Impact of legislation to reduce the drink-drive limit on road traffic accidents and alcohol consumption in Scotland: a natural experiment study

Article history

The research reported in this issue of the journal was funded by the PHR programme as project number 14/186/58. The contractual start date was in March 2018. The final report began editorial review in May 2018 and was accepted for publication in January 2019. The authors have been wholly responsible for all data collection, analysis and interpretation, and for writing up their work. The PHR editors and production house have tried to ensure the accuracy of the authors’ report and would like to thank the reviewers for their constructive comments on the final report document. However, they do not accept liability for damages or losses arising from material published in this report.

Declared competing interests of authors

Emma McIntosh and Andy Jones are members of the National Institute for Health Research Public Health Research programme funding board.

Copyright statement

© Queen’s Printer and Controller of HMSO 2019. This work was produced by Lewsey et al. under the terms of a commissioning contract issued by the Secretary of State for Health and Social Care. This issue may be freely reproduced for the purposes of private research and study and extracts (or indeed, the full report) may be included in professional journals provided that suitable acknowledgement is made and the reproduction is not associated with any form of advertising. Applications for commercial reproduction should be addressed to: NIHR Journals Library, National Institute for Health Research, Evaluation, Trials and Studies Coordinating Centre, Alpha House, University of Southampton Science Park, Southampton SO16 7NS, UK.

2019 Queen’s Printer and Controller of HMSO

Introduction

It is widely recognised that drink driving is a leading cause of road traffic accidents (RTAs). In 2000, the European Commission estimated that one-quarter of road accident fatalities were a result of alcohol consumption. 1 Although there have been large reductions in accidents involving drink driving in Great Britain over recent decades (deaths and serious injuries related to drink driving fell by three-quarters between 1980 and 20102), drink-drive accidents still accounted for 13% of all road deaths in 2012. 3 It has been estimated that drink-driver injury accidents cost the Scottish economy £80M per year. 4 There is evidence that accident risk increases with blood alcohol concentration (BAC), such that the relative risk of being involved in a fatal crash as a driver is 4–10 times greater with a BAC in the range 0.05–0.07 g/dl than for a zero BAC. 5

The first country to introduce a legal BAC limit to combat drink driving was Norway in 1936, where it became illegal to drive with a BAC level of ≥ 0.05 g/dl. Since then, many other countries and jurisdictions have followed this ‘Scandinavian model’ to deter drink driving. Legal BAC limits are in place in countries and regions across Europe, North America, Japan and Australasia, with most countries having BAC limits of 0.05 or 0.08 g/dl. 6 The British Road Safety Act (BRSA)7 introduced a legal limit of 0.08 g/dl in 1967, which is still in place today. The exception is Scotland, where the BAC limit was reduced to 0.05 g/dl on 5 December 2014.

The desired effect of a legal BAC limit or a reduction of the legal BAC limit is to deter people from drink driving. However, in an early study assessing the impact of the BRSA, it was found that single-vehicle night-time collisions returned to pre-law levels after an initial marked and statistically significant decline. 7 It has been suggested that the reason for this observation is that drivers initially have a heightened perception of risk of being caught drink driving, followed by a realisation that the actual risk is lower than initially thought. 6 Hence, it has been argued that for a law or a change in law to be successful, publicity and public education has to be maintained, and that random breath testing (RBT) measures are required in conjunction with other efforts. 8

Effectiveness of reducing the blood alcohol concentration limit from 0.08 to 0.05 g/dl

