CHAPTER 3: OBJECTIVE 2, WHAT ARE THE ENVIRONMENTAL CONDITIONS ASSOCIATED WITH DRIVER CHOICE OF ENGAGEMENT IN SECONDARY TASKS OR DRIVING WHILE DROWSY? WHAT ARE THE RELATIVE RISKS OF A CRASH OR NEAR-CRASH WHEN ENGAGING IN DRIVING INATTENTION WHILE ENCOUNTERING THESE ENVIRONMENTAL CONDITIONS? (continued)

Table 3.19. Odds ratio point estimates and 95 percent confidence intervals for the interaction of drowsiness and traffic density.

Type of Traffic Density	Odds Ratio	Lower CL	Upper CL
LOS A: Free Flow	4.67	3.02	7.21
LOS B: Flow with Some Restrictions	4.81	2.70	8.58
LOS C: Stable Flow – Maneuverability and Speed are more Restricted	3.63	2.01	6.54
LOS D: Flow is Unstable – Vehicles are unable to pass with temporary stoppages	4.29	1.88	9.80
LOS E: Unstable Flow- Temporary restrictions, substantially slow drivers	3.71	1.93	7.13

Note: numbers in bold font indicate that the point estimate is significantly different than normal, baseline driving (or an odds ratio of 1.0).

Table 3.20. Odds ratio point estimates and 95 percent confidence intervals for the interaction of complex secondary tasks and traffic density.

Type of Traffic Density	Odds Ratio	Lower CL	Upper CL
LOS A: Free Flow	4.67	2.32	9.38
LOS B: Flow with Some Restrictions	3.67	1.65	8.19
LOS C: Stable Flow – Maneuverability and Speed are more Restricted	3.80	1.68	8.58
LOS D: Flow is Unstable – Vehicles are unable to pass with temporary stoppages	1.75	0.61	5.01
LOS E: Unstable Flow- Temporary restrictions, substantially slow drivers	2.45	1.01	5.93

Note: numbers in bold font indicate that the point estimate is significantly different than normal, baseline driving (or an odds ratio of 1.0).

Table 3.21. Odds ratio point estimates and 95 percent confidence intervals for the interaction of moderate secondary task and traffic density.

Type of Traffic Density	Odds Ratio	Lower CL	Upper CL
LOS A: Free Flow	0.95	0.63	1.45
LOS B: Flow with Some Restrictions	0.69	0.39	1.23
LOS C: Stable Flow – Maneuverability and Speed are more Restricted	0.69	0.38	1.26
LOS D: Flow is Unstable – Vehicles are unable to pass with temporary stoppages	0.31	0.13	0.76
LOS E: Unstable Flow- Temporary restrictions, substantially slow drivers	1.18	0.59	2.34

Note: numbers in bold font indicate that the point estimate is significantly different than normal, baseline driving (or an odds ratio of 1.0).

Surface Condition

The surface condition of roadways has been identified as a frequent contributing factor for crashes and near-crashes. Reductionists used the video and driving performance sensors to assess the status of the roadway surfaces. This analysis was conducted to determine whether drivers engaged in inattentive driving on roads with poor surface conditions. Table 3.22 shows the frequency of the drowsiness and secondary-task-related events and baseline epochs for all six surface condition types. Nearly all of the events and epochs occurred on dry pavement.

Table 3.22. The frequency of drowsiness- and secondary-task-related epochs that occurred at each roadway surface condition level.

Surface Condition	Frequency of Drowsiness-Related Crash and Near-Crash Events	Frequency of Secondary-Task-Related Crash and Near-Crash Events	Frequency of Drowsiness-Related Baseline Epochs	Frequency of Secondary-Task -Related Baseline Epochs
Dry	98	197	666	3681
Wet	13	29	83	395
Icy	1	1	0	3
Snowy	0	0	6	35
Muddy	0	0	0	0
Other	0	0	0	1
Total	112	227	755	4115

Figure 3.6 shows the percentages of drowsiness-related, secondary-task-related, and total baseline epochs that occurred for each type of surface condition. Nearly 90 percent of all drowsiness-related, secondary-task-related, and total baseline epochs occurred on dry pavement, while very low percentages occurred on icy, snowy, and muddy roads. Nearly identical patterns were observed for percent of drowsiness-related and total number of baseline epochs, as well as for secondary-task-related and total number of baseline epochs. This indicates that drivers did not choose to engage in secondary tasks or drive drowsy as a function of the surface condition of the roadway.

