Abstract
Purpose:
A previous trial found lower alcohol use risk during follow-up among adolescent primary care patients receiving computer-facilitated Screening and provider Brief Advice (cSBA) compared to treatment-as-usual (TAU). We tested whether the effect was mediated by alcohol-related perceived risk of harm (PRoH).
Methods:
We analyzed data from the cSBA trial on 12- to 18-year-old patients at 9 New England practices (n = 2,096, 58% females). The trial used a quasi-experimental pre–post design with practices being their own controls (TAU followed by cSBA). Because prior alcohol experience could modify effects, we stratified analyses by baseline past 12-month drinking. Among baseline nondrinkers, we tested baseline to 3-month trajectories in PRoH of “trying alcohol” as an effect mediator for drinking at 3- and 12-month follow-up. Similarly, among those with prior drinking, we examined baseline to 3-month trajectories in PRoH of “weekly binge drinking” as an effect mediator for drinking and binge drinking. We used the Hayes product of coefficients mediation approach.
Results:
Among baseline nondrinkers (n = 1,449), cSBA had higher PRoH compared to TAU for “trying alcohol,” and higher PRoH in turn was associated with lower follow-up drinking risk. PRoH mediated their cSBA effect at 12 months, but not 3 months. Among adolescents with prior drinking (n = 647), cSBA had higher PRoH for “weekly binge drinking,” which was associated with lower drinking risk at both follow-ups, and lower binge drinking risk at 3 months. PRoH mediated their cSBA effect on drinking at both follow-ups, and binge drinking at 3 months.
Conclusion:
A computer-facilitated primary care intervention enhanced adolescents’ perceived alcohol risks which in turn was associated with lower drinking risk.
Keywords: Adolescent alcohol use, Primary care, Preventive medicine, Screening and brief intervention, Brief advice
Underage drinking continues to be a major public health concern in the U.S. Although recent trends indicate that underage drinking is decreasing, the 2019 Youth Risk Behavior Survey (YRBS) found that 29% of high school students drank alcohol in the previous 30 days [1]. This population consumes less alcohol overall than adults; however, on the occasions they do drink, they often drink in larger quantities. Over 90% of alcohol consumed by young people is through many drinks over a few hours [2]. This practice, known as binge drinking, is often defined as 5 or more drinks for males and 4 or more drinks for females on one occasion, and 14% of 2019 YRBS respondents reported this behavior at least once in the past 30 days [1]. However, the National Institute on Alcohol Abuse and Alcoholism recommends lower binge drinking thresholds for children and adolescents (i.e., >3 drinks for 9- to 17-year-old girls and 9- to 13-year-old boys; ≥4 drinks for 14- to 15-year-old boys; ≥5 drinks for 16 years and older boys [3]), as they tend to have greater blood alcohol concentrations than adults after drinking the same amount of alcohol [4]. Alcohol intoxication among adolescents is associated with increased risks of injuries, physical and sexual assault, impaired judgment, and changes in brain development [1]. Between 2011 and 2015, the Centers for Disease Control estimated that excessive drinking was responsible for more than 3,500 deaths among people below 21 years of age each year [1]. Moreover, earlier alcohol use initiation increases the risk of developing alcohol use disorder in the future [5], making anticipatory guidance/primary prevention efforts to delay the initiation of drinking among adolescents critical for reducing future morbidity and mortality. In 2019, 2.3 million adolescents initiated alcohol use in the past 12 months, and annual initiates of alcohol use have remained constant from 2012 to 2018 [6].
Recent public health efforts to address underage drinking include promoting Screening and Brief Intervention (SBI) by medical providers, particularly in the primary care setting. An increasing proportion of adolescents see a primary care provider yearly [7,8] and many have trusting, longitudinal relationships with their providers [9]. Previous studies have shown strong acceptability, feasibility, and promising effects of primary care–based SBI on adolescent substance use outcomes [10,11]. In one of the first large multisite trials among adolescents in primary care, we examined the effect of a computer-facilitated Screening and provider Brief Advice (cSBA) intervention for adolescent patients delivered during routine primary care visits. cSBA consisted of a computer program which adolescents completed in the clinic just prior to seeing their provider. The computer program administered the screening questions (the CRAFFT, a valid and reliable screening tool used to identify problematic substance use in adolescents [12]), and then immediately provided adolescents with individualized feedback and psychoeducation illustrating the health-related risks of substance use (average completion time 5 minutes). Primary care providers then received a printed form with the patient’s screening results and “talking points” to guide a 2- to 3-minute provider-teen conversation about substance use during the time-alone portion of the visit. The cSBA trial was conducted in primary care practices in both the U.S. and the Czech Republic, and we found that the cSBA system, compared to treatment-as-usual (TAU) significantly increased patient receipt of brief counseling about alcohol during a primary care visit in both countries. During a 12-month follow-up period, cSBA was associated with lowered risk of alcohol use in the U.S. sample, but not in the Czech sample [12].
