Injunctive Norms and Driving Under the Influence and Riding With an Impaired Driver Among Young Adults in Washington State

Abstract

Purpose:

Alcohol- and cannabis-impaired driving behaviors remain a public health concern especially among young adults (i.e., ages 18–25). Limited updates to prevention efforts for these behaviors may be due, in part, to limited understanding of malleable psychosocial predictors. The current study assessed associations between perceived injunctive norms (i.e., acceptability) of driving under the influence of alcohol (DUI-A) and cannabis (DUI-C), and riding with a driver under the influence of alcohol (RWI-A) and cannabis (RWI-C) in Washington State young adults.

Methods:

Participants included 1,941 young adults from the 2019 cohort of the Washington Young Adult Health Survey. Weighted logistic regressions assessed the associations between peer injunctive norms and impaired driving-related behaviors.

Results:

A weighted total of 11.5% reported DUI-A, 12.4% DUI-C, 10.9% RWI-A, and 20.9% RWI-C at least once in the past 30 days. Overlap between the outcomes was observed, indicating some young adults had engaged in multiple impaired driving-related behaviors. After controlling for substance use frequency, weighted logistic regressions indicated more positive perceived injunctive norms were associated with nearly 2 ½ times higher odds of DUI-A, 8 times higher odds of DUI-C, 4 times higher odds of RWI-A and six and a half times higher odds of RWI-C.

Discussion:

Results increase the understanding of how injunctive norms–a potentially malleable psychosocial factor–are associated with four impaired driving-related outcomes. Prevention programs that focus on assessing and addressing the norms of these outcomes individually and collectively, such as normative feedback interventions and media campaigns, may be helpful in reducing these behaviors.

Although vehicle crashes have been a leading cause of death among young adults (YAs; i.e., individuals aged 18–25) for more than the past 30 years [1], there has been limited progress in recent years on empirically supported and evaluated prevention efforts to reduce crashes, specifically in regard to impaired driving. While public health efforts have led to reductions in alcohol-impaired crash fatalities since the 1980’s, these declines have stagnated with one in four crash fatalities involving alcohol in 2019 alone [2].

With numerous states continuing to legalize cannabis for medical and nonmedical use [3], the landscape of impaired driving has changed in the past decade. Similar to alcohol, cannabis has been shown to impair driving ability [4,5]. While some research has provided mixed findings [6], a large body of epidemiological and experimental research has indicated increased crash risk with cannabis use [7-9]. This research indicates innovative prevention efforts in reducing driving under the influence of alcohol (DUI-A) and cannabis (DUI-C) are needed. However, psychosocial risk factors, such as injunctive norms, that may be malleable to prevention efforts and associated with impaired driving behaviors remain understudied. Increasing understanding of such factors may lead to the adaptations of empirically supported prevention efforts [10], such as creation of normative feedback interventions [11-13], to reduce impaired driving behaviors via reducing social acceptability of impaired driving behaviors, and in turn number of crashes from impaired driving.

Perceived injunctive norms, or perceived approval (or disapproval) of a behavior by others [14], are a powerful and malleable predictor of many health risk behaviors including alcohol use, cannabis use, and consequences associated with their use in several studies with college or young adult samples [15-19]. Though limited, previous research also suggests injunctive norms may be an important factor to assess regarding impaired driving behaviors. Studies with college students have indicated those who perceived the typical college student as more approving of DUI-A had increased likelihood of DUI-A themselves [20,21]. A roadside sample of drivers found those who reported people who are important to them would be disappointed if they DUI-C were less likely to report future intentions to DUI-C [22]. Another college sample showed significant correlations between injunctive norms of DUI-C and intentions to DUI-C but failed to find a significant association between injunctive norms and intentions after controlling for factors such as perceived behavioral control and personal attitudes [23].

