This is an open-access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work, first published in the Journal of Medical Internet Research, is properly cited. The complete bibliographic information, a link to the original publication on http://www.jmir.org/, as well as this copyright and license information must be included.
Preventing smoking initiation among adolescents is crucial to reducing tobacco-caused death and disease. This study focuses on the effectiveness of a Web-based computer-tailored smoking prevention intervention aimed at adolescents.
The intent of the study was to describe the intervention characteristics and to show the effectiveness and results of a randomized controlled trial. We hypothesized that the intervention would prevent smoking initiation among Dutch secondary school students aged 10-20 years and would have the largest smoking prevention effect among the age cohort of 14-16 years, as smoking uptake in that period is highest.
The intervention consisted of a questionnaire and fully automated computer-tailored feedback on intention to start smoking and motivational determinants. A total of 89 secondary schools were recruited via postal mail and randomized into either the computer-tailored intervention condition or the control condition. Participants had to complete a Web-based questionnaire at baseline and at 6-month follow-up. Data on smoking initiation were collected from 897 students from these schools. To identify intervention effects, multilevel logistic regression analyses were conducted using multiple imputation.
Smoking initiation among students aged 10-20 years was borderline significantly lower in the experimental condition as compared to the control condition 6 months after baseline (OR 0.25, 95% CI 0.05-1.21,
The findings of this study suggest that computer-tailored smoking prevention programs are a promising way of preventing smoking initiation among adolescents for at least 6 months, in particular among the age cohort of 14-16 years. Further research is needed to focus on long-term effects.
International Standard Randomized Controlled Trial Number (ISRCTN): 77864351; http://www.controlled-trials.com/ISRCTN77864351 (Archived by WebCite at http://www.webcitation.org/6BSLKSTm5).
Of every three young smokers, one will die as a result of their tobacco use [
In adolescents, nicotine dependence develops rapidly during experimentation, often before adolescents start smoking on a daily basis [
Several effective, mostly school-based, adolescent smoking prevention programs have been developed [
Computer-tailored interventions provide feedback adapted to the user’s individual characteristics and needs [
Few studies have focused on computer-based tailored programs addressing adolescent smoking prevention. Both Prokhorov and colleagues [
This paper focuses on a Web-based smoking prevention and cessation program aimed at Dutch adolescents, called “Smoke Alert”, which consisted of a Web-based questionnaire and fully automated, computer-tailored feedback. Smokers were provided feedback messages about how to stop smoking and non-smokers could learn how to refrain from smoking. The Smoke Alert program addressed both smoking cessation and prevention, as adolescents in schools for this age category can be both smokers and non-smokers. The intervention presented in this study was an improved version of the Smoke Alert program that was described in an earlier study and had shown positive effects on smoking cessation [
The main aim of this paper is to describe the intervention characteristics and to show the results of the randomized controlled trial on its effectiveness for the prevention of smoking among Dutch adolescents. This trial was conducted among students ranging from 10-20 years of age in order to detect whether implementation could be recommended for different age groups, since usage statistics showed that a wide age range of students participated in the previous version of the Smoke Alert program. We hypothesized that smoking initiation rates would be lower in the experimental condition at 6-month follow-up, as compared to the control condition (hypothesis 1). By targeting social influences and providing skills for refusing cigarettes, we expected the smoking prevention program to be most effective for adolescents in a context in which some of their peers already smoke [
Intervention effectiveness was studied by means of a cluster randomized controlled trial and encompassed the implementation of Smoke Alert in the experimental condition (at school). The intervention was being tested against a no-intervention control condition. Allocation ratio was 1:1 and respondents from both conditions filled out a Web-based questionnaire at baseline and at 6-month follow-up, assessing smoking behavior, intention to start smoking, age, gender, and educational level. The trial is registered in the ISRCTN Register (ISRCTN77864351).
