Numerous studies have illustrated how risky driving behavior plays a major role in the occurrence of road traffic crashes (RTCs), leading to the consensus that changes in driver behavior is one of the key points for traffic safety interventions.^1–7 One of the approaches considered to be effective and long-lasting is to change attitudes towards traffic safety.^8,9,10 The rationale of this attitudes–behavior–outcome model assumes that risk-taking attitudes correlate with risk-taking behavior, which in turn predict RTCs. Although the potential of this model is often advocated and sometimes used in traffic safety campaigns,^11–13 the causal pathway between attitudes, driving behavior and RTCs risk is rarely studied.¹⁴ This gap in knowledge is due to the fact that most studies that have shown a strong relationship between drivers’ attitudes and their behavior on the roads^8,15–18 did not include data on RTCs. The few studies that considered drivers’ attitudes to RTCs produced inconsistent findings. A study by Assum among Norwegian drivers¹⁴ found that the association between attitudes and accident risk disappears when controlling for mileage and for driver’s age. Another recent study from Norway¹⁹ found that attitudes towards rule violations and speeding influenced involvement in risky driving behaviors, which had an effect on near accidents and accidents. Unfortunately, accident data were recorded retrospectively and over lifetimes, making it difficult to disentangle the temporal sequence in the associations between attitudes, risky driving, and crashes. We conducted a study in a large cohort of employees and recent retirees (the GAZEL cohort) to test the hypothesis that behavioral predictors of serious RTCs were correlated with unfavorable attitudes towards traffic safety. The three specific objectives of this study were: (1) to identify behavioral predictors of serious RTCs, (2) to identify patterns of attitudes towards traffic safety, and (3) to examine the relationship between these specific attitudinal patterns and behavioral predictors of serious RTCs.

MATERIALS AND METHODS

Participants

The participants were current employees or recent retirees of the French national electricity and Gas Company, Electricité De France–Gaz De France (EDF-GDF), who volunteered to participate in a research cohort, known as the GAZEL cohort. The GAZEL cohort was established in 1989 and originally included 20 624 subjects working at EDF-GDF, comprising men aged 40–50 years and women aged 35–50 years at baseline. Since 1989, this cohort has been followed up by means of yearly self-administered questionnaires and by data collection from the company’s human resources and medical departments. The objectives and methods of the cohort have been described in detail elsewhere.²⁰

MATERIALS

Driving behavior and road safety questionnaire 2001

In 2001, a driving behavior and road safety (DBRS) questionnaire was mailed to the 19 894 living members of the GAZEL cohort. This questionnaire was previously piloted on 500 randomly selected participants. The answers and comments of the 330 respondents were used to finalize it.

Risky driving behaviors were assessed using five questions: The participants were asked to report their maximum speed in built-up areas, on rural roads, and on highways (a circle had to be drawn on a scale for each location with marks labelled with multiples of ten; when the circle was placed between two marks, say 40 and 50, it was coded 45). Drinking and driving was assessed with the question: “In the 12 past months, have you ever driven after drinking too much alcohol?” Driving while sleepy was assessed using the question: “In the 12 past months, have you ever driven while sleepy?” Participants were also asked whether they had ever used personal connections to avoid paying the fine for a traffic violation (traffic ticket fixing). Participants were considered to have risky behavior if they reported answering a phone call (with or without a car kit), whatever the driving circumstances and/or if they reported not stopping the vehicle before starting a phone call. Drivers were also asked whether they considered they could violate the Highway Code if they deemed a rule improper. The type of principal vehicle owned in January 2001 was noted and included in one of five categories according to its maximum speed.

Attitudes towards traffic regulations were assessed by asking participants whether they agreed or disagreed with a set of 19 assertions related to traffic regulation and enforcement issues debated in France in 2001. Assertions were positive or negative and were related to speed limitations, seat belt use, drinking and driving, and the overall traffic regulation system. Assertions were constructed in neutral language to avoid suggesting the response.¹⁷ Positive and negative attitudes were coded as 0 and 1, respectively.