Although previous studies have offered evidence supporting SBI efficacy, few published studies have reported on their mechanisms of action. Finding evidence supporting an intervention’s hypothesized mechanism of action may enhance confidence in its effectiveness, advance its adoption into clinical practice, make adolescent alcohol policy recommendations, and help guide future research. Because cSBA was designed to enhance adolescent awareness of substance use–associated harms within the context of a health visit, we hypothesized that a potential mechanism of the observed cSBA effect on adolescent alcohol use outcomes was through increased perceived risk of harm (PRoH). Perceptions about one’s susceptibility to, and severity of, undesirable consequences have long been hypothesized as influences of health behaviors as part of the widely studied Health Belief Model [13,14]. Trends in national survey data show that rates of adolescent alcohol and cannabis use tend to be lower when adolescents’ PRoH from use is higher [15]. Other studies of adolescent risk behaviors such as e-cigarette use and distracted driving found similar inverse relationships between perceived risk and behavior [16,17]. The current study conducted a secondary analysis of data from the original cSBA trial to examine whether PRoH mediated the effect of cSBA on adolescent alcohol use outcomes seen in the U.S. sample. We hypothesized that PRoH from alcohol use would be more likely to stay high or increase from baseline to follow-up among patients receiving cSBA compared to those receiving TAU, and that higher PRoH would in turn be associated with a lower likelihood of using alcohol.
Methods
Study design and population
This study is a secondary analysis of data from the U.S. trial testing cSBA among 12- to 18-year-olds (N = 2,096, 58% females) recruited from 9 New England practices between November 2005 and October 2008 [12]. The original trial methods have been described in detail elsewhere [12], and are summarized briefly here. The trial used a quasi-experimental design in which practices served as their own controls.
During the 18-month TAU phase, providers were instructed to continue with their usual practice and patients completed study assessments only. At the point of cross-over into the 18-month cSBA phase, all providers completed a 1-hour intervention training which included a demonstration of the cSBA computer program, a review of an example Report Form with the provider “talking points,” and watching a 20-minute video modeling provider brief advice. Patients in the cSBA phase completed the cSBA computer program after the baseline assessment battery and before the clinical encounter. The Institutional Review Board of Boston Children’s Hospital and each participating site approved the study protocol.
Measures
Participants in both study phases completed identical study assessments at baseline (before seeing their provider), and at 3- and 12-month follow-up (72.3% and 73.7% follow-up rates, respectively), as previously described [12,18-20]. Study measures have been previously described; they assessed participant demographics, substance use, PRoH from use of a substance, and perceived problematic substance use involvement of parents, siblings, and peers (subscales from the Personal Experience Inventory) [21] as potential confounders of the intervention effect. At baseline, all measures, except for substance use, were self-administered using a laptop computer; to assess alcohol use during the past 90 days (both days of use and number of drinks on each drinking day), research assistants conducted a modified Timeline Follow-Back interview [22]. Subsequently, follow-up assessments were conducted by phone interview, under private conditions.
PRoH associated with alcohol use was measured using 2 self-administered questions derived from the Monitoring the Future survey: “How much do you think people risk harming themselves (physically or in other ways) if they…”, “…try one or 2 drinks of an alcoholic beverage” (“trying alcohol”), and “…have 5 or more drinks once or twice each weekend” (“weekly binge drinking”). The response options were “no risk,” “slight risk,” “moderate risk,” and “great risk” [23].
Statistical analysis
To first verify our hypothesis that PRoH from alcohol use differed significantly based on previous alcohol use experience, we used the chi-squared test to compare baseline PRoH of “trying alcohol” and “weekly binge drinking” between adolescents that reported past 12-month alcohol use at baseline and those that did not. With the analysis results confirming our hypothesis, we conducted all subsequent analyses in SAS Studio 3.7 stratified by baseline report of past 12-month alcohol use (use/no use).
Prior to conducting mediation analyses, we first used the chi-squared test and t-test to identify any significant baseline demographic differences between the cSBA and TAU groups within the baseline alcohol use subgroups. Any variables that showed a significant difference at a p-value <.05 were entered as control variables in subsequent intervention effect analyses. For each of the 2 PRoH mediator variables (“trying alcohol” [PRoH-TA] and “weekly binge drinking” [PRoH-WBD]), we constructed ordinal trajectory variables using baseline and 3-month follow-up data. We focused on this timeframe because we hypothesized that the cSBA effect on PRoH would be strongest near the time of intervention receipt. Each variable initially had 4 categories indicating whether perceived risk had stayed high (responses of moderate or great risk at both timepoints), increased (no/low risk at baseline; moderate/great risk at follow-up), decreased (moderate/great risk at baseline; no/low risk at follow-up), or stayed low (no/low risk at both timepoints).
Because of highly skewed data, we dichotomized days of alcohol use reported at 3- and 12-month follow-ups into “any” or “none” in the past 90 days. Similarly, we constructed dichotomous 3- and 12-month “any binge drinking” variables using the National Institute on Alcohol Abuse and Alcoholism–recommended age and gender-specific definitions for children and adolescents [3].