Figure 3.6. Percentage of secondary-task-, drowsiness-related and total baseline epochs for all surface conditions.

Odds ratio calculations were conducted to determine whether the near-crash/crash risks associated with driving drowsy or while engaging in complex or moderate secondary tasks were different as a function of poor surface conditions. Table 3.23 presents the odds ratios calculated for driving drowsy on dry, wet, and icy surface conditions. (Odds ratios were not calculated for the other surface conditions because there were either no baseline epochs or no crash or near-crash events observed for these conditions.) Driving while drowsy on either dry or wet roadways increased near-crash/crash risk by at least three times over that of driving alert on a dry or wet roadway.

The odds ratios for engaging in complex secondary tasks on dry roadways increased near-crash/crash risk by four times over that of driving alert on dry roadways (Table 3.24). The relative near-crash/crash risk of engaging in complex secondary tasks on wet roadways was neither significantly different from 1.0 nor significantly different than driving alert on a wet roadway. This result is also not intuitive, but may be due in part to slower speeds and increased headway distances commonly occurring on rainy roadways.

A similar pattern was found for engaging in moderate secondary tasks, which was found to not be as risky as driving drowsy or while engaging in complex secondary tasks (Table 3.25). Dry and wet roadways were also not significantly riskier than one another, suggesting that the interaction found for the complex secondary task and surface condition is unique to complex-secondary-task engagement.

Table 3.23. Odds ratio point estimates and 95 percent confidence intervals for the interaction of drowsiness and surface condition.

Type of Surface Condition	Odds Ratio	Lower CL	Upper CL
Dry	4.52	3.39	6.03
Wet	3.17	2.03	4.95
Icy	N/A	N/A	N/A

Note: numbers in bold font indicate that the point estimate is significantly different than normal, baseline driving (or an odds ratio of 1.0).

Table 3.24. Odds ratio point estimates and 95 percent confidence intervals for the interaction of complex secondary tasks and surface condition.

Type of Surface Condition	Odds Ratio	Lower CL	Upper CL
Dry	4.44	2.88	6.84
Wet	1.03	0.58	1.80
Icy	N/A	N/A	N/A

Note: numbers in bold font indicate that the point estimate is significantly different than normal, baseline driving (or an odds ratio of 1.0).

Table 3.25. Odds ratio point estimates and 95 percent confidence intervals for the interaction of moderate secondary tasks and surface condition.

Type of Surface Condition	Odds Ratio	Lower CL	Upper CL
Dry	0.85	0.65	1.12
Wet	0.73	0.47	1.15
Icy	N/A	N/A	N/A

Note: numbers in bold font indicate that the point estimate is significantly different than normal, baseline driving (or an odds ratio of 1.0).

ROADWAY INFRASTRUCTURE

Traffic Control

The type of traffic control device that a driver needed to heed either 5 seconds prior to or during the course of the crash or near-crash was recorded by trained data reductionists for the events. If a driver needed to heed a traffic control device during the 6-second baseline segment, the reductionist also marked it accordingly. Otherwise, the reductionists recorded No Traffic Control.

Table 3.26 presents the frequency of drowsiness- and secondary-task-related events and baseline epochs where the driver was heeding a particular traffic-control device. Most of the events and epochs were marked as No Traffic Control.

Table 3.26. The frequency of secondary-task-related crash and near-crash events and baseline epochs that were recorded for each type of traffic-control device.

Type of Traffic Control Device	Frequency of Drowsiness-Related Crash and Near-Crash Events	Frequency of Secondary Task-Related Crash and Near-Crash Events	Frequency of Drowsiness-Related Baseline Epochs	Frequency of Secondary-Task-Related Baseline Epochs
Traffic Signal	13	42	40	614
Stop Sign	2	5	3	73
Traffic Lanes Marked	2	4	28	273
Yield Sign	0	0	2	18
Slow or Warning Sign	0	0	2	7
No Passing Sign	0	0	0	1
One-way road	0	0	0	8
Officer or Watchman	0	0	0	3
No Traffic Control	91	169	676	3,609
Other	3	3	4	15
Total	108	223	755	4,114

Note: inattention is defined as only those events where drivers were involved in secondary tasks or were severely drowsy.

