Objective. To determine whether a brief session of behavior change counseling (BCC), offered to injured adolescents in the emergency department (ED) as a therapeutic intervention, could be used to change injury-related risk behaviors and the risk of reinjury.
Study Design. A randomized, controlled trial.
Participants. Adolescents between 12 and 20 years old who were undergoing treatment for an injury in the ED and who were cognitively able to participate in the intervention.
Setting. An urban ED at a level 1 pediatric trauma center.
Intervention. Study participants completed a baseline risk behavior prevalence assessment. Participants were then randomly assigned to receive BCC or routine ED care. Those in the treatment group underwent a brief session of BCC with a study social worker focused on changing an identified injury-related risk behavior (seatbelt use, bicycle helmet use, driving after drinking, riding with an impaired driver, binge drinking, or carrying a weapon). Participants were recontacted 3 months and 6 months after enrollment to assess the prevalence of positive behavior change and the interim occurrence of medically treated injuries.
Results. We enrolled 631 participants (78% of those eligible) and obtained follow-up for 76% at 3 months and 75% at 6 months. The relative risk of a positive behavior change with respect to seatbelt use was 1.34 (95% confidence interval [CI]: 1.00, 1.79) at 3 months, favoring the intervention group. The relative risk for the same outcome was 1.47 (95% CI: 1.09, 1.96) at 6 months. A positive change in bicycle helmet use was 1.81 (95% CI: 1.02, 3.18) times more likely at 3 months and 2.00 (95% CI: 1.00, 4.00) times more likely at 6 months in the intervention group. There was no effect of the intervention on changes in other target behaviors. Over the 6-month follow-up period, the risk of reinjury requiring medical attention did not differ between treatment groups.
Conclusions. Brief BCC can be delivered to adolescents undergoing treatment for injury in the ED and can be used to address injury-related risk behaviors. The intervention was associated with a greater likelihood of positive behavior change in seatbelt and bicycle helmet use. This effect lasted over 6 months of follow-up. BCC was not associated with changes in other risk behaviors and could not be shown to significantly reduce the risk of reinjury.
Injuries are the leading cause of death and disability among adolescents and young adults in the developed world.1,2 Many common injury mechanisms—such as motor vehicle occupant injuries, bicycle injuries, and injuries attributable to firearms or intentional injuries—are associated with specific risk behaviors. Interventions to change these risk behaviors will be required to reduce the toll taken by associated injury events.3,4
Many adolescents engage in 1 or more of these risk behaviors and teens, as a group, are at high risk for injury. Teen drivers and their passengers are more likely than adults to experience a motor vehicle collision.5 Although we know that seatbelt use reduces the risk of serious or fatal injury in a motor vehicle collision by as much as 73%,6 over 16% of adolescents report that they “rarely” or “never” use safety belts.7 Bicycle related injuries are also prevalent among young people, and head injury accounts for three quarters of bicycle-injury deaths.8 The use of bicycle helmets can reduce the risk of brain injury by as much as 88%.8,9 Unfortunately, over 85% of teen bicyclists report “rarely” or “never” wearing a helmet when riding.7
Alcohol use is another behavior that increases the risk of pedestrian injuries, falls, drowning, and burns and is strongly associated with homicide and suicide. The 30-day prevalence of “binge drinking” among American teens is over 30%.7 The association between alcohol use and motor vehicle collisions is also well known. Alcohol combined with teen driving inexperience increases the risk substantially.10 In national surveys, 13% of teens report driving after drinking alcohol at least once in the preceding 30 days, and almost 35% report riding with an impaired driver.7
Finally, carrying a weapon is associated with fighting at school and may be a risk factor for injury.11 The national 30-day prevalence of weapon-carrying in teens is estimated to be 17%.7 Among adolescents, these risk behaviors cluster in individuals. Teens who engage in 1 risk behavior are more likely to engage in others.12,13 Promoting behavior change in teens can be especially difficult, as adolescents are often ambivalent about the need for this change.
Motivational interviewing, or behavior change counseling (BCC), is a patient-centered, directive therapy used to address risky behaviors. It has been tested with success in medical settings.14–19 BCC interventions are brief discussions that strategically direct the patient to explore why one might change a target behavior, rather than prescribing skills for how to change.20 The goal is to achieve behavior change by increasing the patient’s motivation to change.21 Key activities in a BCC session include avoiding argument, showing empathy, enhancing self-efficacy, and developing a perceived discrepancy between the patient’s real and ideal behavior.22 This therapeutic technique is well described and can be identified, codified, and learned.23
BCC has been tested with adolescents in 5 randomized trials. In 2 studies, conducted in emergency department (ED) settings while adolescents were being treated for injuries, BCC reduced alcohol-related harm and tobacco use.14,18 In 3 studies not conducted in the ED, BCC reduced drinking among adolescents about to enter intensive substance abuse treatment, among college binge drinkers, and among seventh- to ninth-grade school athletes.19,24,25 No studies exist that address the use of BCC to reduce injury-related risk behaviors other than alcohol use.
