Alcohol use and risk of vehicle accidents: cross-sectional study in the city of São Paulo, Brazil

ABSTRACT BACKGROUND: Harm to other people caused by an individual under the influence of alcohol (UIA) can occur in a variety of relationship situations between the drinker and these other people. OBJECTIVES: To estimate the prevalence of the risk of vehicle accidents (RVA) involving people who are UIA, according to sociodemographic variables, respondent alcohol use and gender. DESIGN AND SETTING: Cross-sectional study, in which a household survey was carried out on a cluster-stratified representative sample of urban residents in the city of São Paulo. METHODS: The final sample was composed of 1,155 subjects aged 18-59 years, who were interviewed using the GENACIS Harm-to-Others questionnaire. Individuals were defined as having been harmed if an affirmative response was given to at least one of the questions that refers to RVA involving people who had been UIA in the last twelve months. Post-stratification weights were calculated to adjust for the study design and for no response. Since the outcome was binary, logistic regression was used in multivariable analysis. RESULTS: The final response rate was 58.6%. The overall prevalence of RVA was 13.6% (95% confidence interval, CI 11.0-16.7): 16.6% among men and 10.8% among women. After the logistic regression, age remained as a protective factor (odds ratio, OR 0.95) and binge drinking as a risk factor (OR 2.03). CONCLUSION: This study showed that binge drinking was associated with RVA.


INTRODUCTION
Harm to other people caused by individuals who are under the influence of alcohol (UIA) can happen in many ways, to different degrees and to people with different relationships with the drinker. Harm to other people due to alcohol consumption can include intentional or unintentional injury and can affect a whole group or single individuals, who may be spouses, relatives, friends, non-relatives living in the same home, co-workers or even strangers. [1][2][3] Studies have shown that traffic accidents and alcohol consumption are directly related, thereby causing serious consequences to individuals, families and the community. [4][5][6][7][8] Brazil is one of the countries with the highest rates of traffic accidents relating to alcohol consumption, with huge social costs. [4][5][6][7][8] A recent review by Araújo et al. 9 assessed several studies on traffic accidents and alcohol use in Brazil, and showed that there was high prevalence of alcohol use and an association between this and the severity of injuries. A Brazilian survey conducted in 2012 indicated that the prevalence of individuals who reported driving after consuming alcohol in the last year was 27.3% among men and 7.1% among women. 10 Epidemiological studies evaluating alcohol use relating to fatalities in the state of São Paulo, the Federal District and the city of Porto Alegre found alcohol in the blood in 45%, 43% and 32% of the subjects, respectively. Among non-fatal accident victims treated in trauma and emergency care centers in Uberlandia, state of Minas Gerais, 31.8% of these patients tested positive for blood alcohol content, and they more frequently required hospitalization (70.4% versus 37.9%; P < 0.05). 11,12 A study carried out in the city of Rio de Janeiro showed that 42.5% of fatal traffic accidents involved individuals with some level of blood alcohol concentration, among which 66.2% had levels between 0.6 and 2.0 g/L of alcohol in the blood. Four to eight doses of alcohol, on average, in the last hours before the accident may indicate excessive alcohol consumption. However, this study in Rio de Janeiro found that the frequency of blood alcohol tests was low: only 34

OBJECTIVE
The aim of the present study was to estimate the prevalence of the risk of vehicle accidents (RVA) involving people who were UIA, according to sociodemographic characteristics and respondent alcohol use.

Sample
This study comprised a cross-sectional survey of the general population that was carried out on a cluster-stratified representative sample of urban residents of the city of São Paulo. The participants were aged 18 to 59 years. The sample was probabilistic. Stratified sampling procedures were used in three stages: census tract, household and resident. Considering that only one resident in each household would be interviewed, 1,250 households were included in the sample. These households were distributed in 50 census tracts. In each census tract, 25 households were surveyed and, in each household, a single resident. To select this resident, we used the procedure proposed by Kish. 18 Considering that empty dwellings would be found and that the refusal rate among the residents was likely to be 20%, 1,600 households were drawn, i.e. 32 in each census tract. If the home was found to be unoccupied, a maximum of two further visits were made, to try to find a resident. The interviews were conducted between September 2014 and January 2015. The percentage of the households that received the interviewers was 71.0%, but 17.5% of the households refused to participate, which resulted in a final response rate of 58.6%.

Setting
The study site was the city of São Paulo, which is the capital of the RVA was the dependent variable in this study. Individuals were defined as having been harmed if an affirmative response was given to at least one of the questions referring to vehicle accidents that involved people who were UIA, in the last twelve months. The dependent variable was constructed through the following questions: • Have you ever been a passenger of a driver who had drunk a lot?
• Have you had a car accident (or other motor vehicle accident) because of someone else's drinking?
• Did you feel at risk in the car when someone else was driving, because of how much the person had drunk?
• Were you injured in a car accident because of how much someone else had drunk?

Independent variables
The instrument used for data-gathering was the GENACIS Harm-to-Others questionnaire (available upon request): • Sociodemographic variables All sociodemographic characteristics, including the variables of gender, age, color, income and education were included in this study and were assumed to be independent for the analysis.
• Respondent alcohol use The respondents were questioned about their alcohol consumption. They were classified as abstinent (had never made use of alcohol in their lifetime), former drinker (had not used alcohol in the last 12 months) or current drinker (had had alcohol intake in the last 12 months. The drinkers reported their consumption in terms of quantity and frequency on a typical day. Binge drinking was also investigated, and this was considered to consist of consumption of four or more drinks for women and five or more drinks for men on one occasion in the last 12 months. For the analysis, the number of doses on a typical day, any occurrences of binge drinking and the score from the alcohol use disorders identification test (AUDIT) were included, for current drinkers only.

