CHAPTER 4: OBJECTIVE 3, DETERMINE THE DIFFERENCES IN DEMOGRAPHIC DATA, TEST BATTERY RESULTS, AND PERFORMANCE-BASED MEASURES BETWEEN INATTENTIVE AND ATTENTIVE DRIVERS. HOW MIGHT THIS KNOWLEDGE BE USED TO MITIGATE THE POTENTIAL NEGATIVE CONSEQUENCES OF INATTENTIVE DRIVING BEHAVIORS? COULD THIS INFORMATION BE USED TO IMPROVE DRIVER EDUCATION COURSES OR TRAFFIC SCHOOLS? (continued)
Spearman correlations were conducted to determine whether there were any linear relationships between the frequency of involvement in inattention-related events and survey responses/test scores for both the high- and low-involvement groups. Table 4.20 presents only those test scores/survey responses that were significant.
Note that none of the low-involvement group’s correlations were significant with only accident involvement approaching significance at a 0.06 probability level. The rest of the significant correlation coefficients were for the high-involvement group. Those scores or responses that demonstrated a linear relationship with inattention-related crash and near-crash involvement were Driver Age, Driving Experience, and Neuroticism Scale. Driver age has been found in the past to be highly inversely related to crash involvement. Given that most of the drivers probably received their driver’s license in the United States at approximately age 16, these two responses are probably highly correlated with each other. The neuroticism scale has been found in previous research to correspond to drivers self-reported crash involvement; this is an interesting finding in that this demonstrates high correlation to actual crash and near-crash involvement.
Note: Numbers in bold font indicate statistical significant using a 0.05 probability value.
As part of the exploratory nature of these analyses, a second analysis using three groups was also conducted. With three groups, some separation between the two tails of the distribution is present so that any differences in those drivers who are the most and least involved in inattention-related crashes and near-crashes may be more easily distinguished. The drivers were grouped into three levels of involvement in inattention-related crashes and near-crashes: low, moderate, and high involvement. These groups were based upon the number of inattention-related crashes and near-crashes that each driver was involved (Figure 4.11). “Low involvement” refers to those drivers who were not involved in any or were involved in one inattention-related crash and/or near-crash. The “moderate involvement” group was involved in two to four inattention-related crashes or near-crashes. The “high involvement” group was involved in five or more inattention-related crashes or near-crashes. Therefore, “high involvement” refers to those drivers with high numbers of inattention-related crashes and/or near-crashes and “low involvement” refers to those drivers with none or only one inattention-related crash and/or near-crash.
Univariate analyses of variance (ANOVA) tests were conducted using the three levels of inattention-related event involvement. All survey responses and test scores that were appropriate were used as dependent variables. Only those ANOVA tests that were significant will be reported in the following section. Table 4.21 provides the descriptive statistics for the drivers assigned to low-, medium-, and high-involvement groups.
The results of the univariate ANOVA tests using three involvement groups indicated that five of the test scores that were significantly different for the two-group analysis also proved to be significantly different for the three-group analysis. These five test scores/demographic data were mean driver age, years of driving experience, self-reported traffic violations, agreeableness, and conscientiousness. Two other test scores were found to be significantly different using three groups that were not significantly different using two groups: these two test scores were daytime sleepiness score and self-reported accident involvement. The three-group scores on extraversion and openness to experience were not significantly different even though these tests were significantly different with only two groups.
These results indicate that the extremely low- and extremely high-involvement groups were significantly different from each other for daytime sleepiness scores. For self-reported accident involvement, the two extreme groups were actually not significantly different from each other rather the moderate-involvement group actually reported significantly more accidents than did the high-involvement or the low-involvement groups. It could be hypothesized that this was an artifact of age in that the high-involvement drivers were, on average, 25 years old whereas the low- and moderate-involvement driver groups had an average age of 39 and 38, respectively.
Separating the drivers into three groups failed to find significant differences for the two personality inventory scales of extraversion and openness to experience. This result may be explained statistically in that by separating the drivers into three groups reduces the statistical power of the sample due to the decreased numbers of drivers in each group.
Most of the statistical tests that were significant with only two groups were also significant with three groups. All univariate analysis results are presented in Table 4.22. Given the exploratory nature of these analyses, conducting two analyses (a two-group and a three-group) was an important step in understanding these data. Both analyses have benefits. The two-group analysis, with a larger number of drivers per group, has better statistical power whereas the three-group analysis provides more separation between the extreme drivers. The significant results demonstrated that very few differences existed between the two- and three-group analyses; therefore, the results that were observed are stable and reliable for the driving population.
Correlations were also conducted for each group of involvement. Correlations were performed using the frequency of involvement in inattention-related crashes and near-crashes versus driver survey responses or test battery scores. The significant results are shown in Table 4.23. Several more tests obtained or approached significant results with three groups. The Dula Dangerous Driving: Aggressive Driving Index, the Dula Dangerous Driving Overall Index, Neuroticism, Agreeableness, and Conscientiousness all demonstrated significant correlations for the high-involvement group only. The neuroticism scale also obtained significance for the moderate-involvement group. The Driving Stress Inventory: Thrill-Seeking Scale reached significance for the low-involvement group but no other group.
Note: Numbers in bold font indicate statistical significant using a 0.05 probability value
A logistic regression was conducted to determine whether multiple data sources, all obtained from demographic data, test battery results, and performance-based tests, could be used to predict whether a driver was either highly involved in inattention-related crashes and near-crashes or not. Only the seven variables that demonstrated significant differences in involvement level for the above tested t-tests or ANOVAs were used in the analysis. These variables were:
None of the correlation coefficients for any of the above variables or test battery results was greater than ±0.4, which is considered to be a small to moderate effect size in the behavioral sciences. Nevertheless, these variables were used in the logistic regression analysis.