In France, where the BAC limit was reduced from 0.08 to 0.05 g/dl (which from now on will be referred to BAC8to5) in 1996, a study in the Haute-Savoie province showed that by 1997 the annual number of alcohol-related accidents causing death had reduced by over one-third. 9 In addition, the effectiveness of the BAC8to5 has been demonstrated in the states of Queensland and New South Wales in Australia. In Queensland, after the reduction of the BAC limit was made law in 1982, a 14% reduction in serious accidents and an 18% reduction in fatal accidents was observed. In New South Wales, where the law change was in 1980, reductions of 7% and 8% were observed, respectively. 10 A study from Austria, evaluating the change in law in 1988, showed a 9.4% reduction in the percentage of drink-driving accidents related to the total number of accidents. 11 In Denmark, the BAC8to5 happened in 1998 and a study showed that the percentage of drivers who reported that they would have no alcohol or restrict themselves to one drink within 2 hours before driving increased from 71% to 80% in the first year of the change in law. However, this study did not observe a reduction in alcohol-related crashes. 12 A meta-analysis, which did not take into account the quality of the individual studies included, estimated that a lowering of the BAC limit to 0.05 g/dl is associated with a 11% reduction in fatal alcohol-related crashes. 13

The scientific quality of these studies examining the effectiveness of the BAC8to5 is variable, with mostly before-and-after-type study designs being employed,9,11 which have known flaws. The Danish study12 used survey data on drink-driving behaviour, which are susceptible to self-reporting biases. The Australian study,10 the main focus of which was to evaluate RBT, was of better quality, with RTA data spanning a long period, and adjustments being made for seasonal, temporal trend, road usage and economic factors in a time-series model.

From a critique of the literature, the highest quality studies are ones that used European data from 15 EU countries to evaluate the change in law or introduction of law to 0.05 g/dl BAC limits,14 and the aforementioned evaluation from Australia. 10 The European study created a panel data set across countries and time, and employed difference-in-differences (DiD)-type methods via generalised linear models to estimate intervention effects. The natural variation in when and where BAC changes occurred allowed the researchers to estimate before-and-after RTA rates, in intervention and control groups. The results show that the 0.05 g/dl BAC limit is associated with a 4.5% and 7.4% reduction in total death rates (using population and per distance driven denominators, respectively). However, a major limitation of this study was that the unit of time was years and the change in law or introduction of law to 0.05 g/dl limits for the individual countries was rarely the start of a calendar year. Therefore, a crude fractional correction had to be applied to the time-varying binary covariate that would switch on in the year the BAC law changed. Furthermore, it was not possible to adjust for seasonal effects and other contextual factors that varied within a year and between countries. Finally, the outcome measure was not all RTA rates, but only those RTAs that caused death. The Australian study10 evaluated BAC law changes that occurred in two states over 30 years prior to the intervention under study and it is reasonable to suggest that the large effect sizes observed may be more difficult to achieve given the large improvements over time in road safety and drink driving becoming increasingly socially unacceptable.

In the studies summarised above, it has been assumed that any BAC intervention effect is homogeneous across the population. Evidence that a reduced BAC limit is effective not only at the BAC levels involved (i.e. in Scotland, 0.05–0.08 g/dl) but at all BAC levels6 gives credence to a homogeneous effect. If this is true, this would mean that a BAC intervention reduces absolute levels of socioeconomic deprivation (SED) inequality if RTA prevalence differs by SED group. A higher RTA prevalence in more deprived groups is plausible because the amount of alcohol individuals drink is positively associated with the probability of drink driving,15 and high levels of SED are associated with high levels of drinking. 16 To our knowledge, no study has evaluated whether or not SED could influence the magnitude of the BAC intervention effect in this manner.

Cost-effectiveness of blood alcohol concentration limits

There have been a limited number of studies investigating whether BAC limit laws, or changes to BAC limit laws, is cost-effective policy-making. A key message from a Lancet review paper on effective and cost-effective policies to reduce the harm caused by alcohol was that drink-driving countermeasures (namely, reduction in BAC limits and enforcement via breath-testing campaigns) are cost-effective. 17 The modelling exercises in this paper estimated the cost per disability-adjusted life-year (DALY) saved for different World Health Organization sub-regions, assuming effect sizes taken from one of the systematic reviews already mentioned. 6 In a study that estimated the costs and effects of a range of enforcement strategies for reducing the burden of RTAs in developing countries, drink-driving legislation (and enforcement via breath-testing campaigns) was shown to be a cost-effective strategy, saving as many DALYs as bicycle helmet use. 18 A recent cost–benefit analysis (CBA) was undertaken to inform policy-makers in New Zealand about the merits of the BAC8to5. The analysis indicated that the policy would produce a net benefit to the economy, with a net present value of NZ$200M. 19

As well as there being a lack of studies in general, to our knowledge no study has concurrently measured effectiveness and cost-effectiveness using the same data sources.