The comparisons between the percent of drowsiness-related, secondary-task-related, and total number of baseline epochs for each type of traffic-control device are shown in Figure 3.7. The percentages are very similar across the board, which indicates that drivers are not choosing to engage in secondary tasks or drive while drowsy differently when encountering any of these traffic control devices. This is not to say that drivers were not engaging in secondary tasks while safely sitting at a stop sign or traffic light. This type of analysis could not be performed because the vehicle needed to be moving during the 6 seconds of the epoch for that segment to qualify as a baseline epoch (as discussed in Chapter 1: Introduction and Method).

Figure 3.7. Percentage of secondary-task-related, drowsiness-related, and total number of baseline epochs for each type of traffic control device.

Odds ratios were calculated to determine whether engaging in complex or moderate secondary tasks or driving while drowsy while encountering any of these traffic control devices increased an individual’s near-crash/crash risk (Tables 3.27 through 3.29). The odds ratio calculations for drowsiness suggest that drowsiness, by itself, increases an individual’s risk of being involved in a crash or near-crash by at least 2.7 times over that of an alert driver encountering the same traffic-control device (Table 3.27). None of the traffic-control devices were significantly more risky in the presence of drowsiness than any other traffic-control device.

The odds ratios for complex-secondary-task engagement were similar. Engaging in complex secondary tasks in the presence of a traffic signal, stop sign, or no traffic-control device increased near-crash/crash risk by at least three times over that of an alert driver at a similar traffic-control device (Table 3.28). Stop signs or traffic signals were not significantly riskier than no traffic-control devices. Odds ratios for other traffic-control devices were not available due to low statistical power.

The odds ratios for moderate secondary task engagement were not significantly different from 1.0 except for traffic signal (Table 3.29). The odds ratio for traffic signals actually showed a protective effect, suggesting either that the traffic signal was perhaps able to redirect drivers’ attention to the forward roadway or that the presence of a traffic signal was highly correlated with increased traffic, which redirected drivers’ attention to the forward roadway. Overall, engaging in moderate secondary tasks is not as risky as driving drowsy or engaging in complex secondary tasks in the presence of any of these traffic-control devices.

Table 3.27. Odds ratio point estimates and 95 percent confidence intervals for the interaction of drowsiness and each type of traffic-control device.

Type of Traffic-Control Device	Odds Ratio	Lower CL	Upper CL
Traffic Signal	2.71	1.90	3.85
Stop Sign	5.55	2.71	11.36
Traffic Lanes Marked	5.57	2.43	12.78
No Traffic Control	4.83	3.60	6.48

Note: numbers in bold font indicate that the point estimate is significantly different than normal, baseline driving (or an odds ratio of 1.0).

Table 3.28. Odds ratio point estimates and 95 percent confidence intervals for the interaction of complex secondary tasks and each type of traffic-control device.

Type of Traffic-Control Device	Odds Ratio	Lower CL	Upper CL
Traffic Signal	3.14	2.15	4.58
Stop Sign	3.27	1.38	7.75
No Traffic Control	4.02	2.47	6.54

Note: numbers in bold font indicate that the point estimate is significantly different than normal, baseline driving (or an odds ratio of 1.0).

Table 3.29. Odds ratio point estimates and 95 percent confidence intervals for the interaction of moderate secondary tasks and each type of traffic-control device.

Type of Traffic-Control Device	Odds Ratio	Lower CL	Upper CL
Traffic Signal	0.41	0.28	0.59
Stop Sign	0.73	0.34	1.56
Traffic Lanes Marked	2.29	0.98	5.31
No Traffic Control	0.92	0.70	1.22

Note: numbers in bold font indicate that the point estimate is significantly different than normal, baseline driving (or an odds ratio of 1.0).

Relation to Junction

The relation to junction variable was also adapted from the GES Crash Database to refer to whether the driver was in close proximity to a roadway junction. If the onset of a crash or near-crash occurred in or near an intersection, merge ramp, or interchange, the event was recorded as such; otherwise it was recorded as a non-junction. Likewise, if the vehicle passed through an intersection, interchange, or entered a merge ramp during the 6-second segment of the baseline epochs, then the appropriate relation to junction variable was recorded. Otherwise, non-junction was recorded for that baseline epoch. The different types of junctions used by data reductionists are presented in Table 3.30 along with the frequency of secondary-task- and drowsiness-related events and baseline epochs. Note that most events and epochs were not near roadway junctions (i.e., they were “non-junction”).

