Computer-Facilitated Substance Use Screening and Brief Advice for Teens in Primary Care: An International Trial

Abstract

OBJECTIVE:

Primary care providers need effective strategies for substance use screening and brief counseling of adolescents. We examined the effects of a new computer-facilitated screening and provider brief advice (cSBA) system.

METHODS:

We used a quasi-experimental, asynchronous study design in which each site served as its own control. From 2005 to 2008, 12- to 18-year-olds arriving for routine care at 9 medical offices in New England (n = 2096, 58% females) and 10 in Prague, Czech Republic (n = 589, 47% females) were recruited. Patients completed measurements only during the initial treatment-as-usual study phase. We then conducted 1-hour provider training, and initiated the cSBA phase. Before seeing the provider, all cSBA participants completed a computerized screen, and then viewed screening results, scientific information, and true-life stories illustrating substance use harms. Providers received screening results and “talking points” designed to prompt 2 to 3 minutes of brief advice. We examined alcohol and cannabis use, initiation, and cessation rates over the past 90 days at 3-month follow-up, and over the past 12 months at 12-month follow-up.

RESULTS:

Compared with treatment as usual, cSBA patients reported less alcohol use at follow-up in New England (3-month rates 15.5% vs 22.9%, adjusted relative risk ratio [aRRR] = 0.54, 95% confidence interval 0.38–0.77; 12-month rates 29.3% vs 37.5%, aRRR = 0.73, 0.57–0.92), and less cannabis use in Prague (3-month rates 5.5% vs 9.8%, aRRR = 0.37, 0.17–0.77; 12-month rates 17.0% vs 28.7%, aRRR = 0.47, 0.32–0.71).

CONCLUSIONS:

Computer-facilitated screening and provider brief advice appears promising for reducing substance use among adolescent primary care patients.

What’s Known on This Subject:

Primary care settings provide an important venue for early detection of substance use and intervention, but adolescent screening rates need improvement. Screening and brief interventions appear effective in reducing adult problem drinking but evidence for effectiveness among adolescents is needed.

What This Study Adds:

A computer-facilitated system for screening, feedback, and provider brief advice for primary care can increase adolescent receipt of substance use screening across a variety of practice settings, and shows promise for reducing adolescents’ use of alcohol and cannabis.

More than 40% of US adolescents are current alcohol drinkers, and more than 20% use cannabis (marijuana) or another drug.¹ This is a serious national problem because substance use is strongly linked to the leading causes of adolescent mortality and many other health problems.^2–5 Primary care offices are promising venues for screening, prevention, and early intervention.⁶ The American Academy of Pediatrics recommends that health care providers screen all adolescents for substance use as part of routine preventive care.^7,8 Adherence to this recommendation, however, is low.^9,10 Stated reasons include lack of time and personnel to perform the screening, unfamiliarity with screening tools, lack of training in how to deal with positive screens, and lack of effective interventions.¹¹

The CRAFFT is a valid and reliable screener for adolescent medical patients, and is brief enough to be practical for busy medical offices.^12–14 “CRAFFT” is a mnemonic acronym formed by the first letters of key words in the test’s 6 yes/no questions (Fig 1). Each “yes” scores 1 point; a total score of ≥2 has a sensitivity of 0.80 and specificity of 0.86 for identifying substance abuse or dependence.¹³ Although the CRAFFT can be conducted by clinician-interview or self-administered questionnaire, adolescents report being more likely to provide honest answers on questionnaires, even when they know the provider will receive the results.¹⁵

FIGURE 1.

The CRAFFT screening interview.