A backward selection technique was used to first identify those variables that make significant partial contributions to predicting whether a driver involvement was low or high. This procedure produced a logistic regression equation with two variables: Driver Age and Agreeableness. The resulting significant regression coefficients and relevant statistics are shown in Table 4.24.
A forward selection technique was then used to ensure that both of these variables were making significant partial contributions to the prediction equation. The results of this test resulted in the same regression equation, indicating that both Driver Age and Agreeableness are both predictive of a driver’s level of involvement in inattention-related crashes and near-crashes.
The correlation coefficients for both Driver Age and Agreeableness were both negative, indicating that as Age or Agreeableness increases, involvement in inattention-related crashes and/or near-crashes will decrease. The odds ratio estimates, as calculated as part of the logistic regression, for Driver Age was 0.96 (Lower Confidence Limit = 0.92 and Upper Confidence Limit = 1.0), which was not significantly different from 1.0. The odds ratio estimate for Agreeableness was similar at 0.94 (Lower Confidence Limit = 0.89 and Upper Confidence Limit = 0.99). These results indicate a slight protective effect in that as an Age or Agreeableness score increases, there will be a decrease in involvement in inattention-related crashes and near-crashes.
DISCUSSION. HOW MIGHT THESE RESULTS BE USED TO MITIGATE THE POTENTIAL NEGATIVE CONSEQUENCES OF INATTENTIVE DRIVING BEHAVIORS AND COULD THIS INFORMATION BE USED TO IMPROVE DRIVER EDUCATION COURSES OR TRAFFIC SCHOOLS?
As part of this analysis, the health screening, questionnaires, and driving performance-based tests were all analyzed to determine if the scores obtained on any of these measures correlated or could determine differences in high- or low-involvement in inattention-related crashes and near-crashes. There were seven variables that produced significant t-tests: Driver Age, Driving Experience, number of moving violations in the past 5 years, and four of the personality scales from the NEO Five-Factor Inventory: Extroversion, Openness to Experience, Agreeableness, and Conscientiousness. When three groups were used, Daytime Sleepiness Rating and Accident Involvement also identified significant differences between groups. For the correlation analysis, several test batteries were significant with three groups that were not significant when using two groups of drivers. A logistic regression was conducted to determine if any of these seven variables were predictive of driver inattention. The results of this analysis indicate that Driver Age and Agreeableness both demonstrated some predictive nature to driver involvement in inattention-related crashes and near-crashes.
The results of the logistic regression indicate that none of the demographic data or test scores, except for Driver Age and the Agreeableness score from the NEO Five-Factor Inventory, demonstrate predictive abilities to pre-determine which drivers may be at greater risk of inattention-related crashes and near-crashes. Predictive qualities aside, obtaining significant differences and significant correlations using highly variable human performance data demonstrates that many of these surveys and test batteries do provide useful information about the driving population.
The significant results of Driver Age, for both the logistic regression and the t-tests, indicate that drivers’ education of the dangers of distraction and drowsiness while driving is critical. Note that the younger drivers were over-represented in inattention-related crash and near-crash involvement (Figure 4.2). The significant results in Driving Experience are not surprising as this variable is highly correlated with Driver Age.
The significant t-tests and ANOVAs detecting that the high-involvement drivers were significantly younger than the other groups suggests that younger drivers are over-involved in inattention-related crashes and near-crashes. These results lend some support to those states who have already implemented graduated driver’s licensure programs to restrict specific types of driver distraction. The results from this analysis also lend support to those studies that have already shown that these actions may in fact reduce younger drivers’ involvement in crashes and near-crashes (Hedlund and Compton, 2005). As part of graduated licensure programs, some states have restricted the number of passengers in the vehicle and other states have banned hand-held-device use for teenage drivers. Conducting a naturalistic driving study with teen drivers would be the next research step to determine frequency of engagement in inattention-related tasks and the impact of inattention on driving.
It is very interesting that the self-reported variable, number of traffic violations received in the past 5 years, indicated that high-involvement drivers also had a higher frequency of traffic violations than the low-involvement drivers. This result suggests that those drivers who are attending traffic schools due to multiple traffic violations may indeed be those drivers who are more highly involved in inattention-related crashes and near-crashes. This also suggests that driver inattention is a topic that needs to be addressed in traffic school training. Based on results from other chapters in this report, one item of training may be to assist drivers in their decisions of when to engage in a secondary task, for example. Near-crash/crash risks are much higher in intersections, wet, snowy, or icy roadways, and in moderate traffic density that is moving faster than 25 miles per hour, etc. These are not times in which to engage in a secondary task if it is not necessary that a driver do so. Results from other chapters in this report suggest that eyeglances greater than 2 seconds away from the forward roadway increase near-crash/crash risk. Teaching drivers how to scan the roadway environment but returning to the forward roadway at least once every 2 seconds may also be useful information to incorporate into traffic school and driver’s education programs. More research is required to determine how to best present this information and how to optimally incorporate this information into a driver training program.
The results of this analysis indicated that Driver Age, Driving Experience, self-reported traffic violations and crashes, daytime sleepiness ratings, and personality inventory scores indicated significant differences between the high- and low-involvement drivers for both two and three groups of involvement in inattention-related crashes and near-crashes. Given the exploratory nature of these analyses, two separate analyses were conducted using two groups of involvement and three groups of involvement.