Screening and Brief Advice to Reduce Adolescents’ Risk of Riding With Substance-Using Drivers

Abstract

Objective:

Alcohol- and drug-related car crashes are a leading cause of death for adolescents in the United States. This analysis tested the effects of a computer-facilitated Screening and Brief Advice (cSBA) system for primary care on adolescents’ reports of driving after drinking or drug use (driving) and riding with substance-using drivers (riding).

Method:

Twelve- to 18-year-old patients (N = 2,096) at nine New England pediatric offices completed assessments only during the initial 18-month treatment-as-usual (TAU) phase. Subsequently, the 18-month cSBA intervention phase began with a 1-hour provider training and implementation of the cSBA system at all sites. cSBA included a notebook-computer with self-administered screener, immediate scoring and feedback, and 10 pages of scientific information and true-life stories illustrating substance-related harms. Providers received screening results, “talking points” for 2 to 3 minutes of counseling, and a Contract for Life handout. Logistic regression with generalized estimating equations generated adjusted relative risk ratios (aRRR) for past-90-day driving and riding risk at 3- and 12-month follow-ups, controlling for significant covariates.

Results:

We found no significant effects on driving outcomes. At 3 months, cSBA youth were less likely than TAU to report riding with a drinking driver (aRRR = 0.70, 95% CI [0.49, 1.00]), and less likely to report riding with a driver who had used cannabis or other drugs (aRRR = 0.46, 95% CI [0.29, 0.74]). The effect was even greater (aRRR = 0.34, 95% CI [0.16, 0.71]) for riding with drinking drivers who were adult family members. All effects dissipated by 12-month followup.

Conclusions:

Screening and pediatrician brief advice shows promise for reducing adolescents’ risk of riding with substance-using drivers.

Thousands of adolescents in the United States die in motor vehicle crashes each year, and alcohol and drug use are major contributors (DuPont, 2011; Rehm et al., 2009). In 2015, one in four fatally injured teen drivers (26%) had a blood alcohol concentration (BAC) greater than .01 g/dL, and one in five (21%) had a BAC greater than .08 g/dL (National Highway Traffic Safety Administration [NHTSA], 2017a). Among younger motor vehicle fatalities (ages 14 and younger), one in six (17%) were driven by an alcohol-impaired driver (NHTSA, 2017b). In recent years, the rates of adolescents reporting recent (past 2 weeks) driving after using cannabis (12%) and riding with a driver who had used cannabis (19%) have steadily increased and now surpass alcohol-involved driving (9%) and riding (17%) rates (O’Malley & Johnston, 2013). Therefore, it is important to identify strategies that can effectively identify and mitigate these potentially fatal risky behaviors.

Youth should be screened for riding with an alcohol- or drug-using driver (hereafter, “riding”) beginning at age 12 years (Ewing et al., 2015) and for alcohol- and druginvolved driving (hereafter “driving”) at age 15 when driving usually begins. The CRAFFT, a widely recommended assessment developed specifically for use in pediatric medical settings, offers a brief screen for driving/riding risk (Knight et al., 2002). CRAFFT is a mnemonic acronym composed of the first letters of the key words in its six questions. The “C” question is, “Have you ever ridden in a CAR driven by someone (including yourself) who was ‘high’ or had been using alcohol or drugs?”

Several recent studies have validated substance use screens that only include questions about the frequency of past-12-months use (Clark et al., 2016; Levy et al., 2016a). A recent American Academy of Pediatrics Clinical Report recommends this approach (Levy et al., 2016b), but limiting screening to an assessment of patients’ personal use results in a lost opportunity to identify and intervene with those at risk because of riding with a substance-using driver, regardless of their own substance use.

To inform this discussion, we performed a secondary analysis of data from a 2012 study testing a computerfacilitated screening and pediatrician brief advice (cSBA) system based on the CRAFFT (Harris et al., 2012). Patients self-administered the screen on a computer in a private location before the physician visit, immediately received their CRAFFT score and risk level, and viewed 10 pages presenting science and true-life stories illustrating the health risks of alcohol and drug use. The system produced a report for the pediatrician with the screening results and a brief list of “talking points” designed to prompt 2 to 3 minutes of counseling.

All patients received a printed copy of the Contract for Life (Students Against Destructive Decisions [SADD], 2016), which asks adolescents never to drive after substance use or accept a ride from a substance-using driver. Instead, they agree to call their parent(s) for a safe ride home. Parents, in turn, agree to provide safe and sober transportation home and postpone any discussion until the following day. Pediatricians were instructed to ask all patients to discuss the Contract for Life with their parent(s) and to follow up with the provider if additional discussion was needed. This strategy preserved the confidentiality of high-risk patients while serving as a prevention strategy for both high- and low-risk patients.