We hypothesized that adolescents and young adults undergoing treatment for an injury in the ED would benefit from a brief BCC intervention to reduce injury-related risk behavior. We chose to intervene in the ED after an injury because we believed that the emotional distress and proximity to the injury might create a “teachable moment” of increased susceptibility to messages about behavior change. Our objective was thus to test the effectiveness of BCC in the reduction of injury-related risk behaviors and reinjury rate, when delivered in the ED to injured youth.
We conducted a randomized, controlled trial between February 1999 and July 2000. The study was conducted in the ED of an urban public hospital that also serves as the only level 1 pediatric trauma center for a large geographic area in the Pacific Northwest. Eligible participants were 12 to 20 years old, admitted to the ED for treatment of an injury, and cognitively able to participate in the intervention. We excluded minors whose guardians could not be located, those who were medically unstable, those whose presenting complaint was child abuse, domestic violence, or rape, and those who did not speak English. Youth who were intoxicated on admission were eligible for later enrollment if their mental status cleared before discharge. In a few cases, youth were enrolled after admission to a hospital ward if their medical condition in the ED precluded contact with the interventionist. Eligibility was not limited to those injured as a result of one of the targeted risk behaviors—any injured youth could be enrolled.
Study staff stationed in the ED identified all eligible participants during a 15:00–23:00 shift on weekdays and 12:00–24:00 shift on weekends. After obtaining participant and parental consent (for those under 18), the interventionist recorded basic demographic information and the participant completed a self-administered baseline behavior prevalence questionnaire. Administration of the questionnaire and all subsequent study procedures were conducted in confidence with parents absent from the room. We asked about the past 30-day prevalence of the following:
bicycle helmet use, among those who rode a bike;
driving after drinking, among those who drove;
riding with an impaired driver;
binge drinking (defined as ≥4 drinks on 1 occasion for females and ≥5 drinks on 1 occasion for males); and
carrying a weapon (a knife, club, or gun).
Participants were then randomly assigned to treatment or intervention groups using computer-generated cards filed in sequentially numbered, opaque, sealed envelopes. A statistician not otherwise involved in the study produced the cards. Interventionists opened the envelopes in the ED. Treatment assignment cards and envelopes were returned to study investigators to monitor for irregularities. Because the pattern and mutability of risk behaviors was likely to vary across the age span of adolescence, randomization was blocked by age group (12–14, 15–17, or 18–20 years) to allow meaningful subgroup analyses.
Youth assigned to the control group received standard medical care but no specific risk-reduction messages or counseling. Those assigned to the intervention group underwent a 20-minute session of BCC with a trained therapist. The intervention was conducted at the bedside in the ED while the participant was awaiting medical care. The interventionist reviewed the participant’s risk behavior questionnaire and selected 1 topic for discussion. Interventionists were instructed to select as a topic a risk behavior endorsed by the participant. Participants who endorsed no risk behaviors were asked to select a topic of most relevance to them.
The techniques of behavior change counseling used in the intervention have been well described elsewhere.20,22,23 Briefly, the interventionist began by establishing rapport with the subject. The interventionist then negotiated an agenda (in selecting a topic for discussion) and assessed the participant’s readiness to change the selected behavior (as conceptualized in Prochaska’s “stages of change”21). Most of the interview focused on increasing the perceived importance of the behavior change or bolstering the participant’s self-efficacy with respect to making desired changes. Throughout the interaction, the interventionist used specific techniques to deal with resistance from the participant and to impart additional information about the risk behavior in an impartial manner. Interventionists made an effort to conclude every interview on a positive and encouraging note.
The interventionists in this study were master’s-level social workers. Each received 8 hours of formal instruction in the philosophy and techniques of BCC, including role-playing exercises. Throughout the study, interventionists attended weekly supervision sessions with the study psychologist to develop and refine their skill with BCC. A subset of 30 interventions was also audio-taped and reviewed by the study psychologist to ensure that interventionists were conducting therapy that adhered to the spirit of BCC. These sessions were actively selected to insure that each interventionist was monitored.
