bs_bs_banner
RESEARCH REPORT
doi:10.1111/add.12383
Balance—a pragmatic randomized controlled trial of an online intensive self-help alcohol intervention Håvar Brendryen1, Ingunn Olea Lund1, Ayna Beate Johansen1, Marianne Riksheim1, Sverre Nesvåg1,2 & Fanny Duckert1 SERAF—Norwegian Centre for Addiction Research, University of Oslo, Oslo, Norway1 and Alcohol and Drug Research Western Norway, Stavanger University Hospital, Stavanger, Norway2
ABSTRACT Aims To compare a brief versus a brief plus intensive self-help version of ‘Balance’, a fully automated online alcohol intervention, on self-reported alcohol consumption. Design A pragmatic randomized controlled trial. Participants in both conditions received an online single session screening procedure including personalized normative feedback. The control group also received an online booklet about the effects of alcohol. The treatment group received the online multi-session follow-up program, Balance. Setting Online study in Norway. Participants At-risk drinkers were recruited by internet advertisements and assigned randomly to one of the two conditions (n = 244). Measurements The primary outcome was self-reported alcohol consumption the previous week measured 6 months after screening. Findings Regression analysis, using baseline carried forward imputation (intent-to-treat), with baseline variables as covariates, showed that intervention significantly affected alcohol consumption at 6 months (B = 2.96; 95% confidence interval = 0.02–5.90; P = 0.049). Participants in the intensive self-help group drank an average of three fewer standard alcohol units compared with participants in the brief self-help group. Conclusions The online Balance intervention, added to a brief online screening intervention, may aid reduction in alcohol consumption compared with the screening intervention and an educational booklet. Keywords Behaviour change intervention, early intervention, harmful drinking, hazardous drinking, mobile telephone text messages. Correspondence to: Håvar Brendryen, SERAF, Postboks 1039 Blindern, 0315 Oslo, Norway. E-mail: [email protected] Submitted 18 December 2012; initial review completed 22 February 2013; final version accepted 30 September 2013
INTRODUCTION Online alcohol interventions can be important in combating alcohol misuse [1–10], but more insight is needed into how different types of online intervention should be implemented into the real world [7,11,12]. One possibility is to chain the brief and intensive self-help intervention formats together into a stepped-care framework [13]. This may exhaust the potential of online automated interventions, prior to introducing more resource-intensive treatment requiring person-to-person communication. Therefore, the feasibility and effectiveness of this public health approach to alcohol misuse needs to be investigated within a naturalistic online setting. Intensive self-help interventions are often based on cognitive–behavioural and self-control principles, and © 2013 Society for the Study of Addiction
may comprise goal-setting, self-monitoring, refusal skills, behavioural contracting, emotion regulation and relapse prevention [5–10]. Such interventions are usually multisession programs and pre-suppose multiple visits for the user to take full advantage of the treatment. In contrast, standard practice for brief interventions, which is the most common format for online self-help, is to provide a screening with personalized normative feedback, and sometimes motivational enhancement components, within a 5–20-minute session [1–4,14]. Brief interventions require little from the users, and it is a viable and effective treatment option alone, but effect sizes tend to be small [3]. However, brief interventions also represent an opportunity for recruiting select people to intensive selfhelp. A recent meta-analysis found that the effect sizes were higher in trials of intensive self-help interventions compared to trials of brief interventions [4]. In previous Addiction, 109, 218–226
RCT of an online alcohol intervention
trials, intensive self-help was compared to passive control conditions [5–7] (i.e. neither personalized normative feedback nor motivational enhancement was provided to the control group), or the recruitment or treatment procedure required either contact beyond the internet or person-to-person communication [8–10]. To date, no trials have compared brief intervention alone to brief plus intensive self-help in an online setting [1–4]. This paper reports on a pragmatic randomized controlled trial comparing two versions of the online intervention ‘Balance’ [15]: the brief intervention-only versus the full version comprising both the brief and intensive self-help. We hypothesized improved treatment outcome in the intensive self-help condition compared to brief intervention only, 2 and 6 months after baseline.