The annual GAZEL cohort questionnaire

Sociodemographic data from the cohort database included sex, year of birth, occupational category. Low-quantity alcohol drinkers were defined as men (women) reporting 1–13 (1–6) drinks over 1 week as opposed to high-quantity alcohol drinkers who reported 14–27 (men) and 7–20 (women) drinks over the same period. Regular drinkers were defined as those who reported consuming alcohol on ⩾3 days per week, as opposed to episodic drinkers who reported drinking alcohol on <3 days per week.

Traffic safety-related data from the annual GAZEL cohort questionnaire

Questions on mobility and accidents of the past year were added to the annual GAZEL cohort questionnaire sent in January 2002, January 2003 and January 2004 providing data for the years 2001–3, respectively. Drivers were asked how many kilometers they had driven a 4–wheel or 2–wheel motorized vehicle in the last 12 months to estimate their mileage. The nature of travel (work, private) and driver or passenger status were taken into account. The participants were also asked to report whether or not they had been involved in one or more RTC in the past 12 months. A distinction was made between serious RTC, defined as a crash that required at least a medical consultation, heavy material crashes when at least one vehicle was towed and light material crashes. Only serious RTC were taken into account in the analyses.

Statistical analysis

Because the mean and variance of the number of serious RTCs in the 2001–3 period were not different in our study,^21,22 we fitted generalized linear Poisson regression models²⁶ with time-dependent covariates to estimate the rate ratios of serious RTCs in the 2001–3 period associated with reported risky driving behaviors in 2001. The effects of potential confounders were tested using three nested models. In model 1, the effect of each risky driving behavior on the risk of serious RTCs was assessed separately as an independent variable. RRs in model 2 were further adjusted for age, sex, occupational category in 2001, the type of vehicle owned in 2001, alcohol consumption (a five-category time-dependent covariate) and mileage (a time-dependent covariate). As recommended by Janke,²⁴ a logarithmic transformation was applied to annual mileage. All risky driving behaviors significantly associated with the risk of serious RTCs in model 2 were included in model 3. Those that remained largely associated with the risk of serious RTCs were selected for the next steps of the study.

Principal component analysis with orthogonal (varimax) rotation²⁵ was used to determine underlying dimensions of attitudes towards traffic safety among participants. An item was considered to load on a given component if the factor loading was ⩾0.5. The factor scores (REGR factor score) from the first four attitudinal factors of the principal component analysis were saved. Factor scores are the scores of each participant (row) for each factor (column). To compute the factor score for a given participant for a given factor, the participant’s (automatically) standardized score for each variable is multiplied by the corresponding factor loading of the variable for the given factor, and the products are summed. This procedure runs automatically in most statistical packages such as SPSS, SAS etc. These factor scores were divided into three distinct levels using the centiles (scores at or below the 25th centile mean less negative towards traffic safety and scores at or above the 75th centile mean more negative attitude towards traffic safety) and then regressed on the risky driving behaviors selected in the previous step. All odds ratios were adjusted for age, sex and socioprofessional category.

RESULTS

We received responses to the 2001 DBRS questionnaire from 14 226 of the 19 894 living members of the GAZEL cohort. Some were excluded (n = 26) because of sociodemographic data discrepancies with the general cohort database. Another 326 respondents were excluded because they did not drive in 2001 and 50 others because their data on driving behaviors were missing. A further 377 participants were also excluded because they did not respond to any of the three succeeding annual GAZEL cohort questionnaires of the year 2001–3. Among the remaining 13 447 participants (10 300 men, 3147 women), 328 reported one serious road traffic accident (RTA), 8 reported two serious RTA and 1 reported three serious RTA in the 2001–3 period; 76 other serious RTA were excluded because they were reported to have occurred before the completion of the 2001 DBRS questionnaire. These 13 447 participants represented 67.6% of the 19 894 members of the GAZEL cohort that were alive in 2001. Their mean age was 56.4 years with a range of 48–62 on 1st January 2001.