Association between blood alcohol concentration and population drinking

In a National Institute for Health and Care Excellence (NICE) review of the effectiveness of laws to limit BAC, a conceptual framework was developed to set out the rationale as to why a reduction of BAC from 0.08 to 0.05 g/dl could work. A logic model was developed, based on a set of assumptions about the links between the BAC intervention, change in drink-driving behaviours and risk of RTAs. 20 Interestingly, this work emphasised not only the importance of the perception of risk of being detected and punished but also the potential of a BAC intervention to reduce overall alcohol consumption.

It has been shown that population drinking (i.e. alcohol consumption per capita) is positively associated with a large number of outcomes (e.g. liver cirrhosis, suicide, violence). 21 A recent study, by Norström and Rossow, hypothesised that population drinking would also be associated with driving while under the influence of alcohol. The study’s area-based analyses, using Norwegian and Swedish data, demonstrated a strong and statistically significant association; as alcohol consumption per capita increases or decreases, so does the incidence of driving under the influence of alcohol. 22 We are not aware of any studies that have evaluated whether or not a legal reduction in the BAC limit (for a whole country) has led to a reduction in that country's population drinking. This could be a wider, unintended, outcome that can be attributed to change in BAC legislation. This would be beneficial to public health because alcohol consumption per capita is positively associated with alcohol-related harms. 23

The change of the drink-drive limit in Scotland

The reduction of the BAC limit in Scotland from 0.08 to 0.05 g/dl on 5 December 2014 is the intervention under study in this report. Although there is an evidence base that establishing a 0.05 g/dl BAC limit, and changing the limit from 0.08 to 0.05 g/dl, is effective in reducing RTAs, the scientific quality of the studies is poor, and the best study,14 in our view, still had limitations with regard to the unit of time used and lack of adjustment for seasonal and other contextual effects.

No previous study has investigated whether or not the effectiveness of a change in BAC legislation varies by levels of SED. If, as the evidence suggests, that any intervention effect is observed across the entire population of drink drivers, then it would be expected that absolute levels of SED inequality would be reduced because of the different RTA prevalence across SED groups.

The study is also interested in whether or not a legal reduction in the BAC limit leads to a reduction in that country's population drinking. In early 2015, a Scottish newspaper, in separate articles, reported that pub operators were indicating that overall sales had been ‘hit’ by the change in drink-driving legislation,24 with representatives of the Scottish Licensed Trade Association (SLTA) saying that the change would be worse for business than the smoking ban,25 and the Bank of Scotland's chief economist saying that the change had stunted the growth of the licensed trade. 26 In addition, it was heard, anecdotally, of individuals stating that they had reduced their alcohol drinking consumption in the evening before driving the next morning. In this study, the intervention effect was measured for a population drinking outcome measure using high-quality market research off- and on-sales data with coverage across GB, the off- and on-sales data having been used in past evaluations of Scotland's alcohol strategy. These evaluations are carried out by the Monitoring and Evaluating Scotland's Alcohol Strategy (MESAS) team in NHS Health Scotland.

At the start of this research, in conjunction with MESAS, a theory of change was developed for the change in drink-drive limit in Scotland. This theory of change (Table 1) helped to contextualise and scope the research plan for this project.

There is a lack of economic evaluation studies in the BAC literature. In a review paper on cost-effective policies to reduce the harm caused by alcohol,17 a large number of assumptions had to be made, partly as a result of pulling together evidence from various sources. The cost-effectiveness analysis that is proposed in this study allows the joint distribution of costs and effects to be consistently measured, as the same data sources for measures of effectiveness and cost are being drawn on.