METHODS Participants and procedure Participants were recruited by banner advertisements in online newspapers. When clicking on the advertisement, participants were routed to a website with information on the study. They were informed that participants would be assigned randomly to groups receiving different online tools that all started with screening and feedback on alcohol habits. They had the opportunity to explore the screening and feedback before deciding to participate in the study. They were informed that providing an e-mail and telephone number during the screening implied consent to participate. Participants were informed that by consenting they would receive additional follow-up, including surveys after 2 and 6 months. There was no reimbursement for participation. From the information page the participants were routed to the online screening procedure. After the personalized normative feedback, participants were asked to provide their e-mail address and telephone number if they wanted further information and follow-up, and consented to participate in the study. Trial participants thus signed up in the same way that people in a non-research setting would sign up for the treatment. To be eligible for inclusion into the study the participant had to be an at-risk drinker aged 18 years or older; complete the baseline assessment with no missing items; and provide a valid e-mail address and Norwegian mobile telephone number. At-risk drinking was defined as a score of 3 or higher on The Fast Alcohol Screening Test (FAST) [16]. A computerized automatic simple randomization procedure was performed throughout the recruitment period (from April to November 2011), assigning 244 participants either to the intensive or the brief self-help condition. The randomization took place immediately after each participant had provided an e-mail address. An e-mail was sent to participants © 2013 Society for the Study of Addiction
219
randomized to the intensive self-help condition on the consecutive day with a link and an invitation to the first follow-up session, whereas participants randomized to the brief self-help condition were routed to the booklet. Data were collected using online questionnaires at baseline and at 2 and 6 months post-baseline times. At follow-up points, an e-mail with a link to the questionnaire was sent to participants. Two e-mail reminders were sent to non-responders at both points. At the 6-month follow-up, participants not responding to e-mails were contacted by telephone. The telephone interviewer was blinded to allocation. Participants in the experimental condition logged on to the online sessions with a unique username and password, allowing for a continuous, unobtrusive and reliable measure of program exposure [17,18]. Apart from the telephone interview, there was no person-to-person interaction between participant and experimenters.
Interventions In a previous paper [15], we described the treatment rationale of ‘Balance’, an online intervention that combines the brief and intensive self-help formats. It starts with a screening and feedback session. The purpose is to provide personalized normative feedback; support an informed choice of whether to change or not; and recruit individuals to the intensive self-help intervention. The feedback compared the reported drinking habits to the recommended gender-matched low-risk drinking guidelines. Additionally, they were compared with national gender-matched averages; for example, ‘the average Norwegian male consumes more than 5 drinks less than once a month’, and ‘two thirds of Norwegian males report they have never experienced a blackout’. At the end of the feedback session, users identified as engaging in at-risk drinking are recommended to sign up for the intensive self-help program. The intensive self-help comprises 62 online sessions that are released one by one in a predetermined sequence during 6 months. The control group received an e-booklet, issued by the Norwegian Directorate of Health that covers general information about alcohol and potential risks and harms of drinking. Neither the screening session nor the booklet contains advice on how to achieve a change in drinking behaviour. The rationale for delivering the booklet was not to improve the brief intervention, but to provide something that appeared as a plausible follow-up to the control participants. This was necessary because the participants, prior to the screening, were told that they would receive more alcohol-related information and follow-up after they signed up for the study. Addiction, 109, 218–226
220
Håvar Brendryen et al.
Table 1 Overview of the online sessions across program phase. Program structure and phases Brief intervention Single session with screening (i.e. FAST) and personalized normative feedback Active behaviour change phase Fifty-six sessions: one new session available daily for 8 weeks. Daily logging of alcohol consumption and self-efficacy. Supportive mobile telephone text messages available on demand Follow-up phase Six sessions: one session per week for 4 weeks (four sessions), then once every fourth week for the remaining period (two sessions); supportive mobile telephone text messages available on demand For each interactive online-session an e-mail reminder was sent to the subjects. Within the 6 months of follow-up, the Balance intervention comprises a total of 62 follow-up sessions (not counting the screening session). FAST = Fast Alcohol Screening Test.