• AUDIT
The AUDIT 21 is an instrument with 10 items that was designed to identify the patterns of alcohol use. It has been validated for use in Brazil. 22 When respondents get a score of eight points or more, their consumption is considered to be at a problematic level of use. Although this test includes questions that identify binge drinking, its aim is to identify subjects who already present a problematic pattern of use, as defined by this instrument.

Analytical model
Associations between the outcome and independent variables were evaluated using the Rao-Scott test. 23 Each observation was weighted accordingly to compensate for the effects of delineation and of no response. Firstly, we ran bivariate analyses, to select eligible variables for inclusion in multivariable analysis.
We used multivariable logistic regression analysis, with backward stepwise variable selection. 24 All the variables that showed associations with outcome presenting P ≤ 0.25 were included in the models. The results from these analyses are presented in the form of odds ratios (OR), using 95% confidence intervals (CI). Only variables for which the associations presented P < 0.05 were retained in the final model. We did separate analyses for men and women.
The data were analyzed using the Stata 12.0 statistical software (StataCorp LP, College Station, TX, USA).

Ethics
The proposal for this study was approved by the Research Ethics

RESULTS
The Among women, occurrences of vehicle accidents involving people who were UIA in the last 12 months were associated with being a current drinker (16.3%) and high frequency of alcohol consumption in the last 12 months (P = 0.001).
For the multivariable analysis, it was decided to build a single model for both genders together. All the variables selected were included in the model and they were retained if P > 0.05. In order to highlight the importance of binge drinking, we presented two models ( Table 3). In the first model, only gender and age were included, and this showed the risk associated with male gender (OR = 1.64) and the protective effect of increasing age. In the second model, binge drinking was included and this nullified the risk associated with male gender. It was observed that the binge drinking variable was a risk factor (OR = 2.38) and that its introduction changes the OR values for gender, thus changing its significance.
Also, the inclusion of binge drinking did not prejudice the role of age, which remained a protective factor.

DISCUSSION
Social harms from drinking are inherently interactional.
For social harm to be recognized as occurring, there needs to be not only drinking behavior that is seen as problematic, but also a reaction by someone other than the drinker. The harm may occur to the drinker or to other people. 25 Alcohol surveys have usually asked drinkers themselves about problems from their own drinking. However, it is important to investigate people who have been harmed by other people's drinking. 26 The total percentage of individuals who reported occurrences of RVA in the present study was 12.73%. Using data from the 2014 to 2015 United States National Alcohol Survey, Karriker-Jaffe et al. 27 found that fewer than 1% of the respondents had been in traffic accidents caused by someone who had been drinking, in the last year.  The main outcome of the present study was that a significant association was found between higher prevalence of risks of vehicle accidents (RVA) and binge drinking, caused by other people's consumption. In other words, when drinkers practice binge drinking, they put themselves at risk, regardless of whether or not they are the driver. A causal link has been shown between consumption of alcoholic beverages and occurrences of traffic accidents, considering the effects of alcohol on individuals' perception, vision, reflexes, consciousness and behavior. 28  In the present study, it was observed that age was a protective factor, such that RVA decreased with increasing age. The practice of driving after excessive alcohol consumption showed highest frequency in the 25 to 34-year age group (4.0% among men and 0.7% among women). 7 Duailibi et al. 8  There is no consensus regarding the minimum response rate for population surveys, taking into account the characterization of non-responders, or regarding the extent to which this loss relates to the information sought by the study. Since the subject of RVA is a sensitive issue, we can assume that there is a risk minimization factor: probably the risk is greater than what has been found.
Our data differ from those of some other studies that indicated high abstinence rates in Brazil, especially in populations with lower levels of education. 33,34 It is possible that this result is related to the questionnaire that was used. The GENACIS Harm-to-Others questionnaire assesses the frequency of alcohol consumption through detailed questions about different types of alcoholic beverages, with the aim of minimizing biases in memory, misrepresentation, false responses or non-acceptance, which are common in surveys on embarrassing or intimate issues.
Nonetheless, the figures obtained in the present study are close to the national data, which indicate that the prevalence of drinkers is around 50%. 10 Also, as previously mentioned, the city of São Paulo is characterized by a peculiar situation, in that it is located in the region with highest lifetime rate of alcohol use (80.4%). 15 The strengths of this study, on the other hand, are that it was conducted over a short period, was stratified according to neighborhoods and had a representative sample of residents of the city of São Paulo, between 18 and 59 years of age. Thus, the results can be generalized to other large urban centers and probably do not reflect the patterns of rural areas of Brazil.
The problems arising from consumption of alcoholic drinks by drivers have been studied internationally in a broad manner.
In Brazil, the risk of vehicle accidents associated with alcohol use can be considered to be a major public health problem, with a major impact on morbidity and mortality, especially in the young male population. Therefore, preventive interventions need to focus on these problems, for instance through promoting strict enforcement of drinking and driving laws to prevent this combination of behaviors.

CONCLUSIONS
GENAHTO is the first project to assess the harm that alcohol use causes to other people in diverse societies. 16 The current study aimed to estimate the prevalence of the risk of vehicle accidents involving people who were under the influence of alcohol, and it attempted to fill the gap regarding the lack of studies on people who are victims in accidents involving alcohol consumption by other people. It is important to highlight that we found that individuals who practiced binge drinking put themselves in RVA situations, regardless of whether they are the driver or not. Until now, this impact has not been considered in establishing public policies regarding alcohol consumption. This indicates that there is a need for further studies in this important area, and the results from the present study could be useful for guiding preventive strategies.