Counseling about Proper Use of Motor Vehicle Occupant Restraints and Avoidance of Alcohol Use while Driving: A Systematic Evidence Review for the U.S. Preventive Services Task Force Free

Motor vehicle–related injuries are the leading cause of death among individuals between 3 and 33 years of age in the United States (1) and are a large source of morbidity for the nearly 3 million people who sustain nonfatal injuries annually (2). Increasing the correct use of occupant restraint devices and decreasing alcohol-related driving (that is, driving while under the influence of alcohol or riding with drivers who are under the influence of alcohol) are among the most important strategies to effectively reduce motor vehicle–related fatalities (3–8). Overall, occupant restraint use has been increasing and is considered a public health success (9). All 50 states currently have laws requiring child safety seats for infants and children, and 49 states and the District of Columbia have adult seat belt laws (10). Although belt-positioning booster seats reduce the risk for injury by nearly 60% for children 4 to 7 years of age (11) compared with seat belts, 22 states do not have any laws pertaining to booster seats. All 50 states, the District of Columbia, and Puerto Rico have laws that make it illegal to drive with a blood alcohol concentration of 0.08 g/dL or higher (9), and rates of alcohol involvement among fatal crashes have decreased during the past 2 decades (12).

Despite widespread regulation and overall increases in safer motor vehicle–related behaviors, recent crash data show that more than 50% of fatalities were among unrestrained occupants and nearly 40% involved alcohol (2). Primary care providers and their staff have many opportunities to intervene with patients about these health behaviors already known to reduce the risk for motor vehicle occupant injuries (MVOIs). Children and adolescents younger than 15 years of age average more than 2 visits per year to office-based physicians, and older adolescents and adults average 2 to 8 visits per year (13). Additional public health strategies, such as closing gaps in current laws (14) and implementing evidence- and population-based approaches (8, 15, 16), will be important to make further improvements in motor vehicle safety behaviors. These strategies could include components delivered by primary health care providers or their staff.

Our objective was to systematically assess the evidence on the effectiveness of primary care counseling among people of all ages to increase the correct use of age- and weight-appropriate occupant restraint devices and reduce alcohol-related driving. The Oregon Evidence-based Practice Center (EPC) conducted the review to assist the U.S. Preventive Services Task Force (USPSTF) in updating its 1996 recommendation (17). The full evidence report is available at www.preventiveservices.ahrq.gov. This article summarizes the review's findings.

In conjunction with members of USPSTF, we developed an analytic framework (Appendix Figure 1) and 4 key questions to guide our evidence review.

Key question 1 addressed the direct effect of counseling interventions on actual health risk reductions, whereas key questions 2 and 3 addressed effects on intermediate behavioral outcomes known to lead to health risk reduction. This report did not examine the evidence for the efficacy of health risk reduction for the targeted MVOI-related safety behaviors, because the USPSTF found strong evidence for those relationships in 1996 (17). Correct use is defined by age, weight, and location as recommended by traffic safety organizations (18). Appendix Table 1 describes recommended occupant restraint devices for children younger than 9 years of age. Children younger than 13 years of age should ride in the rear of the vehicle. Safety belts with straps across both the lap and shoulder are recommended for children who have outgrown booster seats, as well as for adolescents and adults.

We considered all studies that were included in the 1996 USPSTF recommendation, and we conducted 5 additional literature searches that were limited to English-language studies. For the key questions pertaining to occupant restraint use (1 and 2), we searched for relevant studies in MEDLINE, Cochrane Central Register of Controlled Trials, Cochrane Database of Systematic Reviews, PsycINFO, CINAHL, and Traffic Research Information Service (TRIS) published from 1992 to July 2005. We also searched the bibliographies of 4 systematic evidence reviews that addressed the effectiveness of counseling for occupant restraints in pediatric populations (16, 19–21). For the key questions addressing counseling about driving while under the influence of alcohol (1 and 3), we considered trials that were included in 3 recent systematic evidence reviews (22–24) and searched MEDLINE, Cochrane Central Register of Controlled Trials, Cochrane Database of Systematic Reviews, PsycINFO, CINAHL, and TRIS for studies published from 2002 to September 2005 to update the searches conducted for those reports.