All participants were recontacted by telephone 3 months and 6 months after enrollment. A research interviewer (not the enrolling interventionist) blinded to the participant’s intervention status readministered the risk behavior prevalence questionnaire. Youth were also asked about the interim occurrence and circumstances of injuries requiring medical care. We abstracted information from each participant’s medical record. This included the nature and severity of the index injury; whether that injury was related to a targeted risk behavior; and whether the participant had been treated for any subsequent injury in the trauma center over the 6 months after enrollment.
We conducted an intention-to-treat analysis; all participants for whom follow-up measures were available were analyzed with the group to which they were randomized. We assessed the occurrence of reinjury during the 6-month follow-up period. Injuries requiring medical care were identified by participant self-report and by a review of medical records at the trauma center. We considered all injuries requiring medical care as an outcome but also examined the incidence of those injuries attributable to risk behaviors targeted in our intervention. Initial sample size estimates called for enrollment of 400 participants in each treatment group to have 0.80 power to detect a diminution of the 6-month reinjury rate from 10% to 4.5% at the conventional 0.05 significance level.
Because behavior change was the other outcome of interest, we classified participants with respect to their change in specific risk behaviors at each follow-up assessment. Participants could be classified as having: 1) positive behavior change (based on Likert scores, frequency reports, or dichotomized behavior reports that improved between baseline and follow-up); 2) no behavior change; or 3) negative behavior change (based on Likert scores, frequency reports, or dichotomized behavior reports that worsened between baseline and follow-up).
We report these outcomes as risk ratios (with 95% confidence intervals [CIs]) that compare the risk of a positive behavior change among those randomized to receive the intervention to the risk among those assigned to the control group. In some cases we also report a risk difference which allowed us to estimate the “number needed to treat”—that is, the number of participants who would need to receive the intervention to effect 1 instance of the specified behavior change. In addition, we used multivariable logistic regression models, in which behavior change was the dichotomous outcome, to test for modification of the effect of the treatment by age, gender, or initial injury severity.
Because of the design of the study, behavioral outcomes are compared across entire treatment groups. For example, we report the prevalence of positive behavior change with respect to seatbelt use in the intervention group as a whole compared with the control group as a whole. This is done although only a proportion of the intervention group actually received an intervention about seatbelt use. As such, this is an inherently conservative analysis.
We could not examine the effect of a specific risk-behavior intervention within the randomized study because participants were not randomized by risk behaviors endorsed. However, we reasoned that participants endorsing a behavior at baseline formed a subgroup most likely to receive a motivational interview specific to that behavior if assigned to the treatment group. Therefore, we also report behavior change outcomes among intervention group participants who endorsed a behavior at baseline and received an intervention specific to that behavior and compare these with participants in the control group who also endorsed the behavior at baseline.
Participants received a gift certificate for $15 at a music store on enrollment and another gift certificate after completing the 6-month interview. The study design, study questionnaires, and study procedures were reviewed and approved by the Human Subjects Research Review Committee of the University of Washington.
The flow of eligible and enrolled participants through randomization and outcome assessment is shown in Fig 1. We identified 1252 potential study participants in the ED. Of these, 440 were excluded based on established criteria. We were unable to contact 93 potential participants (11%) before they were discharged from the ED. Another 88 (11%) refused consent. We thus enrolled 631 participants (78% of all eligible).
Characteristics of the study sample after randomization, including demographic and index injury information as well as the baseline prevalence of risk behaviors, are shown in Table 1. There were no differences between groups in age, gender, school attendance, insurance status, or initial injury severity. At baseline, the distribution of target risk behaviors was similar in the control and intervention groups. The most prevalent risk behavior was lack of bicycle helmet use among those who reported riding a bicycle. Failure to use a seatbelt “always” was also common.
We obtained telephone follow-up for 79% of intervention participants and 73% of control participants at 3 months (P = .15). At 6 months, 74% of participants in the intervention group and 76% of those in the control group were recontacted (P = .44). Compared with those whom we were able to contact, nonrespondents were more likely to be male (71% vs 62%; P =.01) and more likely to be in the 18- to 20-year age group (50% vs 30%; P <.01). The baseline demographic, behavior, and injury characteristics of the nonrespondents did not differ systematically by treatment group assignment.
