RESEARCH AND PRACTICE

Reduction in Fatalities, Ambulance Calls, and Hospital Admissions for Road Trauma After Implementation of New Traffic Laws

Jeffrey R. Brubacher, MD, MSc, Herbert Chan, PhD, Penelope Brasher, PhD, Shannon Erdelyi, BSc, Edi Desapriya, PhD, Mark Asbridge, PhD, Roy Purssell, MD, Scott Macdonald, PhD, Nadine Schuurman, PhD, and Ian Pike, PhD

Motor vehicle crashes (MVCs) are a major Objectives. We evaluated the public health benefits of traffic laws targeting public health problem that disproportionately speeding and drunk drivers (British Columbia, Canada, September 2010). affects youths and young adults.1 Worldwide, Methods. We studied fatal crashes and ambulance dispatches and hospital more than 3300 people per day are killed in admissions for road trauma, using interrupted time series with multiple non- 2 road trauma, and many more are disabled. equivalent comparison series. We determined estimates of effect using linear Crashes are caused by numerous factors, in- regression models incorporating an autoregressive integrated moving average cluding faulty vehicles and poor road design, error term. We used neighboring jurisdictions (Alberta, Saskatchewan, Washington but the majority are attributed to driver-related State) as external controls. factors, especially alcohol-impaired driving,3,4 Results. In the 2 years after implementation of the new laws, significant speeding,5,6 and driver distraction.7,8 In Canada, decreases occurred in fatal crashes (21.0%; 95% confidence interval [CI] = 15.3, driving with a blood alcohol concentration 26.4) and in hospital admissions (8.0%; 95% CI = 0.6, 14.9) and ambulance calls (BAC) of 0.08% or more is a federal criminal (7.2%; 95% CI = 1.1, 13.0) for road trauma. We found a very large reduction in alcohol-related fatal crashes (52.0%; 95% CI = 34.5, 69.5), and the benefits of offense. Most provinces have regulations that the new laws are likely primarily the result of a reduction in drinking and allow for administrative sanctions (fines or driving. license suspension) for drivers with a BAC of Conclusions. These findings suggest that laws calling for immediate sanctions 0.05% or more. Speed limits are set by the for dangerous drivers can reduce road trauma and should be supported. (Am J provinces and enforced through administrative Public Health. 2014;104:e89–e97. doi:10.2105/AJPH.2014.302068) sanctions. Before September 2010, drunk drivers in British Columbia were subject to fines and which add to but do not replace the older laws, and for speeding and related offenses under administrative driving prohibitions of 24 hours drivers with a BAC of 0.05% or more may the old versus new traffic laws. for a BAC of 0.05% or more and of 90 days for be subject to license suspension of 3, 7, or 30 British Columbia’s laws were designed to a BAC of 0.08% or more or failure to provide days; vehicle impoundment of 3, 7, or 30 days deter drinking and driving by increasing the a breath sample. The 90-day driving ban came (discretionary with 3- or 7-day license suspen- severity, certainty, and speed of punishment for into effect after a 21-day period that allowed sions, mandatory for 30-day license suspen- drunk drivers with a BAC of 0.05% or more. the driver to find alternative means of trans- sions); and possible referral to remedial programs Deterrence is effective in laws against traffic portation or to appeal the decision. There were (including ignition interlock) for third-time violations,11 and increased likelihood of pun- no vehicle impoundments for drunk drivers. In offenders. Drivers with a BAC of 0.08% or ishment acts as a greater deterrent than in- addition to administrative penalties, drivers more are subject to 30-day vehicle impound- creased severity.12,13 The speed with which with a BAC of 0.08% or more and those who ment, 90-day license suspension, and possible a punishment is applied is an additional de- refused to provide a breath sample were sub- referral to remedial programs (including igni- terrent.14 For drivers with a BAC of 0.08% or ject to possible penalties under the Criminal tion interlock). In addition, drunk drivers are more, the administrative sanctions under the Code of Canada. Before September 2010, the subject to fees that include towing costs, vehicle new laws are actually less severe than those penalty for street racing was vehicle impound- storage costs, and a processing fee to have their associated with a criminal code conviction for ment for 2 days for a 1st offense and for 30 license reinstated. License prohibitions and drunk driving. However, the new sanctions are days for 2nd and subsequent offenses in addi- vehicle impoundments are issued immediately applied immediately at roadside and are less tion to fines and possible criminal code charges. at the roadside on the basis of results of subject to legal challenge, making punishment In September 2010, British Columbia in- a handheld breathalyzer. The penalties for more swift and certain. Not all drunk drivers troduced harsher penalties for drunk drivers excessive speeding, street racing, and stunt are deterred by threat of punishment,15 but and for excessive speeding (> 40 km/h over driving include fees and mandatory vehicle through vehicle impoundment and license the speed limit) as well as for drivers caught impoundment of 7 to 60 days.9,10 Table 1 suspension, British Columbia’s new laws re- street racing or stunt driving. Under these laws, summarizes penalties for drinking and driving move more dangerous drivers from the roads

October 2014, Vol 104, No. 10 | American Journal of Public Health Brubacher et al. | Peer Reviewed | Research and Practice | e89 RESEARCH AND PRACTICE

TABLE 1—Comparison of British Columbia’s Old and New Speeding and Street Racing Laws