In 1996, the USPSTF recommendation did not specifically address the effectiveness of counseling patients about riding with someone who was under the influence of alcohol (key question 3) or the harms of counseling (key question 4). To cover these 2 areas, we searched MEDLINE, Cochrane Central Register of Controlled Trials, Cochrane Database of Systematic Reviews, PsycINFO, CINAHL, and TRIS for studies published from 1966 to July 2005 and MEDLINE and TRIS for studies published from 1966 to September 2005, respectively. Although no key questions were related to cost, we searched the National Health Service Economic Evaluation Database for data published from the database's inception through July 2005. Literature searches are described in detail in Appendix Table 2 and were supplemented with outside source material from experts in the field.

Two authors reviewed each abstract for potential inclusion by using the inclusion and exclusion criteria described in Appendix Table 3. We conducted 5 searches to cover the separate focus of each key question, and we reviewed all abstracts for potential inclusion for any of the key questions. For all key questions, we included English-language reports of randomized, controlled trials (RCTs) or nonrandomized, controlled clinical trials (CCTs) and comparative observational studies that included patients of any age and were conducted in the United States or other similarly developed countries. Any intervention that included behavioral counseling as 1 of its components was considered. Studies were required to report 1 of the behavioral or health outcomes specified in our key questions and analytic framework or cost-effectiveness outcomes. We excluded studies rated as having poor quality on the basis of the criteria described in the following section.

To be within the scope of the USPSTF, interventions needed to be feasible for, or conducted in, a primary care setting or be available for primary care referral. Criteria for deciding whether the intervention was feasible for a primary care setting were developed previously by members of the Oregon EPC and the USPSTF. These criteria included 4 domains: 1) how the participant was identified, 2) who delivered the intervention, 3) how the intervention was delivered, and 4) where the intervention was delivered. Appendix Table 4 contains a more detailed description of these domains. For an intervention to be feasible for primary care referral, we required that it be conducted in a health care setting or be widely available in the community at a national level (such as a car seat–fitting station within a hospital). We excluded studies that enrolled selected populations (for example, injured or intoxicated patients recruited from an emergency department) that were not representative of patients normally seen in primary care.

Our review did not include programs that counseled risky or harmful alcohol users to reduce alcohol consumption, which was reviewed previously for the USPSTF (22). Rather, we required that alcohol-related counseling interventions target general primary care patient populations of any age and specifically advise patients to reduce drinking and driving (not just reduce overall use of alcohol).

Using the USPSTF's study design–specific criteria (25, 26), 2 authors rated the quality of all included studies and those excluded because of quality issues. For randomized, controlled trials, criteria included 1) the initial assembly of comparable groups (based on adequate randomization, including first concealment and whether potential confounders were distributed equally among groups); 2) maintenance of comparable groups (including attrition, crossover, adherence, and contamination); 3) important differential loss to follow-up or overall high loss to follow-up; 4) equal, reliable, and valid measurements (includes masking of outcome assessment); 5) clear definition of interventions; 6) all important outcomes considered; and 7) an intention-to-treat analysis. For nonrandomized, controlled trials or cohort studies, the initial assembly of comparable groups was judged on the basis of consideration of potential confounders, with either restriction or measurement for adjustment. In the analyses of results of nonrandomized studies, adjustment for confounders was a quality criterion. The USPSTF Methods Work Group has defined a 3-category rating of “good,” “fair,” and “poor” on the basis of these criteria. In general, a good-quality study meets all criteria well. A fair-quality study does not meet, or is not clear that it meets, at least 1 criterion but has no known important limitation that could invalidate its results. A poor-quality study has important limitations. The specifications are not meant to be rigid rules. Rather, they are intended to be general guidelines. Individual exceptions, when explicitly explained and justified, can be made. Appendix Table 5 describes the USPSTF quality criteria in detail.