To evaluate potential for a mediation effect, we first examined the bivariate association between study phases and PRoH trajectories, and between PRoH trajectories and 3- and 12-month alcohol use outcomes, using the Wald chi-squared test in SUDAAN v. 11.0.3 (Research Triangle Institute) to account for correlated error arising from our site cluster-sampling design. Because of small cell sizes for the “decreased” group, we collapsed the “stayed low” and “decreased” categories into a single category. In mediation analyses, we analyzed these PRoH trajectory variables as ordinal variables with scores of 1 for “stayed low”/”decreased,” 2 for “increased,” and 3 for “stayed high.” The “stayed high’ group was assigned the highest score because, for both PRoH-TA and PRoH-WBD variables and among both baseline drinkers and nondrinkers, this group had the highest mean PRoH score at 3-month follow-up (e.g., among baseline nondrinkers, mean PRoH-TA scores for “stayed high,” “increased,” and “stayed low”/”decreased” were 3.43, 3.24, and 1.82, respectively; mean PRoH-WBD scores were 3.82, 3.61, and 2.99, respectively).
We then used the Hayes PROCESS macro for SAS (version 3.4), model template 4 [24], to conduct simple product of coefficient mediation analysis examining baseline to 3-month PRoH trajectories as a mediator of cSBA’s effect on alcohol use outcomes at 3- and 12-month follow-up, while controlling for baseline demographic differences between cSBA and TAU groups. The 3- and 12-month mediation models only included participants who had completed 3- and 12-month follow-ups.
Among adolescents reporting no past 12-month alcohol use at baseline, we analyzed PRoH-TA as a mediator of the effect of cSBA on “any drinking” during 3- and 12-month follow-up, controlling for age, gender, race/Hispanic ethnicity (non-Hispanic white vs. other), parent education level (≥college graduate vs. other), provider type (attending physician vs. nurse practitioner or physician assistant), and visit type (well-visit vs. other). We did not examine binge drinking as an outcome in this group due to insufficient numbers.
Among those with past 12-month alcohol use at baseline, we analyzed PRoH-WBD in 2 mediation models, one predicting “any drinking” and the other predicting “any binge drinking” at 3- and 12-month follow-up, controlling for race/Hispanic ethnicity, provider type, and visit type, which were the only variables that differed between the cSBA and TAU groups at baseline in this subsample.
To evaluate the sensitivity of the analysis results to nonresponse at follow-up, we conducted multiple imputation (n = 10 datasets) in SAS Studio 3.7 and recomputed PRoH trajectory groups and alcohol use/binge drinking outcomes at 3- and 12-month follow-up using the imputed dataset. Analysis of the imputed dataset yielded the same findings as the nonimputed dataset; therefore, we report the original results based on the nonimputed dataset here.
Results
Sample characteristics
At baseline, 30.9% (647/2,096) of study participants reported drinking alcohol in the last 12 months. Compared to those reporting no drinking, those reporting drinking were more likely to be older and female; to have substance-using parents, siblings, and peers; to use other substances; and less likely to have 2 parents at home (Table 1). As hypothesized, those reporting past 12-month alcohol use at baseline perceived alcohol use to be less risky compared to adolescents reporting no past 12-month use at baseline; the percentages reporting “no” or “slight” risk of harm were 80.6% versus 69.4% (p < .0001), respectively, for PRoH-TA, and 23.1% versus 19.6% (p < .0001), respectively, for PRoH-WBD. We report all subsequent results stratified by baseline past 12-month alcohol use status.
Table 1.
Comparison of TAU versus cSBA group characteristics by baseline report of any past 12-month alcohol use
| No past 12-month alcohol use |
Any past 12-month alcohol use |
|||||||
|---|---|---|---|---|---|---|---|---|
| All, n (%) | TAU, n (%) | cSBA, n (%) | Chi-squared p-value |
All, n (%) | TAU, n (%) | cSBA, n (%) | Chi-squared p-value |
|
| Total | 1,449 | 709 | 740 | 647 | 359 | 288 | ||
| Age (mean ± SD) | 14.75 (1.9) | 14.88 (2.0) | 14.63 (1.8) | .01 | 16.79 (1.3) | 16.75 (1.4) | 16.84 (1.2) | .38 |
| Males | 666 (46.0) | 293 (41.3) | 373 (50.4) | <.001 | 210 (32.5) | 116 (32.1) | 95 (32.6) | .93 |
| Race/ethnicity | .01 | .06 | ||||||
| Non-Hispanic white | 915 (63.2) | 450 (63.5) | 465 (62.8) | 438 (67.7) | 239 (66.6) | 199 (69.1) | ||