Table 3.30. The frequency of drowsiness- and secondary-task-related events and epochs that were recorded for each type of relation to junction.

Type of Relation to Junction	Frequency of Drowsiness-Related Crash and Near-Crash Events	Frequency of Secondary-Task-Related Crash and Near-Crash Events	Frequency of Drowsiness-Related Baseline Epochs	Frequency of Secondary-Task-Related Baseline Epochs
Intersection	17	42	30	257
Intersection-Related	11	22	28	232
Entrance/Exit Ramp	7	11	15	65
Parking Lot	0	5	4	112
Driveway/Alley Access	0	3	2	15
Interchange	1	2	1	10
Rail Grade Crossing	0	0	0	0
Other	0	0	1	12
Non-Junction	75	140	674	3,412
Total	111	226	755	4,115

Note: inattention is defined as only those events where drivers were involved in secondary tasks or were severely drowsy.

Figure 3.8 presents the percentages of drowsiness-related, inattention-related, and total number of baseline epochs occurring at each of the junction types. Note that non-junction accounted for 84 percent of the secondary-task-related baseline epochs as well as of the total baseline epochs. There were very small differences between the percentages of secondary-task-related and total number of baseline epochs, suggesting that there are only small differences between the percentages of time spent engaging in secondary tasks whereas encountering these junctions and how often drivers encounter these types of junctions. There were slight differences in the percentage of drowsiness-related epochs and total epochs, suggesting that a higher percentage of drowsiness-related epochs occurred at non-junctions than at or near intersections. This may suggest that drivers may be more relaxed (under-stimulated) and may succumb to drowsiness effects more often while navigating through less-demanding environments.

Figure 3.8. Percentage of secondary-task-related, drowsiness-related, and total number of baseline epochs for each relation to junction.

To determine whether any of these types of junctions present higher near-crash/crash risks for inattentive drivers, the odds ratios for each were calculated (Tables 3.31 through 3.33). The results for the drowsiness-related odds ratios indicate that near-crash/crash risk increased by at least three times for drivers who were navigating intersections, entrance ramps, and interchanges than for those drivers who were alert at similar junctions (Table 3.31). Also, driving while drowsy in general (i.e., non-junction) increases a driver’s near-crash/crash risk by as much as five times over that of an alert driver encountering similar roadway junctions.

Engaging in complex secondary tasks while in a parking lot or near an intersection increased near-crash/crash risk over that of an alert driver at the junction type (Table 3.32). Somewhat surprisingly, the odds ratio for an intersection did not demonstrate an increased near-crash/crash risk. Drivers may be more careful or even avoid engaging in complex tasks during intersections as these are visually and cognitively demanding environments. The odds ratio for engaging in complex secondary tasks in a parking lot was very high, with an increased near-crash/crash risk of nine times over that of an alert driver in a parking lot. This is somewhat higher than was expected, however, there is a wide confidence interval surrounding this point estimate.

The odds ratios for engaging in moderate secondary tasks showed a similar pattern to complex secondary tasks, in that the odds ratio for intersection was lower than for intersection-related or parking lot (Table 3.33). While the pattern is similar, generally the odds ratios for moderate secondary tasks are not significantly different from 1.0, with the exception of intersection. This suggests that engaging in moderate secondary tasks is not as risky as engaging in complex secondary tasks or driving while drowsy in the presence of these types of roadway junctions.

Table 3.31. Odds ratio point estimates and 95 percent confidence intervals for the interaction of drowsiness and each type of relation to junction.

Type of Relation to Junction	Odds Ratio	Lower CL	Upper CL
Intersection	3.48	2.17	5.59
Intersection-Related	6.82	4.10	11.35
Entrance/Exit Ramp	3.21	1.81	5.71
Interchange	5.86	2.39	14.35
Non-Junction	5.02	3.65	6.90

Note: numbers in bold font indicate that the point estimate is significantly different than normal, baseline driving (or an odds ratio of 1.0).

Table 3.32. Odds ratio point estimates and 95 percent confidence intervals for the interaction of complex secondary tasks and each type of relation to junction.