For all included studies, 1 primary reviewer abstracted relevant information into standardized evidence tables and a second author checked the abstracted data. If the investigators disagreed on study content or quality, a third investigator reviewed the study and the final quality rating was based on agreement between 2 of the 3 reviewers. Studies receiving a final quality rating of “poor” (n = 23) were excluded. Major quality problems in studies rated as poor included noncomparable groups at baseline, attrition greater than 40%, and nonblinded outcome assessment by the interventionists or nonstandardized outcome assessment. Because many trials had several methodological problems but were not clearly biased, we rated some included studies as “fair to poor quality.” In general, fair-quality studies reported or matched on some important baseline characteristics, measured outcomes by observation, specified correct use, and had lower attrition. Fair- to poor-quality studies often did not report baseline characteristics, used self-reported outcomes, did not specify correct use, and had higher attrition rates.

We could not conduct quantitative synthesis for any key question because of heterogeneity of intervention methods, populations addressed, and settings. Instead, we qualitatively synthesized our results within categories, focusing first on the age of the population for which MVOI safety behaviors were addressed and second on the setting in which the population was identified and in which the intervention was delivered. Detailed qualitative summaries are reported in the full evidence report and are summarized in this review. For interventions targeting child safety seat use, results were also stratified by whether the program included a demonstration of correct child safety seat use or increased access through a free or discounted distribution program. We calculated absolute differences with 95% CIs for use of restraints between the intervention group and the control group, when sufficient data were reported, by using the RISKDIFF option of the FREQ procedure in SAS, version 8.2 (SAS Institute, Cary, North Carolina). This procedure uses a normal approximation to the binomial distribution to construct asymptotic CIs.

This research was funded by the Agency for Healthcare Research and Quality (AHRQ) under a contract to support the work of the USPSTF. Members of the USPSTF participated in the initial design and reviewed interim results and the final evidence review. The AHRQ had no role in study selection, quality assessment, or synthesis, although AHRQ staff reviewed interim and final evidence reports and distributed the initial evidence report for external review of content by outside experts, including representatives of professional societies and federal agencies. The final published systematic evidence review on which this paper is based was revised on the basis of comments from these external reviewers.

We reviewed 1289 abstracts and 155 complete articles for all key questions (Appendix Figure 2). Seventeen studies (9 RCTs [27–35] and 8 CCTs [36–43]) reported in 17 articles met our inclusion criteria (Tables 1 and 2): 7 from the 1996 USPSTF review, 6 from other systematic reviews or outside sources, and 4 from searches that were conducted for this review. No study that met our inclusion criteria was related to counseling about alcohol-related driving (key question 3) or the harms of counseling (key question 4). Appendix Tables 6, 7, 8, 9, 10, 11, 12, and 13 contain detailed evidence on all included studies. Narrative descriptions of individual included studies and a list of excluded articles describing reasons for exclusion are available in the full evidence report (22). Table 3 summarizes the overall quality of evidence according to USPSTF criteria (20) for each key question.

One large, fair-quality, group-level CCT (n = 286 676) reported the direct effects of behavioral counseling on the incidence of MVOIs among children from birth to 5 years of age (36) (Table 1). In the trial, interventions targeting child safety seat use included behavioral counseling components that were delivered in inpatient and primary care settings. These components were tested in the context of multiple, community-wide approaches to reducing MVOIs and other injuries. Investigators measured MVOIs through a hospital surveillance system during the year before the study and for the 2 years during which the injury prevention programs were conducted. During the 2 years of the programs, MVOI rates in the intervention communities decreased, whereas those in the control communities increased. The odds ratio of risk for MVOI during the preintervention period, compared with the intervention period, in the intervention communities was 2.78 times (95% CI, 1.66 to 4.66 times) as large as that for the control communities, after adjustment for socioeconomic status. Given the trial's nature, however, the effect of the clinical counseling components on MVOI reduction cannot be separately determined from community-based approaches. In addition, this trial was conducted in the early 1980s, when the Massachusetts state legislature was debating a child automobile restraint bill and the baseline use of child restraints was 49%, which is much lower than the current prevalence of use. The effects of clinical counseling about child safety seat use on MVOI in the current, widely regulated environment remain untested. Similarly, we found no study that reported health outcomes of counseling interventions targeting the use of booster seats or safety belts for older children, adolescents, or adults or of interventions targeting alcohol-related driving for any age group.