| Non-Hispanic black | 169 (11.7) | 70 (9.9) | 99 (13.4) | 169 (7.4) | 30 (8.4) | 18 (6.3) | ||
| Non-Hispanic Asian | 104 (7.1) | 50 (7.1) | 54 (7.3) | 47 (7.3) | 27 (7.5) | 20 (6.9) | ||
| Hispanic | 159 (11.0) | 74 (10.4) | 85 (11.5) | 71 (11.0) | 32 (8.9) | 39 (13.5) | ||
| Non-Hispanic other | 102 (7.0) | 65 (9.2) | 37 (5.0) | 43 (6.6) | 31 (8.6) | 12 (4.2) | ||
| Parent education level | ||||||||
| College degree or higher | 703 (49.7) | 299 (43.5) | 404 (55.6) | <.0001 | 270 (44.1) | 152 (45.2) | 118 (42.6) | .51 |
| Parents at home | .20 | .20 | ||||||
| One parent | 383 (26.7) | 195 (27.8) | 188 (25.5) | 195 (30.8) | 116 (32.9) | 79 (28.0) | ||
| Two parents | 1,017 (70.8) | 488 (69.6) | 529 (71.8) | 407 (64.2) | 215 (61.1) | 192 (68.1) | ||
| None or foster home | 37 (2.5) | 18 (2.6) | 19 (2.6) | 32 (5.0) | 21 (6.0) | 11 (3.9) | ||
| Parent substance use | 174 (12.0) | 86 (12.2) | 88 (11.9) | .88 | 148 (22.9) | 84 (23.4) | 64 (22.3) | .74 |
| Sibling substance use | 146 (10.1) | 79 (11.1) | 64 (8.6) | .11 | 246 (38.1) | 120 (33.5) | 112 (39.0) | .13 |
| Peer substance use | 690 (47.7) | 340 (48.1) | 350 (47.3) | .76 | 575 (89.2) | 318 (88.8) | 257 (89.6) | .77 |
| Visit type | ||||||||
| Well-visit | 1,315 (91.9) | 606 (87.1) | 709 (96.5) | <.001 | 504 (78.9) | 245 (69.0) | 259 (91.2) | <.001 |
| First visit | 149 (10.4) | 75 (10.7) | 74 (10.1) | .69 | 71 (11.2) | 40 (11.4) | 31 (11.0) | .89 |
| Female provider | 896 (62.4) | 423 (60.5) | 473 (64.3) | .14 | 451 (70.6) | 240 (67.6) | 211 (74.3) | .07 |
| Provider type | <.001 | .01 | ||||||
| Attending | 954 (66.9) | 497 (71.6) | 457 (62.4) | 363 (57.7) | 209 (60.1) | 154 (54.8) | ||
| NP or PA | 230 (16.1) | 82 (11.8) | 148 (20.2) | 122 (19.4) | 53 (15.2) | 69 (24.6) | ||
| Trainee | 243 (17.0) | 115 (16.6) | 128 (17.5) | 144 (22.9) | 86 (24.7) | 58 (20.6) | ||
| Other past 12-month substance usea | ||||||||
| Tobacco | 73 (5.0) | 47 (6.6) | 26 (3.5) | .007 | 271 (41.9) | 156 (43.5) | 115 (39.9) | .37 |
| Cannabis | 50 (3.5) | 30 (4.2) | 20 (2.7) | .11 | 266 (41.1) | 144 (40.1) | 122 (42.4) | .56 |
| Other drugs | 0 (.0) | 0 (.0) | 0 (.0) | – | 52 (8.0) | 28 (7.8) | 24 (8.4) | .79 |
cSBA = computer-facilitated Screening and primary care provider Brief Advice; NP = nurse practitioner; PA = physician assistant; SD = standard deviation; TAU = treatment as usual.
At baseline assessment.
Within each of these subgroups, there were significant differences between the cSBA and TAU groups at baseline (Table 1). Among nondrinkers, the cSBA group tended to be younger, less likely to be female, to report tobacco use, more likely to be non-Hispanic black, and to have a parent with a college degree. Among both nondrinkers and those with prior drinking, the cSBA group was more likely to be seeing an attending physician, and to be recruited at a well-visit. Subsequent mediation analyses controlled for these baseline differences.
Bivariate associations
In bivariate analyses, baseline nondrinking adolescents receiving cSBA were significantly more likely to have PRoH trajectories that “stayed high” or “increased” for both PRoH-TA and PRoH-WBD compared to those receiving TAU (Table 2). Among adolescents with prior drinking, the cSBA effect was significant only for PRoH-WBD.
Table 2.
Group comparison of trajectories from baseline to 3-month follow-up of perceived risk of harm from trying alcohol and weekly binge drinking, stratified by baseline report of any past 12-month alcohol use
| Perceived risk of harm from trying alcohol |
Perceived risk of harm from weekly bingea drinking |
||||||||
|---|---|---|---|---|---|---|---|---|---|
| Stayed high, n (%) | Increased, n (%) | Stayed low/decreased, n (%) |
p-valueb | Stayed high, n (%) | Increased, n (%) |
Stayed low/decreased, n (%) |
p-valueb | ||
| No past 12-month alcohol use at baseline | No past 12-month alcohol use at baseline | ||||||||
| TAU, n = 482 | 112 (23.2) | 121 (25.1) | 249 (51.7) | .009 | TAU, n = 495 | 378 (76.4) | 58 (11.7) | 59 (11.9) | .011 |
| cSBA, n = 521 | 134 (25.7) | 166 (31.9) | 221 (42.4) | cSBA, n = 533 | 401 (75.2) | 90 (16.9) | 42 (7.9) | ||
| Any past 12-month alcohol use at baseline | Any past 12-month alcohol use at baseline | ||||||||
| TAU, n = 236 | 15 (6.4) | 52 (22.0) | 169 (71.6) | .143 | TAU, n = 236 | 143 (60.6) | 32 (13.6) | 61 (25.8) | .022 |
| cSBA, n = 194 | 23 (11.9) | 39 (20.1) | 132 (68.0) | cSBA, n = 201 | 143 (71.1) | 27 (13.4) | 31 (15.4) | ||
cSBA = computer-facilitated Screening and primary care provider Brief Advice; TAU = treatment as usual.
