Are Countries’ Drink-Driving Policies Associated with Harms Involving Another Driver’s Impairment?

Abstract

Introduction:

International drink-driving policy research generally focuses on aggregate outcomes (e.g., rates of crashes, fatalities) without emphasizing secondhand alcohol-related vehicular harms. In contrast, we investigate associations between drink-driving policies and harms involving another driver’s impairment.

Methods:

Alcohol’s harms to others (AHTO) survey data from 12 countries (analytic N=29,616) were linked to national alcohol policy data from the World Health Organization. We examined separately associations of two 12-month driving-related AHTOs (passenger with an impaired driver; vehicular crash involving someone else’s drink driving) with three national drinking-driving policies--legal blood alcohol concentration (BAC) limits, use of random breath testing, use of sobriety checkpoints, and comprehensive penalties for drink-driving (community service, detention, fines, ignition interlocks, license suspension/revocation, mandatory alcohol treatment, vehicle impoundment, and penalty point system), plus two alcohol tax variables (having excise taxes and value-added tax [VAT] rate). Multilevel logistic regression addressed clustering of individuals within countries and subnational regions, while adjusting for individuals’ gender, age, marital status, risky drinking, and regional drinking culture (% male risky drinkers in sub-national region).

Results:

Controlling for national-, regional-, and individual-level covariates, comprehensive penalties were significantly and negatively associated with both outcomes; other vehicular policy variables were not significantly associated with either outcome. A society’s VAT rate was negatively associated with riding with a drunk driver. Regional male drinking culture was positively associated with riding with an impaired driver, but was not significantly associated with being in a vehicular crash due to someone else’s drinking. In both models, being male, being younger, and engaging in risky drinking oneself each were positively associated with vehicular harms due to someone else’s drinking.

Conclusions:

Although results are associational and not causal, comprehensive penalties may be promising policies for mitigating driving-related harms due to another drinker. Higher VAT rate might reduce riding with a drunk driver.

Introduction

Policy approaches to reducing drink driving (sometimes termed drunk driving, driving under the influence [DUI], or alcohol-related driving impairment), and the study of the associated policy effects, have a long history. In part this is because individually-oriented campaigns and providing deterrent information have a mixed record, though some information programs conveying the likelihood of enforcement activities in a local area have shown some promise (Johnson and Clapp, 2011). In the US, a series of studies of states with various per se legal BAC limits for drivers undertaken in the 1980s and 1990s (summarized by Hingson et al. (1999)) led to a federally-driven reduction in legal BACs across the states. In some low and middle income (LAMI) countries such as Cambodia, where increases in alcohol-involved fatal crashes have been observed, researchers, while calling for enhanced enforcement of existing drink-driving laws, also emphasize the importance of combining these efforts with mutually-reinforcing approaches such as mass media campaigns and affordable taxi services (Bachani et al., 2017).

Internationally, drink-driving policy research generally has focused on aggregate outcomes (e.g., rates of alcohol-impaired crashes, total crashes or fatalities). For example, Brubacher et al. (2014) examined reductions in fatalities, ambulance calls, and hospital admissions for road trauma after introduction of new traffic laws in British Columbia, Canada. In the US, Voas, Tippetts & Fell (2000) studied the relationship of alcohol safety laws with involvement of drinking drivers in fatal crashes. A Brazilian study discussed by Andreuccetti et al. (2016) noted that both establishing and lowering BAC limits for drivers reduced road traffic injuries and the proportion of DUI reports, but the authors also identified loopholes in the law and difficulties with enforcement, not uncommon in developing countries. One of the more comprehensive multinational studies, with country-level data from 133 countries, examined the relationship between a country’s per capita alcohol volume and a pattern of drinking score (a country’s predominantly heavy per-occasion drinking), and various outcomes including driving-related ones (Rehm et al., 2001). A country’s heavy per-occasion drinking pattern and drink driving policies have recently been studied further (Cheng and Pien, 2018). In this study, drink driving policies were categorized as preemptive measures (e.g., random breath testing and BAC driving limits), penalties (e.g., detention and fines), and mandatory treatment and ignition interlocks (Cheng and Pien, 2018). Cheng and Pien (2018) found mandatory treatment and ignition interlock policies, as well as the strength of comprehensive preemptive measures, were less prevalent in countries with more risky drinking patterns; thus they advocated the introduction of these policies in such countries.