The objective of this analysis was to examine the effect of cSBA at 3- and 12-month follow-ups on adolescents’ reports of substance-involved driving/riding and, specifically, riding with a substance-using adult family member.

Method

This was a secondary analysis of data collected during 2005–2009. Detailed methods are published elsewhere (Harris et al., 2012). Briefly, we conducted the study in nine primary care offices in New England. We compared cSBA to treatment as usual (TAU) in a quasi-experimental comparative effectiveness trial in which each site served as its own historical control group. At the start of the study, we instructed providers to continue their usual standard of care, and for the next 18 months we recruited and tested the TAU group. At crossover, we held a 1-hour provider training that comprised a demonstration of the cSBA system; a review of screening reports for low-, medium-, and high-risk patients; and a video that showed how to use the talking points and Contract for Life document in brief counseling. We then initiated the cSBA computer system at all sites and recruited and tested the cSBA group over the final 18 months.

Participants were 12- to 18-year-old patients arriving for routine primary care visits who provided informed assent/ consent. They completed the baseline assessment battery before the provider visit and identical assessments at 3 and 12 months after the visit. They received a $15 merchandise certificate after completing each measurement. The Institutional Review Boards of Boston Children’s Hospital and all other sites approved the study protocol.

Measures

The measures have been previously described in detail (Harris et al., 2017). Participants completed a computer selfadministered questionnaire assessing demographics; visit/provider characteristics; substance use by peers, siblings, and parents (scales derived from the Personal Experience Inventory; Winters, 1993); and a branching questionnaire assessing past-90-day driving and riding risks for alcohol and cannabis or other drugs. Only participants ages 16 years or older received the driving risk items. Participants indicating any riding were asked follow-up questions to determine if they rode with a driver who was age 21 or older and lived in the home. Youth answering affirmatively were categorized as riding with an adult family member. To assess for historical confounding within this asynchronous study design, we measured differences in the frequency of past-year exposure to substance use–related messages received outside of the study intervention and found no differences.

Data analyses

We conducted all analyses using SUDAAN® v.11.0 with site as the nest variable to appropriately account for correlated error arising from our multisite cluster-sampling design. We used an “intent-to-treat” approach, analyzing cSBA adolescents regardless of whether they reported receiving provider advice. We used chi-square tests for categorical variables and t tests for continuous variables to assess baseline equivalence of the two groups and to compare characteristics of those retained versus lost to follow-up. We dichotomized race (White non-Hispanic vs. other), parents in home (two vs. other), parent education level (≥college graduate vs. other), and type of visit (well visit vs. other) to ensure adequate cell sizes. All analyses of driving outcomes included data only for those ages 16 years or older (N = 1,068, 51% of total sample).

Because even one driving or riding instance could result in tragedy, we computed risks for any driving or riding with a drinking or cannabis/other-drug-using driver by collapsing responses into none/any. For analysis of the intervention effect at 3- and 12-month follow-ups, we used logistic regression modeling with generalized estimating equations to compute adjusted relative risk ratios (aRRR) for cSBA compared with TAU, controlling for baseline driving and riding variables, other identified baseline between-group differences, and the multisite sampling design.

Model-based missing data imputation was conducted to evaluate potential response bias in our follow-up effect estimates. Estimates of effect using the imputed data set were generally slightly stronger than the non-imputed, and our findings did not change, so we report non-imputed results.

Results

Sample characteristics

A total of 2,096 of 2,435 eligible patients (86.1%) completed baseline assessments. The mean (SD) age was 15.8 (2.0) years, 58.2% were girls, 64.6% were White/non-Hispanic, 11.0% were Hispanic, 7.2% were Asian/non-Hispanic, 10.4% were Black/non-Hispanic, and 6.9% self-categorized as other. At 3- and 12-months follow-up, retention rates for TAU were 70.7% and 72.4%, respectively, and for cSBA 74.0% and 75.1%, respectively. Compared with those retained at 3 months, noncompleters were older (M_age = 15.8 vs. 15.6, F = 10.4, p < .01); less likely to be White non-Hispanic (54% vs. 68%), χ²(1) = 41.9, p < .01; and more likely to come from a single-parent household (38% vs. 28%), χ²(1) = 181, p < .01; to have ever used alcohol (42% vs. 36%), χ²(1) = 5.5, p < .01 or cannabis (28% vs. 20%) χ²(1) = 19.4, p < .01; and to report having a substance-using parent (19% vs. 14%); χ²(1) = 7.4, p < .01. We found the same differences between 12-month completers and noncompleters. However, the profiles of noncompleters did not differ between cSBA and TAU.