Examining associations between injunctive norms and impaired driving behaviors separately for both alcohol and cannabis is important as perceptions and use of these two substances as well as reports of DUI-A and DUI-C differ. Recent trends for YAs show decreases in approval and use of alcohol but increases in approval and use of cannabis [24]. Similarly, perceptions of the of risk of cannabis use have decreased whereas perceptions of risk of alcohol use have remained unchanged [25]. Regarding impaired driving behaviors, previous research indicates individuals are likely to report DUI-C as safer than DUI-A [26-28] and among those that use each substance, a higher proportion report DUI-C compared to DUI-A [29]. Further, it is important to assess the effects of injunctive norms of riding with an alcohol- or cannabis-impaired driver (RWI-A, RWI-C, respectively) separately from DUI-A and DUI-C. RWI is a unique risk behavior with passengers consisting of about 13% of fatalities from alcohol-related crashes alone [30]. Reports of RWI-A and RWI-C among adolescents and YAs are typically higher than reports of DUI-A and DUI-C [31,32]. Further, previous research with college students suggests more permissive peer injunctive norms of RWI-A are associated with increased willingness to RWI-A and decreased intentions to use safe alternative transportation [33] as well as a greater frequency of RWI-A behavior [34] and a positive correlation between injunctive norms of RWI-C and intentions to RWI-C [23].

Despite previous findings, as well as the potential utility of injunctive norms in prevention efforts, research has yet to examine injunctive norms of both DUI and RWI where impairment is caused by alcohol or cannabis separately and how these are associated with impaired driving-related outcomes within the same sample. Additionally, much of the young adult research is conducted with samples of college students, limiting generalizability of the findings. Similarly, studies including only samples of drivers with a valid driver’s license exclude an important segment of the population who drive without having a valid driver’s license, also limiting generalizability of the findings. Moreover, RWI is applicable to all young adults in Washington state (WA), not just those who have a valid driver’s license. Thus, the current study sought to examine the association between perceived injunctive norms of impaired driving-related behaviors and impaired driving-related outcomes (i.e., DUI-A, DUI-C, RWI-A, and RWI-C) in a general sample of YAs in WA.

Methods

Participants

Participants (N = 1941) were recruited as part of the 2019 cohort of the Washington Young Adult Health Survey. Participants were WA residents aged 18-25 recruited via direct mailing using approved contact information provided by the Washington State Department of Licensing and online advertising who completed a web-based survey of substance use, risk factors, and health behaviors. A full description of recruitment, screening, and data collection procedures is provided elsewhere [35]. The University of Washington Institutional Review Board reviewed and approved all measures and procedures.

Measures

Demographics assessed included age, sex assigned at birth, race and ethnicity, employment status (i.e., work 35 hours or more a week), and college status (i.e., currently in a 4-year college). Region of WA (i.e., East, Northwest, Southwest) and Rural-Urban Commuting Area (RUCA) codes [36] were determined based on participants’ addresses. RUCA codes that defined metropolitan areas (i.e., RUCA codes 1-3) were recoded to compare to nonmetropolitan areas (i.e., RUCA codes 4-10). To improve generalizability to WA YAs, post-stratification weights were created. Strata were based on an individual’s sex assigned at birth, geographic region, and race/ethnicity. The proportion of YAs in WA within each stratum, based on 2010 US Census data, was divided by the proportion of YAs in the sample in that stratum [35]. These weights were applied to all analyses. All variables used for weighting were also included as covariates in the models. Race and ethnicity variables were recoded with Non-Hispanic White–largest segment of the sample–as the reference group compared to Hispanic any race, Non-Hispanic Asian, and Non-Hispanic other race, which included participants who reported being American Indian or Alaskan Native (1.29% of total sampled), Black or African American (1.91%), multiracial (6.08%), Native Hawaiian or Pacific Islander (0.26%) or Other race not listed (0.46%). These response options were combined because of the relatively small sample size in each category. See Table 1 for weighted and unweighted percentages of all demographic variables.

Table 1.

Unweighted and weighted sample Demographics

	Unweighted %	Weighted %
Race/Ethnicity
Asian or Asian American	10.87	7.70
Caucasian/White	59.92	66.37
Hispanic	19.22	14.95
Other or Multiracial	9.99	10.98
American Indian or Alaskan Native	1.29	1.57
Black or African American	1.91	4.05
More than one Race	6.08	4.65
Native Hawaiian or other Pacific Islander	0.26	0.56
Other Race	0.46	0.15
Sex Assigned at Birth
Female	68.06	48.41
Male	31.94	51.59
College Status
Enrolled in 4-Year University	30.71	29.01
Not Enrolled in 4-Year University	68.47	70.99
Missing	0.82
Work Status
Full-time	33.59	63.94
Not Full-time	65.79	36.06
Missing	0.62
Urbanicity (RUCA Codes)
Metro	88.61	88.23
Non-Metro	11.39	11.77
Region
Northwest	44.31	43.31
East	26.94	27.97
Southwest	28.75	28.72
	Unweighted M (SD)	Weighted M (Std Error)
Age	21.78 (2.30)	21.87 (0.06)

RUCA = rural urban commuting area.