Penalties Driving Offense Before September 2010a After September 2010b,c

BAC level 0.0–0.049 12-h driving prohibition for drivers with restricted license 0.05–0.079 (1st offense) 24-h driving prohibition 3-d driving prohibition Possible 3-d vehicle impoundment Fees (£ $600) 0.05–0.079 (2nd offense within 5 y) 24-h driving prohibition 7-d driving prohibition Possible 7-d vehicle impoundment Fees (£ $780) 0.05–0.079 (3rd offense within 5 y)d 24-h driving prohibition 30-d driving prohibition 30-d vehicle impoundment Fees (£ $1330) ‡ 0.08 or refuses breath testd 90-d driving prohibition beginning 21 d after the offense 90-d driving prohibition (immediate) (this 21-d period was to allow driver to arrange alternate 30-d vehicle impoundment transportation or to appeal the decision) Fees (£ $1450) Possible Criminal Code of Canada charges Speeding and stunt driving Speeding £ 40 km over limit Fines and demerit points No change Speeding > 40 km per hour or stunt drivinge Fines and demerit points 7-d vehicle impoundment (first offense) Fees £ $210 Speeding > 40 km per hour or stunt drivinge Fines and demerit points 30-d vehicle impoundment (second offense within 2 y) Fees £ $700 Speeding > 40 km per hour or stunt drivinge Fines and demerit points 60-d vehicle impoundment (third offense within 2 y) Fees £ $1200 Street racing 1st offense Fines and demerit points (street racing only) 7-d vehicle impoundment 48-h vehicle impoundment Fees £ $210 2nd offense within 2 y Fines and demerit points (street racing only) 30-d vehicle impoundment 30-d vehicle impoundment Fees £ $700 3rd and subsequent offense within 2 y Fines and demerit points (street racing only) 60-d vehicle impoundment 30-d vehicle impoundment Fees $1200

Note. BAC = blood alcohol concentration; IRP = immediate roadside prohibitions. aAll penalties that existed before September 2010 remain in effect. Starting September 2010, IRPs provided an additional tool for police to use at their discretion. bSince September 2010, police usually apply immediate penalties under the new laws. However, they still have the option of using the preexisting laws instead. Therefore, the 3 typical outcomes for drivers with BAC ‡ 0.05% or who refuse a breath test are as follows: (1) most get an IRP, (2) some get a 24-h (or 12-h) driving prohibition, and (3) some get a 90-d driving prohibition plus criminal code charges. cNote that the penalties for excessive speeding (> 40 km over limit), street racing, and stunt driving are identical under the new laws. dDrivers with 3 offenses for BAC = 0.05%–0.079% or 1 offense for BAC > 0.08% may be referred to the responsible driver program, the ignition interlock program, or both. eThe new laws expanded the vehicle impound program to introduce and define “stunt driving,” something that was not specifically covered under the old laws.

quicker. High-risk drunk drivers can also be measures16 and that public perception of police 5 years preceding the new laws to 1853 in the mandated to undergo therapy, install an alco- commitment to traffic enforcement had in- subsequent 12 months (Figure1). The new laws hol interlock device on their vehicle, or both. creased.17 During the first 12 months, British were associated with change in driver behav- These laws were promoted through public Columbia police used the new laws to issue ior. In the 2012 British Columbia roadside awareness and educational campaigns and re- 22 734 roadside license prohibitions for surveys, 6.5% of evening and nighttime drivers ceived considerable media coverage. Opinion drinking and driving. At the same time, the tested positive for BAC compared with 9.9% in and roadside surveys in 2012 found that number of criminal code convictions for im- 2010 before the new laws were introduced.16 majority of drivers were aware of the new paired driving fell from 8221 per year in the Telephone surveys found that the percentage

e90 | Research and Practice | Peer Reviewed | Brubacher et al. American Journal of Public Health | October 2014, Vol 104, No. 10 RESEARCH AND PRACTICE

of British Columbia drivers reporting driving In this study, we investigated the effects of the research ethics board. Briefly, we used an while over the legal limit was above the new laws on total fatal crashes, hospital ad- interrupted time-series approach to model the national average in 2009 and 2010 but fell missions, and ambulance calls. We drew on impact of the new laws on crashes and medical below the national average in 2011and 2012.18 police, ambulance, and hospital data from events. We assessed the intervention effects The new laws included harsher penalties for British Columbia and on police data from using a linear regression model with an autor- excessive speeding but, unlike the case with adjacent jurisdictions in which similar legisla- egressive integrated moving average (ARIMA) drinking and driving, the number of sanctions tive changes were not implemented. error term. for speeding or for excessive speeding did not increase (Figure 1). METHODS Data Sources We previously reported a reduction in We analyzed British Columbia data from alcohol-related fatal crashes in the 2 years after We used deidentified administrative data MVC fatalities from police reports (2000--- introduction of British Columbia’s new laws.19 and obtained approval from our institutional 2012), hospital admissions (2005---2011), and ambulance calls (2004---2012). We obtained a MVC fatality data from Alberta, Saskatchewan, Intervention and Washington State police reports (2000--- Criminal code 2012). We did not have access to hospitaliza- 90-day ADPs 1500 90-day IRPs tion or ambulance data from any jurisdiction 3-, 7-, or 30-day IRPs other than British Columbia. 1000 Road fatalities. The British Columbia traffic 500 accident system contains details of all police- reported crashes in British Columbia, including 0 factors that police believe contributed to the 2008 2009 2010 2011 crash (e.g., speeding, impaired driving). The Time traffic accident system is reconciled with coro- ’ b Intervention ners data to ensure that it captures all fatal crashes (i.e., those who died on a public road and within 30 days of the crash). Nonfatal 20 000 crashes may not be captured because police 15 000 attendance at nonfatal crashes is discretionary. For this reason, we studied only fatal crashes.

