across datasets. These differences highlight potential variations in participant characteristics or driving environments, which may influence the measurement of EOR and its relationship to crash risk. Future studies should examine the underlying factors ensuring that comparisons across datasets account for methodological and contextual differences.

Curry et al. (2022) confirmed that teen drivers with ADHD have a higher crash risk than their peers without ADHD. In their recent exploratory study, inattention—one of the core symptoms of ADHD—was identified as a driver-related proximate factor contributing to the heightened crash risk for teens with ADHD. The difference between inattention and other proximate factors (e.g., driving errors or risky driving behavior) are nuanced. It has been suggested that further research is needed to elucidate the nuances and identify specific proximate factors to better understand why drivers with ADHD are more likely to crash than their peers without ADHD.

The findings of previous studies raise the possibility of differences in the crash risk of various glance durations between teen drivers with and without ADHD. The results of this study indicate that teens with ADHD did not exhibit higher crash risk associated with various eye-glance durations. Future research should explore alternative indicators beyond total EOR duration and single longest EOR, such as the number of glances, to better capture inattention-related deficits. Teen drivers with ADHD will frequently and briefly scan all targets in the field of view, including in and outside vehicle, potentially increasing the crash risk. Whether the frequency of off-road glances mediates the relationship between driving experience and crash risks is an important topic of interest.

Limitations

There are several limitations to these analyses, some due to the constraints of NDSs in general, and some in terms of the analyses themselves.

First, the small sample size of participants with ADHD in the SHRP 2 NDS may not provide a stable estimation, making it challenging to analyze ADHD-related deficits in relation to crash risk and inattention. At the same time, the SHRP 2 NDS dataset had very little data collected during teensʼ first 6 months of driving. This limitation made analyses difficult when using the SHRP 2 NDS to assess behavior by early driving experience. On the other hand, the SPDS NDS had sufficient data to evenly parse out two groups of driving phases, with 77 participants in the first 6 months and 68 in the second 6 months. This balanced distribution allowed for more structured comparison of driving behaviors over time.

Total EOR and single longest EOR did not appear to effectively measure the ADHD-related crash risk. The direct measurement of crash risk using ORs for different EOR durations did not show a significant difference between the ADHD and non-ADHD groups. Thus, the extent to which differences in glance behavior impact crash risk among teens with ADHD remains unknown. Mixed-effect logistic regression models yield critical new insights into the relationship

between the crash risk due to inattention and key factors, such as driving experience, age, and ADHD. These analyses assess ORs across four or five thresholds of glance duration. The use of predefined thresholds (e.g., ≥ 1 s, ≥ 2 s, ≥ 3 s, ≥ 4 s, and ≥ 5 s) for categorizing EOR duration may oversimplify the relationship between inattention and crash risk; however, additional data collection will be needed to better identify and understand how the risk of EOR modulates across different roadway contexts.

This study is one of the first to fully investigate prevalence of teen driversʼ EOR behaviors and the associated CNC risk of these behaviors. Teens are at a heightened CNC risk from EOR during their first 6 months of driving independently as compared to later independent driving. Parents and driverʼs education instructors could improve teen driving outcomes by reminding/enforcing teen driver attention to the forward roadway, especially during their early driving experience.