All 17 included trials reported on the use or correct use of age- and weight-appropriate restraints. Most of these were published during the late 1970s or 1980s, before the widespread enactment of child safety seat or safety belt legislation. Overall, the trials addressed patient populations of all ages, with the most extensive literature involving infants and toddlers up to 4 years of age (6 RCTs [27–32] and 7 CCTs [36–42]) (Table 1). One RCT (33) focused on booster seat use among children 4 to 7 years of age, and 3 trials targeted safety belt use among older children, adolescents, or adults (1 CCT [43] and 2 RCTs [34, 35]) (Table 2). Across all trials, the tested behavioral counseling interventions comprised a wide range of educational approaches, including counseling by clinicians, written materials, films on automotive safety, live demonstrations of child safety seat use, and group-level informational sessions. Five trials included both educational components and the distribution of a free or reduced-cost child safety seat (29, 31, 39, 40) or booster seat (33). One trial included education plus behavioral reinforcement components (28). Interventions were conducted in primary care outpatient clinics and inpatient maternity wards (where clinicians encounter mothers of newborn infants) or met the criteria for being feasible for delivering in primary care or for referral for primary care. Table 1 describes trials among infants and children up to 4 years of age and stratifies the studies by setting and time of delivery (antepartum period vs. during well-child care). Interventions also varied in other characteristics, such as who delivered the counseling or the intensity of the counseling. All trials had methodological flaws (7 fair-quality and 10 fair- to poor–quality trials), and none were of good quality according to the USPSTF criteria. Because trial characteristics were highly heterogeneous and study quality was limited, we did not quantitatively combine results across any of the trials.

Results varied widely among the trials targeting increased use of child safety seats among infants and toddlers up to 4 years of age. The largest absolute differences in use between intervention and control groups (47% to 72%) were reported in 3 of 4 trials that assessed interventions of combined education with distribution of a child safety seat (2 RCTs [29, 31] and 1 group-level CCT [39]). In these trials, the interventions were delivered either during the end of pregnancy or during the peripartum hospitalization to newly postpartum women. These large differences in use or correct use were measured at the initial follow-up either immediately after hospital discharge (29, 31) or 9 months after the intervention (39). The effects were diminished at subsequent follow-up measurements (1% to 43%). Several methodological limitations may have led to an overestimate of effect size among these trials. The 2 RCTs included very small samples (n = 14 and 30), and the third trial measured outcomes by self-report, did not specify correct use, and excluded 13% of infants eligible for the intervention group because their caregivers did not accept the car seat loan. In contrast to these findings, a fourth trial that also assessed an intervention of education plus child safety seat distribution reported a small difference in correct use of child safety seats between the intervention and control groups (absolute difference, 5% to 7%) (40).

Of the remaining trials targeting children up to 4 years of age, 4 trials (3 fair-quality [36, 37] or fair- to poor-quality [38] CCTs and 1 fair- to poor-quality RCT [27]) evaluated counseling by pediatricians during well-child care. Two trials measured follow-up at 2 months and reported an increase in restraint use, ranging from an absolute difference between the intervention and control group of 13% to 21% (37, 38). Trials that reported initial or repeated follow-up later than 2 months reported that use was similar between the intervention and control groups (absolute difference, 2% to 6%) (27, 36, 37). The remaining trials targeting child safety seat use varied in setting, when the intervention was delivered, and results, and each had multiple methodological flaws (all were rated fair to poor quality) (28, 30, 32, 42). Among the trials targeting increased use of infant or child safety seats, a demonstration of correct child safety seat use was an intervention component of trials reporting larger increases in use or correct child safety seat use. More specifically, among the 8 trials that included a demonstration of correct child safety seat use (28, 29, 31, 37, 39–42), 6 reported increases in child safety seat use (absolute differences, 17.8% to 72%) (28, 29, 31, 37, 39, 42). Five trials did not include a demonstration of correct use as part of the intervention (27, 30, 32, 36, 38), and only 1 of these (a fair- to poor-quality CCT [38]) reported an increase in use.