While some of the studies just summarized implicitly include harms to non-driving individuals (such as passengers or pedestrians)—e.g., (Brubacher et al., 2014)—they have not emphasized secondhand alcohol-related vehicular harms stemming from someone else’s impaired driving. In contrast, in this international study we examine associations between drink-driving policies and two harms due to the alcohol-attributed impairment of a driver other than the respondent—in the last 12 months being a passenger with an impaired driver, and experiencing a vehicular crash involving someone else’s drink driving.

Consequences of secondhand effects of alcohol-related impairment are widespread (Callinan et al., 2016; Room et al., 2019a) and often serious (Greenfield et al., 2009; Laslett et al., 2020; Nayak et al., 2019). In a 2005 US national adult population survey, Greenfield et al. (2009) found that 44% reported ever being a passenger with a driver who had had too much to drink (3.3% in the last 12 months), while 8% reported ever being in a motor vehicle accident because of someone else’s drinking (but only 0.3% in the prior year). Indeed, ever having ridden with a drunk driver was the most prevalent of six heterogeneous harms from other drinkers also including assault, vandalism (involving home, car or property), family problems or marital difficulties, and financial trouble. More recently, Nayak et al. (2019) estimated that in the US, 6.9 million women (5.5%) and 8.3 million men (7.0%) reported driving-related harms caused by other drinkers. The extent of the issue emerges clearly in one of the few comprehensive appraisals of strategies for reducing driving-related injuries and death available, produced by a recent blue-ribbon panel (National Academies of Sciences, 2018). The following cogent statement is included among its key facts: “Like smoking, there are second-hand effects of alcohol-impaired driving in which the injured have no voice in the harmful decision. Almost 40 percent of [US] alcohol-impaired driving fatalities in 2015 were victims other than the drinking driver; for comparison, about 8.5 percent of smoking-related deaths [US] were due to second-hand smoke in 2015.” (National Academies of Sciences, 2018, p 1.4). A multinational counterpart is not to our knowledge available.

Here, using a standardized, multinational survey dataset on alcohol’s harms to others (AHTO), we analyze data from 12 countries where AHTO was ascertained from the victim’s perspective (Wilsnack et al., 2018). We investigate the association between driving-relevant and taxation-based national alcohol policies and rates of experiencing, in the prior 12 months, each of two types of specific secondhand effects—riding as a passenger with a driver who had been drinking too much, and having a motor vehicle accident due to someone else’s drinking—considering certain plausible societal and individual-level confounders. We also examine the possible role that heavy drinking culture in subnational regions may play in relation to individual driving-related harms from other drinkers, testing whether, across countries, the riskiness of drinking in a respondent’s region of residence might play a role in a person’s exposure to driving-related harms due to other drinkers.

Materials and Methods

Data Sources

Survey data (total N=29,786) were collected in 12 countries varying in level of development: Australia, Chile, Denmark, India, Ireland, Lao People’s Democratic Republic (Lao PDR), New Zealand, Nigeria, Sri Lanka, Thailand, United States, and Vietnam. Table 1 summarizes characteristics of the surveys analyzed here; the surveys used a standard instrument developed for a Thai Health Foundation/World Health Organization supported project. More details are presented elsewhere (Callinan et al., 2016; Kaplan et al., 2017; Wilsnack et al., 2018). The AHTO surveys differed somewhat in sampling frame (national vs certain large areas within a country), sampling method (population probability sampling vs household member replacement sampling), age range and average age of samples, and modes of administration (telephone vs face-to-face interviews). Because of variations in sampling methods and recording of non-respondents, response rates (or in two cases, cooperation rates) of surveys with probability sampling ranged from 37% to 99%, with a median of 83%. Based on available subnational regional designations and risky drinking data, the sample available for analysis was N=29,786. For the multivariate analyses including covariates, the analytic sample was reduced to N=29,616.

