Lowering the legal alcohol limit in Belgium: Potential effects on the number of traffic victims.

BACKGROUND: Since 1994, the legal limit of Blood Alcohol Concentration (BAC) is 0.5 g/L for the general drivers' population in Belgium. Since 2015, this limit has been lowered to 0.2 g/L for professional drivers. So far, no specific limitation has been adopted for novice drivers in Belgium. Recently, two bills were submitted to the House of Representatives: the first one proposed to impose a zero-limit for every driver, the second one proposed to restrict this zero-limit only to novice drivers. OBJECTIVE: The present study evaluated the potential impact of the two bills that aim to reduce the legal BAC limit from 0.5 to zero, either for all drivers or for novice drivers only. We elaborated three scenarios related to the BAC categories for which lowering the legal BAC limit to zero would affect drink-driving behaviour: "Targeted BAC levels" scenario assumed that the new policy would impact only the specifically targeted BAC category, i.e., drivers in the category BAC below 0.5 g/L. "Extended impact" scenario corresponded to the "Targeted BAC levels" scenario to which we added a "halo effect" on drivers in the BAC category "0.5 g/L ≤ BAC < 0.8 g/L". "Broad range impact" scenario corresponded to the "Extended impact" scenario to where we extended the "halo effect" to drivers in the BAC category "0.8 g/L ≤ BAC < 1.2 g/L". METHODS: The effect estimates were based on firstly, scientific literature on risks related to drink-driving at different BAC-levels, secondly data on crashes in Belgium and thirdly data on drink-driving behaviour in Belgium and in other European countries. RESULTS: In case of a zero limit for all drivers, an annual reduction can be expected of 10 to 17 fatalities (i.e., a decrease of 2.4% to 3.9%), 8 to 20 severe injuries (i.e., a decrease of 0.3% to 0.8%) and 135 to 315 slight injuries (i.e., a decrease of 0.4% to 0.8%). In case a zero limit is only applied to novice drivers, an annual decrease can be expected by 2 to 4 fatalities (i.e., a decline of 3.7% to 6.2%), 8 to 16 serious injuries and 135 to 262 slight injuries (i.e., a decline of 1.7% to 3.2% in both cases). DISCUSSION: There is no clear evidence on which of the three scenarios would be the most plausible. As the relative risk of a car crash increases strongly with the BAC level, the success of either measure will strongly depend on its ability to also affect drink driving at concentrations that are forbidden already. This also means that most of the casualties could be prevented if compliance with current rules increased.

Forecasting German crash numbers: The effect of meteorological variables.

At the end of each year, the German Federal Highway Research Institute (BASt) publishes the road safety balance of the closing year. They describe the development of accident and casualty numbers disaggregated by road user types, age groups, type of road and the consequences of the accidents. However, at the time of publishing, these series are only available for the first eight or nine months of the year. To make the balance for the whole year, the last three or four months are forecasted. The objective of this study was to improve the accuracy of these forecasts through structural time-series models that include effects of meteorological conditions. The results show that, compared to the earlier heuristic approach, root mean squared errors are reduced by up to 55% and only two out of the 27 different data series yield a modest rise of prediction errors. With the exception of four data series, prediction accuracies also clearly improve incorporating meteorological data in the analysis. We conclude that our approach provides a valid alternative to provide input to policy makers in Germany.

A systematic cost-benefit analysis of 29 road safety measures.

Economic evaluations of road safety measures are only rarely published in the scholarly literature. We collected and (re-)analyzed evidence in order to conduct cost-benefit analyses (CBAs) for 29 road safety measures. The information on crash costs was based on data from a survey in European countries. We applied a systematic procedure including corrections for inflation and Purchasing Power Parity in order to express all the monetary information in the same units (EUR, 2015). Cost-benefit analyses were done for measures with favorable estimated effects on road safety and for which relevant information on costs could be found. Results were assessed in terms of benefit-to-cost ratios and net present value. In order to account for some uncertainties, we carried out sensitivity analyses based on varying assumptions for costs of measures and measure effectiveness. Moreover we defined some combinations used as best case and worst case scenarios. In the best estimate scenario, 25 measures turn out to be cost-effective. 4 measures (road lighting, automatic barriers installation, area wide traffic calming and mandatory eyesight tests) are not cost-effective according to this scenario. In total, 14 measures remain cost-effective throughout all scenarios, whereas 10 other measures switch from cost-effective in the best case scenario to not cost-effective in the worst case scenario. For three measures insufficient information is available to calculate all scenarios. Two measures (automatic barriers installation and area wide traffic calming) even in the best case do not become cost-effective. Inherent uncertainties tend to be present in the underlying data on costs of measures, effects and target groups. Results of CBAs are not necessarily generally valid or directly transferable to other settings.

The European road safety decision support system on risks and measures.