Over 6 months of follow-up, 14.0% of the intervention group reported reinjury, whereas 15.8% of the control group required medical care for a second or subsequent injury event. The calculated risk ratio is 0.88 (95% CI: 0.59, 1.32). When the self-reported mechanism of reinjury was examined, 14 of the 82 reinjury events were related to 1 or more of the 6 target behaviors. Of these injuries, 6 occurred in the intervention group (2.1%) while 8 occurred among control group participants (2.9%), giving a risk ratio of 0.73 (95% CI: 0.26, 2.07). No statistically significant effects on reinjury rate were documented.
Both the intervention and the control group demonstrated a reduction in risk behavior prevalence after ED treatment for an injury, an effect noted among adults in other studies.16 For a program effect to be detected, it would have to exceed this natural trend toward positive behavior change that was seen even among participants who received no intervention (Fig 2).
The relative risk of a positive change for specific behaviors is shown in Table 2. Intervention group members were more likely to report a positive change in seatbelt and bicycle helmet use. These effects were noted at 3 months and persisted among those interviewed at 6 months. We found no significant effect of the intervention on driving after drinking, riding with an impaired driver, binge drinking, or carrying weapons. The rate of positive change in these behaviors was not associated with treatment assignment whether behavior was measured dichotomously (as reported) or on a continuous frequency scale (data not shown). We also examined the risk of adverse effects (a negative change in risk behavior) according to treatment group assignment. The intervention was not associated with an increased prevalence for any of the target behaviors.
The positive effect of the intervention on seatbelt use was modified by the participant’s age (P = .018 by likelihood ratio testing). Seatbelt use was most profitably addressed among participants 15 to 17 years old; at the 3-month follow-up, the intervention was associated with a relative risk of positive behavior change of 2.45 (95% CI: 1.46, 4.10) in this group as compared with 1.69 (95% CI: 0.75, 3.84) among 12- to 14-year-olds and 0.79 (95% CI: 0.41, 1.50) in 18- to 20-year-olds. There was no effect modification by age for any other behavioral outcome nor did we detect interaction between the intervention effect and the participant’s gender.
In a posthoc analysis, we also noted that intervention effects for seatbelt use and bicycle helmet use were modified by the severity of the initial injury. Initial injuries of greater severity (as measured by the Injury Severity Score) were associated with a greater likelihood of positive behavior change in both the intervention and control groups. However, the additional beneficial effect of the intervention was evident primarily among participants with the least severe injuries (Fig 3). Individuals with more severe injuries seemed to receive no benefit from the intervention beyond the behavior change associated with the injury itself.
When participants in the control group who endorsed a behavior at baseline are compared with participants in the intervention group who endorsed the same behavior at baseline and who underwent BCC specific to that behavior, the intervention’s effect on seatbelt and helmet use seems stronger. Table 3 demonstrates the relative risk of a positive behavior change with respect to seatbelt and bicycle helmet use. The number of participants whom one would need to treat to achieve 1 positive behavior change ranged from 1.8 to 5.5 for these outcomes. The intervention had no effect on other target behaviors, even when analysis was restricted to those receiving a specific intervention.
Behavior change counseling has been shown to be an effective and time-efficient therapeutic tool when used to address addictive behaviors and other health risks.26 Because identifiable behaviors are associated with the risk of injury among adolescents, these behaviors are logical targets for a therapeutic intervention. We hypothesized that offering a BCC intervention at a teachable moment, when youth are undergoing treatment for an injury, could reduce the prevalence of these behaviors.
We found that BCC conducted with injured youth in the ED was associated with a positive and sustained change in seatbelt and bicycle helmet use. The likelihood of positive behavior change with respect to seatbelt use among those receiving the intervention was 47% greater than among participants in the control group. The rate of positive behavior change in bicycle helmet use was doubled. We estimate that BCC would need to be used with 2 to 3 injured youth to effect 1 positive change in bicycle helmet use and with ∼5 injured youth to produce 1 positive change in seatbelt use. These behavior changes were durable over 6 months of follow-up.
We found no effect of the intervention on driving after drinking, riding with an impaired driver, binge drinking, or carrying a weapon. The risk of reinjury did not differ between treatment groups and, although the number of injuries attributable to 1 of the 6 target behaviors was lower in the intervention group, the total number of injury events associated with any target behavior was small, and the difference was not statistically significant.
One limitation of this study was inherent in the research design. We did not limit enrollment to adolescents who screened positive for an injury-related risk behavior. We made this decision for 2 reasons. First, we recognize the fluid nature of risk behavior in the teen years. Adolescents with no risk behaviors at baseline may acquire risk behaviors over time. In terms of the theory that informed our intervention, these teens might be considered to be in a “maintenance” stage of behavior change. The intervention could be used to support this desirable lack of risk behaviors. Our study design thus allowed us to look at the effect of intervention on the rate of negative behavior change. Second, we did not feel that we could enroll only those adolescents who screened positive for a risk behavior without jeopardizing their confidentiality when obtaining parental consent.