The intensive self-help program was delivered to the treatment group through multiple interactive sessions, reminder e-mails and mobile telephone text messages (Table 1). The central concept of Balance is to support continued self-regulation throughout the behaviour change process [15,19]. There are four key aspects of the program; the first is focus on goal-setting and tracking of alcohol consumption on a day-to-day basis. The second is on relapse prevention; for example, when clients report drinking more than their pre-set target, they receive personalized content aimed at preventing a full-blown relapse [20]. The third is emotion regulation [21], where content and assignments from positive psychology and from cognitive behavioural therapy are used [22–24]. Finally, the intervention covers alcohol education (i.e. the same topics as in the booklet provided to the control group). Balance uses tunnelled information architecture [25]. The program withholds and gradually releases sessions in a predetermined sequence. This facilitates learning moments that are short and many rather than few and lengthy, and are in accordance with the need for continuous self-regulation. The awareness of one’s own attempt to self-regulate and change behaviour will be stimulated by the distributed and frequent contact points, which represent a tacit way of telling the clients that behaviour change is a process that calls for sustained effort [26–28]. In a recent trial, participants randomized to a freedom of choice condition were more satisfied but learned less compared to participants in a tunnel design condition [29], demonstrating that restricting the freedom of choice can improve learning in e-health interventions. Although restricting choice, a © 2013 Society for the Study of Addiction
tunnel design does not dictate passive users, as it is well suited to foster interactive dialogues. Hence, the sessions include interactive tasks, cognitive behavioural assignments and quizzes. As part of a lapse prevention system, participants are given the option of scheduling a supportive mobile telephone text message tailored to their reported need on that particular day (motivation, mood, or self-efficacy). An average session consists typically of 1000 words, split between 10 and 15 screens. Each session takes from 3 to 10 minutes to complete, depending on the clients’ depth of processing and speed of reading. Completing the 62 follow-up sessions thus requires up to 10 hours of reading. Unless actively withdrawing, the participant will receive an e-mail reminder each time a new session is released. There is a danger that the many e-mails create annoyance among participants with low motivation to complete the program. The treatment rationale of the intervention is described thoroughly in a separate paper [15]. The Norwegian version of the program is currently publicly available [30].
Measures at baseline Total weekly alcohol consumption was defined as the sum of drinks reported for each of the previous 7 days. The scale for each day ranged from zero to 10, giving a weekly consumption range from zero to 70. In Norway, a standard alcohol unit is equivalent to 12 g of pure alcohol. The Fast Alcohol Screening Test (FAST) is a brief version of The Alcohol Use Disorder Identification Test (AUDIT) [16]. The four items of FAST was administered at baseline. The first item of FAST, which measures frequency of binge episodes during the last year, was modified to reflect the Norwegian official guidelines for low-risk drinking. A binge drinking episode was defined as more than four or more than six drinks in one session for females and males, respectively. FAST also includes items that assess how often during the last year responders have experienced memory problems, failed to do what is normally expected from them due to alcohol consumption, and if anyone has been concerned about their drinking. FAST scores range from zero to 16, each item score ranges from zero to 4 and a score of 3 or higher suggests at-risk drinking. Gender and age were the only socio-demographic characteristics that were assessed. A broader baseline assessment was not included due to practical and methodological issues. By preserving the screening procedure in the way it was already used in practice, ecological validity would be higher compared to a more complex screening assessment. Addiction, 109, 218–226
RCT of an online alcohol intervention
Assessed for eligibility (n = 276)
221
Excluded (n = 32):
Enrolment
Not at-risk drinker (n = 20) No valid e-mail (n = 6) No valid mobile phone number (n = 2) < 18 years (n = 4)
Analyses
Follow up
Allocation
Randomised (n = 244)
Figure 1 Flowchart
Allocated to brief plus intensive self-help (n = 125)
Allocated to brief self-help (n = 119)
Lost to follow-up 2 months (n = 65) 6 months (n = 52)
Lost to follow-up 2 months (n = 28) 6 months (n = 25)
Intent-to-treat analyses (n = 125)
Intent-to-treat analyses (n = 119)
Complete case analyses at 2 months (n = 60)
Complete case analyses at 2 months (n = 91)
Complete case analyses at 6 months (n = 73)
Complete case analyses at 6 months (n = 94)
Measures at 2- and 6-month follow-up
RESULTS
Total weekly alcohol consumption was the primary and only outcome reported herein. This was measured in the same way at the 2- and 6-month follow-up as at baseline.
There were 276 unique registrations, 32 of which did not fulfil the inclusion criteria (Fig. 1). This left 244 participants to be randomized into two experimental conditions. Table 2 shows participant characteristics at baseline. The response rate of the web surveys fell from the 2- to the 6-month follow-up. With additional telephone-based follow-up at the 6-month point, the total response rate increased (Table 3). Significantly more participants in the control group responded to the survey both at 2 months (χ2 = 19.8, P < 0.001) and 6 months (χ2 = 11.0, P < 0.001). There were no significant differences between non-responders and responders on any baseline variables at any of the follow-up assessments for the sample as a whole, or for experimental conditions separately. Table 4 shows the results of the series of regression analyses comparing alcohol consumption per week between the control and treatment groups. While the results at two months was inconclusive, the final model for the main comparison at 6 months showed that intervention affected alcohol consumption significantly [B = 2.96; 95% confidence interval (CI) = 0.02–5.90; P = 0.049]. Specifically, participants in the intensive
Analyses All the analyses were based on the standard alpha level of 0.05 (two-tailed). Assuming an effect size of Cohen’s d = 0.35, a sample size of 260 was necessary to achieve an 80% chance (power) of detecting an effect at the P < 0.05 level of significance. Linear regression analyses were used to compare outcomes across conditions for each of the two follow-up points. The 6-month follow-up is considered the prime outcome. The primary comparisons applied the intent-to-treat principle, in which all missing values were substituted with baseline values. Complete case analyses were performed as secondary comparisons. Three linear regression models were performed. The first model included experimental condition as the only predictor; the second model included baseline weekly alcohol consumption as a covariate; and the third model included all the baseline variables (i.e. baseline alcohol consumption, FAST, age and gender). © 2013 Society for the Study of Addiction
Addiction, 109, 218–226
222
Håvar Brendryen et al.