To assess impaired driving-related behaviors participants reported how many times, in the past 30 days, they drove a car or other vehicle 1) after consuming alcohol (DUI-A), and 2) within 3 hours after using cannabis (DUI-C); and they had been a passenger in car or other vehicle when the driver 3) “probably had too much to drink to be driving” (RWI-A), and 4) had used cannabis within 3 hours of driving (RWI-C). Response options were dichotomized to create past month impaired driving-related outcomes, with 0 indicating none in the past 30 days, and one indicating any (one or more) occasions in the past 30 days.

Perceived injunctive norms of all four impaired driving-related behaviors were assessed. Participants were asked, “How acceptable or unacceptable it is for someone your age in your community to…” drive a car or other vehicle after 1) probably having too much to drink to be driving and 2) within 3 hours of using cannabis; and be a passenger in a car or other vehicle 3) when the driver probably had too much to drink to be driving and 4) when the driver had used cannabis within 3 hours of driving. Response options included 1 (Totally Unacceptable), 2 (Somewhat Unacceptable), 3 (Somewhat Acceptable), and 4 (Totally Acceptable). Since impaired driving outcomes were dichotomized as none versus any, and the ultimate prevention goal is to have zero occasions of impaired driving behaviors [37], these injunctive norm variables were dichotomized into 0 Totally Unacceptable and one all other response options (i.e., Somewhat Unacceptable, Somewhat Acceptable, and Totally Acceptable). Alternate coding, described in the statistical analysis section, was considered for sensitivity analyses.

Past month frequency of alcohol and cannabis use was assessed by asking participants to report the number of days they drank alcohol and the number of days they used cannabis in the past 30 days.

Statistical analysis

To assess the associations between injunctive norms and impaired driving-related behaviors, a series of logistic regression models were conducted using Proc Surveylogistic [38] in SAS software version 9.4 [39]. This was done in two steps. First, Model one regression analyses included injunctive norms while also controlling for the aforementioned demographic covariates (i.e., sex assigned at birth, geographic region, and race/ethnicity, age, college status, work status, and urbanicity). To understand the association between injunctive norms and outcomes while accounting for the possible associations between these variables and the frequency of substance use, Model two regressions also included frequency of use of the substance related to the outcomes (e.g., in the models examining DUI-A and RWI-A, we added frequency of alcohol use as a covariate).

A series of sensitivity analyses were also conducted. First, injunctive norms variables using the full scale of response options were considered as possible ordinal predictors. To assess this, ordinal logistic regressions predicting each of the injunctive norms were conducted. All models indicated violations of proportional odds. Due to this violation, the prevention goal of having zero impaired driving occurrences, and to ease the interpretability of results, injunctive norm variables were deemed as most appropriate to be dichotomized into 0 Totally Unacceptable and one all other response options. Second, Model two regressions were performed utilizing truncated samples of participants who reported using the relevant substance in past 30 days (e.g., DUI-A was examined only among participants who reported drinking alcohol at least once in the past 30 days).

Results

Weighted percentages for response options for impaired driving-related outcomes and injunctive norms are reported in Table 2. A total of 11.5% reported DUI-A, 12.4% reported DUI-C, 10.9% reported RWI-A, and 20.9% reported RWI-C at least once in the past 30 days. Results indicate overlap between the outcomes indicating that some participants who reported one impaired driving-related behavior also reported other impaired driving-related behaviors. For example, among participants who reported DUI-C, 22.7% also reported RWI-A and 65.6% reported RWI-C. More than three-fourths of WA YAs perceived DUI-A and RWI-A as Totally Unacceptable, whereas only about one-third reported DUI-C and RWI-C as Totally Unacceptable.