Table 1.

Overview of 12 AHTO country surveys included in the analysis

Country	Year	N	Available N	Sampling scope	Age range	Interview mode	Response rate	Funding sourcea	# regions
Australia	2008	2,649	2,379	National	18+	LL CATI	50%b	Alc Res & Educ Foundd	5
Chile	2013	1,500	1,424	7 cities & hinterland	18–64	Face-to-face	72%	WHO-Thai Health	5
Denmark	2011	5,133	5,133	National	15–79	Web/telephone	64%	Aarhus University	5
India	2013/14	3,403	3,351	Karnataka state	18+	Face-to-face	97%	WHO-Thai Health	4
Ireland	2015	2,005	2,005	National	18+	LL CATI	37%	Health Ministry	4
Lao PDR	2012/13	1,257	1,212	3 regions	15–64	Face-to-face	99%	WHO-Thai Health	3
New Zealand	2008/9	3,068	2,611	National	12–80	LL CATI	64%	NZ Health Res. Council	5
Nigeria	2013	2,270	2,248	6 states/capital territory	18–64	Face-to-face	Not availablec	WHO-Thai Health	3
Sri Lanka	2013	2,475	2,353	National	18–64	Face-to-face	93%	WHO-Thai Health	6
Thailand	2012/13	1,695	1,695	5 regions	18–64	Face-to-face	94%	WHO-Thai Health	6
USA	2014/15	2,830	2,327	National (50 states + DC)	18+	LL & Cell CATI	60%b	NIAAAd	9
Vietnam	2012/13	1,501	1,479	6 provinces	18–64	Face-to-face	99%	WHO-Thai Health	6

^aAdditional national funding sources also involved

^bCooperation rate

^cCannot be calculated because data on non-response not collected

^dAlcohol Research and Educational Foundation; NIAAA: National Institute on Alcohol Abuse and Alcoholism

Abbreviations: CATI: computer-assisted telephone interviewing; LL:landline phone; PDR: People’s Democratic Republic

These survey data were linked to national alcohol policy variables directly relevant to drinking driving that were compiled by the World Health Organization (WHO) for the Global Information System on Alcohol and Health (GISAH) (World Health Organization, 2016). These data reflect policies in 2012, concurrent with or preceding the AHTO data collection conducted in 2012–14, with the exception of Australasian data that had been collected in 2008 (Australia and New Zealand).

Measures

National:

Three categories of national drinking-driving policy measures were considered. BAC limit for the general population (multiplied by 100 for interpretability), with a higher value denoting a less stringent BAC policy. Random breath testing is a dichotomous variable indicating a country’s having (vs. not having) this policy. Likewise, sobriety check points indicates a country’s having the policy (1=yes, 0=no). Comprehensive Penalties for drunk driving is a factor score derived from six dichotomous variables (1=yes, 0=no) each indicating a country’s having the given law (vs. no law): community service, ignition interlocks, license suspension or revocation, mandatory alcohol treatment, vehicle impoundment, and a penalty point system for drinking driving (together having a Cronbach’s alpha of 0.92). Principal component analysis of these six items yielded a single factor. Two additional related items—having laws for detention and fines—were initially considered for this domain but were not included in the construct due to lack of variability in these policies across the 12 countries in our analysis (all countries had these policies). We also considered as two general population measures two alcohol taxation variables, value added tax (VAT) rate (%) and excise taxes (a composite of the sum of three dichotomous variables (1=yes, 0=no), each indicating a country’s having excise tax on beer, wine, and spirits.