Behavioral predictors of serious road traffic crashes

The crude risk of serious RTA in the 2001–3 period increased significantly with self-reported high speed on highways, risky use of cellular phone and with self-reported frequency of drinking and driving, sleepy driving and violation of a traffic law deemed to be improper (table 1). When adjusted for potential confounders (model 2), the risk of serious RTCs was associated with exceeding speed limitations on rural roads and on highways, risky use of cellular phone, sleepy driving, and violation of a traffic law deemed to be improper. In the final multivariate model (model 3) that included all significant predictors from model 2, maximum speed more than 90 km/h on rural roads, risky use of cellular phone, and sleepy driving were the three self-reported behaviors that remained significantly associated with the risk of serious RTCs.

Pattern of attitudes towards traffic safety

The principal component analysis followed by orthogonal rotation identified six factors with Eigen values >1. We retained the first four factors formed at least by two variables explaining 37.8% of the total variance (table 2). The first factor compared participants in favor of enforcement to those not in favor. The second factor compared participants in favor of speed limitations to those not in favor. The third factor compared participants in favor of continuous assessment of driving ability to those not in favor. Finally, the fourth factor compared participants in favor of alcohol prohibition on roads to those not in favor.

Who were participants with negative attitudes towards traffic safety?

Higher mean scores for negative attitudes towards enforcement (factor 1) were found among men, relatively younger participants, those who reported high-quantity episodic alcohol consumption, those who reported higher annual mileage and those who reported having the most powerful vehicles (table 3). Higher mean scores for negative attitudes towards speed limitations (factor 2) were found among women, unskilled workers, participants who reported high-quantity episodic alcohol consumption, those who reported higher annual mileage and those who reported having the most powerful vehicles. Higher mean scores for negative attitudes to continuous assessment of driving ability (factor 3) were found among unskilled workers, participants who reported low-quantity episodic alcohol consumption and those with intermediate annual mileage. Finally, higher mean scores for negative attitudes towards alcohol prohibition on roads (factor 4) were found among older participants, managers, participants who reported high-quantity episodic alcohol consumption and those who reported intermediate annual mileage and having intermediate-power vehicles.

Negative attitudes and behavioral predictors of serious RTCs

Results from multivariate logistic regression model (table 4) show that negative attitudes towards enforcement and speed limitation were strongly associated with speeding and risky use of phones whereas negative attitudes towards alcohol prohibition were associated with sleepy driving.

DISCUSSION

In this study, risky driving behavior such as exceeding speed limitations on rural roads (90 km/h), sleepy driving, and risky use of cellular phones were found to be predictors of serious RTCs among participants of the GAZEL cohort in the 2001–2003 period. These predictors were found to be significantly correlated with negative attitudes towards enforcement, speed limitations and alcohol prohibition on roads.

Strengths of the study

These findings support the view that individuals with a high propensity for driving behaviors associated with an increased risk of RTCs were more likely to have negative attitudes towards traffic safety.^19–22 In contrast to many studies, the present study’s participants were from a large cohort followed-up on a regular basis and we were able to record incidents of RTCs and to account for a wide range of potential confounders, some of them being updated each year. The prospective design of our study allowed us to control the temporal relationship between the driving behaviors and the risk of subsequent serious RTCs.