The European Road Safety Decision Support System (roadsafety-dss.eu) is an innovative system providing the available evidence on a broad range of road risks and possible countermeasures. This paper describes the scientific basis of the DSS. The structure underlying the DSS consists of (1) a taxonomy identifying risk factors and measures and linking them to each other, (2) a repository of studies, and (3) synopses summarizing the effects estimated in the literature for each risk factor and measure, and (4) an economic efficiency evaluation instrument (E3-calculator). The DSS is implemented in a modern web-based tool with a highly ergonomic interface, allowing users to get a quick overview or go deeper into the results of single studies according to their own needs.

Impact of alcohol checks and social norm on driving under the influence of alcohol (DUI).

This study investigated the influence of alcohol checks and social norm on self-reported driving under the influence of alcohol above the legal limit (DUI). The analysis was based on the responses of 12,507 car drivers from 19 European countries to the SARTRE-4 survey (2010). The data were analysed by means of a multiple logistic regression-model on two levels: (1) individual and (2) national level. On the individual level the results revealed that driving under the influence (DUI) was positively associated with male gender, young age (17-34), personal experience with alcohol checks, the perceived likelihood of being checked for alcohol, perceived drunk driving behaviour of friends (social norm) and was negatively associated with higher age (55+). On a national level, the results showed a negative association with a lower legal alcohol limit (BAC 0.2g/l compared with BAC 0.5g/l) and the percentage of drivers checked for alcohol. DUI was positively associated with the percentage of respondents in the country that reported that their friends drink and drive (social norm). The comparison of the results obtained on national and individual levels shows a paradoxical effect of alcohol checks: Countries with more alcohol checks show lower DUI (negative association) but respondents who have been personally checked for alcohol show a higher chance of DUI (positive association). Possible explanations of this paradox are discussed. The effects of the social norm variable (perceived drunk driving behaviour of friends) are positively associated with DUI on both levels.

Latent risk and trend models for the evolution of annual fatality numbers in 30 European countries.

In this paper a unified methodology is presented for the modelling of the evolution of road safety in 30 European countries. For each country, annual data of the best available exposure indicator and of the number of fatalities were simultaneously analysed with the bivariate latent risk time series model. This model is based on the assumption that the amount of exposure and the number of fatalities are intrinsically related. It captures the dynamic evolution in the fatalities as the product of the dynamic evolution in two latent trends: the trend in the fatality risk and the trend in the exposure to that risk. Before applying the latent risk model to the different countries it was first investigated and tested whether the exposure indicator at hand and the fatalities in each country were in fact related at all. If they were, the latent risk model was applied to that country; if not, a univariate local linear trend model was applied to the fatalities series only, unless the latent risk time series model was found to yield better forecasts than the univariate local linear trend model. In either case, the temporal structure of the unobserved components of the optimal model was established, and structural breaks in the trends related to external events were identified and captured by adding intervention variables to the appropriate components of the model. As a final step, for each country the optimally modelled developments were projected into the future, thus yielding forecasts for the number of fatalities up to and including 2020.

Comparing single vehicle and multivehicle fatal road crashes: a joint analysis of road conditions, time variables and driver characteristics.

The difference between single vehicle crashes and multivehicle crashes was investigated in a collection of fatal crashes from six European countries. Variables with respect to road conditions, time variables, and participant characteristics were studied separately at first and then jointly in a logistic multiple regression model allowing to weigh different accounts of single vehicle as opposed to multivehicle crash occurrence. The most important variables to differentiate between single and multivehicle crashes were traffic flow, the presence of a junction and the presence of a physical division between carriageways. Heavy good vehicles and motorcycles were less likely to be involved in single vehicle crashes than cars. Moreover crashes of impaired drivers with more passengers were more likely to be single vehicle crashes than those of other drivers. Young drivers, rural roads, nights and weekends were all shown to have a higher proportion of single vehicle crashes but in the multivariate analysis these effects were demonstrated to be mediated by the road conditions named above.

Multilevel analysis in road safety research.

Hierarchical structures in road safety data are receiving increasing attention in the literature and multilevel (ML) models are proposed for appropriately handling the resulting dependences among the observations. However, so far no empirical synthesis exists of the actual added value of ML modelling techniques as compared to other modelling approaches. This paper summarizes the statistical and conceptual background and motivations for multilevel analyses in road safety research. It then provides a review of several ML analyses applied to aggregate and disaggregate (accident) data. In each case, the relevance of ML modelling techniques is assessed by examining whether ML model formulations (i) allow improving the fit of the model to the data, (ii) allow identifying and explaining random variation at specific levels of the hierarchy considered, and (iii) yield different (more correct) conclusions than single-level model formulations with respect to the significance of the parameter estimates. The evidence reviewed offers different conclusions depending on whether the analysis concerns aggregate data or disaggregate data. In the first case, the application of ML analysis techniques appears straightforward and relevant. The studies based on disaggregate accident data, on the other hand, offer mixed findings: computational problems can be encountered, and ML applications are not systematically necessary. The general recommendation concerning disaggregate accident data is to proceed to a preliminary investigation of the necessity of ML analyses and of the additional information to be expected from their application.