Charges, No. 10 000 Fatal crashes related to alcohol or speeding were our outcome of interest, and fatal crashes 2008 2009 2010 2011 unrelated to alcohol or speeding served as an Time internal control. c Similarly, we obtained monthly counts of total Intervention fatal crashes and of fatal crashes involving 1200 alcohol or speeding from police reports of 1000 neighboring jurisdictions (Alberta, Saskatchewan, 800 and Washington State). We used the analyses of 600 these data as an external comparison to account

Charges, No.400 Charges, No. for regional trends and conditions that might explain the observed effects of the new laws in 2008 2009 2010 2011 British Columbia. Time Hospital admissions and ambulance dispatches. All hospital admissions in Canada Note . ADPs = administrative driving prohibitions; IRPs = immediate roadside prohibitions. In part (a), an abrupt decrease are recorded in the Discharge Abstract Database, occurred in the number of criminal code charges and 90-day ADPs for impaired driving after the introduction of the new laws which includes mechanism of injury but has no accompanied by an even greater increase in the number of IRPs issued under the new laws (note that, under the old laws, drivers with criminal code charges were typically also given a 90-day driving prohibition). In parts (b) and (c), the new information on factors that caused the crash. In laws were not associated with a change in the number of charges for speeding or for excessive speeding. However, under the this study, we used admissions for road trauma as new laws, excessive speeding resulted in mandatory vehicle impoundment for 7 days. a measure of hospital utilization by road injury FIGURE 1—Monthly charges for (a) impaired driving, (b) speeding, and (c) excessive victims in British Columbia. We defined “road speeding: British Columbia, September 2007–September 2011. trauma admissions” as those with International Classification of Diseases, 10th Revision (ICD-10),20

October 2014, Vol 104, No. 10 | American Journal of Public Health Brubacher et al. | Peer Reviewed | Research and Practice | e91 RESEARCH AND PRACTICE

external cause of injury codes ranging from factor for any driver involved in the crash. fit was deemed inadequate, we used residual V02.1 (pedestrian injured in collision with 2- Crashes with multiple contributory factors were diagnostics to identify improved error struc- or 3-wheeled motor vehicle, traffic accident) included in all relevant analyses. We conducted tures. We selected the model with the lowest through V89.9 (person injured in unspecified subgroup analyses that excluded crashes in- Akaike information criterion that also satisfied vehicle accident). We also included sequelae of volving both alcohol and speeding. We ex- residual assumptions. The error model was transport accidents (Y85, Y850, Y959). cluded cases in which relevant contributory not constrained to be the same across jurisdic- We excluded non---traffic accident events factors were cited only for pedestrians or pedal tions or outcomes. such as boarding or alighting from a vehicle, cyclists. We fitted all models using conditional-sum- events that occurred on industrial premises, We analyzed event rates using an interrup- of-squares to find starting values. Final esti- events that did not involve a motorized vehicle, ted time-series approach. We fitted a multivar- mates were obtained by maximum likelihood. and events in which the only vehicle involved iate linear regression model that included an We performed all regression analyses with R was a streetcar, railway vehicle, specialized indicator variable for the intervention (new version 3.0.1 (R Foundation for Statistical agricultural vehicle, airplane, or watercraft. We law) effect, an internal control variable, and Computing, Vienna, Austria). The intervention included traffic injury involving snowmobiles and a seasonal ARIMA error term to control for effect was treated as abrupt and permanent. all-terrain vehicles because these vehicles are seasonality, autocorrelation, and data instabil- We considered P < .05 to be statistically sig- subject to trafficlawswhendrivenonapublic ity. Internal controls served as a surrogate for nificant; all tests were 2-sided. road. We used hospital admissions for non-MVC unmeasured confounders and trends unrelated trauma (i.e., all other injury codes) as controls. to the intervention. Controls were scaled and RESULTS The British Columbia Ambulance Service centered for model convergence. We originally provides emergency medical service for the included the effects of other events, such as Between January 2000 and December entire province. Each call for emergency med- a court challenge and temporary suspension of 2012, 4507 fatal crashes occurred in British ical assistance is assigned 1 of 32 Advanced the new laws (November 2011),22 in the re- Columbia, including 1121 related to alcohol Medical Priority Dispatch System numbers gression model. However, these events did not and 1600 related to speeding (163 to exces- using standardized questions. Multiple calls for have a statistically significant effect on crash sive speed). In 589 fatal crashes, both alcohol the same event are combined into a single rates, and we omitted them from the final and speeding were involved. We found overall dispatch record. We used ambulance dis- model. downward trends and seasonal patterns with patches for traffic and transportation incidents To complement ARIMA analyses, which peaks during summer months for both alcohol- (code 29) as a measure of ambulance service estimate permanent changes in mean crash and speeding-related fatal crashes. Figure 2 use by road injury victims. We used ambulance rate, we performed supplementary analyses shows monthly rates of fatal alcohol-related calls for non-MVC trauma as controls. to identify breakpoints (significant changes in crashes over the study period. the slope of the crash rate trendline) and to Table 2 summarizes the changes in occur- Analysis estimate the year in which any detected rence rates for each outcome after the new laws For each outcome, we aggregated counts by breakpoint occurred. To do this, we fitted went into effect. In the 24 months after British month and converted them to rates. For regression models to annual crash rates with Columbia’s new laws were implemented, sta- crashes, rates were per 1 000 000 licensed a segmented relationship for the yearly trend. tistically significant reductions occurred in total drivers. Hospital admissions and ambulance The number of breakpoints was automatically fatal crashes and in fatal crashes related to calls included victims of all ages and were selected but constrained to be less than 3.23,24 alcohol and to speeding. Segmented regression converted to rates per 1 000 000 British Data transformations were made to meet the analyses detected 1 breakpoint in the annual Columbia residents. We adjusted rates to ac- assumptions of the model. First, we took the trend for alcohol-related fatal crash rates in count for the number of days in each month. logarithm of crash rates to stabilize the variance 2010 (95% confidence interval [CI] = 2009, We chose the time before intervention a priori to when necessary. Second, we used differencing 2011). Breakpoints for speeding-related and accommodate data availability and to exceed the to achieve a stationary time series. If necessary, all-cause fatal crash rates occurred in 2005, recommended minimum number of observa- we calculated the difference from one period to before the new laws. tions required for time-series analyses.21 the next to remove trends over time. We also In subgroup analyses that excluded crashes For the purpose of this study, we considered used seasonal differencing to eliminate any involving both alcohol and speeding, we fatal crashes to be alcohol related if police cited strong seasonal patterns. found a 46.5% reduction (95% CI = 23.1, alcohol involvement, alcohol suspected, or An iterative procedure was used to identify 70.0) in alcohol-related crashes and a non- ability impaired by alcohol as a contributory the order of the ARIMA error model. We statistically significant reduction of 21.1% factor for any driver of a motorized vehicle selected the initial model using a stepwise (95% CI = –4.4, 46.7) in speeding-related involved in the crash. We deemed crashes to be algorithm to minimize the corrected Akaike crashes. During the same period, we ob- speeding related if police cited exceeding the information criterion. We explored correlation served a smaller reduction in alcohol-related speed limit, excessive speed, driving too fast for structures via plots of the autocorrelation and fatal crashes in Alberta (32.7%) and much conditions, or unsafe speed as a contributory partial autocorrelation functions. If the model smaller reductions in Saskatchewan and