Participants in the present study were students from secondary schools in the Netherlands. The eligibility criteria for participants were: age between 10 and 20 years; having computer/Internet literacy; having sufficient command of Dutch; no previous exposure to the earlier version of Smoke Alert [
The Smoke Alert program was based on the I-Change Model, or the Integrated Model for exploring motivational and behavioral change [
The previous version of Smoke Alert [
The questionnaire and content of the feedback messages of Smoke Alert were updated versions of previously used questionnaires and feedback, derived from evidence-based interventions on smoking prevention and cessation [
To measure intention to start smoking, students were asked to select a statement that best described their situation, with options ranging from “I know for sure I won’t ever start smoking” to “I think I will start smoking within 1 month”. Three social cognitive concepts were measured according to the I-Change Model: namely, attitude towards smoking, perceived social influence, and self-efficacy not to smoke. Attitudes were assessed by 9 items that measured the pros and cons of smoking, for instance: “If I smoked, I would feel more confident”, “If I smoked, it would cost me a lot of money”, etc. Perceived influences from the social environment were measured by 2 items that assessed social modeling. Self-efficacy was measured with 6 items via which students could indicate how sure they were that they could remain a non-smoker in certain situations. These situations can be divided into 2 types: stressful situations (eg, feeling nervous) and social situations (eg, at a party, when friends smoke) [
The respondents used their unique log-in information, provided by their teachers, to access the intervention website at school. Students in the experimental condition received their feedback on the computer screen immediately after filling out the questionnaire. The advice consisted of a home page, containing an introduction and a 30-second animated video, as well as several subpages, each providing feedback on a specific determinant (for a screenshot of the home page, see
The first subpage was dedicated to beliefs about smoking (ie, attitude). The students’ beliefs were considered as a balance indicating whether he or she perceived more or less advantages than disadvantages of smoking. The students’ opinion of each belief was stated and commented on. These messages had the general intention of countering beliefs about the positive effects of smoking (eg, smoking will make me feel relaxed, smoking will make me popular) and to strengthen beliefs about the negative effects (eg, smoking will cost me a lot of money). The second subpage addressed the perceived social influence. Based on students’ answers, they were informed about the negative influence of smokers in their environment. When the student indicated having a lot of smokers in their environment, the idea of smoking as a “normal activity” was counteracted by stating that the majority of people in the Netherlands do not smoke. When most people in the environment of the student were non-smokers, the feedback confirmed that smoking is not the norm in the Netherlands. The third subpage was dedicated to self-efficacy. For situations where the student expected difficulties in remaining a non-smoker, strategies were offered to help the student to get through these situations without initiating smoking (eg, thinking about the reasons for being a non-smoker). The final subpage focused on action plans. The feedback reflected on every action plan the student had indicated he or she would use in situations where someone would offer a cigarette. Examples of action plans were provided when a student was not planning to use a certain action plan. The main message regarding action plans was: by preparing yourself for the situation when someone offers you a cigarette, you will be more confident and it will be easier to refuse the cigarette.
Examples of the computer-tailored feedback messages are provided in
Examples of feedback messages [
Feedback type | Message |
Intention feedback |
|
Attitude feedback |
|
Social influence feedback |
|
Self-efficacy feedback |
|
Action plans feedback |
|
The I-Change Model [
The primary outcome measure was smoking behavior defined as smoking at least occasionally. Respondents were asked to pick a statement that best described them out of 9 smoking-related statements [
Intention to start smoking was measured by asking students to select a statement that best described their situation, with the following response options: (1) “I know for sure I won’t ever start smoking”, (2) “I think I won’t ever start smoking”, (3) “I think I will start smoking in the future”, (4) “I think I will start smoking within 5 years”, (5) “I think I will start smoking within 1 year”, (6) “I think I will start smoking within 6 months”, and (7) “I think I will start smoking within 1 month”. Adolescents were also asked to report their age (in years), gender (1=“male”, 0=“female”) and educational level: high (senior general secondary education / pre-university education=1) or low (practical education / lower secondary professional education=0).
Power analysis was based on the assumption that 2% of the experimental condition would initiate smoking 6 months after baseline, whereas among the control condition the national prevalence rate of ever smoking was expected to increase by 7% at the age of 15, the expected mean age at follow-up. To be able to detect this difference with a power of .80 at 5% significance level (two-sided testing), assuming that 95% of the schools in the control condition have uptake rates between 0.8% and 56%, corresponding with an intraclass correlation (ICC) of .34 on the logit scale, 54 schools and 702 non-smokers should be included in the study. Accounting for the efficiency loss due to unequal amounts of students per school, the number of schools was raised by 10% [
The schools were randomly assigned to the experimental or control condition. Randomization was performed automatically by computer software that was developed specifically for the execution of Web-based computer-tailored programs [
All analyses were done using MLwiN (multilevel modelling for Windows), since adolescents were nested within schools. Ignoring this nesting structure may inflate type I errors and lead to too narrow confidence intervals for treatment effects [
Students’ participation in both conditions was voluntary, respondents were guaranteed anonymity, and it was explained that they could withdraw participation at any time. This study was part of a larger study on the effectiveness of the Smoke Alert study for which ethical clearance was obtained [
In total, 89 schools signed up for participation, resulting in a total of 10,500 students. At baseline, 83 out of 89 schools responded to the questionnaire, 4 schools indicated that they no longer had time to participate, and 2 schools did not explain their non-response to the baseline questionnaire. A total of 6078 students completed the baseline questionnaire, 1099 did not meet the inclusion criteria, resulting in a total of 4979 non-smokers that remained for participation in the follow-up measurement.
Mean age of the respondents at baseline was approximately 14 years (SD 1.1), with age ranging between 10 and 20 years (
The CONSORT flowchart (
Attrition analysis showed that lower educated students were significantly more likely to drop out compared to higher educated students (OR 0.37, 95% CI 0.19-0.70,
Baseline sample characteristics of non-smoking adolescents (n=4979), recruited in 2011.