Investigators of 1 fair- to poor-quality RCT evaluated an intervention to increase the use of booster seats among primarily low-income, African-American children 4 to 7 years of age (33). The intervention consisted of brief (5-minute) educational counseling by a certified car seat technician in an emergency department setting, delivered either with or without a free booster seat, and was compared with a usual care control group. On the basis of self-reported data at 1 month after the intervention was delivered, 98% of families in the education plus distribution group reported using a booster seat, compared with 5.5% of families in the other 2 groups (control or education only) combined (P < 0.001). The trial had several methodological problems that could have introduced bias, including high overall attrition (35%), differential attrition across treatment groups (40%, 39%, and 25%), self-reported outcomes, and analysis of only the completers. In addition, families who reported using a booster seat when presenting to the emergency department were not eligible for inclusion. Therefore, the magnitude of benefit from the education plus distribution programs in a general primary care population cannot be directly determined from these findings.

Investigators of 3 trials evaluated safety belt use among children, adolescents, or adults. Of these, 1 fair-quality CCT reported short-term improvement in observed seat belt use in the intervention group compared with the control group immediately after the intervention (38% vs. 5%; P < 0.001), but investigators analyzed only children and adolescents who were not wearing seat belts when arriving to the visit (43). Investigators of the remaining 2 trials measured seat belt use at 6 to 36 months of follow-up and reported that use was similar between intervention and control groups at these times (34, 35). Among these trials, the most intensive seat belt intervention was tested in the Dartmouth Prevention Project (34), a large, fair-quality, cluster-randomized RCT evaluating an office-based structured prevention intervention delivered to 3145 fifth- and sixth-grade children. Counseling by a pediatrician or nurse practitioner during well-child care visits was supplemented by a contract for a family policy, reinforcement of the message at subsequent office visits over 36 months, written materials mailed to the home, and telephone calls alternately targeting the parent and child. No statistically significant differences were found in the proportion of children who reported always wearing seat belts during the previous month at the 12-, 24-, or 36-month follow-up (odds ratios, 0.87, 0.96, and 0.89, respectively [all 95% CIs include 1.0]).

For all age groups, the volume and quality of research were inadequate to quantitatively address questions about essential elements of efficacious interventions, other positive outcomes from behavioral counseling interventions addressing seat belt use, or maintenance of MVOI safety behaviors after behavioral counseling interventions. The complexity of the type of restraint device needed (for example, child safety seat, booster seat, or safety belt) and receptivity of the patient because of differences in age influence the type of intervention that may be needed. These issues were not specifically addressed in the studies.

Our searches found no studies of primary care interventions evaluating behavioral counseling in general populations to reduce driving while under the influence of alcohol or riding with drivers who are under the influence of alcohol.

Our searches found no studies of adverse effects of counseling to use age- and weight-appropriate restraints or reduced driving while under the influence of alcohol or riding with drivers who are under the influence of alcohol.

The evidence for reducing the risk for injury and death when using recommended motor vehicle occupant restraints has been previously demonstrated to be strong. The current prevalence of restraint use is near the Healthy People 2010 goal of 100% use for infants and is more than 90% for children age 1 to 3 years (44). Incorrect use, however, remains common in these age groups and diminishes the level of protection provided. Restraint use is less prevalent among children 4 to 7 years of age, among whom premature advancement to seat belts increases the risk for injuries and among adolescents and adults (20% to 25% nonuse in these age groups) (45).