To meet the needs of both providers and patients, we developed a computer-facilitated screening and brief advice (cSBA) system consisting of a computerized screening and educational component before the visit, and provider advice during the visit. There is substantial evidence from studies conducted in the United States,^16–22 and other countries^17,20,23,24 supporting the effectiveness of screening and brief physician advice among adult primary care patients, especially in the reduction of harmful drinking and its associated consequences (eg, motor vehicle crashes, emergency department visits).^16–19 It is unknown, however, whether these findings are generalizable to younger patients, as there have been fewer studies among adolescents in primary care.^25,26 Existing studies suggest that primary care screening and brief interventions can positively impact adolescent health issues, such as tobacco use,^27–30 nutrition and physical activity,^31,32 and depression.^33,34 A large longitudinal study of 14-year-old primary care patients²⁶ found that screening and brief provider counseling significantly increased helmet use but did not reduce adolescent alcohol or drug use, suggesting the need to explore supplemental strategies to enhance effectiveness, such as the computerized education component that occurs before the provider visit in the cSBA system.

The primary objective of this study was to evaluate the immediate and short- and long-term effects of the cSBA system for adolescents in primary care. Immediate effects included providers’ brief counseling behaviors during the visit and adolescents’ reactions to it. Short- and long-term effects were adolescents’ use of alcohol and cannabis 3- and 12-months after the visit. We hypothesized that, compared with treatment as usual (TAU), more cSBA patients would report receiving provider advice, rate the quality of the advice as high, and report less substance use at the 3-month follow-up; however, without reinforcement, we hypothesized that the effect would be reduced by the 12-month follow-up. A secondary objective was to assess the separate effects of the cSBA intervention on preventing initiation of substance use by nonusers, and on promoting cessation among users.

Methods

We conducted the study at 9 primary care offices in 3 New England states, and in 10 pediatric generalist offices in Prague, Czech Republic (for more detail, see Supplemental Information). We used a quasi-experimental before-after design to compare cSBA and TAU. We first held a 1-hour orientation at each site to explain the study’s purpose, procedures, and safety protocol, and instructed providers to continue their usual practices during the TAU phase. For the ensuing 18 months, we recruited and assessed the TAU group. At the crossover point, we conducted a 1-hour provider training and initiated the cSBA protocol at all sites, then recruited and tested the cSBA group during the final 18 months.

Recruitment procedures were identical at each site during both study phases. Patients aged 12 to 18 years arriving for routine care, who were medically and emotionally stable on the day of the visit, able to read and understand the cSBA program, and available for follow-ups, were eligible. Czech Republic adolescents are seen for well-visits biannually, so we primarily recruited 13-, 15-, and 17-year-old patients there. Adolescents who participated in the TAU phase were excluded from the cSBA phase. In both TAU and cSBA phases, research assistants (RAs) contacted families before the visit to explain the study purpose, procedures, and confidentiality protection, and instructed interested patients to arrive 30 minutes early for their appointment. Upon arrival, RAs privately obtained informed participant assent (<18 years) or consent (≥18 years). Parents gave informed consent either in person, by phone, or sent a signed consent form in with the patient (Czech Republic). Participants then completed the baseline assessment and, for those receiving the intervention, the cSBA program, before seeing the provider. Participants received a merchandise certificate for completing each assessment ($5 United States, 200 Kč [$10–$12] Czech Republic). The institutional review boards of Children’s Hospital Boston, all New England sites, and the Charles University Second Faculty of Medicine Ethics Committee (Prague, Czech Republic) approved the study protocol.

Intervention Protocol

The cSBA intervention began with a self-administered screening (Fig 1) that asks about lifetime and past-12-month use of substances followed by the CRAFFT questions. The CRAFFT screen had an embedded skip pattern, so that patients with no history of substance use completed the CAR question only. If the CRAFFT was completed, the program immediately displayed the individual’s CRAFFT score and risk-level (low, medium, high) on a thermometerlike graphic. All cSBA adolescents then viewed the same 10 pages of scientific information and true-life stories illustrating the health risks of substance use, which we created based on feedback from focus groups of adolescents who reported finding these types of information most compelling. All cSBA adolescents completed the same computer program before the medical visit, and the average completion time was 5 minutes. Providers received a report form with the screening results, risk level, and 6 to 10 “talking points” designed to prompt a 2- to 3-minute provider/teen conversation about the health effects of substance use, and that recommended abstinence. The talking points on health risks were the same regardless of patients’ substance use status but advice was individualized to either “not start” or “stop” using substances (see Supplemental Information). Provider training included a demonstration of the cSBA program, review of a sample provider report, and a 20-minute video demonstrating provider brief counseling. For the Czech Republic study, we translated, back-translated, culturally adapted, and validated the CRAFFT screen.³⁵ We substituted 1 cSBA informational page on nonmedical use of prescription drugs with 1 on volatile inhalants, which is far more common there, and we translated/back-translated all other study materials.