Estimation of fatality and injury risk by means of in-depth fatal accident investigation data.

OBJECTIVE: In this article the factors affecting fatality and injury risk of road users involved in fatal accidents are analyzed by means of in-depth accident investigation data, with emphasis on parameters not extensively explored in previous research. METHODS: A fatal accident investigation (FAI) database is used, which includes intermediate-level in-depth data for a harmonized representative sample of 1300 fatal accidents in 7 European countries. The FAI database offers improved potential for analysis, because it includes information on a number of variables that are seldom available, complete, or accurately recorded in road accident databases. However, the fact that only fatal accidents are examined requires for methodological adjustments, namely, the correction for two types of effects on a road user's baseline risk: "accident size" effects, and "relative vulnerability" effects. Fatality and injury risk can be then modeled through multilevel logistic regression models, which account for the hierarchical dependences of the road accident process. RESULTS: The results show that the baseline fatality risk of road users involved in fatal accidents decreases with accident size and increases with the vulnerability of the road user. On the contrary, accident size increases nonfatal injury risk of road users involved in fatal accidents. Other significant effects on fatality and injury risk in fatal accidents include road user age, vehicle type, speed limit, the chain of accident events, vehicle maneuver, and safety equipment. In particular, the presence and use of safety equipment such as seat belt, antilock braking system (ABS), and electronic stability program (ESP) are protection factors for car occupants, especially for those seated at the front seats. CONCLUSIONS: Although ABS and ESP systems are typically associated with positive effects on accident occurrence, the results of this research revealed significant related effects on accident severity as well. Moreover, accident consequences are more severe when the most harmful event of the accident occurs later within the accident chain.

Risk and protection factors in fatal accidents.

This paper aims at addressing the interest and appropriateness of performing accident severity analyses that are limited to fatal accident data. Two methodological issues are specifically discussed, namely the accident-size factors (the number of vehicles in the accident and their level of occupancy) and the comparability of the baseline risk. It is argued that - although these two issues are generally at play in accident severity analyses - their effects on, e.g., the estimation of survival probability, are exacerbated if the analysis is limited to fatal accident data. As a solution, it is recommended to control for these effects by (1) including accident-size indicators in the model, (2) focusing on different sub-groups of road-users while specifying the type of opponent in the model, so as to ensure that comparable baseline risks are worked with. These recommendations are applied in order to investigate risk and protection factors of car occupants involved in fatal accidents using data from a recently set up European Fatal Accident Investigation database (Reed and Morris, 2009). The results confirm that the estimated survival probability is affected by accident-size factors and by type of opponent. The car occupants' survival chances are negatively associated with their own age and that of their vehicle. The survival chances are also lower when seatbelt is not used. Front damage, as compared to other damaged car areas, appears to be associated with increased survival probability, but mostly in the case in which the accident opponent was another car. The interest of further investigating accident-size factors and opponent effects in fatal accidents is discussed.

Who is dominating the Dutch neighbourhood? On the role of subsyllabic units in Dutch nonword reading.

To assess the role of the subsyllabic units onset-nucleus (ON; spark) and rime (spark) in Dutch visual word recognition, we compared lexical decisions to four groups of nonwords in which the existence of ONs and rimes was orthogonally manipulated. Nonwords with existent ONs and/or rimes were rejected more slowly and less accurately. ON and rime neighbours thus influence Dutch nonword reading to the same extent. Simulations with the interactive activation model (McClelland & Rumelhart, 1981) revealed that this model with left-to-right coded representations could not replicate the effects found in the lexical decision data whereas an adapted version with representations of onset, nucleus, and coda could. Effects of the larger units ON and rime emerged from activation patterns created by the smaller units onset, nucleus, and coda.

Rhyming words and onset-rime constituents: an inquiry into structural breaking points and emergent boundaries in the syllable.

Geudens and Sandra, in their 2003 study, investigated the special role of onsets and rimes in Dutch-speaking children's explicit phonological awareness. In the current study, we tapped implicit phonological knowledge using forced-choice similarity judgment (Experiment 1) and recall of syllable lists (Experiment 2). In Experiment 1, Dutch-speaking prereaders judged rime-sharing pseudowords (/fas/-/mas/) to sound more similar than pseudowords sharing an equally sized nonrime unit (/fas/-/fak/). However, in a syllable recall task (/tepsilonf/, /ris/, /nal/), Dutch-speaking prereaders were as likely to produce recombination errors that broke up the rime (/tepsilons/) as to produce errors that retained the rime (/repsilonf/). Thus, a rime effect was obtained in a task that highlighted the phonological similarity between items sharing their rimes, but this effect disappeared in tasks without repetition of rime units. We conclude that children's sensitivity to rimes depends on similarity relations and might not reflect a fixed perceived structure of spoken syllables.