“Weekly binge drinking” refers to having 5 or more drinks once or twice each weekend.
Adjusted Wald F-statistic p-value from the Wald chi-squared test in SUDAAN version 11.0.3.
PRoH trajectories were associated with alcohol use outcomes at 3-month follow-up in both subgroups of adolescents, with “stayed high” and “increased” trajectory groups having significantly lower rates of reporting any drinking and any binge drinking compared to the “stayed low/decreased” trajectory group (Table 3). These associations were extinguished by 12-month follow-up.
Table 3.
Bivariate association between baseline to 3-month follow-up trajectories of perceived risk of harm and past 90-day alcohol use outcomes at 3- and 12-month follow-up, separately for adolescents without and with past 12-month alcohol use at baseline
| Perceived risk of harm from trying alcohol |
Perceived risk of harm from weekly bingea drinking |
||||||||
|---|---|---|---|---|---|---|---|---|---|
| Stayed high, n (%) | Increased, n (%) |
Stayed low/ decreased, n (%) |
p-valueb | Stayed high, n (%) |
Increased, n (%) |
Stayed low/decreased, n (%) |
p-valueb | ||
| No past 12-month alcohol use at baseline | Any past 12-month alcohol use at baseline | ||||||||
| Any past 90-day drinking, n (%) | Any past 90-day drinking, n (%) | ||||||||
| 3 months, n = 1,003 | 7 (2.8) | 8 (2.8) | 30 (6.4) | .027 | 3 months, n = 436 | 142 (49.8) | 30 (50.8) | 68 (73.9) | <.0001 |
| 12 months, n = 969 | 12 (5.9) | 20 (7.0) | 43 (9.0) | .305 | 12 months, n = 398 | 126 (58.1) | 37 (63.8) | 81 (65.9) | .336 |
| Any past 90-day bingec drinking, n (%) | |||||||||
| 3 months, n = 437 | 75 (26.2) | 21 (35.6) | 42 (45.7) | .004 | |||||
| 12 months, n = 398 | 84 (38.7) | 25 (43.1) | 50 (40.7) | .818 | |||||
“Weekly binge drinking”” refers to having 5 or more drinks once or twice each weekend.
Adjusted Wald F-statistic p-value from the Wald chi-squared test in SUDAAN version 11.0.3.
The past 90-day “binge” drinking outcome measure was defined using the age- and gender-specific National Institute on Alcohol Abuse and Alcoholism guidelines for children and adolescents (i.e., for girls, ≥3 drinks in a day for 9–17years old and ≥4 for 18 and older; for boys, ≥3 drinks in a day for 9–13 years old, ≥4 for 14–15, and ≥5 for 16 and older).
Mediation analysis results
Figures 1 and 2 show path diagrams presenting the mediation modeling results stratified by baseline past 12-month alcohol use status. Adjusted path coefficients are shown for the intervention effect on each PRoH mediator variable; the effect of the PRoH variable on the alcohol use outcome variable of interest, controlling for study arm; the direct effect of the intervention on the alcohol use variable, controlling for the effect of PRoH; and the indirect effect of the intervention on alcohol use through PRoH (thus, the sum of the direct and indirect effects equals the total intervention effect). Models adjusted for baseline differences between study arms.
Figure 1.
Past 90-day alcohol use outcomes among adolescents reporting no past 12-month alcohol use at baseline: results of mediation analysis examining baseline to 3-month trajectories in perceived risk of harm from “trying alcohol” as a mediator of the cSBA effect at 3-month (A) and 12-month (B) follow-up. Description: Path diagram showing coefficients from mediation analyses. aAll mediation models adjusted for age, gender, race/Hispanic ethnicity (non-Hispanic white vs. other), parent education level (≥college graduate vs. other), type of visit (well-visit vs. other), and provider type (attending physician vs. other). *p < .05.
Figure 2.
Past 90-day alcohol use and binge drinking outcomes among adolescents reporting past 12-month alcohol use at baseline: results of mediation analysis examining baseline to 3-month trajectories in perceived risk of harm from “weekly binge drinking” as a mediator of the cSBA effect at 3-month (A) and 12-month (B) follow-up. aAll mediation models adjusted for race/Hispanic ethnicity (non-Hispanic white vs. other), type of visit (well-visit vs. other), and provider type (attending physician vs. other). *p < .05, **p<.01.
Among adolescents with no past 12-month drinking at baseline, cSBA continued to predict higher PRoH-TA trajectories in adjusted analyses (Figure 1A,B). Higher PRoH was in turn associated with significantly lower odds of reporting any drinking by 3- and 12-month follow-ups (e.g., Figure 1A shows that a 1-unit increase in the PRoH-TA trajectory score, with study arm controlled, was associated with 43% lower odds of past 90-day alcohol use at 3 months). However, PRoH-TA was not a significant mediator of the cSBA effect at 3-month follow-up, as indicated by a nonsignificant indirect effect and significant direct effect (Figure 1A). In contrast, PRoH-TA did mediate the cSBA effect on past 90-day alcohol use at 12-month follow-up, as shown by the significant indirect effect and nonsignificant direct effect (Figure 1B). That is, cSBA was associated with a .13-unit higher PRoH-TA trajectory score compared to TAU, and it was through this higher PRoH-TA trajectory score that cSBA was associated with a significantly lower odds of past 90-day alcohol use reported at 12 months (indirect effect = product of coefficients for the cSBA effect on PRoH-TA [.13] and for the PRoH-TA effect on alcohol use [−.34]). The direct effect coefficient can be interpreted as, at the same PRoH-TA trajectory score, cSBA is associated with 19% lower odds of past 90-day alcohol use at 12 months compared to TAU, but this difference is not statistically significant.