Measures

The baseline assessment began with a past-90-day, modified timeline followback^14,36 interview, separately recording frequency of use of each substance. Participants then self-administered a computerized questionnaire that assessed demographics and perceived substance use by peers, siblings, and parents (scales derived from the validated Personal Experience Inventory^37,38). RAs recorded the office visit type and the provider’s gender and type. To evaluate potential historical confounding owing to the asynchronous study design, we asked respondents how often in the past 12 months they had heard information about alcohol or drugs in the news, in their school or community, or from friends or family.

Immediately after the visit, adolescents completed a post-visit checklist that assessed whether the provider gave advice not to use alcohol and drugs, their satisfaction with the visit, the likelihood of following their provider’s advice, and their rating of the way their provider gave the advice. At 3- and 12-month follow-up visits, participants completed assessments identical to the baseline either in person or by phone (>95%, because of the difficulty of scheduling in-person assessments). RAs ensured that participants could speak privately before conducting phone assessments. This difference in data collection mode occurred in both study arms equally and should not change any between-group effects.

Data Analyses

We conducted all analyses using SUDAAN v.10.0 (Research Triangle Institute, Research Triangle Park, NC) with site as the nest variable to account for correlated error arising from our site cluster-sampling design. To assess immediate effects, we computed the proportions of patients who reported receiving provider advice not to use, receiving information regarding the health risks of substance use, being “very satisfied” with their visit, and being “very likely” to follow their provider’s advice generally, and, among those receiving advice about substance use, the percent rating the provider’s advice as “very good” or “excellent.”

The timeline followback–derived frequency-of-use variables were highly skewed, so we used a dichotomized use/no use as our primary short- and long-term outcome variable. Our a priori hypothesized outcome variables were rates of any use, initiation, and cessation. We defined initiation as any use at follow-up among those reporting no past-12-months use at baseline, and cessation as no use at follow-up among those reporting any past-12-months use at baseline. We compared any past-90-day use at the 3-month follow-up and any past-12-month use at the 12-month follow-up. We stratified analyses by country because of some different demographic variables, and by substance (alcohol, cannabis). We did not analyze use of drugs other than cannabis because of low prevalence (≤2%).

We used an “intent-to-treat” approach, analyzing cSBA adolescents regardless of whether they reported receiving provider advice. We excluded follow-up assessments completed >2 months late. We used χ² tests for categorical variables and t tests for continuous variables to assess baseline group equivalence. We dichotomized race (white non-Hispanic versus other [United States only]), parents in home (2 versus other), parent education level (≥college graduate versus other), and type of visit (well-visit versus other) to ensure adequate cell sizes. For analysis of the intervention effect on initiation and cessation by 3- and 12-month follow-up, we used logistic regression modeling with generalized estimating equations (GEE) to compute adjusted relative risk ratios (aRRR) for cSBA compared with TAU, controlling for demographics, peer/family substance use, visit/provider characteristics, and the multisite sampling design. These analyses inherently controlled for baseline use through stratification of participants into baseline nonuser and user groups. We ran separate models for 3- and 12-month outcomes because of the different timeframes examined (past 90 days versus past 12 months). To examine the intervention effect on use at follow-up, we used 2 types of analysis. We conducted logistic regression analysis (with GEE to account for within-site clustering) to compare use probabilities between groups at each follow-up, with baseline data for each outcome variable entered as a covariate in the model. This method of longitudinal data analysis corresponds to a Markov chain transition model where subsequent observations are conditional on previous observations.³⁹ We also conducted a repeated measures analysis that included data from all 3 time points in mixed effects regression analyses, which modeled subject-specific coefficients as random effects and used a generalized linear model owing to the binary outcome (use or no use). The findings from the mixed effects modeling and GEE logistic regression analyses were no different, so we are presenting the latter results only.