Research questions

The research answered four primary questions:

1. Has the change in drink-driving legislation in Scotland been effective (i.e. a reduction in RTAs)?

2. Has the change in drink-driving legislation in Scotland led to changes in the relative, and absolute, RTA rates and can these changes be correlated to the level of SED?

3. Has the change in drink-driving legislation in Scotland led to a reduction in population alcohol consumption?

4. Has the change in drink-driving legislation in Scotland provided good value for money (i.e. has the legislation been cost-effective)?

Introduction

In this chapter, the overall research design is summarised and then the methods used to answer each research question under study are described.

Research design

A natural experimental design was employed to measure the causal effect of the change in BAC legislation in Scotland from 0.08 to 0.05 g/dl on 5 December 2014 (i.e. the intervention, BAC8to5). The control group was England and Wales, that is, the countries in GB that still have a 0.08 g/dl BAC law. The data for the intervention and control groups used to measure effectiveness came from the same data sources and cover the same study period (4 years in duration, 2 years pre and post change in legislation, January 2013–December 2016). The study followed the Medical Research Council’s guidelines on best practice for conducting natural experiments27 and referred to the Strengthening the Reporting of OBservational studies in Epidemiology reporting guidelines for observational studies. 28

Research question 1: has the change in drink-driving legislation in Scotland been effective (i.e. a reduction in road traffic accidents)?

The approach to answering this question was to measure the percentage change in the weekly level of RTAs attributable to BAC8to5 by quantifying the change observed in pre- and post-intervention periods and contrasting these between intervention and control groups.

Research question 2: has the change in drink-driving legislation in Scotland led to changes in the relative, and absolute, road traffic accident rates and can these changes be correlated to the level of socioeconomic deprivation?

To answer this research question, the study followed the same methods as outlined for research question 1.

Research question 3: has the change in drink-driving legislation in Scotland led to a reduction in population alcohol consumption?

The approach to answering this question was to measure the percentage change in alcohol consumption per capita attributable to the BAC8to5 by quantifying the change observed in pre- and post-intervention periods, and contrasting between intervention and control groups.

Research question 4: has the change in drink-driving legislation in Scotland provided good value for money (i.e. has the legislation been cost-effective)?

The economic evaluation ran in parallel to the effectiveness analyses. Three frameworks for economic evaluation (cost-effectiveness, cost–utility and CBA) were considered to reflect different outcomes and, as such, were dependent on identification and measurement of differences in effectiveness. If differences in effectiveness were not detected, the economic evaluation would revert to a cost analysis.

Approach 1: use of traffic count data as they are (include low-quality data)

In this approach, the study made no attempt to ‘clean’ the data and just used them as they are, regardless of data quality (or missing data altogether). Figure 1 plots traffic flow over time using this approach, showing an increasing amount of traffic over time in Scotland but not in England and Wales.

FIGURE 1.

Traffic flow counts in (a) Scotland and (b) England and Wales: including missing values and poor-quality data. Note that the solid black line indicates a change in legislation date and the dashed black line indicates a pseudo-change in legislation date.

Approach 2: low-quality and missing values imputed by algorithm

In this approach, if a count was not flagged as ‘high quality’ it was recorded as missing. A missing value was imputed by the average of the previous and the following day. If this was not possible because either of those values were missing, the average of 2 days back and forward was used. For the remaining missing data, the algorithm moved to 7-day steps to account for any day-of-the-week effects. Figure 2 plots traffic flow over time using this approach, showing an increasing amount of traffic over time in Scotland, but a decrease in England and Wales.

FIGURE 2.

Traffic flow counts in (a) Scotland and (b) England and Wales: missing values and poor-quality data imputed by average values. Note that the solid black line indicates a change in legislation date and the dashed black line indicates a pseudo-change in legislation date.