e92 | Research and Practice | Peer Reviewed | Brubacher et al. American Journal of Public Health | October 2014, Vol 104, No. 10 RESEARCH AND PRACTICE

Intervention the exception of a probable spillover effect in Alberta, found no evidence of a reduction in 11 Fatal alcohol Fatal non−alcohol and non−speeding total fatal crashes. 10 Another possible explanation for our find- 9 ings is that changes in reporting practice oc- curred that coincided with implementation of 8 the new traffic laws. Police reports, in particu- 7 lar, may be affected by changes in reporting 6 practice because police do not report all 25,26 fi 5 crashes, and police classi cation of the 000 Licensed Drivers cause of crashes may be somewhat subjective. 4 However, we found no evidence of a change in 000

3 police reporting practice around the time of this 2 study. Moreover, we studied only fatal crashes, Rate/1 which police are obligated to investigate and 1 which are reconciled with coroners’ data to 0 ensure no cases are missed. In addition, we 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 found reductions in outcomes reported from 3 Time independent sources (police, ambulance, and Note. The vertical line indicates the first full month of implementation of the new laws in October 2010 (intervention hospital). Ambulance dispatch reports and effect = –1.41 fatal alcohol crashes/1 000 000 licensed drivers). The dotted curves represent the smoothed annual hospital admissions for road trauma capture average of monthly rates from the fitted model for fatal alcohol crashes (dark curve) and from observed rates for fatal all events that are correctly classified, and we nonalcohol and nonspeeding crashes (light curve). found no evidence of a change in reporting FIGURE 2—Monthly alcohol-related fatal crash rate per 1 000 000 licensed drivers in British of these events. In particular, the Advanced Columbia: January 2000–December 2012. Medical Priority Dispatch System codes used to classify ambulance dispatches and the ICD-10, injury codes for hospital admissions had not Washington State (1.2% and 9.8%, respec- fatal crashes per 1 000 000 drivers and 5.71 changed during the time of the study. Consid- tively); the latter 2 were not statistically hospital admissions and 47.29 ambulance ering all these factors, our findings are highly significant. Similarly, fatal crashes related to dispatches per 1 000 000 residents. Our find- unlikely to be explained by changes in report- speeding decreased by 30.6% (95% CI = ings of reductions not only in fatal crashes but ing practice. 17.0, 41.9) in Washington State. We found also in ambulance calls and hospital admissions Other researchers have also found that a smaller, nonsignificant decrease in fatal for road trauma strengthen the conclusion that both criminal and administrative traffic laws speeding crashes in Alberta (14.4%) and British Columbia’s laws were effective in re- reduce drunk driving fatalities. Asbridge et al. a trivial, non---statistically significant increase ducing road trauma. An alternate explanation studied drunk driving fatalities and found an in Saskatchewan (2.5%). is that these benefits are attributable to factors 18% reduction after Canada criminalized driv- During the study period, British Columbia other than the new laws. We feel that this is ing with a BAC of 0.08% or more in 196627 had 3553 hospital admissions and 32 225 unlikely for several reasons. First, we used and a 14.5% reduction after Ontario imple- ambulance calls for road trauma per year. Monthly internal controls to account for trends in mented license suspensions for drivers with rates of road trauma---related hospital admis- need for trauma care or the number of fatal a BAC of 0.08% or more.28 In a study of US sions and ambulance calls are illustrated in crashes. Second, with the exception of fines for state laws, Wagenaar and Maldonado-Molina29 Figures 3 and 4. We found an 8.0% (95% CI = drivers using cell phones (February 2010), found that administrative license suspension 0.6, 14.9) reduction in hospital admissions and British Columbia did not introduce any other reduced alcohol-related fatal crash involvement a 7.2% (95% CI = 1.1, 13.0) reduction in significant traffic laws or policy changes in by 5%, whereas postconviction policies (e.g., ambulance calls attributable to road trauma. 2010 or 2011. The cell phone laws are un- mandatory jail time for those convicted of driving likely to have had a significant impact on while impaired) had no discernible effects. Sim- DISCUSSION crash fatality rates because an analysis of ilarly, Traynor30 reviewed US trafficlawsand distraction-related fatal crashes (not shown) found that states with more restrictive laws British Columbia’s new traffic laws were showed no intervention effect. Furthermore, regulating teen driving and driving under the associated with decreases in fatal crashes only 11 crash fatalities were attributed to use influence had lower road fatality rates. (21.0%) and in hospital admissions (8.0%) of electronic equipment between 2004 and The benefits of British Columbia’s new and ambulance calls (7.2%) for road trauma, 2012. Finally, we looked at crash fatality out- laws are likely primarily the result of preven- corresponding to monthly decreases of 2.21 comes in neighboring jurisdictions and, with tion of crashes caused by drinking and driving.