We used all available data to examine potential nonresponse bias. We performed missing data imputation by using multivariable regression and receiver operating characteristic curves to determine the optimal probability cut point (see Supplemental Information for more detail). The imputed analyses results were similar to the nonimputed, so we are reporting results from nonimputed models.

Results

In New England, 2106 (86.5%) of 2435 eligible patients completed baseline assessments, with TAU participation rate slightly lower than cSBA (z-score = −4.01, P < .01) (Fig 2). TAU participants were older; more likely to be female; of “other” race; to have a parent who did not graduate college; and to report having a parent, sibling, or peer who uses substances (Table 1). They were less likely to be presenting for a well visit, and to have seen a nurse practitioner or a female provider. We controlled for these differences in all subsequent analyses. In Prague, 100% of eligible patients (589) participated, with no significant baseline between-group differences. There were no between-group differences in either country in the frequency of past-year exposure to substance-related messages outside the study. Follow-up rates were >70% in New England and >80% in Prague.

FIGURE 2.

Study design, recruitment, and retention in New England and in Prague, Czech Republic.

TABLE 1.

Baseline Demographics and Visit Characteristics

	New England			Prague, Czech Republic
	ALL N (%) (N = 2096)	TAU n (%) (n = 1068)	cSBA n (%) (n = 1028)	ALL N (%) (N = 589)	TAU n (%) (n = 297)	cSBA n (%) (n = 292)
Age (mean ± SD)	15.8 ± 2.0	15.9 ± 2.0	15.6 ± 2.0	15.0 ± 1.6	15.0 ± 1.6	15.0 ± 1.6
Females	1220 (58.2)	659 (61.7)	561 (54.6)	278 (47.2)	139 (47.6)	139 (46.8)
Race/Ethnicitya
White non-Hispanic	1353 (64.6)	689 (64.5)	664 (64.6)	589 (100)	297 (100)	292 (100)
Hispanic	230 (11.0)	106 (9.9)	124 (12.1)	0 (0.0)	0 (0.0)	0 (0.0)
Asian non-Hispanic	151 (7.2)	77 (7.2)	74 (7.2)	0 (0.0)	0 (0.0)	0 (0.0)
Black non-Hispanic	217 (10.4)	100 (9.4)	117 (11.4)	0 (0.0)	0 (0.0)	0 (0.0)
Other non-Hispanic	145 (6.9)	96 (9.0)	49 (4.8)	0 (0.0)	0 (0.0)	0 (0.0)
Parents at home
Two parents	1424 (69.2)	703 (67.3)	721 (71.0)	379 (65.0)	184 (63.2)	195 (66.8)
One parent or other	635 (30.8)	341 (32.7)	294 (29.0)	204 (35.0)	107 (36.8)	97 (33.2)
Parents’ highest education level
College/University degree or higher	973 (48.0)	451 (44.1)	522 (52.0)	192 (33.1)	92 (31.7)	100 (34.5)
High school/Secondary schoolb graduate	832 (41.0)	427 (41.7)	405 (40.3)	217 (37.4)	111 (38.3)	106 (36.6)
Did not complete high school/secondary school	81 (4.0)	46 (4.5)	33 (3.5)	90 (15.5)	44 (15.2)	46 (15.9)
Don’t know	141 (7.0)	99 (9.7)	42 (4.2)	81 (140)	43 (14.8)	38 (13.1)
Visit type
Well visit	1819 (87.9)	851(81.0)	968 (95.0)	589 (100)	297 (100)	292 (100)
First visit	220 (10.7)	115 (11.0)	105 (10.3)	0 (0)	0 (0)	0 (0)
Female provider	1349 (64.9)	663 (62.9)	684 (67.1)	522 (88.8)	263 (88.9)	259 (88.7)
Parent substance usec	322 (15.4)	170 (15.9)	152 (14.8)	64 (10.9)	32 (11.0)	32 (10.8)
Sibling substance usec	392 (18.7)	205 (19.2)	187 (18.2)	77 (13.2)	41 (13.9)	36 (12.5)
Peer substance usec	1265 (60.5)	658 (61.8)	607 (59.1)	396 (67.3)	204 (68.9)	192 (65.8)

^aIn the Czech Republic, 97% were Czech nationality and 3% other.