Approach 3: low-quality and missing values imputed by multiple imputation method

In this approach, the study used multiple imputation to impute missing values for the traffic counts (as for approach 2, if the count was not flagged as ‘high quality’ it was recorded as missing). Using this model-based multiple imputation approach allows any patterns in the data to be reflected in the imputed data sets. The study used the ‘Amelia’ package in R (The R Foundation for Statistical Computing, Vienna, Austria)43 to carry out the imputations, as this routine is specifically designed for the imputation of time series data. The study created 20 imputed data sets for each region in Scotland and in England and Wales, using spline terms with up to four knots to model the underlying time trend. The data sets were then aggregated, and a mean value calculated for each missing data point prior to use in the main statistical analysis. Figure 3 plots traffic flow over time using this approach, showing an increasing amount of traffic over time in Scotland and in England and Wales.

FIGURE 3.

Traffic flow counts in (a) Scotland and (b) England and Wales: missing values and poor-quality data imputed by the multiple imputation method. Note that the solid black line indicates a change in legislation date and the dashed black line indicates a pseudo-change in legislation date. W, week.

The study considered this third approach to be the best, as it applies the most statistically robust approach to low-quality/missing data. Furthermore, the general increasing trend in traffic flow over time that results (see Figure 3) fits with DfT reporting. 29 Therefore, the study used these estimates as traffic flow denominators for RTA rates in the statistical analyses (see Chapter 5).

Introduction

In this chapter, the reporting of the results is structured by the following four research questions:

1. Has the change in drink-driving legislation in Scotland been effective (i.e. a reduction in RTAs)?

2. Has the change in drink-driving legislation in Scotland led to changes in the relative, and absolute, RTA rates and can these changes be correlated to the level of SED?

3. Has the change in drink-driving legislation in Scotland led to a reduction in population alcohol consumption?

4. Has the change in drink-driving legislation in Scotland provided good value for money (i.e. has the legislation been cost-effective)?

Research question 1: has the change in drink-driving legislation in Scotland been effective (i.e. a reduction in road traffic accidents)?

Plots of the time series of the outcome measure, with a vertical line superimposed to denote the time when the BAC8to5 started, are shown in Figure 4. It can be seen, in general, that weekly RTA rates are higher in England and Wales than in Scotland. This is true both for rates calculated with traffic flow (see Figure 4b) and for population (see Figure 4c) denominators. For example, broadly, the rates are between 5 and 9 RTAs per 1000 traffic count in Scotland and in the corresponding figures the rates are between 6 and 10 for England and Wales (see Figure 4b). From visual inspection of the plots for Scotland, no clear interruption of the time series (i.e. a BAC8to5 effect) is apparent after the change in drink-drive legislation. In the pre-intervention periods, the level of RTAs in Scotland was stable but for England and Wales an increase was observed.

FIGURE 4.

Weekly RTA counts for (a) Scotland and (b) England and Wales, RTA rates with traffic flow as the denominator for (c) Scotland and (d) England and Wales, RTA rates with population as the denominator for (e) Scotland and (f) England and Wales between January 2013 and December 2016. Note that the solid black line indicates a change in legislation date and the dashed black line indicates a pseudo-change in legislation date. W, week.

In Table 2, the age, sex and SED demographics of the RTAs based on the demographics of the drivers involved in the accidents are described. Although it can be seen that the demographic distributions are very similar between the intervention and control groups (also the case when using a different demographic assignment; see Table 8), this is reassuring in terms of having chosen an appropriate control group; however, it would be erroneous to infer anything about the demographics of those involved in RTAs using these statistics. To assist in this regard, further descriptive analysis shows that the mean [standard deviation (SD)] number of drivers per RTA was 1.72 (SD 0.73) in Scotland and 1.84 (SD 0.71) in England and Wales. The corresponding figures for the number of casualties per RTA was 1.29 (SD 0.77) for Scotland and 1.33 (SD 0.82) for England and Wales.