Among adolescents reporting past 12-month alcohol use at baseline, PRoH-WBD was a significant mediator of cSBA’s effects on both any alcohol use and binge drinking outcomes at 3-month follow-up (Figure 2A). In adjusted models, receipt of cSBA predicted higher PRoH-WBD, which in turn predicted lower rates of any past 90-day drinking and binge drinking (Figure 2A). At 12-month follow-up, PRoH-WBD was not associated with odds of past 90-day binge drinking and was not a significant cSBA effect mediator (Figure 2B).
Discussion
This study provides evidence that the cSBA approach can enhance PRoH related to alcohol use among adolescent primary care patients, as intended, and that enhanced perceived risk may mediate cSBA’s effects on alcohol use outcomes. Among adolescents with no prior alcohol use, cSBA was associated with higher rates of adolescents with “increased” PRoH (both PRoH-TA and PRoH-WBD) from baseline to 3-month follow-up, compared to TAU. Among those with prior use, cSBA had a higher rate of patients maintaining high PRoH-WBD from baseline to 3 months. Higher PRoH has been shown in numerous studies, including prospective studies, to be associated with lower risk for substance use among adolescents [25,26].
As hypothesized, higher perceived risk, in turn, was associated with lower adjusted odds of any past 90-day alcohol use at both 3- and 12-month follow-up among adolescents with and without prior alcohol use experience in our sample. Higher PRoH-WBD also predicted lower adjusted odds of past 90-day binge drinking at 3-month follow-up among adolescents with prior alcohol use, an effect extinguished by 12 months. We advise caution in inferring the direction of causality of the 3-month effects as the PRoH trajectory variables covered the same reporting period as the alcohol use variables, and PRoH can be changed by behavioral experience (risk reappraisal) [27]. Indeed, we found that adolescents reporting prior alcohol use at baseline had substantially lower levels of PRoH overall compared to those without prior alcohol use. Although previous studies of PRoH and adolescent substance use tend to share the same limitation due to their cross-sectional designs [16,28,29], there is emerging evidence for a prospective effect of PRoH on use [25,27]. A recent large longitudinal study of adolescents aged 10–15 years found that greater awareness of alcohol or cigarette use–related harm at one data collection wave predicted greater likelihood of being persistent nonusers of the substance 1 year later [25].
Finally, we found significant mediation of cSBA effects by PRoH for odds of past 90-day drinking at 12 months among baseline nondrinkers, and at both 3 and 12 months among those with prior drinking. The lack of statistically significant direct effects in these mediation models indicates that the intervention affected alcohol use outcomes largely through their effect on PRoH. The lack of a significant mediation effect at 3 months among baseline nondrinkers may be due, in part, to small cell sizes for any past 90-day alcohol use at 3 months (only 45 out of 1,003). Among adolescents with prior drinking at baseline, PRoH was a significant mediator for past 90-day binge drinking at 3 months, but not at 12 months. The dissipation of the effect by 12 months is unsurprising given that perceived risk of substance use tends to decline as adolescents age [30], highlighting the need for age-appropriate reinforcement strategies over time.
To our knowledge, this is among the first studies to report on the effect of primary care–based screening and provider brief advice on adolescents’ perceptions of risk related to alcohol use, an evidence-based protective factor in adolescents’ substance use risk. It is encouraging that a computer-facilitated protocol that required just 5 minutes of patient time before the visit, and 2–3 minutes of provider time with the patient had such robust effects on adolescents’ perceptions about alcohol use, which in turn were significantly associated with alcohol use behavior during 12-month follow-up. A recent analysis of cross-sectional survey data from a national sample of adolescents aged 12–17 years found that, among those visiting a healthcare provider in the past year, those that reported being asked about alcohol use during a healthcare encounter had significantly higher likelihood of perceiving that daily binge drinking entailed a great risk, even after controlling for potentially confounding demographic, psychosocial, and behavioral characteristics, including parental discussions about substance use and receipt of school-delivered health education about substance use [31].
Healthcare providers are a trusted information source for adolescent patients [9]. Hence, these encouraging findings support the primary care visit as an important teachable moment for providers to advise adolescents about substance use during private conversations that place substance use in a nonpejorative health risk context. For adolescents who have not yet started drinking, providers can prevent initiation by offering anticipatory guidance and positive feedback while conveying an openness to further discussion in future visits. For adolescents who report drinking, providers can provide a brief Motivational Interviewing-style intervention [32] to prevent further drinking and hazardous drinking patterns. For both types of patients, providers should, after asking the patient for permission, share current information about the health risks of alcohol use, especially for the developing brain, and a clear recommendation that delaying use is best for their health.