Region within a Society:

As a proxy for regional-level risky drinking culture, viewed as a potential moderator of national policy effects, the proportion of male risky drinkers in each region was calculated. Although women’s risky drinking varies more by country than men’s, and gender ratios of heavy drinking also differ across regions within countries (Grittner et al., 2020), for depicting regional drinking, male risky drinking was selected to characterize drinking culture. This variable was created by aggregating to the regional level within a given society the individual-level self-reported risky drinking variable, defined as drinking at least 60g ethanol on a single occasion at least monthly (vs. less frequently), based on a 12-month heavy drinking per occasion question: “how often did you have at least 5 drinks (or 60 gm) of any kind of alcoholic beverage in a single day, that is, any combination of cans, bottles or glasses of beer, glasses of wine, drinks containing liquor, or homemade alcoholic beverages of any kind?” (followed by categorical frequency responses from once a day, to 1 to 3, times a month, less than once a month, and never in the last 12 months). Using the median proportion of risky drinkers across regions in the 12 countries, the regions were then divided between low-risk (<.270), medium-risk (.270–.404) and high-risk drinking cultures (>.404) for use in analyses.

Individual

Two driving-related AHTO outcomes were constructed as dichotomous variables based on the 12-month questions: (a) “have you been a passenger with a driver who had too much to drink” (i.e., riding with an impaired driver) or (b) “has someone who has been drinking been responsible for a traffic accident you were involved in?” (i.e., a vehicular crash due to someone else’s drinking). Demographic characteristics included in the models were age, gender, and marital status (married or living with a partner versus any other status, including never married, divorced and widowed). The respondent’s own self-reported risky drinking (based on the heavy drinking occasion measure described above) was also included as a covariate. Education (high school graduation or higher education level vs. less than a high school level) was not significantly associated with the outcomes in preliminary analyses, and thus it was not included in the final models.

Statistical Analysis

Two three-level logistic regression models were estimated separately for each driving-related AHTO outcome that addressed nesting of individuals within sub-national regions and within societies (primarily countries), while adjusting for an individual’s gender, age, marital status, and own risky drinking behavior; and regional drinking culture (percent male risky drinkers in the region). Analyses were weighted to adjust for non-response and selection probability.

Results

Distribution of Risky Drinking Regions

Table 2 shows diverse drinking cultures across the 12 countries. The proportion of male risky drinkers was the highest in Chile (45%), followed by Ireland (43%) and Denmark (41%), with the lowest in the U.S. (14%) and Nigeria (21%). The table also presents the range of within-country regional proportions of male risky drinkers of each country.

Table 2.

Regional Proportion of Risky Drinkers (≥ 60 g EtOH at least monthly), Low- Medium- and High-risk regions, and drink-driving policies by country

Country	Proportion of male risky drinkers		Number of regions			National Drink Driving Policiesb				National alcohol tax policies
	Mean (Standard Error)	Range	Low-risk region < 0.270a	Medium-risk region .270 −.404	High-risk region > 0.404a	BAC limit	Random Breath testing 1=yes, 0=no	Sobriety check points 1=yes, 0=no	Comprehensive penalties (Range=0–6)	VAT rate (%)	Excise Tax (Range=1–3)
Australia	0.38 (0.0004)	0.29 – 0.41	0	4	1	0.05	1	0	6	10	2
Chile	0.45 (0.0008)	0.43 – 0.55	0	0	5	0.03	1	0	2	19	3
Denmark	0.41 (0.0000)	0.41 – 0.41	0	0	5	0.05	1	1	5	25	3
India	0.35 (0.0025)	0.23 – 0.59	1	2	1	0.03	1	0	1	0	3
Ireland	0.43 (0.0013)	0.36 – 0.51	0	2	2	0.05	1	1	2	23	3
Lao PDR	0.23 (0.0020)	0.13 – 0.28	2	1	0	0.08	0	0	0	10	1
New Zealand	0.32 (0.0004)	0.29 – 0.36	0	5	0	0.08	1	1	6	15	3
Nigeria	0.21 (0.0008)	0.17 – 0.25	3	0	0	0.05	0	0	1	5	3
Sri Lanka	0.27 (0.0027)	0.13 – 0.51	4	1	1	0.08	1	0	0	20	3
Thailand	0.27 (0.0018)	0.15 – 0.35	2	4	0	0.05	0	1	1	7	3
USA	0.14 (0.0006)	0.10 – 0.18	9	0	0	0.08	0	1	6	0	3
Vietnam	0.26 (0.0042)	0.08 – 0.51	4	0	2	0	0	1	2	10	3
All Countries	0.32a (0.0008)	0.08 – 0.59	25	19	17	Mean=0.54 SD=02	Mean=0.65 SD=0.48	Mean=0.54 SD=0.50	Mean=3.17 SD=2.35	Mean=12.61 SD=9.03	Mean=2.83 SD=0.48