Driving after drinking or drug use

At baseline, 4.7% of youth ages 16 years old and older reported past-90-days driving after drinking, 5.2% after cannabis/other drug use, and 8.9% after alcohol or drug use.

Riding with a substance-using driver

At baseline, 22.4% of youth (all ages) reported past-90-days riding with a drinking driver, 17.8% with a cannabis/drug-using driver, and 30.7% with a driver using alcohol or drugs. The TAU group had a significantly higher baseline rate compared with cSBA of reporting riding with a driver who had used cannabis/other drugs (Table 1).

Table 1.

Unadjusted group prevalence rates and adjusted relative risk ratios (aRRRs) for past-90-day driving and riding risk behaviors at baseline and 3- and 12-month follow-ups

	TAU	cSBA
Variable	n/N (%)	n/N (%)	aRRR [95% CI]a	p
Drove after drinking (among ≥16-year-olds)
Baseline	35/667 (5.2)	24/591 (4.1)	1.05 [0.57, 1.92]	.71
3 months	8/451 (1.8)	6/459 (1.3)	0.77 [0.25, 2.37]	.64
12 months	15/468 (3.2)	10/442 (2.3)	1.00 [0.64, 1.57]	.80
Drove after using cannabis or other drugs (among ≥16-year-olds)
Baseline	23/667 (3.4)	42/591 (7.1)	0.79 [0.44, 1.40]	.48
3 months	14/451 (3.1)	8/459 (1.7)	0.63 [0.25, 1.63]	.34
12 months	18/468 (3.8)	14/442 (3.2)	1.01 [0.36, 2.85]	.84
Rode with any driver who had been drinking
Baseline	278/1,068 (26.0)	192/1028 (18.7)	0.84 [0.65, 1.07]	.20
3 months	103/755 (13.7)	60/761 (7.9)	0.70 [0.49, 1.00]	.05
12 months	109/773 (14.4)	79/772 (10.3)	0.81 [0.59, 1.11]	.26
Rode with an adult family member who had been drinking
Baseline	101/1,068 (9.5)	71/1028 (6.9)	0.77 [0.56, 1.06]	.11
3 months	37/775 (4.9)	11/761 (1.4)	0.34 [0.16, 0.71]	.001
12 months	33/773 (4.4)	28/772 (3.5)	0.88 [0.51, 1.50]	.62
Rode with any driver who had used cannabis or other drugs
Baseline	228/1,068 (21.4)	145/1028 (14.1)	0.75 [0.58, 0.98]	.05
3 months	82/755 (10.9)	43/761 (5.7)	0.46 [0.29, 0.74]	.005
12 months	80/773 (10.6)	64/772 (8.4)	0.88 [0.61, 1.29]	.36
Rode with an adult family member who had used cannabis or other drugs
Baseline	27/1,068 (2.5)	10/1028 (1.0)	0.47 [0.22, 1.02]	.06
3 months	9/755 (1.2)	4/761 (0.5)	0.54 [0.15, 1.87]	.19
12 months	6/773 (0.8)	6/772 (0.8)	1.29 [0.34, 4.84]	.71

Notes: TAU = treatment as usual; cSBA = computer-facilitated screening and brief advice; CI = confidence interval.

^aAll baseline aRRR models adjusted for baseline past-12-month substance use, age, gender, race, and parent education, with precision estimates adjusted for the multisite sampling design. All 3- and 12-month follow-up models also included baseline driving/riding variables as control variables.

Nearly 1 in 10 (8.2%) reported being driven in the past 90 days by an adult family member who had been drinking, 1.8% by an adult family member who had been using drugs, and 9.3% by an adult family member who had used alcohol or drugs.

Among youth who reported any riding, 44.2% had not used any substances themselves during the past 12 months.

Intervention effect for driving

There were no significant intervention effects on driving outcomes at either 3- or 12-month follow-ups, likely because of small numbers (Table 1).

Intervention effect for riding

Interestingly, past-90-day riding rates showed substantial drops from baseline to follow-up in both cSBA and TAU. After adjustment for baseline riding and other covariates, cSBA youth had, compared with TAU, a 30% lower risk of riding with a drinking driver and 54% lower risk of riding with a driver using drugs at the 3-month follow-up, but the effects dissipated by 12 months (Table 1). The apparent effect size at 12 months for alcohol (19%), however, may warrant further testing in a larger sample.

