Our findings reveal an extremely high prevalence of self-reported drinking and driving among a high education segment the Spanish participants in the SUN cohort study. Only slightly more than half our participants (53%) reported "never" drinking and driving, and up to 30% of the participants acknowledged to drink and drive "sometimes". Although our cohort is not expected to be representative of the Spanish educated population, this high prevalence seems to be also spreading to the general Spanish population if one is to use the 2002 SARTRE-3 unpublished findings or studies from other countries [4, 8, 12, 14–19].
In particular, and even though our sample is not comparable on educational level to those in some US studies [4, 12, 14] our findings confirm the impact of age and gender on the prevalence of this practice. Our findings also confirm the impact of educational level seen in other studies [15, 18]. Probably, the effect of age on increasing the prevalence is what justifies that our findings differ from those in studies with younger populations, such as those derived from university students [16, 17], including an European study , although it may also be that the wording of the questions used in those studies (e.g., "within the past 30 days how many times [have you] driven when [you] have had perhaps too much to drink") elicits more conservative answers. It also confirms the trends observed in 4 of 7 countries around the world .
More importantly, the strengths of cohort studies reside on the investigation of relationships between variables and the follow-up data that they can provide. Our study found that drinking and driving was also related to other unsafe practices. Not surprisingly, binge drinking and daily drinking were also more common amongst those who reported drinking and driving, as seen in US studies , where 80% of drink and drive episodes are reported by those who also binge drink , or where drunk drivers are also more than twice likely to report daily drinking . In our study, drivers who have drunk were also less likely to use their safety belts, a finding also reported in [4, 16, 19]. Although we did not investigate this issue, it is noteworthy to mention here that drivers who had been drinking were less likely to restrain their children . Rarely investigated in other analytical studies, our findings also highlight the higher prevalence of drinking and driving amongst those who are more exposed to traffic since they drive the most.
The elevated prevalence rates of drinking and driving of the educated population included in our study may seem in conflict with other studies, where high income and education levels have proved to have a protective effect on alcohol-related fatal motor vehicle crashes . However, those studies have not investigated whether this effect is so because more educated and/or wealthier individuals drink less or because of other issues such as better health status pre accident, better vehicles, driving in better roads (e.g., highways) or receiving better health care.
Never reported before is our finding that health professionals (physicians or nurses) are more likely to report drinking and driving than other university graduates even after controlling for possible confounders. The role of these health professionals in educating the population regarding the health consequences of drinking and of drinking and driving has been long advocated for [9, 26, 27]. Yet, their capability to do so may be impaired due to their own lifestyles. Health professionals act as models for their patients and this can be seen, for example, in smoking cessation programmes. Several reports suggest that when an appropriately-trained physician provides counselling and guidance, smoking cessation success rates can be increased [28–30]. However, it has been found that anywhere from 30–63% of smokers had not received cessation counselling from their physicians within the last year . Physicians' own smoking patterns and quitting behaviours are important because physicians serve as models for their patients and play a key role in the reinforcement of smoke-free health facilities . It can be difficult to be a patient's model and have more self-confidence in the success of patients' smoking counselling if the frequency of smoking among doctors is high, as it has been shown in Italy . In fact, as frequency of smoking decreases in doctors they try to convince more people to stop smoking. In a comparison between Finnish (smoking prevalence of 6.7% in males) and Estonian physicians (smoking prevalence 18.6% in males), Finnish doctors with a much lower frequency of smoking, are more likely to believe that it is their responsibility to help patients stop smoking . Some similar attitudes have been observed regarding health education provided by nurses. Nurses who smoke are less motivated to provide cessation support for patients, have less positive attitudes to the value of smoking cessation, are less likely to have received smoking cessation training and are less likely to want further training . How does this tobacco-based evidence translate into the motor vehicle safety world remains a subject for further investigation, but our current study suggests the possibility that traffic safety practices may not be well addressed by health professionals.
Also never reported for Europe is our finding that the prevalence of self-reported drinking and driving seems to be diminishing in more recent years, at least among women. Whether this is a true trend or simply a reflection of an increase in socially desirable answers to the question related to drinking and driving remains to be investigated. It could also be the case that subjects enrolling in the cohort in more recent years are substantially different from the earlier participants. However, since our sampling framework has not changed and neither has changed our invitation protocol, we do not believe this to be an explanation for this finding. Analyses of US data from 1983 to 2003 suggest that there has been a decrease in the overall percent of drinking and driving among motor vehicle drivers and motorcyclist . This reduction is particularly true among younger individuals, which has turned into a shift in the peak age with the highest rates of drinking and driving from the 20–24 years old to those between 40–44 years old. This evidence about the reduction in the prevalence of drinking and driving over time is mentioned in several other publications as possible reasons to explain the reductions of motor vehicle alcohol-related deaths in the US up to 2001 . However, other studies are pointing out to increases in alcohol impaired driving both in the US  and in European countries , such as Finland . As our cohort continues to grow due to its open enrolment nature, we will continue to monitor these trends.
As with other public health problems, an array of interventions to reduce the prevalence of alcohol amongst road users has been set in practice around the world, including legislation prohibiting driving at or over certain levels of alcohol in blood, or the recommendation that advice on this matter is provided in the context of routine general medical visits . Most of these measures are in place in Spain. However, in light of these findings and others in the literature, it seems as if much more needs to be done.