^aMedian proportion of male risky drinkers = 0.407

^bSee Definitions in Methods

Estimating Associations between National Policies and Driving-related AHTO

Table 2 shows drinking driving policies for each country. Controlling for the individual characteristics and regional male risky drinking, comprehensive penalties for drink driving (aOR = 0.637; 95% CI: 0.536 – 0.758, p<.0001) and VAT rate (aOR = 0.938; 95% CI: 0.894 – 0.985, p<.05) were significantly and negatively associated with having been a passenger with an impaired driver (see Table 3, left column). Comprehensive penalties (aOR = 0.685; 95% CI: 0.569 – 0.823, p<.0001) was also significant for vehicular crash due to someone else’s drinking (also in Table 3, right column). Other policy variables were not significant for either outcome.

Table 3.

Associations between drink-driving policies and driving-related harms from others’ drinking: Multilevel logistic regression models

	Riding with Drunk Driver		Crash due to Other Drunk Driver
	n=29,616		n=29,616
VARIABLES	AOR	95% CI	AOR	95% CI
INDIVIDUAL-LEVEL
Female	0.753†	(0.576 – 0.985)	0.697**	(0.532 – 0.913)
Age	0.972****	(0.966 – 0.979)	0.980****	(0.972 – 0.989)
Married/living with partner	0.786**	(0.660 – 0.937)	0.847	(0.658 – 1.090)
Risky drinking by victim	2.766****	(2.212 – 3.605)	1.604****	(1.336 – 1.926)
COUNTRY-LEVEL
BAC Limit	1.077	(0.918 – 1.263)	0.961	(0.800 – 1.154)
Random breath testing	2.172	(0.406 – 11.619)	2.342	(0.775 – 7.079)
Sobriety check point	3.826	(0.608 – 24.077)	1.739	(0.691 – 4.376)
Comprehensive penalties	0.637****	(0.536 – 0.758)	0.685****	(0.569 – 0.823)
Value-added tax	0.938*	(0.894 – 0.985)	0.959	(0.910 – 1.011)
Excise tax	0.739	(0.430 – 0.271)	0.970	(0.593 – 1.589)
REGIONAL-LEVEL
Medium risk region	1.210	(0.784 – 1.866)	1.381	(0.768 – 2.483)
High risk region	1.772*	(1.135 – 2.766)	1.808	(0.597 – 5.480)

Regional Covariates

High-risk regional drinking culture was positively associated with riding with an impaired driver (aOR = 1.717; 95% CI: 1.135 – 2.766, p < .05), but regional drinking culture was not significantly associated with experiencing a vehicular crash due to someone else’s drinking.

Individual Covariates

Men and younger people were more exposed to both types of drinking-driving-related harms. Married or partnered individuals tended to be less exposed than others (mostly single people) to riding with an impaired driver, but marital status was not significant for vehicular crash due to someone else’s drinking. Risky drinking by the survey respondent was associated with a large increase in the odds of being exposed to these specific driving-related risks/harms from other drinkers.

Discussion

The relationships seen for Comprehensive Penalties—community service, ignition interlocks, license suspension or revocation, mandatory alcohol treatment, vehicle impoundment, and penalty point system for drinking driving—suggest these policies may have promise for potentially mitigating vehicular-related harms from other drinkers, being associated with lower experience of both riding with an impaired driver and having a vehicular crash due to another drinking driver. Random breath testing and sobriety checkpoints were associated with lower odds of riding with a drunk driver, but this was not significantly the case for crashes due to another drinking driver. These policy strategies are nevertheless showing a negative association with the more prevalent risk of riding with a drunk driver, and thus may also have some promise. Having a low BAC limit showed no significant relationship with either driving-related AHTO.