The modelling results are shown in Table 3. The change in drink-drive legislation was associated with a 2% relative decrease in RTA counts in Scotland [relative risk (RR) 0.98, 95% CI 0.91 to 1.04; p = 0.53]. However, the pseudo-change in legislation was associated with a 5% decrease in RTA counts in England and Wales (RR 0.95, 95% CI 0.90 to 1.00; p = 0.05). Similar results were observed when modelling RTA rates, both with the traffic flow denominator and with the population denominator. For RTA rates with traffic flow as the denominator, the DiD-type estimate indicates a 7% increase in rates for Scotland relative to England and Wales (unadjusted RR 1.07, 95% CI 0.98 to 1.17; p = 0.1). Adjustment for age, sex and SED changed these results minimally, as would be expected given the comparability of the demographics described in the last paragraph. Appendix 1, Table 9, shows the corresponding table (for models d and f only) based on a different demographic assignment: youngest age, less frequent sex and least deprived SED group. This did not substantially change the modelling results.

Research question 2: has the change in drink-driving legislation in Scotland led to changes in the relative, and absolute, road traffic accident rates and can these changes be correlated to the level of socioeconomic deprivation?

Table 4 shows the modelled estimates of the BAC8to5 effect size by quintiles of SED, which have been obtained from models fitted with the appropriate interaction terms. To illustrate, for RTAs in which the postcode of the driver indicates the highest level (20%) of SED in Scotland, the change in drink-drive legislation was associated with a 0% relative decrease in RTA counts in Scotland (RR 1.00, 95% CI 0.92 to 1.10; p = 0.93). As can be seen by scanning the results in Table 4, for both the intervention and the control group, effect sizes were relatively homogeneous across SED levels and this is reinforced by the p-values for the tests of interactions (p = 0.72, RTA counts; p = 0.71, RTA rates with traffic flow as the denominator; and p = 0.72, RTA rates with population as the denominator). For England and Wales (pseudo-intervention), the corresponding p-values are p = 0.58 for RTA counts, p = 0.58 for RTA rates with traffic flow as the denominator and p = 0.59 RTA rates with population as the denominator. Appendix 1, Table 10, shows the corresponding table based on assigning the RTA with demographics of the driver based on the following rules: youngest age, less frequent sex and least deprived SED group. This did not substantially change the modelling results.

Research question 3: has the change in drink-driving legislation in Scotland led to a reduction in population alcohol consumption?

Plots of the time series of the outcome measure, with a vertical line superimposed to denote the time when the BAC8to5 started, are shown in Figure 5. From visual inspection of the plots for Scotland, no clear interruption of the time series (i.e. a BAC8to5 effect) is apparent after the change in drink-drive legislation. Strong seasonal patterning is shown with large peaks and smaller troughs at the end and start of a calendar year, respectively.

FIGURE 5.

Weekly (a) off-trade (Scotland), (b) off-trade (England and Wales), (c) on-trade (Scotland) and (d) on-trade (England and Wales) alcohol sales between January 2013 and December 2016. Note that the solid black line indicates a change in legislation date and the dashed black line indicates a pseudo-change in legislation date) (a) off-trade alcohol sales (Scotland), (b) off-trade alcohol sales (England and Wales), (c) on-trade alcohol sales (Scotland) and (d) on-trade alcohol sales (England and Wales), W, week.

The modelling results are shown in Table 5. The change in legislation was associated with a 0.3% relative decrease in per capita off-trade sales (–0.3%, 95% CI –1.7% to 1.1%; p = 0.71) and a 0.7% decrease in per capita on-trade sales (–0.7%, 95% CI –0.8% to –0.5%; p < 0.001). The corresponding results for the effect of the pseudo-change in legislation in England and Wales indicate increases in per capita off- and on-trade sales. The statistical testing provided no evidence against the null hypothesis that the residuals were normally distributed and, therefore, the SARIMA models fit the data well. Furthermore, adjustment for a covariate reflecting country-specific Aldi and Lidl market-share percentages marginally changed only the effect sizes reported for on-trade sales outcomes.

Research question 4: has the change in drink-driving legislation in Scotland provided good value for money (i.e. has the legislation been cost-effective)?

With the results from the earlier analyses revealing no significant change in effectiveness, the economic evaluation proceeded as a cost analysis focusing on the resource impacts of the legislation.

Resource data were obtained from the Scottish Government (SG) on the financial impact of the change in drink-drive legislation. The data could be classified as legislation and programme management costs and publicity costs. Law enforcement details and costs from Police Scotland were not publicly accessible. The details of the costing elements are further broken down in Figure 6.