Table 2 Baseline characteristics and drinking outcome at 2 and 6 months. Treatment n = 125
Control n = 119
Baseline Female Age (years) FAST Drinks/week Two months Drinks/week, complete case Drinks/week, intent-to-treat Six months Drinks/week, complete case Drinks/week, intent-to-treat
n (%) 38 (30) Mean ± SD 39 ± 14 6.3 ± 3.0 19.8 ± 14.0
37 ± 13 6.2 ± 2.8 19.4 ± 12.8
14.9 ± 15.6 16.3 ± 13.5
17.3 ± 13.0 18.1 ± 13.4
12.7 ± 12.0 15.4 ± 13.6
17.3 ± 16.0 17.9 ± 15.7
43 (36)
Fast Alcohol Screening Test (FAST) ranges from 0 to 16; a score greater than 2 indicates at-risk drinking. SD = standard deviation.
Table 3 Number of responders in treatment (n = 125) and control group (n = 119) at specified data collections. Data collection
Treatment
Control
Two months (web) Six months (web) Six months (telephone) Six months (web+telephone)
60 (48.0%) 44 (35.2%) 29 (23.2%) 73 (58.4%)
91 (76.5%) 78 (65.5%) 16 (13.4%) 94 (79.0%)
self-help group drank approximately three drinks fewer compared to participants in the brief self-help group. This effect corresponds to a Cohen’s d of 0.20, which is considered a small effect size. The mean and standard deviations for drinking outcome across condition at 2 and 6 months are reported in Table 2. Of the 244 people in the study, 109 reported drinking 10 or more drinks on any day during the three assessment weeks, suggesting a ceiling effect. Treatment adherence As randomization took place at the end of the screening session, all randomized participants completed the normative feedback session. For technical reasons, we have no data for the control group’s use of the e-booklet. Of the 125 participants allocated to the active treatment condition, half (n = 63) dropped out, completing fewer than three sessions. Of the 63 ‘early dropouts’, 38 did not even log on to the first follow-up session. Of the 62 ‘stayers’ (≥3 sessions), 31 completed 20 sessions or more and 10 individuals stayed for the full 6 months of the program (62 sessions). In other words, treatment dose was distributed unequally. Moreover, post-hoc analyses showed a substan© 2013 Society for the Study of Addiction
tial overlap between being an early treatment dropout and not responding to the questionnaires: at the 2-month follow-up, the measurement attrition rates were 75 versus 29% for the early dropouts and the stayers, respectively. At the 6-month follow-up, the measurement attrition rates were 56 versus 27% for the early dropouts and stayers, respectively. The response rates among those completing three sessions or more was similar to that observed in the control group. Treatment adherence and its relation to baseline characteristics and treatment outcome will be analysed in detail in a subsequent paper.