Characteristic | Total |
Experimental condition (E) (n=2469) | Control condition (C) |
|
Age in years, mean (SD) |
|
13.7 (1.1) | 13.7 (1.0) | 13.7 (1.1) |
Male, n (%) |
|
2518 (50.57%) | 1220 (49.41%) | 1298 (51.71%) |
|
||||
|
Low | 2744 (55.11%) | 1374 (55.65%) | 1370 (54.58%) |
|
High | 2235 (44.89%) | 1095 (44.35%) | 1140 (45.42%) |
Intention to start smoking, mean (SD) |
|
1.55 (0.7) | 1.58 (0.7) | 1.52 (0.7) |
Participant flow chart.
Of the 392 students with complete data in the experimental condition, 15 (3.8%) initiated smoking 6 months after baseline. Of the 505 complete cases in the control condition, 28 (5.5%) initiated smoking.
Interactions were added to this model to analyze whether the effect of the program was gender, education, and age dependent. No significant interactions were found regarding these covariates (
Predictors of smoking initiation at 6-month follow-up.
Predictor | OR | 95% CI |
|
Intervention (1=yes, no=0) | 0.25 | 0.05-1.21 | .09 |
Gender (male=1, female=0) | 1.21 | 0.63-2.32 | .56 |
High educational level (1=yes, no=0) | 0.53 | 0.20-1.37 | .19 |
Age | 0.16 | 0.94-1.42 | .17 |
Intention to start smoking (never=1, within 1 month=7) | 2.51 | 1.62-3.89 | <.001 |
Next, in order to test our second hypothesis, an analysis was done for the age cohort of 14-16 years. There were 385 complete cases (E: n=176; C: n=209), with 24 students (11.5%) in the control condition reporting initiation compared to 10 students (5.7%) in the experimental condition.
Predictors of smoking initiation at 6-month follow-up for students aged 14-16 years.
Predictor | OR | 95% CI |
|
Intervention (1=yes, no=0) | 0.22 | 0.05-1.02 | .05 |
Gender (male=1, female=0) | 1.69 | 0.75-3.84 | .21 |
High educational level (1=yes, no=0) | 0.43 | 0.11-1.63 | .21 |
Age | 2.09 | 1.11-3.94 | .02 |
Intention to start smoking (never=1, within 1 month=7) | 2.96 | 1.83-4.78 | <.001 |
This paper describes a cluster randomized controlled trial examining the effectiveness of a computer-tailored intervention on smoking prevention, called Smoke Alert, aimed at adolescents. This trial was conducted among students aged 10-20 years in order to detect whether implementation could be recommended for a wide age range of students. We hypothesized that smoking initiation rates would be lower in the experimental condition at 6-month follow-up, as compared to the control condition. The results offered some support for our first hypothesis revealing that students in the control condition reported higher smoking initiation at 6-month follow-up. The results provided significant support for our second hypothesis, as the data for the 14-16 year age group showed a significant effect with lower smoking initiation rates in the experimental condition.
The results of this study support earlier findings that Web-based computer-tailored programs can be an effective tool in the prevention of smoking among youth [
Smoking initiation was lower in the experimental condition among both higher and lower educated students. This is encouraging, since educational level is one of the strongest predictors of smoking behavior [
Schools serve as an important access point to reach many adolescents. Hence, it is recommended to incorporate the intervention within the regular curriculum at school. This way, even the least motivated adolescents will participate and complete the intervention. As has been noted previously, implementation challenges at school contribute to the decay of prevention program effects over time [
There are several important limitations to consider in interpreting the results of this study. First, all measurements were self-reports. Biochemical validation may not be necessary or advisable in studies like the current study using Internet data collection without face-to-face contact [
Second, of the 1380 schools approached, only 89 agreed to participate, which may reveal an overall negative climate toward smoking prevention in the Netherlands and/or to participating in experimental studies. Most often, however, reasons for non-participation mentioned were lack of time and lack of interest, which is often the case in many schools in the Netherlands, since health promotion is not integrated into the Dutch school curriculum [
Third, we experienced high but equal loss to follow-up in both the experimental and control condition. The attrition at student level was 82% and outnumbered the expected dropout rate of 50%. High attrition is a well-known feature of many studies of eHealth interventions [
Handling dropout was performed using multiple imputation, thus preserving as many cases for analysis on the intervention effects as possible. Multiple imputation is considered the best method for imputing missing values [
Web-based computer-tailored interventions are a promising way of preventing smoking initiation among adolescents for at least 6 months, in particular among the age cohort of 14-16 years. The findings of the present study illustrate the need for smoking prevention programs beyond the 12-14 year age group that is traditionally targeted by these programs. Long-term assessment is needed to determine if the preventive effect of Web-based computer-tailored interventions is sustained in the years following program delivery.
Home page showing computer-tailored feedback with animated video.
CONSORT-EHEALTH checklist V1.6.2 [
This study was supported by grants from the Dutch Ministry of Health, Welfare and Sport.
Hein de Vries is scientific director of Vision2Health, a company that licenses evidence-based innovative computer-tailored health communication tools.