October 2014, Vol 104, No. 10 | American Journal of Public Health Brubacher et al. | Peer Reviewed | Research and Practice | e93 RESEARCH AND PRACTICE

TABLE 2—Effects of the New British Columbia Legislation

Intervention Effect Change in Monthly Event Error Model Estimate (SE)a P % Change (95% CI) Rate per 1 000 000

All-cause fatal MVCs b British Columbia ARIMA (0,0,0)(2,0,0)12 –0.236 (0.036) < .001 –21.0 (–26.4, –15.3) –2.21 b Alberta ARIMA (0,0,0)(3,1,1)12 –0.140 (0.035) < .001 –13.1 (–18.9, –6.9) –1.55

Saskatchewan ARIMA (0,0,0)(0,1,1)12 0.087 (0.892) .922 0.6 (–10.7, 11.8) 0.09

Washington State ARIMA (3,0,0)(1,1,1)12 –0.549 (0.477) .25 –5.7 (–15.4, 4.0) –0.55 Alcohol-related fatal MVCs

British Columbia ARIMA (2,0,0)(0,1,1)12 –1.408 (0.241) < .001 –52.0 (–69.5, –34.5) –1.41 b Alberta ARIMA (0,0,0)(0,1,1)12 –0.396 (0.103) < .001 –32.7 (–45.0, –17.6) –1.03

Saskatchewan ARIMA (0,0,0)(1,1,1)12 –0.069 (0.840) .935 –1.2 (–29.8, 27.4) –0.07

Washington State ARIMA (0,0,3)(0,1,1)12 –0.374 (0.344) .276 –9.8 (–27.4, 7.8) –0.37 Alcohol-only related fatal MVCs British Columbia Normal –0.594 (0.153) < .001 –46.5 (–70.0, –23.1) –0.59 Alberta ARIMA (2,0,0) –0.436 (0.224) .051 –24.3 (–48.7, 0.1) –0.44

Saskatchewan ARIMA (0,0,0)(2,0,2)12 –0.214 (0.571) .708 –6.4 (–39.9, 27.1) –0.21

Washington State ARIMA (0,0,2)(0,1,1)12 –0.139 (0.197) .48 –7.7 (–29.1, 13.7) –0.14 Speeding-related fatal MVCs

British Columbia ARIMA (0,0,0)(2,0,0)12 –1.264 (0.279) < .001 –33.3 (–47.7, –18.9) –1.26 b Alberta ARIMA (0,0,0)(0,0,1)12 –0.156 (0.091) .085 –14.4 (–28.4, 2.2) –0.45

Saskatchewan ARIMA (0,0,6)(0,1,1)12 0.090 (0.511) .861 2.5 (–25.1, 30.1) 0.09 b Washington State ARIMA (1,0,0)(1,0,1)12 –0.365 (0.091) < .001 –30.6 (–41.9, –17.0) –1.13 Speeding-only-related fatal MVCs

British Columbia ARIMA (0,0,2)(2,0,0)12 –0.500 (0.308) .105 –21.1 (–46.7, 4.4) –0.50 Alberta Normal –0.006 (0.181) .972 –0.4 (–21.0, 20.3) –0.01 Washington State Normal –0.277 (0.115)b .016 –24.2 (–39.5, –5.0) –0.41 b Road trauma related hospital admissions: British Columbia ARIMA (2,1,1)(0,1,1)12 –0.083 (0.040) .036 –8.0 (–14.9, –0.6) –5.71 b Road trauma related ambulance calls: British Columbia ARIMA (0,1,1)(3,1,1)12 –0.075 (0.033) .022 –7.2 (–13.0, –1.1) –47.29

Note. ARIMA = autoregressive integrated moving average; CI = confidence interval; MVC = motor vehicle crash. aEstimates are expressed in monthly rate per 1 000 000 or, if the data were log-transformed, as log rate per 1 000 000. In all cases, the estimated effect is also expressed as % change and as change in monthly rate per 1 000 000. bData were log-transformed.

Consistent with this premise, we found Columbia’s new laws were put into effect, to report impaired drivers.37,38 Increased news a marked reduction (52.0%) in fatal alcohol- Alberta politicians were debating similar coverage of drinking and driving may have related crashes. The segmented regression laws.32 In December 2011, in an example of encouraged greater participation in those pro- analysis finding that alcohol-related fatal policy diffusion,33---35 Alberta passed laws with grams. Furthermore, the adjacent nature of the crashes began to decline more rapidly in 2010 immediate roadside penalties for drunk 2 provinces, and easy migration of workers strengthens the evidence linking the new law to drivers. Alberta’s new laws were put into effect between British Columbia and Alberta, sug- a reduction in alcohol-related MVC fatalities. incrementally between July and September gests that many drivers might have carried Neither Saskatchewan nor Washington had 2012. It is not surprising that British Colum- their knowledge of British Columbia laws into declines in fatal alcohol-related crashes. bia’s laws received considerable media cover- Alberta either from habit or from not knowing Alberta, however, had a 32.7% reduction in age in Alberta, with headings such as “Alta. the policy differences between the 2 provinces. fatal alcohol-related crashes that we attribute to May Stiffen Drinking and Driving Laws”32 and Finally, it is possible that Alberta police infor- a spillover effect from the changes in British “B.C. Drunk Driving Law Sobers Alberta.”36 mally increased their own enforcement efforts Columbia. Spillover effects are common in Many Albertans likely changed their drinking in response to reports from British Columbia. studies of law and public policy and are closely and driving behavior as a result of press British Columbia’s new laws also targeted related to the degree of population mobility coverage. In addition, several municipalities in speeding, with harsher penalties for excessive between jurisdictions.31 At the time British Alberta have programs that encourage citizens speeding, street racing, and stunt driving. We