^bIncludes secondary school or gymnasium for Czech sample.

^cPercentage reporting any “agree” response to scale items assessing youth-reported parent substance use, sibling substance use, and peer substance use.

Provider brief advice rates doubled in New England, and quadrupled in Prague (Table 2). More providers in both countries advised patients without substance use not to start, than advised patients with substance use to stop. Compared with TAU, more cSBA adolescents rated the provider advice as “Excellent” or “Very Good,” and being “very likely” to follow the provider’s advice and “very satisfied” with the visit.

TABLE 2.

Adolescents’ Reports of Provider’s Brief Counseling Behaviors and Ratings of Visit

Provider Counseling Behaviors	New England				Prague, Czech Republic
	n	TAU n (%) (n = 1015)	cSBA n (%) (n = 1044)	aRRRa,b (95% CI)	n	TAU n (%) (n = 296)	cSBA n (%) (n = 292)	aRRRa,b (95% CI)
Advised about alcohol	2059	441 (42.2)	707 (69.7)	1.57 (1.44–1.71)	586	70 (23.6)	213 (73.2)	3.10 (2.52–3.80)
Not to startc	1280	278 (45.6)	478 (71.3)	1.53 (1.38–1.70)	207	25 (25.8)	78 (70.9)	2.72 (1.90–3.90)
To stopd	779	85 (19.6)	153 (44.3)	2.19 (1.70–2.84)	379	19 (9.6)	76 (42.0)	4.36 (2.76–6.90)
Advised about cannabis and drugs	2059	470 (45.0)	720 (70.9)	1.50 (1.38–1.63)	588	72 (24.3)	231 (79.1)	3.27 (2.67–4.00)
Not to startc	1609	366 (46.0)	568 (69.9)	1.47 (1.34–1.61)	457	49 (21.0)	167 (74.2)	3.56 (2.75–4.59)
To stopd	449	63 (25.5)	109 (54.0)	2.18 (1.60–2.97)	129	16 (16.1)	34 (50.7)	3.17 (1.68–5.97)
Addressed health risks of alcohol	2058	336 (32.2)	673 (66.3)	2.00 (1.80–2.23)	588	36 (12.2)	168 (57.5)	4.79 (3.48–6.59)
Addressed health risks of cannabis and drugs	2057	334 (32.1)	657 (64.7)	2.01 (1.81–2.24)	588	37 (12.5)	160 (54.8)	4.49 (3.27–6.18)
“Excellent”/ “Very Good” rating of provider informatione	1162	318 (70.5)	545 (76.6)	1.09 (1.01–1.18)	300	27 (38.6)	45 (63.0)	1.67 (1.22–2.30)
“Very” likely to follow provider advice	2057	554 (53.1)	603 (59.5)	1.13 (1.04–1.23)	588	65 (22.0)	94 (32.2)	1.50 (1.31–1.98)
“Very” satisfied with visit	2057	646 (61.9)	679 (66.9)	1.07 (1.00–1.15)	588	115 (38.9)	127 (43.5)	1.12 (0.92–1.37)

^aaRRR with TAU as the reference group.

^bUS logistic regression models were run using SUDAAN v. 10.0 software to account for the multisite sampling design, and adjusted for age, gender, race, parent education level, provider type, provider gender, well visit, first visit, any lifetime smoking, any lifetime alcohol or drug use. Czech Republic models adjusted for age and gender only as there were no other differences between experimental groups.