The available scientific literature provides fair evidence that, among infants and children up to 4 years of age, behavioral counseling interventions have been effective in increasing short-term correct use of infant and child safety seats at the time of hospital discharge or within 2 months after initially delivering the intervention. Effects have subsequently diminished, in many cases because use increases over time in groups without intervention. Many of the successful interventions included a demonstration of correct safety seat use. The largest effect sizes were seen among the trials that included a safety seat distribution program through a reduced-cost loan or giveaway program. Several interventions that did not include distribution programs, however, were also effective, at least in the short term.

Most of the studies included in this review had multiple methodological flaws, and no study was of good quality according to the USPSTF criteria. Some of the better-quality trials that were most relevant to the primary care setting were nonrandomized, controlled trials conducted during the late 1970s to 1980s, when many states first began passing child seat restraint laws. Some studies reported observed prevalence of correct use of child safety seats in fewer than 10% of the study population. Awareness and attitudes about restraint use differed greatly from current awareness and attitudes. Experts in the field, including authors of previous evidence reviews, have expressed concern about the limited quality and lack of recent studies in this body of evidence (19, 20, 46), especially given the magnitude of public health burden.

Because of recently revised safety recommendations, only 28 states currently have laws that apply to children in booster seats, and most of these do not cover all children up to age 8 years (10). We found few relevant data describing the effectiveness of primary care clinician counseling of children or parents on the use of booster seats. One recent trial conducted in an emergency department setting demonstrated a large increase in self-reported use among low-income families that received both education and a free booster seat (33). The intervention, however, was delivered by certified car seat technicians who had undergone intensive training, and it required the distribution of a free booster seat to be effective. Translating these findings to the primary care setting is not yet tested. Few adequate-quality trials have evaluated counseling to increase seat belt use among older children, adolescents, and adults.

We found no research addressing the effect of behavioral counseling interventions delivered to unselected patients in primary care to reduce alcohol-related driving or riding with an impaired driver. The USPSTF, however, has previously recommended screening and brief interventions for alcohol misuse in primary care (47), and these interventions may also improve alcohol-related MVOIs. Among the trials included in the systematic review on primary care screening and interventions for risky and harmful alcohol use that were prepared to support the USPSTF recommendation (22), 1 RCT found a reduction in the self-reported rate of driving after drinking among risky or harmful alcohol users who received an intervention to reduce alcohol use (48). Another RCT of a similar intervention reported a reduction in motor vehicle crashes with nonfatal injuries in a subanalysis of adults 18 to 30 years of age (49). More than 80% of alcohol-impaired driving episodes are reported by people who also reported binge drinking (50). Thus, screening all patients for alcohol misuse and then intervening with risky and harmful users (instead of counseling all primary care patients about reducing alcohol-related driving) may be the best evidence-based approach that is currently available for primary care clinicians.

Interventions to reduce MVOIs are complex, involving both regulatory approaches and behavioral counseling through community-based and clinical settings. Furthermore, the complexity of behavioral counseling approaches changes as the individual ages (complex restraint systems for infants and children, inherent risk-taking behavior for adolescents, and established behaviors for adults). Although primary care behavioral counseling interventions to increase correct age- and weight-appropriate restraint use among infants and toddlers may increase short-term use of restraints, these effects may diminish over time. Effective interventions included education, demonstration of correct use, and child safety seat distribution programs and were tested during a time of growing cultural support and increasing regulatory requirements for child safety restraint use. Data from primary care studies were lacking for interventions to increase the use of belt-positioning booster seats for children 4 to 8 years of age, an area where interventions are needed because of lower use and gaps in current child safety seat legislation. Similarly, no interventions targeting drivers 16 to 24 years of age, a known high-risk group, were available. Data describing the effects of behavioral counseling interventions among adults were also limited. Across age groups addressed, recent or good-quality trials for any MVOI-related safety behaviors were lacking. Many of the available studies were conducted when restraint use was less common, and the studies in populations with higher baseline use did not show improvements in restraint use, suggesting a possible ceiling effect. Misuse of child safety restraints remains common and diminishes their effectiveness.