e94 | Research and Practice | Peer Reviewed | Brubacher et al. American Journal of Public Health | October 2014, Vol 104, No. 10 RESEARCH AND PRACTICE

a Intervention

850

800

750

700 000 BC Residents

650 000

600 Rate/1 2005 2006 2007 2008 2009 2010 2011 2012 Time b Intervention

90

80

70

60 000 BC Residents 50 000

40

30 Rate/1 2005 2006 2007 2008 2009 2010 2011 2012 Time

Note. BC = British Columbia. The vertical line indicates the first full month of implementation of immediate roadside prohibition programs in October 2010 (intervention effect = –5.7 admissions for road trauma per 1 000 000 BC residents). The dotted curve represents the smoothed annual average of monthly rates from observed rates in part a and from the fitted model in part b. FIGURE 3—Monthly hospital admissions per 1 000 000 population for (a) all trauma causes except road trauma and (b) road trauma: British Columbia, January 2004–March 2012.

found a 33.3% reduction in fatal speeding explain why British Columbia’s laws had evaluations of traffic laws rely exclusively on crashes in British Columbia. In a subgroup a smaller effect on speeding fatalities. Neither police data,43---46 which often have limited analysis that excluded crashes related to Alberta nor Saskatchewan had reductions in information on injury severity and may not both alcohol and speeding, the reduction in speeding-related fatal crashes. It is interesting capture all injury crashes.25,26 To account fatal speeding-related crashes was smaller to note that Washington State had a statistically for trends and unmeasured confounders, (21%) and was not statistically significant at the significant reduction in speeding-related fatal we used ARIMA models21 and internal controls. nominal level. However, this analysis was crashes. Beginning in July 2010, Washington We also compared British Columbia traffic based on half as many events, resulting in an began to roll out “target zero teams.”42 These fatalities with those in adjacent jurisdictions. imprecise estimate of the effect (i.e., wide con- teams use a systems approach including educa- Because medical data do not include factors fidence intervals), and important reductions in tion and increased enforcement to target speed- that caused the crash, we cannot confirm that this subgroup cannot be ruled out. Other re- ing, impaired driving, and failure to use seatbelts. the observed reduction in ambulance calls and searchers found that both higher fines and The reduction in speeding-related crashes seen in hospital admissions for road trauma was attribut- visible enforcement reduce the incidence of Washington may be related to this approach. able to fewer crashes caused by the targeted risk speeding39,40 but that increased enforcement is factors (alcohol, speeding). Finally, deterrence- more effective.41 British Columbia’s new laws Strengths and Limitations based laws often have reduced effectiveness over did not include increased speed enforcement We used multiple data sources to examine time. Longer follow-up will be required to and did not increase sanctions for the majority the effects of British Columbia’s new laws on determine whether the beneficial effects of of speeding drivers. This limited scope may health outcomes and on fatal crashes. Most British Columbia’s new laws are long lasting.

October 2014, Vol 104, No. 10 | American Journal of Public Health Brubacher et al. | Peer Reviewed | Research and Practice | e95 RESEARCH AND PRACTICE

a Intervention

1800

1700

000 BC Residents 1600

000 1500 Rate/1

2004 2005 2006 2007 2008 2009 2010 2011 2012 Time

b Intervention

900

800

700

600 000 BC Residents

000 500

400 Rate/1 2004 2005 2006 2007 2008 2009 2010 2011 2012 Time

Note. BC = British Columbia. The vertical line indicates the first full month of implementation of immediate roadside prohibition programs in October 2010 (intervention effect = –47 calls for road trauma/1 000 000 BC residents). The dotted curve represents the smoothed annual average of monthly rates from observed rates in part a and from the fitted model in part b. FIGURE 4—Monthly ambulance calls per 1 000 000 population for (a) all trauma causes except road trauma and (b) road trauma: British Columbia, January 2004–March 2012.