FIGURE 6.

Details of programmatic-level costs of changing drink-drive legislation.

Law enforcement costs

Law enforcement costs include costs involving police officers, vehicles and roadside equipment used in the enforcement process. 18 However, data on enforcement costs associated with the legislation are not centrally held and require separate negotiation with Police Scotland. However, RoSPA46 provided suggestions for resources required to enforce drink-driving legislation, and these are summarised in Figure 7.

FIGURE 7.

Suggested law enforcement costs from RoSPA (summarised from The Royal Society for the Prevention of Accidents’ Road Safety Factsheet46). a, https://news.gov.scot/news/drink-drive-campaign-success (accessed 24 May 2019); and b, https://roadsafety.scot/topics/drink-driving/ (accessed 24 May 2019).

Summary of principal findings

• The study found that lowering the BAC limit from 0.08 to 0.05 g/dl in Scotland was not associated with a change in the level of RTAs in the first 2 years post change in legislation.

• As well as no overall effect for RTA outcome, no effect modification by SED level was found.

• The study found that lowering the BAC limit from 0.08 to 0.05 g/dl in Scotland was not associated with a change in the level of off-trade alcohol consumption in the first 2 years post change in legislation.

• The study found that lowering the BAC limit from 0.08 to 0.05 g/dl in Scotland was associated with a small relative reduction (< 1%) in the level of on-trade alcohol consumption in the first 2 years post change in legislation.

Findings in context of literature and theory of change

The RTA finding was unexpected, given that the existing evidence base (as summarised in Chapter 1) supported a reduction in RTAs and associated mortality outcomes. A high-quality study14 found a reduction in total death rates between 4.5% and 7.4% (depending on the choice of denominator). Although the outcome is all RTAs and not just those that resulted in deaths, the 95% CIs of the results (see Table 3) are not congruent with effect sizes of that magnitude. Moreover, when the results are expressed relative to the concurrent control group (England and Wales), increases in rates of RTAs are observed.

Referring to the theory of change (see Table 1), which was developed a priori to this research being undertaken, legislation failure was put forward as an explanation if no change in the RTA outcome was found. The legislation could have failed because of the lack of enforcement; a previous study found a significant negative association between enforcement and RTA rates. 9 The European commission stated that a key to the success of changing drink-drive legislation is the accompanying enforcement (i.e. frequent and systematic RBT), supported by public education, publicity and awareness campaigns involving all stakeholders. 51 English police force data have shown that there were 25% fewer RBTs in 2015 compared with 2011. 52 Findings from a parallel qualitative evaluation of the change in drink-drive legislation in Scotland show that an initial substantial investment in public education and media campaigning at the time of the limit reduction in December 2014 was not maintained in 2015 and 2016 (Dr Niamh Fitzgerald, Institute for Social Marketing, University of Stirling, 2018, personal communication). Other plausible reasons for legislation failure are if the majority of drink-driving RTAs are caused by people who continue to ignore the law under the 2014 legislation, or if drink-driving RTAs are caused by people who used to drink-drive within a BAC of 0.05–0.08 g/dl under the old legislation (but obey the 2014 legislation) are a small fraction of all RTAs.

To the study authors’ knowledge, no previous study has evaluated whether or not a legislation change of the BAC limit has led to a reduction in that country's population drinking (and it should be pointed out that the SG never suggested that it would). For off-trade alcohol sales, no association was found, and it is worth emphasising that off-trade sales account for a large proportion of alcohol consumption in Scotland (e.g. approximately 70% of total alcohol sales in 2016). The study observed a small reduction in on-trade alcohol sales; this is in line with findings from the aforementioned qualitative study with publicans/on-licensees reporting small impacts on their businesses. It is worth noting that this small reduction is contrary to early post change in legislation newspaper reporting24 and SLTA viewpoints,25 which suggested larger reductions would be found. In summary, the concern of drinking heavily the night before driving in the morning (which was expressed in the theory of change and was discussed by participants of the public involvement group meeting), and any other mechanisms, have not changed Scotland's alcohol drinking levels.