Although this study’s findings support the value of addressing harm perception in underage drinking prevention efforts, the lower efficacy of cSBA in the Czech Republic highlights the need to tailor harm messaging to the cultural context. Messaging about the health risks of using alcohol may be an effective strategy in the U.S. where there has been a history and culture of messaging on the harms of alcohol use; in the Czech Republic, however, youth drinking is highly normative, so a similar kind of messaging in that type of context may have little effect.
This study had some potential limitations. The original study utilized a nonrandomized, asynchronous design in which there may have been unmeasured group differences that were uncontrolled in the analyses, including historical confounders (e.g., contextual differences between the TAU and cSBA phases). All study sites were in New England, and it is unknown the extent to which the results are generalizable to adolescents in other locations. The study does not address whether changes in alcohol harm perception coincided with similar changes in harm perceptions about other substances. The PROCESS macro used to conduct mediation analyses did not have an option to specify the site-based nested structure of the data. However, the design effects resulting from the site-based cluster-sampling design (the ratio of the standard error accounting for within-cluster correlations to the standard error under the assumption of a simple random sample), calculated using SUDAAN v. 11.0.3, tended to be small (range .94–1.04), suggesting limited impact of correlated error on our study results. Finally, our study relied on self-reported and interviewer-collected data, which may be prone to recall error and social desirability bias.
Nonetheless, this study is an important addition to the growing literature on the efficacy and active components of primary care–based alcohol SBI for adolescents. Our findings provide support for the value of universal alcohol SBI with adolescents during primary care visits, which places the discussion of substance use within a health context, with a trusted health professional. However, more studies are needed to determine the replicability of these findings, and that use a randomized study design, larger samples, and sites outside of New England. Moreover, further work is needed to identify effective ways to reinforce and extend effects over time.
Primary care providers’ advice about the health risks of alcohol in just a few minutes can increase adolescents’ PRoH from drinking, which is in turn associated with reduction in the likelihood of patients’ drinking during the months following the visit. This is a promising strategy that warrants further development and testing.
IMPLICATIONS AND CONTRIBUTION.
Computer-facilitated Screening and provider Brief Advice, designed to enhance adolescents’ awareness of alcohol’s health risks, was previously found to be associated with lower risk of adolescent drinking during follow-up, compared to usual care. Perceived risks of alcohol use significantly mediated the intervention effect among adolescents reporting prior drinking at baseline.
Acknowledgments
The original study was supported by the National Institute on Drug Abuse [R01DA018848 and R01DA018848-03S1]; the National Institute on Alcohol Abuse and Alcoholism [K07 AA013280]; the Maternal and Child Health Bureau of the U.S. Department of Health and Human Services [T20MC07462 and T71NC0009]; the Davis Family Charitable Foundation; the Carl Novotny & Judith Swahnberg Fund; the Ryan Whitney Memorial Fund; and the J.F Maddox Foundation. The study findings were presented at the 2017 Annual Meeting of the International Network on Brief Interventions for Alcohol and Other Drugs (INEBRIA), and the 2020 Annual Meeting of the Association for Medical Education and Research in Substance Abuse (AMERSA).
Footnotes
Conflict of Interest Disclosures: The authors have no potential conflicts of interest to disclose. The authors have no financial relationships relevant to this article to disclose.
Clinical Trial Registration: www.clinicaltrials.gov, NCT0022787.
References
- [1].Underage drinking ∣ CDC. Available at: https://www.cdc.gov/alcohol/fact-sheets/underage-drinking.htm. Accessed January 12, 2021. [Google Scholar]
- [2].National Institute of Alcohol Abuse and Alcoholism. Underage Drinking ∣ National Institute on Alcohol Abuse and Alcoholism (NIAAA). Available at: https://www.niaaa.nih.gov/publications/brochures-and-fact-sheets/underage-drinking. Accessed January 12, 2021.