Type of Relation to Junction	Odds Ratio	Lower CL	Upper CL
Intersection	1.59	0.86	2.97
Intersection-Related	3.32	1.73	6.38
Parking Lot	9.11	3.76	22.07

Note: numbers in bold font indicate that the point estimate is significantly different than normal, baseline driving (or an odds ratio of 1.0).

Table 3.33. Odds ratio point estimates and 95 percent confidence intervals for the interaction of moderate secondary tasks and each type of relation to junction.

Type of Relation to Junction	Odds Ratio	Lower CL	Upper CL
Intersection	0.50	0.31	0.81
Intersection-Related	0.63	0.37	1.44
Entrance/Exit Ramp	1.12	0.61	2.05
Parking Lot	0.65	0.29	1.44
Driveway/Alley Access	2.00	0.64	6.28
Interchange	2.57	0.89	7.46
Non-Junction	0.95	0.70	1.30

Note: numbers in bold font indicate that the point estimate is significantly different than normal, baseline driving (or an odds ratio of 1.0).

SUMMARY

Two primary research questions were addressed in this chapter:

• Do drivers choose to engage in secondary tasks or drive drowsy during more dangerous or adverse environmental conditions?
  
  
• Are any of these environmental conditions riskier than others for inattentive drivers?

Both of these questions were addressed for eight different environmental conditions: ambient lighting, weather, road type, roadway alignment, traffic density, surface condition, traffic-control device, and relation to junction. The results for the first question indicate that far fewer drowsiness-related baseline epochs were observed during the daylight hours than drowsiness-related crashes and near-crashes. Secondly, a greater percentage of drowsiness-related baseline epochs were identified during darkness than drowsiness-related crashes and near-crashes. Drowsiness was also seen to slightly increase in the absence of high roadway or traffic demand. A higher percentage of drowsiness-related baseline epochs were found during free-flow traffic densities, on divided roadways, and areas free of roadway junctions.

The results for the second question were more varied. Each of the eight environmental conditions resulted in odds ratios greater than 1.0 for both drowsiness and engaging in complex secondary tasks. Engaging in moderate secondary tasks rarely resulted in odds ratios significantly greater than 1.0, indicating that these behaviors may not be as risky as driving drowsy or driving while engaging in complex secondary tasks.

In Chapter 2, Objective 1, the odds ratio for risk of driving while drowsy was four to six times that of normal, baseline driving, engaging in complex secondary task was three times, and engaging in moderate secondary tasks was two times that of an alert driver. In this chapter, these total odds ratios decreased when comparing across environmental conditions. While a decrease is to be expected when narrowing the focus of the analysis, it should also be noted all three types of tasks are still riskier than attentive driving.

The baseline dataset also provided some interesting results. For example, drivers are operating their vehicles during the daytime, on dry pavement, and on straight, non-junction roadways a majority of the time. While nighttime driving, adverse weather conditions, intersections, and other difficult roadway geometries increase individual near-crash/crash risk, it is important to note that many crashes and near-crashes occur in the absence of these adverse conditions.

While many of these results are of interest to human factors researchers, roadway designers, and urban planners, it is important to remember that these data were collected only in a metropolitan, urban driving environment (Northern Virginia/Washington, DC, metropolitan area). The results are only generalizable to other urban/metropolitan driving environments and not to the United States driving population in general.

It is important to note that the 20,000 baseline epochs used in these analyses and calculations of relative near-crash/crash risk were not selected based upon any of the above environmental variables. These epochs were selected at random and these environmental conditions were not used in the sampling procedure. Some degree of caution is suggested in the interpretation of these relative near-crash/crash risks given that the baseline epochs were not selected to specifically assess environmental variables.

While population attributable risk percentages were calculated in Chapter 2 when assessing the general effects of the four types of driver inattention, population attributable risk percentages were not calculated for the environmental conditions discussed in the current chapter. Because the environmental conditions were not considered when selecting the baseline sample, a population attributable risk percentage calculation would only be a gross estimate.

Even after collecting data for 12 months on 100 vehicles, there were still many environmental variables with insufficient statistical power to accurately calculate odds ratios. A larger scale naturalistic driving study is needed to not only obtain accurate and valid measures for many of the variables presented in this chapter, but also for more generalizable results to the United States driving population.