Conclusions Brasher is with the Centre for Clinical Epidemiology and access and assisted with methodology. P. Brasher over- Traffic laws that introduced immediate Evaluation, University of British Columbia. Shannon Erdelyi saw statistical analysis and helped with interpretation of is with the Department of Statistics, University of British results. S. Erdelyi performed statistical analysis under license suspension and vehicle impoundment Columbia. Mark Asbridge is with the Department of Com- supervision of P. Brasher and prepared figures and for drunk drivers and vehicle impoundment for munity Health and Epidemiology, Faculty of Medicine, Dal- tables. E. Desapriya contributed relevant background excessive speeding, stunt driving, or street housie University, Halifax, Nova Scotia. Scott Macdonald is information on the importance of media awareness with the Centre for Addictions Research of British Columbia, campaigns and on driver surveys showing that drivers racing were associated with significant reduc- University of Victoria, British Columbia. Nadine Schuurman are aware of the new laws. M. Asbridge contributed tions in fatal crashes and in hospital admissions is with the Department of Geography, Faculty of Environ- to the section on policy diffusion and helped with and ambulance dispatches for road trauma. mental Studies, Simon Fraser University, Burnaby, British interpretation of results. R. Purssell provided back- Columbia. Ian Pike is with the British Columbia Injury ground information on details of British Columbia’s These findings suggest that laws calling for im- Prevention and Research Unit, Faculty of Medicine, University new laws. S. Macdonald provided background infor- mediate sanctions for dangerous drivers can re- of British Columbia. mation on deterrence theory and its application to fi duce road trauma and should be supported. j Correspondence should be sent to Jeffrey R. Brubacher, traf c laws. N. Schuurman assisted with data access Vancouver Coastal Health Research Pavilion, Room 719, and mapping. I. Pike assisted with access to hospital 828 West 10th Avenue, Vancouver, BC, Canada V5Z admission data and with interpretation of hospital 1M9 ([email protected]). Reprints can be ordered at admission data. All authors reviewed and contributed About the Authors http://www.ajph.org by clicking the “Reprints” link. to the final article. Jeffrey R. Brubacher, Herbert Chan, Edi Desapriya, and This article was accepted April 24, 2014. Roy Purssell are with the Department of Emergency Acknowledgments Medicine, Faculty of Medicine, University of British Contributors This research was funded by a grant from the Canadian Columbia, Vancouver. Roy Purssell is also with the British J. R Brubacher oversaw the project as a whole and was Institutes of Health Research and the Canadian Institute for Columbia Centre for Disease Control, Vancouver. Penelope primary author of the article. H. Chan coordinated data Health Information. J. R. Brubacher is funded through

e96 | Research and Practice | Peer Reviewed | Brubacher et al. American Journal of Public Health | October 2014, Vol 104, No. 10 RESEARCH AND PRACTICE