^cRates of advice to not start alcohol or cannabis/drug use were calculated only for adolescents who had no prior use of the substance.

^dRates of advice to stop were calculated only for adolescents who had ever used the substance.

^eAmong adolescents reporting receiving advice about alcohol or drugs.

New England cSBA adolescents reported lower rates of any substance use compared with TAU at both follow-ups (3-months: aRRR=0.62 [95% confidence interval (CI) = 0.44–0.87], adjusted absolute risk difference [ARD] = 9.3%, number needed to treat [NNT] = 11; 12-month aRRR = 0.80 (0.64–0.99), ARD = 7.9%, NNT = 13), largely owing to lower drinking rates at both follow-ups. Among drinkers, there was significantly more cessation of drinking at 3 months, but the effect dissipated by 12 months (Table 3). In contrast, the effect on initiation was not significant at 3 months, perhaps because of small numbers, but robust and significant at 12 months: 44% fewer cSBA adolescents than TAU started drinking during the 12-month study period (ARD = 6%, NNT = 17). We found promising effect sizes in the hypothesized direction for any use and initiation of cannabis at 3 months, but they did not reach statistical significance and dissipated by 12 months.

TABLE 3.

Rates of Self-Reported Alcohol and Cannabis Use, Initiation, and Cessation at Baseline and 3- and 12-mo Follow-up Visits

	Any Past-90-Day Use at 3-Month Follow-up
	New England				Prague, Czech Republic
	N	TAU n (%) (n = 755)	cSBA n (%) (n = 761)	aRRRa,c (95% CI)	N	TAU n (%) (n = 245)	cSBA n (%) (n = 271)	aRRRa,c (95% CI)
Alcohol
Baseline	1516	155 (20.5)	122 (16.0)	0.77 (0.56–1.05)	516	113 (46.1)	129 (47.6)	1.00 (0.80–1.25)
3 Months	1515d	173 (22.9)	118 (15.5)	0.54 (0.38–0.77)e	516	127 (51.8)	126 (46.5)	0.82 (0.63–1.06)
Initiationb	1066	30 (5.9)	17 (3.1)	0.64 (0.32–1.28)	198	10 (11.4)	13 (11.8)	0.96 (0.42–2.21)
Cessationb	450	101 (41.4)	104 (50.7)	1.49 (1.17–1.91)e	318	40 (25.5)	48 (29.8)	1.54 (0.99–2.38)
Cannabisa
Baseline	1516	62 (8.2)	62 (8.1)	1.25 (0.79–1.95)	516	16 (6.5)	14 (5.2)	0.74 (0.35–1.54)
3 Months	1515d	72 (9.5)	56 (7.4)	0.68 (0.40–1.15)	516	24 (9.8)	15 (5.5)	0.37 (0.17–0.77)e
Initiationb	1309	15 (2.3)	8 (1.2)	0.45 (0.18–1.11)	442	8 (3.8)	2 (0.9)	0.22 (0.05–1.04)
Cessationb	206	50 (46.7)	51 (51.5)	1.00 (0.72–1.40)	74	16 (50.0)	29 (69.0)	1.50 (0.97–2.32)
	Any Past-12-Month Use at 12-Month Follow-up
	New England				Prague, Czech Republic
	N	TAU n (%) (n = 758)	cSBA n (%) (n = 765)	aRRRa,c (95% CI)	N	TAU n (%) (n = 266)	cSBA n (%) (n = 264)	aRRRa,c (95% CI)
Alcohol
Baseline	1523	240 (31.7)	194 (25.4)	0.82 (0.64–1.06)	530	163 (61.3)	153 (58.0)	0.89 (0.76–1.03)
12 Months	1523	284 (37.5)	224 (29.3)	0.73 (0.57–0.92)e	530	199 (74.8)	185 (70.1)	0.96 (0.86–1.04)
Initiationb	1089	92 (17.8)	68 (11.9)	0.66 (0.47–0.93)e	216	35 (43.7)	37 (33.3)	0.76 (0.53–1.08)
Cessationb	434	48 (20.0)	38 (19.6)	1.50 (0.93–2.42)	316	9 (5.5)	5 (3.3)	1.18 (0.37–3.73)
Cannabis
Baseline	1522f	101 (13.3)	95 (12.4)	0.77 (0.56–1.05)	529f	36 (13.6)	38 (14.4)	1.02 (0.63–1.64)
12 Months	1522f	133 (17.5)	119 (15.6)	0.85 (0.61–1.19)	529f	76 (28.7)	45 (17.0)	0.47 (0.32–0.71)e
Initiationb	1326	58 (8.8)	52 (7.8)	0.81 (0.54–1.21)	458	47 (20.5)	22 (9.7)	0.47 (0.29–0.76)e
Cessationb	196	27 (26.7)	28 (29.5)	1.01 (0.57–1.78)	74	7 (19.4)	15 (39.5)	2.53 (1.06–6.05)e