As a result of these theoretical and practical concerns, we randomized some young people with no reported risk behaviors. We then compared rates of behavior change according to treatment group assignment, even if most individuals in the intervention group did not receive an intervention specific to the behavioral outcome of interest. This makes detection of change in these low-prevalence risk behaviors very unlikely and may, in part, explain our failure to promote behavior change among participants who drove after drinking, rode with an impaired driver, drank to excess, or carried weapons.
Alternatively, adolescents may view some of these unchanged risk behaviors as self-protective: they may choose to ride with an impaired driver if they themselves are too impaired to drive or they may carry a weapon for self-defense. In addition, behaviors associated with alcohol use and driving typically involve groups of adolescents. Peer effects may make it more difficult to influence these behaviors than individual decisions to use a safety belt or to wear a bicycle helmet. And, finally, risk taking is a developmentally salient activity for at least some adolescents,13,27 suggesting that a proportion of these behaviors may be relatively intransigent.
Our study included adolescents over a broad age range (12–20 years). We recognize that the developmental processes underlying risk behavior change over this period of time and so stratified randomization to allow careful analyses by age group. We found that the intervention was especially effective in promoting a positive change in seatbelt use among 15- to 17-year-old participants. Teens who are just establishing patterns as drivers may find these behaviors to be more malleable than would older drivers with established seatbelt habits.28,29
Recognition of effect modification by age group could allow targeted interventions with subpopulations in whom the intervention is especially efficacious. Young people who are being treated for an injury that is associated with 1 of the target risk behaviors (eg, an unhelmeted bicyclist with a head injury) may constitute another subpopulation particularly receptive to the intervention. Our study did not have adequate power to test this hypothesis.
There are opportunities for additional study. Our interventionists were social workers with a background and ongoing training in the techniques of BCC. However, it is not known how much training or supervision is required to effectively use these techniques.26 Nor do we know how much time must be devoted to a session of BCC to demonstrate positive effects. It may be feasible to teach ED physicians, ED nurses, or ED social workers to perform brief motivational interviews with injured teens treated in the ED setting. The practicality and effectiveness of these approaches should be carefully evaluated.
The concept of a “teachable moment” also warrants additional study. We intervened when participants were at the hospital being treated for an injury under the assumption that adolescents would be especially open to behavior change at that time.14,18 However, BCC can certainly be performed in less acute settings. The added benefit of intervening shortly after an injury event remains an unanswered, empirical question.
We believe that we have shown BCC to be an efficacious and acceptable intervention to positively change 2 important injury-related behaviors among injured youth: seatbelt use and bicycle helmet use. Because we analyzed by treatment assignment rather than by baseline risk behavior, the results reported here are inherently conservative. We believe that a program that screened for seatbelt or bicycle helmet nonuse and offered BCC targeted to these identified risks would be more effective than our results might suggest. Additional study is required to identify subpopulations who might be especially amenable to this intervention, to determine the long-term durability of positive change and to assess the personnel and training requirements that would be required to implement such a program as a routine preventive aspect of trauma care.
This study was supported by an EMS-C Targeted Issues grant from HRSA/MCHB (6 H34 MC 00068-02) and by an NICHD Career Development Award (5 K23 HD01341-02) (Dr Johnston).
- Received September 24, 2001.
- Revision received February 15, 2002.
- Reprint requests to (B.D.J.) 325 Ninth Ave, Box 359960, University of Washington, Seattle, WA 98104. E-mail:
- ↵Miller W, Rollnick S. Motivational Interviewing: Preparing People to Change Addictive Behavior. New York, NY: Guilford Press; 1991
- ↵Rollnick S, Mason P, Butler C. Health Behavior Change: A Guide for Practitioners. Edinburgh, Scotland: Churchill Livingstone; 1999
- ↵Miller W, Zweben A, DiClemente C, Rychtarik R. Motivational Enhancement Therapy Manual: A Clinical Research Guide for Therapists Treating Individuals With Alcohol Abuse and Dependence. Rockville, MD: National Institute on Alcohol Abuse and Alcoholism; 1992
- ↵Aubrey L. Motivational Interviewing With Adolescents Presenting for Outpatient Substance Abuse Treatment [doctoral dissertation]. Albuquerque, NM: University of New Mexico; 1998
- Copyright © 2002 by the American Academy of Pediatrics