a Scholar Award from the Michael Smith Foundation for general deterrence: theory and evidence. Criminology. at: http://www.edmontonsun.com/news/alberta/2010/ Health Research. 2001;39(4):865---892. 11/16/16174126.html. Accessed November 23, 2013. We thank the Insurance Corporation of British 14. Beccaria C. On Crimes and Punishment. New York, 33. Grattet R, Jenness V, Curry TR. The homogenization Columbia for providing access to the data from the British NY: MacMillan; 1963. and differentiation of hate crime law in the United States, Columbia Traffic Accident System used in this article; 15. Houston DJ, Richardson LE Jr. Drinking-and-driving 1978 to 1995: innovation and diffusion in the criminal- Julie Wei, Randy Slemko, and the British Columbia ization of bigotry. Am Sociol Rev. 1998;63(2):286---307. Ambulance Service for providing access to the ambu- in America: A test of behavioral assumptions underlying lance dispatch data used in this article; and Neil Arason public policy. Polit Res Q. 2004;57(1):53---64. 34. Asbridge M. Public place restrictions on smoking in (British Columbia Office of the Superintendent of Motor 16. Beirness DJ, Beasley EE. An evaluation of immedi- Canada: assessing the role of the state, media, science and Vehicles) for providing a detailed explanation of British ate roadside prohibitions for drinking drivers in British public health advocacy. Soc Sci Med. 2004;58(1):13---24. Columbia’s old and new traffic laws. Columbia: findings from roadside surveys. Traffic Inj 35. Shipan CR, Volden C. Policy diffusion: seven lessons Prev. 2014;15(3):228---233. for scholars and practitioners. Public Adm Rev. 2012;72 Human Participant Participation 17. Ministry of Justice, Road Safety Unit, Police Services (6):788---796. This study was approved by the University of British Division. Public attitude survey 2012: British Columbia. 36. Platt M. B.C. drunk driving law sobers Alberta. Columbia Clinical Research Ethics Board. Available at: http://www.pssg.gov.bc.ca/policeservices/ Toronto Sun. September 22, 2010. Available at: publications-index. Accessed February 15, 2014. http://www.torontosun.com/news/canada/2010/09/ 23/15447056.html. Accessed November 12, 2013. References 18. The Road Safety Monitor 2012: Drinking and driving in Canada by region: trends. Available at: 37. Edmonton Police Services. Camrose adopts Curb the 1. World Health Organization. Global Status Report on http://tirf.ca/publications/project_show.php?pid=16. Danger. Available at: http://www.edmontonpolice.ca/ Road Safety 2013: Supporting a Decade of Action. Geneva, Accessed February 15, 2014. News/SuccessStories/CamroseAdoptsCurbTheDanger. Switzerland: World Health Organization; 2013. Available aspx. Accessed November 15, 2013. at: http://www.who.int/violence_injury_prevention/ 19. Macdonald S, Zhao J, Martin G, et al. The impact on road_safety_status/2013/en/index.html. Accessed alcohol-related collisions of the partial decriminalization 38. Drivers RI. Be the difference: community action for February 15, 2014. of impaired driving in British Columbia, Canada. Accid safe roads. Available at: http://bethedifferencealberta.com/ Anal Prev. 2013;59:200---205. report-impaired-drivers. Accessed November 23, 2013. 2. World Health Organization. Global Status Report on fi fi Road Safety: Time for Action. Geneva, Switzerland: World 20. International Classi cation of Diseases, 10th Revision. 39. Jones B, Haas K, Kirk A, Grif th A. Self-reported fi Health Organization; 2009. Available at: http://www. Geneva, Switzerland: World Health Organization; 1992. effectiveness of double- ne zones as a speed control who.int/violence_injury_prevention/road_safety_status/ 21. Box GEP, Jenkins GM. Time-Series Analysis: Forecasting measure. Appl Health Econ Health Policy. 2004;3(1):17---28. 2009. Accessed February 15, 2014. and Control. 2nd ed. San Francisco, CA: Holden-Day; 1976. 40. Vaa T. Increased police enforcement: effects on 3. Borkenstein RF, Crowther RF, Shumate RP, Ziel 22. Mulgrew I. B.C.’s tough new impaired-driving rules speed. Accid Anal Prev. 1997;29(3):373---385. WB, Zylman R. The role of the drinking driver in traffic declared unconstitutional in part, judge rules. Vancouver 41. Ryeng EO. The effect of sanctions and police accidents (the Grand Rapids Study). Blutalkohol. 1974; Sun. November 30, 2011. Available at: http://www. enforcement on driver’s choice of speed. Accid Anal Prev. 11(suppl):7---13. vancouversun.com/news/tough+impaired+driving+ 2012;45:446---454. 4. Blomberg RD, Peck RC, Moskowitz H, Burns M, rules+declared+unconstitutional+part+judge+rules/ 42. Target Zero—Washington State strategic highway Fiorentino D. The Long Beach/Fort Lauderdale relative 5790600/story.html. Accessed December 3, 2011. safety plan 2013. Available at: http://www.targetzero. risk study. J Safety Res. 2009;40(4):285---292. 23. Muggeo VMR, AdelfioG.Efficient change point com. Accessed February 11, 2014. 5. Blows S, Ameratunga S, Ivers RQ, et al. Risky driving detection for genomic sequences of continuous mea- 43. Wagenaar AC, Maldonado-Molina MM, Erickson DJ, habits and motor vehicle driver injury. Accid Anal Prev. surements. Bioinformatics. 2011;27(2):161---166. Ma L, Tobler AL, Komro KA. General deterrence effects 2005;37(4):619---624. 24. Muggeo VM. Estimating regression models with of US statutory DUI fine and jail penalties: long-term Accid Anal Prev 6. Grundy C, Steinbach R, Edwards P, Green J, unknown break-points. Stat Med. 2003;22(19):3055---3071. follow-up in 32 states. . 2007;39:982---994. Armstrong B, Wilkinson P. Effect of 20 mph traffic speed 25. Amoros E, Martin J-L, Lafont S, Laumon B. Actual 44. Andreuccetti G, Carvalho HB, Cherpitel CJ, et al. zones on road injuries in London, 1986-2006: controlled incidences of road casualties, and their injury severity, Reducing the legal blood alcohol concentration limit for interrupted time series analysis. BMJ. 2009;339:b4469. modelled from police and hospital data, France. Eur J driving in developing countries: a time for change? Results and implications derived from a time-series 7. Asbridge M, Brubacher J, Chan H. Cell Phone Use Public Health. 2008;18(4):360---365. analysis (2001---10) conducted in Brazil. Addiction. and Traffic Crash Responsibility: A Culpability Analysis 26. Amoros E, Martin J-L, Chiron M, Laumon B. Road 2011;106(12):2124---2131. of Motor Vehicle Collisions in British Columbia. Paper crash casualties: characteristics of police injury severity presented at: Canadian Multidisciplinary Road Safety Con- misclassification. J Trauma. 2007;62(2):482---490. 45. Blais E, Gagne M-P. The effect on collisions with injuries of a reduction in traffic citations issued by police ference XXI; May 8---11, 2011; Halifax, Nova Scotia, Canada. 27. Asbridge M, Mann RE, Flam-Zalcman R, Stoduto G. officers. Inj Prev. 2010;16(6):393---397. 8. Caird JK, Willness CR, Steel P, Scialfa C. A meta- The criminalization of impaired driving in Canada: analysis of the effects of cell phones on driver perfor- assessing the deterrent impact of Canada’s first per se law. 46. Friedman LS, Barach P, Richter ED. Raised speed mance. Accid Anal Prev. 2008;40(4):1282---1293. J Stud Alcohol. 2004;65(4):450---459. limits, case fatality and road deaths: a six year follow-up using ARIMA models. Inj Prev. 2007;13(3):156---161. 9. Office of the Superintendent of Motor Vehicles. 28. Asbridge M, Mann RE, Smart RG, et al. The effects Road safety rules and consequences of unsafe driving. of Ontario’s administrative driver’s licence suspension Available at: http://www.pssg.gov.bc.ca/osmv/road- law on total driver fatalities: a multiple time series analysis. safety/index.htm. Accessed November 23, 2011. Drugs Education Prev Policy. 2009;16(2):140---151. 10. British Columbia Motor Vehicle Act. Available at: 29. Wagenaar AC, Maldonado-Molina MM. Effects of http://www.bclaws.ca/EPLibraries/bclaws_new/document/ drivers’ license suspension policies on alcohol-related ID/freeside/96318_00. Accessed November 23, 2011. crash involvement: long-term follow-up in forty-six states. 11. Shepherd JP. Criminal deterrence as a public health Alcohol Clin Exp Res. 2007;31(8):1399---1406. strategy. Lancet. 2001;358(9294):1717---1722. 30. Traynor TL. The impact of state level behavioral fi 12. Wright V. Deterrence in criminal justice: evaluating regulations on traf c fatality rates. J Safety Res. 2009; certainty vs severity of punishment. Available at: 40(6):421---426. http://www.sentencingproject.org/doc/Deterrence% 31. Baicker K. The spillover effects of state spending. 20Briefing%20.pdf. Accessed November 11, 2011. J Public Econ. 2005;89(2---3):529---544. 13. Nagin DS, Pogarsky G. Integrating celerity, impul- 32. Landry F. Alta. may stiffen drinking and driving sivity, and extralegal sanction threats into a model of laws. Edmonton Sun. November 17, 2010. Available

October 2014, Vol 104, No. 10 | American Journal of Public Health Brubacher et al. | Peer Reviewed | Research and Practice | e97