Intervention effect for riding with substance-using adult family member

Similarly, in adjusted analysis, we found a 66% lower risk at 3 months for cSBA compared with TAU of riding in a car with an adult family member who had been drinking. This effect dissipated by the 12-month follow-up. For riding with an adult family member who had been using cannabis/other drugs, there were too few reports to draw any meaningful conclusion.

Discussion

This study shows that a brief primary care counseling intervention that includes the Contract for Life is a promising strategy to reduce the short-term risk of adolescents riding with a substance-using driver. The effect sizes of 30% and 54% for alcohol and drug use, respectively, required only a few minutes of clinician time. We found an even larger short-term effect when the driver in question was an adult family member.

Unfortunately, all effects dissipated by the 12-month follow-up, consistent with prior studies of brief interventions in primary care, which found that effect sizes tend to bestrongest at earliest follow-up points, with decay in intervention effects over time (O’Donnell et al., 2014). These results highlight the need for future research to identify practical strategies for use in primary care that reinforce and extend brief intervention effects over time.

The sizeable drop of approximately 50% in riding for the TAU group could have been a result of assessment reactivity (Walters et al., 2009) or a Hawthorne effect of being observed (Murray et al., 1988). However, we also found that about one third of TAU pediatricians advised their patients about driving/riding-related risk behaviors, which was about half the rate reported in the cSBA group. Yet, the drop in the cSBA group was significantly larger. A recent systematic review and meta-analysis of brief alcohol interventions for drinking and driving among youth (Steinka-Fry et al., 2015) identified 12 studies that tested an alcohol intervention of 5 hours or less duration over a 1-month period among 11- to 25-year-olds. Overall, they found a modest (0.15 SD) improvement but significant effect on drinking and driving that was primarily correlated with a post-intervention reduction in heavy use of alcohol. We could find no published studies examining the effects of brief interventions on riding with a substance-using driver (Williams et al., 2007), despite a number of reports showing that this is a serious concern among youth (Beck et al., 2010; Bell et al., 2005; Cartwright & Asbridge, 2011).

A number of potential study limitations should be noted. First, these data were collected in 2005–2009 and may not reflect current adolescent behavior. The assessment battery relied on adolescent self-report, which could introduce recall or social-desirability bias. Although we were not able to identify any reports assessing the reliability of adolescent self-report of riding/driving behaviors, research suggests that computer self-administered confidential questionnaires are associated with higher disclosure of sensitive behaviors (Beck et al., 2014; Newman et al., 2002). To avoid making parents secondary subjects, we did not ask participants directly whether it was a parent who was driving the car after drinking or drug use. In some instances it could have been an over-21 sibling. However, in a previous study, we found a significant correlation between adult family member driving after substance use and other items that measured parents’ substance use (e.g., “I am worried about my parent(s) drinking/drug use”; Harris et al., 2017). Our measure of riding with a driver under the influence of “cannabis or other drugs” does not allow differentiation of specific drugs and required adolescents to subjectively judge what the driver had been using. This study used a quasi-experimental asynchronous comparison group design, which can be subject to historical confounding. Future studies should be randomized trials. Additionally, although we gave all patients the Contract for Life, we do not know how often family members actually received these or made agreements to request and provide safe transportation home. Finally, we did not ask about drivers’ licenses, which would have given a more accurate denominator for our driving rates.

Study strengths included a large and diverse sample from a variety of primary care settings over several U.S. states, and testing an intervention approach that required minimal pediatrician time and preserved the confidentiality of youth at risk.

This study provides initial evidence supporting pediatrician screening and brief counseling for riding risk. It also highlights the importance of screening youth for riding with substance-using drivers. An updated version of the CRAFFT screener (CRAFFT 2.0) begins with simple questions regarding frequency of use during the past 12 months and then uses a skip-pattern to maximize efficiency (Harris et al., 2016). Those who report no use answer the CAR question only, whereas those who report use answer all six CRAFFT questions (www.CRAFFT.org). The computer-facilitated approach takes adolescents on average less than 5 minutes to complete (Showalter et al., 2016) and requires only minutes of a pediatrician’s time, making it a practical and efficient way to reduce risk of car crash injuries and death. Although the number of alcohol-related car crashes has recently declined, the rate of cannabis-related vehicle crashes is rising, especially in states that have legalized its use (Salomonsen-Sautel et al., 2014).

Conclusion

Pediatrician screening and brief counseling appears promising for reducing adolescents’ short-term risk of riding with a substance-using driver, especially when the driver is an adult family member. More studies are needed using randomized designs that test new strategies to extend the effect of the intervention.