With the study’s results showing no significant change in effectiveness, the economic evaluation proceeded as a cost analysis focusing on the resource impacts of the legislation. The cost results show that the financial costs of changing the drink-driving legislation in Scotland were not insubstantial. These resources need to be considered as one component of the wider portfolio of interventions (e.g. minimum unit pricing) to change Scotland's relationship with alcohol. As such, although a traditional micro-economic evaluation provides evidence relating to the costs/cost-effectiveness of the legislation to reduce drink driving, the short- and long-term macro effects of the complete portfolio of interventions are less amenable to formal identification, measurement and valuation.

Challenges, strengths and limitations of the study

Conclusion

The findings indicate that the change in drink-drive legislation in Scotland in December 2014 did not have the expected effect of reducing RTAs in the country, and nor did it change Scotland's alcohol drinking levels.

Wider contributions

Sandy Allan (RoSPA) participated in the steering group meetings and reviewed this final report.

Clare Beeston (NHS Health Scotland) assisted in developing the theory of change.

Denise Brown (Institute of Health and Wellbeing, University of Glasgow), Maya Clayton (NRS), Lauren Schofield (NHS National Services Scotland) and David Walsh (Glasgow Centre for Population Health) all assisted by providing expertise and resources to allow the appropriate measurement of SED for this study.

Niamh Fitzgerald (Institute for Social Marketing, University of Stirling) is principal investigator of a parallel-qualitative evaluation of the change in drink-drive legislation in Scotland (funded by the Chief Scientist Office – HIPS/16/49) and she participated in a steering group meeting and reviewed the discussion of this final report. Furthermore, Niamh facilitated and took part in the public involvement group meeting.

Behnom Havaei-Ahary (Road Traffic Statistics, DfT) assisted in obtaining and providing guidance on the STATS19 data set.

Mark Robinson (NHS Health Scotland) participated in the steering group meetings and reviewed this final report.

Contributions of authors

Jim Lewsey (Reader in Medical Statistics, Institute of Health and Wellbeing, University of Glasgow) led the design and execution of the study, oversaw all analyses and led the preparation of the final report.

Houra Haghpanahan (Research Assistant, Institute of Health and Wellbeing, University of Glasgow) was the lead researcher on the study, leading the preparation of all analyses and contributed to all sections of the final report.

Daniel Mackay (Reader in Public Health, Institute of Health and Wellbeing, University of Glasgow) contributed to the overall design, in particular the design and analysis of all statistical analyses.

Emma McIntosh (Professor of Health Economics, Institute of Health and Wellbeing, University of Glasgow) led the design of, and supervised the, economic evaluation analyses and reviewed and commented on the final report.

Jill Pell (Henry Mechan Professor of Public Health, Institute of Health and Wellbeing, University of Glasgow) contributed to the design of the study.

Andy Jones (Professor in Public Health, Norwich Medical School, University of East Anglia) contributed to the design, provided advice on some of the statistical analyses, and reviewed and commented on the final report.

All authors contributed to the design of the final report and approved the final version.

Publication

Haghpanaha H, Lewsey J, Mackay D, McIntosh E, Pell J, Jones A. An evaluation of the effects of lowering blood alcohol concentration limits for drivers on the rates of road traffic accidents and alcohol consumption: a natural experiment. Lancet 2018;393:321–9.

Data-sharing statement

All data requests should be submitted to the corresponding author for consideration. Access to available anonymised data may be granted following review.

Disclaimers

This report presents independent research funded by the National Institute for Health Research (NIHR). The views and opinions expressed by authors in this publication are those of the authors and do not necessarily reflect those of the NHS, the NIHR, NETSCC, the PHR programme or the Department of Health and Social Care. If there are verbatim quotations included in this publication the views and opinions expressed by the interviewees are those of the interviewees and do not necessarily reflect those of the authors, those of the NHS, the NIHR, NETSCC, the PHR programme or the Department of Health and Social Care.