- [3].National Institute on Alcohol Abuse and Alcoholism. Alcohol Screening and Brief Intervention for Youth: Practitioner’s Guide. Washington, D.C.: National Institutes of Health, Department of Health and Human Services; 2011. [Google Scholar]
- [4].Donovan JE. Estimated blood alcohol concentrations for child and adolescent drinking and their implications for screening instruments. Pediatrics 2009;123:e975–81. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [5].Hingson RW, Heeren T, Winter MR, et al. Age at drinking onset and alcohol dependence. Arch Pediatr Adolesc Med 2006;160:739–46. [DOI] [PubMed] [Google Scholar]
- [6].Substance Abuse and Mental Health Services Administration. Key substance use and mental health indicators in the United States: Results from the 2019 National Survey on Drug Use and Health. Washington, D.C.: (HHS Publication No. PEP20-07-01-001, NSDUH Series H-55); 2020. [Google Scholar]
- [7].Adams SH, Park MJ, Irwin CE, et al. Adolescent and young adult preventive care: comparing national survey rates. Am J Prev Med 2015;49:238–47. [DOI] [PubMed] [Google Scholar]
- [8].Adams SH, Park MJ, Twietmeyer L, et al. Association between adolescent preventive care and the role of the affordable care act.JAMA Pediatr 2018;172:43. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [9].Harris SK, Woods ER, Sherritt L, et al. A youth-provider connectedness measure for use in clinical intervention studies. J Adolesc Heal 2009;44:S35–6. [Google Scholar]
- [10].D’Amico EJ, Parast L, Shadel WG, et al. Brief motivational interviewing intervention to reduce alcohol and marijuana use for at-risk adolescents in primary care. J Consult Clin Psychol 2018;86:775–86. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [11].Tanner-Smith EE, Lipsey MW. Brief alcohol interventions for adolescents and young adults: a systematic review and meta-analysis. J Subst Abuse Treat 2015;51:1–18. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [12].Harris SK, Csémy L, Sherritt L, et al. Computer-facilitated substance use screening and brief advice for teens in primary care: an international trial. Pediatrics 2012;129:1072–82. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [13].Rosenstock IM, Strecher VJ, Becker MH. Social learning theory and the health belief model. Heal Educ Behav 1988;15:175–83. [DOI] [PubMed] [Google Scholar]
- [14].Jones CJ, Smith H, Llewellyn C. Evaluating the effectiveness of health belief model interventions in improving adherence: a systematic review. Health Psychol Rev 2014;8:253–69. [DOI] [PubMed] [Google Scholar]
- [15].Substance Abuse and Mental Health Services Administration C for BHS and Q. Rockville, MD: The NSDUH Report: Trends in Adolescent Substance Use and Perception of Risk from Substance Use; 2013. [PubMed] [Google Scholar]
- [16].Giovacchini CX, Pacek L, McClernon FJ, et al. Use and perceived risk of electronic cigarettes among north carolina middle and high school students. N C Med J 2017;78:7–13. [DOI] [PubMed] [Google Scholar]
- [17].Carter PM, Bingham CR, Zakrajsek JS, et al. Social norms and risk perception: predictors of distracted driving behavior among novice adolescent drivers hhs public access. J Adolesc Heal 2014;54:32–41. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [18].Louis-Jacques J, Knight JR, Sherritt L, et al. Do risky friends change the efficacy of a primary care brief intervention for adolescent alcohol use? J Adolesc Health 2014;54:449–53. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [19].Harris SK, Johnson JK, Sherritt L, et al. Putting adolescents at risk: riding with drinking drivers who are adults in the home. J Stud Alcohol Drugs 2017;78:146–51. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [20].Knight JR, Kuzubova K, Csemy L, et al. Computer-facilitated screening and brief advice to reduce adolescents’ heavy episodic drinking: a study in two countries. J Adolesc Heal 2018;62:118–20. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [21].Winters KC, Henley GA. Personal Experience Inventory (PEI) Test and Manual. Los Angeles, CA: Western Psychological Services; 1989. [Google Scholar]
- [22].Sobell LC, Sobell MB. Timeline Follow-Back. In: Litten RZ, Allen JP, eds. Measuring Alcohol Consumption. Totowa, NJ: Humana Press; 1992:41–72. [Google Scholar]
- [23].Bachman JG, Johnston LD, O’Malley PM. The Monitoring the Future Project After 27 Years: Design and Procedures. Ann Arbor, MI: Institute for Social Research; 2001. [Google Scholar]
- [24].PROCESS macro. The PROCESS macro for SPSS, SAS, and R-PROCESS macro for SPSS and SAS. Available at: https://processmacro.org/index.html. Accessed January 12, 2021.
- [25].Cable N, Roman MMF, Kelly Y. What could keep young people away from alcohol and cigarettes? findings from the UK household longitudinal study. BMC Public Health 2017;17:371. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [26].Kelly Y, Goisis A, Sacker A, et al. What influences 11-year-olds to drink? Findings from the Millennium Cohort Study. BMC Public Health 2016;16:169. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [27].Grevenstein D, Nagy E, Kroeninger-Jungaberle H. Development of risk perception and substance use of tobacco, alcohol and cannabis among adolescents and emerging adults: evidence of directional influences. Subst Use Misuse 2015;50:376–86. [DOI] [PubMed] [Google Scholar]
- [28].Prince MA, Swaim RC, Stanley LR, et al. Perceived harm as a mediator of the relationship between social norms and marijuana use and related consequences among American Indian youth. Drug Alcohol Depend 2017;181:102–7. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [29].Hanauer M, Walker MR, Machledt K, et al. Association between perceived risk of harm and self-reported binge drinking, cigarette smoking, and marijuana smoking in young adults. J Am Coll Heal 2021; 69:345–52. [DOI] [PubMed] [Google Scholar]
- [30].Johnston LD, Miech RA, O’malley PM, et al. Monitoring the Future: National Survey Results on Drug Use 1975-2019. 2019 Overview; Key Findings on Adolescent Drug Use. Ann Arbor, MI: National Institute on Drug Abuse; 2020. [Google Scholar]
- [31].Volcarelli LB, Bagley SM, Hadland S, et al. Healthcare providers’ role in adolescents’ perceived risk of alcohol use. Washington, D.C.: Assoc. Med. Educ. Res. Subst. Abus.; 2017. [Google Scholar]
- [32].Naar-King S, Suarez M. Motivational Interviewing with Adolescents and Young Adults. New York, NY: Guilford Press; 2011. [Google Scholar]