^aNew England logistic models for both 3- and 12-month outcomes adjusted for the multisite sampling design; baseline past-12-month substance use; age; gender; parent education level; type of visit (well visit or other); perceived parent, sibling, and peer substance use; provider gender; and connectedness to provider. Prague models adjusted for the multisite sampling design, baseline past-12-month substance use, age, and gender.

^b“Initiation” models analyzed only participants reporting no past-12-month use at baseline, whereas “cessation” models included only those reporting any past-12-month use at baseline. New England and Prague models adjusted for the same variables listed in footnote a, excluding baseline substance use, which is already accounted for by the stratified analyses.

^caRRRs with TAU as the reference group.

^dThere were missing data for 1 respondent each in 3-month New England alcohol use (cSBA) and cannabis use (TAU).

^eP < .05.

^fThere were missing data for 1 respondent each in 12-month marijuana use for New England and Prague TAU groups.

There was no significant cSBA effect on any substance use or on alcohol use in Prague; however, there were significantly reduced cannabis use rates compared with TAU at both follow-ups (3-month ARD = 6%, NNT = 16; 12-month ARD = 15%, NNT = 7). At the 12-month follow-up, we found significant effects for cSBA on both initiation and cessation of cannabis. Compared with New England, Prague participants reported significantly higher drinking rates over lifetime (aRRR = 2.72, 95% CI 2.44–3.04), past-12-months (aRRR = 3.36, 95% CI 2.95–3.82), and past-90-days (aRRR = 4.84, 4.00-5.87). Cannabis use rates were similar between countries.

Discussion

This study provides preliminary evidence of the efficacy of a structured cSBA system in both increasing primary care provider counseling regarding substance use, and reducing adolescent substance use. The cSBA system doubled the number of adolescents receiving brief counseling and increased patient satisfaction with the provider and the visit; however, providers still reached only ∼70% in the cSBA group. We are unable to say why this number was not higher, nor why more providers gave advice not to start than to stop. The latter may be because of some providers’ desire to avoid confrontation with patients who are using substances.

We also found that, with only 2 to 3 minutes of provider time, the cSBA system reduced, relative to usual care, adolescent alcohol use in New England and cannabis use in Prague, with effects persisting through the 12-month study period. The natural trend for substance use prevalence, as shown in national surveys and seen in the current study, is to increase as adolescents age⁴¹; however, the cSBA group had significantly lower rates of use compared with TAU at each follow-up, resulting in a slower increase over time.

The effects on cannabis use in New England were smaller in apparent size, and all in the hypothesized direction, but they did not reach statistical significance. Cannabis use was far less prevalent than drinking in our sample, resulting in small numbers and lower power. Our sample size may have been inadequate to detect an intervention effect above that of assessment reactivity, which can have a substantial impact on substance use.^40–42

In Prague, alcohol use was not affected. We suspect cultural factors played a substantial role. Alcohol use is highly normative in the Czech Republic, which has one of the highest per capita rates of beer consumption among all countries.⁴³ Czech beer is inexpensive (between $1 and $2 US⁴³), and the drinking age is lower than in the United States (18 years). In contrast, cannabis use is not a cultural norm. Our collaboration with the Czech Republic began with an e-mail from a Prague psychiatrist requesting permission to translate and use the CRAFFT screen in a Ministry of Health project to address a sharp rise in adolescent drug use after the 1989 “Velvet Revolution.”⁴⁴ It is therefore gratifying to see that the Czech cSBA system had a powerful and lasting effect on cannabis use. The reasons for different findings in the United States and Czech Republic need further exploration, but underscore the value of multicultural studies.