In regard to individual factors associated with the harms, the most noteworthy was the respondent’s own risky drinking. As noted in other studies (Fillmore, 1985; Greenfield et al., 2016; Greenfield et al., 2009), the individual’s own risky drinking is associated with one-and-a-half-fold (crashes due to a drunk driver) to two-and-a-half-fold (riding with a drunk driver) higher odds of reporting these vehicular harms. With individual drinking accounted for, high-risk regional drinking culture (vs low-risk) also was associated with putting oneself at risk as a passenger of a driver who has drunk too much. Even some of the vehicular crashes reported could involve a known co-drinker at the wheel, so it is no surprise that the drinking of the potential victim may be also associated this harm due to another drinker. As plausible and expected, residing in a region with a high-risk regional male drinking culture, compared to a low-risk drinking culture, was positively associated with riding with an impaired driver. Conversely, regional male drinking culture was not significantly associated with a vehicular crash due to someone else’s drinking. There may be other factors than drinking cultures that may play roles in vehicular crashes such as road conditions, traffic safety rules, the mix of motor vehicles on the road (e.g., cars, motorcycles, and three-wheelers), which were not controlled for in our models. Our results suggest that more attention should be placed on strategies that reduce normative influences in the drinking culture surrounding an individual. This could mitigate the risk of harm from other drinkers, since places with heavy male drinking patterns may well engender acceptance of drunkenness (Room et al., 2019b), leading to the higher odds of putting oneself at risk by riding with a drunk driver, in victims and perpetrators alike.

Regarding other covariates, women are at lower risk of both vehicular outcomes than men, and younger people have greater risk of both outcomes. On the other hand, being married or partnered appears protective in the case of riding with an impaired driver only, when other factors are accounted for, including age and gender. This too has been noted in other AHTO studies (Greenfield et al., 2009).

Limitations

Clearly this cross-sectional study cannot identify causal relationships. In addition, there may be other country- or regional-level differences that confound the relationship between national drink driving polices and driving-related AHTO, but were unadjusted in our analyses. The odds ratios of policy variables for the crashes models were in the same direction as those in the riding with a drunk driver models, but random breath testing and sobriety checkpoints, for example, showed wider confidence limits suggesting the low base-rate of 12-month crashes due to other drinkers might be underpowered to detect policy associations. As is sometimes the case for cross-national studies, some data (including those for alcohol policy variables) may not have been collected or reported consistently across countries. For example, information about BAC reported in GISAH concerned automobiles, which may not have fully captured the effect of BAC limits in countries (such as Vietnam) where motorcycles are the dominant means of transportation, This may also explain, at least in part, the null results involving BAC limits. Still, offsetting these concerns, the measure of comprehensive penalties showed an even stronger negative association for crashes caused by someone else’s drinking than was the case for riding with a drunk driver, and this had a remarkably narrow confidence interval.

Conclusions

Even with limitations, this study meaningfully contributes to the alcohol policy literature. These analyses are among the first to take secondhand effects of drunk driving into account in a multinational policy framework, while attending to subnational male-dominated drinking culture as a potentially related environmental and normative influence. Results are consistent with the possibility that enacting comprehensive penalties—laws stipulating community service, ignition interlocks, license suspension or revocation, mandatory alcohol treatment, vehicle impoundment, and a penalty point system for drinking driving—may have greatest promise for reducing vehicular harms due to other drinking drivers. Enactment of random breath testing and sobriety checkpoints (RBT-SC) policies might also reduce risks associated with riding with drunk drivers. Although not significant in either harm outcome, the higher rather than lower odds associated with lowering BAC limits might mean that societies respond to drink-driving harms, when legislators perceive troubles on the roads involving heavy drinking and so tend to enact these policies, widely perceived to be effective (Babor et al., 2010), in an effort to curb such alcohol-related harms. All the policy results here are associational only, and are suggestive of hypotheses to be tested using stronger research designs to tease out causal effects.