Table 2.

Percentages of responses for impaired driving-related behaviors and injunctive norms

	Past month DUI-A		Past month DUI-C		Past month RWI-A		Past month RWI-C		Injunctive DUI-A		Injunctive DUI-C		Injunctive RWI-A		Injunctive RWI-C
	No	Yes	No	Yes	No	Yes	No	Yes	Totally un-acceptable	All other responses	Totally un-acceptable	All other responses	Totally un-acceptable	All other responses	Totally un-acceptable	All other responses
Total Sample	88.5%	11.5%	87.7%	12.3%	89.1%	10.9%	79.1%	20.9%	85.3%	14.7%	33.9%	66.1%	76.3%	23.7%	36.6%	63.4%
Past Month DUI-A
No			90.0%	10.0%	93.0%	7.0%	81.8%	18.2%	87.0%	13.0%	36.8%	63.2%	77.6%	22.4%	39.4%	60.6%
Yes			71.0%	29.0%	61.8%	38.2%	60.5%	39.5%	73.2%	26.8%	12.6%	87.4%	68.7%	31.3%	16.7%	83.3%
Past Month DUI-C
No					90.8%	9.2%	85.5%	14.5%	86.3%	13.7%	38.0%	62.0%	77.2%	22.8%	40.8%	59.2%
Yes					77.3%	22.7%	34.4%	65.6%	64.5%	35.5%	16.5%	83.5%	50.8%	49.2%	15.2%	84.8%
Past Month RWI-A
No							83.6%	16.4%	87.9%	12.1%	36.0%	64.0%	79.5%	20.5%	39.3%	60.7%
Yes							42.9%	57.1%	64.5%	35.5%	16.5%	83.5%	50.8%	49.2%	15.2%	84.8%
Past Month RWI-C
No									86.8%	13.2%	40.6%	59.4%	78.9%	21.1%	43.9%	56.1%
Yes									79.7%	20.2%	8.5%	91.5%	66.6%	33.4%	9.3%	90.7%

Prevelances are provided in weighted percentages. All other responses collapse response options: “Somewhat Unacceptable,” “Somewhat Acceptable,” and “Totally Acceptable”; DUI-A = Driving under the influence of alcohol; DUI-C = Driving under the influence of cannabis; RWI-A = Riding with a driver under the influence of alcohol; RWI-C = Riding with a driver under the influence of cannabis; all impaired driving-related behaviors assessed for the past 30 days.

Odds ratios for Models one and two for all outcomes are presented in Figure 1. Compared to those who reported DUI-A was Totally Unacceptable, participants who reported any other (i.e., more favorable) injunctive norm for DUI-A had 2.38 times greater odds (95% C.I. = 1.61, 3.52) of DUI-A. After controlling for frequency of alcohol use, which itself was positively associated with DUI-A (OR = 1.15; 95% C.I. = 1.12, 1.19), the odds ratio reflecting the greater likelihood to DUI-A for those with more favorable injunctive DUI-A norms remained statistically significant and in fact increased (OR = 2.47; 95% C.I. = 1.64, 3.72). Compared to those who reported DUI-C was Totally Unacceptable, participants who reported any other injunctive norm for DUI-C had 11.90 times greater odds (95% C.I. = 6.53, 21.67) to DUI-C. Including cannabis use frequency in the model, higher frequency of use was associated with significantly higher odds to DUI-C (OR = 1.15; 95% C.I. = 1.13, 1.17), and the association between injunctive norms and DUI-C remained significant (OR = 7.93; 95% C.I. = 4.24, 14.83). Compared to those who reported RWI-A was Totally Unacceptable, participants who reported any other injunctive norm for RWI-A had 3.80 times greater odds (95% C.I. = 2.65, 5.44) to RWI-A. When adding frequency of alcohol use into the model, higher frequency of use was associated with 9% higher odds to RWI-A (95% C.I. = 1.06, 1.12), and the association between injunctive norms and RWI-A remained significant (OR = 3.98; 95% C.I. = 2.79, 5.67). Lastly, compared to those who reported RWI-C was Totally Unacceptable, participants who reported any other injunctive norm for RWI-C had 7.73 times greater odds (95% C.I. = 5.34, 11.18) to RWI-C. When adding frequency of cannabis use into the model, greater frequency of use was associated with 5% higher odds to RWI-C (95% C.I. = 1.04, 1.07), and the association between injunctive norms and RWI-C remained significant (OR = 6.58, 95% C.I. = 4.52, 9.60). Sensitivity analyses using truncated datasets provided similar results as analyses using the entire sample (see Table A1).