To date, most studies evaluating screening and brief interventions with adolescents have been conducted in emergency departments,^45–48 college campuses,^21,49–53 or schools.^54–57 The primary care office is a key setting for adolescent screening and brief intervention, with more than 22 million preventive care medical office visits by patients aged 15 to 24 each year,⁵⁸ compared with 19 million emergency department visits.⁵⁹ Also, primary care providers often have long-term relationships with patients and their families, potentially making brief advice more powerful. The use of computers to facilitate the process resulted in increased frequency and quality of physician brief advice, with minimal time burden on providers.

We could find no other published studies on primary care screening and brief intervention for adolescent substance use in English-language journals. We found 1 small study (n = 99) conducted in a single site by De Micheli and colleagues in Brazil.⁶⁰ At 6-month follow-up, they found significantly lower rates of tobacco, alcohol, and cannabis use among youth receiving a prevention intervention. Our study expands this previous work to a larger sample, a variety of practice types, and 2 other countries, enhancing generalizability of the findings and supporting the feasibility of implementation across a range of settings. An additional strength of the current study is that we compared our intervention to TAU that often already included substance use screening and ad hoc advice. This conservative approach may have made it more difficult to detect an intervention effect, but it increased the likelihood that a detected effect is robust.

Our study had potential limitations. We used a nonrandomized, asynchronous study design in which historical trends or other unmeasured group differences may have confounded our results. The 2 groups in New England were not equivalent in baseline substance use, although we controlled for this in data analysis. All study sites were in New England and Prague. Other locations could be different. Our study relied on self-reported and interviewer-collected data, which may be prone to recall error and social desirability bias. Previous studies have shown self-report to be a valid method for measuring substance use among adolescents, however, and it compares favorably with other methods of substance use detection, such as laboratory testing.^14,61,62 Although there was 25% to 30% loss to follow-up in our New England sample, attrition was similar between groups, both in the rates and the profile of those lost to follow-up, and we found that our results did not change after missing data imputation. Finally, we were unable to assess effects on use of drugs other than cannabis because of insufficient numbers.

Future studies should use larger samples and randomized designs, include more information on the negative health effects of substance use, and add new strategies designed to extend the intervention’s effect over time. They should also include strategies to improve intervention fidelity among providers, such as self-report adherence checklists and audiotaping of brief advice with review and feedback. Finally, studies are needed to determine the cost-effectiveness/cost-benefit of cSBA, as well as to elucidate its mechanisms of action so as to promote its effective implementation and dissemination.

Conclusions

Computer-facilitated screening and provider brief advice appears to be a promising strategy for reducing substance use among adolescent primary care patients, although replications of this study are needed with larger samples. The protocol involved only 1 hour of provider training, 5 minutes of patient time before the visit, and 2 to 3 minutes of provider time during the encounter, with some positive effects sustained up to a year later. Providers today face opposing pressures: recommendations to screen patients for more and more problems, and financial realities that require they see more patients quickly.^14,63,64 Use of a computer-facilitated system such as this one may offer a way to improve both patient care and provider efficiency.

Glossary

ARD

absolute risk difference

aRRR

adjusted relative risk ratio

CI

confidence interval

CRAFFT

mnemonic acronym formed by the first letters of key words in the test’s 6 yes/no questions

cSBA

computer-facilitated screening and brief advice

GEE

generalized estimating equations

NNT

number needed to treat

RA

research assistant

TAU

treatment as usual