Individuals with negative attitudes towards enforcement issues and accordingly with a high propensity for risky driving behaviors shared three main distinct characteristics consistent with the literature. They were male, had higher mileage and a more powerful vehicle.^22,29,30 However, the relationship between the attitudinal patterns and other variables including age, socioprofessional category and alcohol consumption presented a more distorted picture. As expected, high-quantity episodic alcohol consumers had a higher mean score for negative attitudes towards alcohol prohibition on roads. However, self-reported frequency of drinking and driving was not found to be a predictor of serious RTCs in this study. One plausible explanation is that alcohol consumption rate is moderate among the participants of the GAZEL cohort as high alcohol consumers were less willing to become members of the cohort.²³ However, it is also possible that some of the heavy drinkers have under-reported their alcohol consumption habits. Another noteworthy result was that sleepy driving, one of the most important predictors of serious RTCs in this cohort, was not greatly associated with negative attitudes towards traffic safety issues after controlling for sociodemographic characteristics. This could be explained either by the fact that sleepy driving does not constitute a violation of the Highway Code or that driver attitudes towards sleepy driving were not assessed in this study.

Potential limitations of the study

Both serious RTCs and attitudes towards traffic safety and driving behaviors were self-reported. Although measures of self-reported risky driving behaviors have been found to have considerable validity in predicting traffic accident risk,^28,29 they cannot possibly account for the whole risk attributable to behavioral factors. It should also be noted that our sample, despite its large size and the inclusion of diverse trades and socioeconomic groups throughout France, mostly represented employed or retired middle-aged drivers. Although this sample characteristic limits the generalizability of our results some what, the findings were consistent with Iversen and Rundmo’s study,¹⁹ which was conducted among a sample of Norwegian drivers, representative in terms of age, gender and education, and found that attitudes towards rule violations and speeding influenced involvement in risky driving behaviors which in turn were associated with near accidents and accidents.

Although scores for negative attitudes in three out of the four attitudinal factors were positively associated with behavioral predictors of serious RTCs, the variance fraction explained by these attitudinal factors was relatively low (ranging from 7 to 10%) probably due to the many other factors that influenced drivers’ behavior, including vehicles, road conditions and other drivers. Moreover, it is important to point out, as suggested by Iversen and Rundmo,¹⁹ that self-reported measures can affect results. The moderate explained variance could also be due to the relative homogeneity of participants’ attitudes towards traffic safety, as suggested by the standard deviations of the scores of negative attitude towards traffic safety (table 2). It is interesting, however, to underscore that although our study was conducted among middle-aged participants considered as the safest group when compared with other age groups, our data are consistent with an attitude–behavior–accident pathway model. Finally, more response options than “agree” and “disagree” for the formulation of attitudinal scale could have enhanced the discrimination effectiveness.

Conclusion and implications for prevention

In summary, drivers’ attitudes to a number of road traffic issues currently discussed were correlated with their reported behaviors which influenced their risk of being involved in traffic accidents. This unsurprising pattern was also described in a study of European drivers’ attitudes towards enforcement³⁰ in which it was shown that drivers penalized for speeding in the three years preceding the surveys were likely to report more frequently than other drivers that they exceeded speed limitations. Accordingly, these drivers were also those who were not concerned about road safety and did not want more enforcement or harsher penalties for offenders. These drivers deemed traffic enforcement measures to be arbitrary, unjust, and unuseful or designed for a purpose other than road safety. A survey of attitudes to road safety conducted in 2003 among a representative sample of British adults revealed that 41% of men surveyed thought that “speed cameras are only used to raise money” and 35% thought there are too many cameras on the roads.³¹

However, as stated by Redelmeier et al⁷ traffic policing effectiveness, in particular traffic enforcement, should depend on the extent to which it gains social support (road users, traffic safety associations…). While the meaningful deterrent threat of enforcement is effective in securing the compliance of “early adopters”,³² it does not work for some other drivers or “late adopters” for whom a more in-depth attitudinal change could be required. In a European study of drivers’ attitudes towards enforcement,³¹ it seemed that some drivers’ experience of fines did not change either their attitudes, or their behavior. Interventions aimed at changing drivers’ negative or distorted opinions of traffic “enforcement” as well as “speed limitations” and “alcohol prohibition on roads” could improve their compliance with road traffic rules.