Figure 1.

Adjusted odds ratios (AOR) and 95% confidence intervals from logistic regressions estimating associations between more permissive injunctive norms of impaired driving-related behaviors and impaired driving-related outcomes in the past 30 days. Model 1 includes all demographic covariates. Model 2 includes all demographic covariates and frequency of use of relevant substance (i.e. alcohol or cannabis).

Discussion

The current study assessed the associations between injunctive norms (i.e., acceptability) of both driving under the influence and riding with a driver under the influence of alcohol and cannabis and reported impaired driving-related behaviors (i.e., DUI-A, DUI-C, RWI-A and RWI-C) among a large sample of YAs in WA. Results indicated that over 10% of YAs in WA reported each of the four impaired driving-related outcomes in the past 30 days. RWI-C was more prevalent than the other behaviors with over 20% reporting engaging in it at least once in the past 30 days. Moreover, there was considerable overlap of the behaviors, indicating some participants engaged in multiple types of impaired driving-related behaviors. The most consistent overlap was with RWI-C, as 40% or more of participants who reported DUI-A, DUI-C, or RWI-A also reported RWI-C. This suggests individuals who have recently engaged in one type of impaired driving-related behavior may have increased willingness to engage other types of impaireddriving related behaviors. This is also consistent with previous research that has showed positive associations between RWI and DUI behaviors in adolescents [40].

A greater percentage of YAs perceived DUI-C as more acceptable compared to DUI-A. This pattern was similar for RWI-C compared to RWI-A. This continues to suggest that cannabis impaired driving-related behaviors are perceived differently from alcohol impaired driving-related behaviors. This finding is similar to national data showing differences in peer norms that suggest an increasing acceptability of cannabis, but not alcohol, use [24]. National Monitoring the Future data show declines in disapproval of cannabis use over the past five years (e.g., suggesting greater acceptability of cannabis use), with rates of disapproval of alcohol use remaining steady. Greater perceived acceptability of cannabis may be accounted for, in part, by a growing number of YAs who report their friends use cannabis. For example, the percentage of YAs who reported most or all of their friends used cannabis has increased from 15% in 2010 to 24% in 2020 [24]. Therefore, prevention efforts may need material that addresses cannabis impaired driving behaviors and alcohol impaired driving behaviors separately as well as jointly, in addition to continuing to address harms associated with substance use given the increasing rates of acceptability. Additionally, evaluations of prevention and intervention programs need to assess these four behaviors individually. This may also suggest tailored programs could be beneficial depending on type of substance use and whether the individual reports DUI, RWI or both.

More permissive injunctive norms were found to consistently be positively associated with each of the impaired driving-related outcomes, even after controlling for frequency of alcohol or cannabis use. This indicates YAs who perceived it was anything other than Totally Unacceptable (i.e., Somewhat Unacceptable, Somewhat Acceptable, Totally Acceptable) for someone their age in their community to engage in each of the outcomes had higher odds of engaging in each of the impaired driving-related behaviors themselves. Prevention efforts highlighting accurate injunctive norms and changing perceptions of DUI and RWI to be unacceptable may be effective at reducing impaireddriving related behaviors. Importantly, odds ratios for the associations between injunctive norms and outcomes were considerably higher for cannabis-related outcomes (i.e., DUI-C, RWI-C) compared to alcohol-related outcomes (i.e., DUI-A, RWI-A), and slightly higher for RWI-A compared to DUI-A. This may indicate that programs directed at changing injunctive norms may have the potential to have a greater impact for cannabis-related outcomes and RWI-A.

While findings indicate injunctive norms may influence DUI and RWI behaviors, additional work is needed to develop and evaluate prevention efforts for this work to be applied to public health practice. Prevention and intervention efforts aimed at changing attitudes related to DUI and RWI could utilize personalized normative feedback which contrasts one’s own belief or attitudes about impaired driving behaviors, particularly alcohol-related driving behaviors, with that of their peers. Personalized normative feedback has shown to change other behaviors, such as alcohol use, through changing norms [12,41]. This could be provided within a motivational interviewing (MI) approach [42]. Previous efforts have utilized brief MI interventions to enable discussion of perceived “not so good” and “good” effects of alcohol and have shown they may reduce DUI-A [43-45]. Incorporating injunctive norms specific to impaired driving-related behaviors may further enhance these intervention effects and expand them to DUI-C and RWI behaviors. Potential avenues these could be added to are driver’s education classes, emergency room visits (especially for individuals injured as a passenger of an impaired driver), and mandated drug and alcohol counseling for individuals who were arrested for DUI. Further, injunctive norms may also be used in larger media campaigns. One county in WA has employed positive descriptive norm (i.e., perception of how many people do not engage in a risk behavior [46,47]) messaging in the past [48], and this platform may be expanded to present injunctive norms. Special consideration should be taken regarding evaluations of such messaging campaigns to ensure they are having the intended effect.

Limitations and Future Directions

There are limitations to the present study. Participants included YAs residing in WA where cannabis is legal for purchase for those 21 years and older. This limits generalizability of findings to individuals in other states, including states with varied levels of the legal status of cannabis. Additional safety behaviors related to driving, including seatbelt use and driving speed, were not assessed. Future work should include these as covariates and explore other influences on injunctive norms, including community-level factors such as availability of public transportation, proximity to alcohol and cannabis outlets, and amount of law enforcement presence. Research should also compare results from these injunctive norms of a similar aged individual within their community with more specific norms on DUI and RWI such as perceived approval from parents, siblings, and significant others. The current study was cross-sectional and longitudinal data assessing the impact of injunctive norms on these behaviors is needed, and their potentially bidirectional association, especially given that injunctive norm development may be impacted by previous DUI and RWI. It is unclear how childhood and adolescent experiences impact YA impaired driving-related injunctive norms, and understanding these longitudinal associations may be important when considering both prevention efforts aimed at pre-driving adolescents, as well as whether the effects of prevention efforts for YAs focused on changing injunctive norms may vary depending on their previous RWI and DUI experiences.

Previous research indicates there is variability in assessment of injunctive norms [49]. The injunctive norms utilized in this study assess acceptability for “someone your age in your community” to engage in the relevant behaviors. Due to the wording of the question, there is the potential for participants to conflate their own approval of these impaired driving behaviors with the perceived approval of this referent group. Therefore, future research should assess both personal approval and injunctive norms simultaneously to confirm the effect of injunctive norms on these behaviors. Further additional work is needed in assessment of injunctive norms given their variability in the literature and subsequent challenges create for understanding their impact on behaviors [49].

Importantly, future work could include qualitative assessments of why cannabis impaired driving behaviors are viewed as safer and more acceptable than alcohol impaired driving behaviors to inform prevention messaging. Lastly, deaths from vehicle crashes, including from impaired drivers, have increased both in WA and nationally during the COVID-19 pandemic.[50, 51] Future research should examine these associations during the peak and subsequent fall of the pandemic and associated changes to daily life to understand differences from pre-pandemic associations.

Conclusion

DUI and RWI are significant and acute public health issues. While most YAs do not engage in these impaired driving-related behaviors, this study finds more than 10% report engaging in DUI-A, DUI-C, RWI-A or RWI-C at least once in the past month. Equally concerning is that while the majority of YAs report alcohol impaired driving-related behaviors as Totally Unacceptable, when impairment was caused by cannabis only about one-third of YAs considered these behaviors as Totally Unacceptable. Future research and prevention/intervention efforts and media campaigns need to continue to find effective ways to address these injunctive norms and reduce these risky driving-related behaviors.

IMPLICATIONS AND CONTRIBUTION.

Between 11% and 21% of a sample of young adults reported driving under the influence or riding with a driver under the influence of alcohol or cannabis.

Young adults who perceived increased acceptability of these behaviors were more likely to engage in these behaviors themselves.