DISCUSSION In this study we compared two types of fully automated online self-help interventions: a traditional brief intervention, which included personalized normative feedback, and an intensive self-help program that also included the brief component. The results partially supported the hypothesis that adding an intensive self-help program to an online brief intervention would increase its effectiveness; that is, an effect was apparent, but statistical significance depended upon analytical strategy, follow-up time and covariates. At the 2-month follow-up the evidence was inconclusive, indicating reduced drinking in the intensive group when examining the complete case analysis, but no difference when using an intent-to-treat approach. By the 6-month follow-up, the intensive group showed improved outcomes with both analytical approaches. None the less, the intent-to-treat analyses were significant only when controlling for baseline characteristics, and the effect size was small. However, the intervention had no apparent negative effects on alcohol consumption, indicating intervention safety regardless of follow-up time, analysis and control for covariates. Due to limitations on time and budget, we were not able to recruit as many participants as we planned to, meaning that this trial can be slightly underpowered to detect true effects. Nevertheless, the primary outcome comparison of this trial, using intentto-treat analysis with baseline carried forward imputation 6 months after baseline, showed that participants in the intensive self-help condition had reduced their weekly consumption by three drinks more than the people who received the brief intervention only. Currently, the most common format for online alcohol programs is brief intervention [1–4]. Brief interventions are effective, but effect sizes tend to be small [3]. In a previous trial [14], two additional sessions of screening and feedback did not enhance the effect of brief intervention. In the current trial, instead of giving more of the same, intensive self-help was provided after a screening and feedback session. The current trial adds to this literature by providing preliminary evidence in support of this Addiction, 109, 218–226
© 2013 Society for the Study of Addiction
5.22 4.51 0.78
4.68 0.76 0.85 0.17 2.61
−1.56
−0.40 0.60
−0.25 0.52
−0.26 −0.02 −2.79
2.06 0.69
2.22 0.64
0.29 0.08 −0.09
95% CI
1.83
B
0.29 0.10 0.95
0.078
RESEARCH REPORT
doi:10.1111/add.12383
Balance—a pragmatic randomized controlled trial of an online intensive self-help alcohol intervention Håvar Brendryen1, Ingunn Olea Lund1, Ayna Beate Johansen1, Marianne Riksheim1, Sverre Nesvåg1,2 & Fanny Duckert1 SERAF—Norwegian Centre for Addiction Research, University of Oslo, Oslo, Norway1 and Alcohol and Drug Research Western Norway, Stavanger University Hospital, Stavanger, Norway2
ABSTRACT Aims To compare a brief versus a brief plus intensive self-help version of ‘Balance’, a fully automated online alcohol intervention, on self-reported alcohol consumption. Design A pragmatic randomized controlled trial. Participants in both conditions received an online single session screening procedure including personalized normative feedback. The control group also received an online booklet about the effects of alcohol. The treatment group received the online multi-session follow-up program, Balance. Setting Online study in Norway. Participants At-risk drinkers were recruited by internet advertisements and assigned randomly to one of the two conditions (n = 244). Measurements The primary outcome was self-reported alcohol consumption the previous week measured 6 months after screening. Findings Regression analysis, using baseline carried forward imputation (intent-to-treat), with baseline variables as covariates, showed that intervention significantly affected alcohol consumption at 6 months (B = 2.96; 95% confidence interval = 0.02–5.90; P = 0.049). Participants in the intensive self-help group drank an average of three fewer standard alcohol units compared with participants in the brief self-help group. Conclusions The online Balance intervention, added to a brief online screening intervention, may aid reduction in alcohol consumption compared with the screening intervention and an educational booklet. Keywords Behaviour change intervention, early intervention, harmful drinking, hazardous drinking, mobile telephone text messages. Correspondence to: Håvar Brendryen, SERAF, Postboks 1039 Blindern, 0315 Oslo, Norway. E-mail: [email protected] Submitted 18 December 2012; initial review completed 22 February 2013; final version accepted 30 September 2013
INTRODUCTION Online alcohol interventions can be important in combating alcohol misuse [1–10], but more insight is needed into how different types of online intervention should be implemented into the real world [7,11,12]. One possibility is to chain the brief and intensive self-help intervention formats together into a stepped-care framework [13]. This may exhaust the potential of online automated interventions, prior to introducing more resource-intensive treatment requiring person-to-person communication. Therefore, the feasibility and effectiveness of this public health approach to alcohol misuse needs to be investigated within a naturalistic online setting. Intensive self-help interventions are often based on cognitive–behavioural and self-control principles, and © 2013 Society for the Study of Addiction
may comprise goal-setting, self-monitoring, refusal skills, behavioural contracting, emotion regulation and relapse prevention [5–10]. Such interventions are usually multisession programs and pre-suppose multiple visits for the user to take full advantage of the treatment. In contrast, standard practice for brief interventions, which is the most common format for online self-help, is to provide a screening with personalized normative feedback, and sometimes motivational enhancement components, within a 5–20-minute session [1–4,14]. Brief interventions require little from the users, and it is a viable and effective treatment option alone, but effect sizes tend to be small [3]. However, brief interventions also represent an opportunity for recruiting select people to intensive selfhelp. A recent meta-analysis found that the effect sizes were higher in trials of intensive self-help interventions compared to trials of brief interventions [4]. In previous Addiction, 109, 218–226
RCT of an online alcohol intervention
trials, intensive self-help was compared to passive control conditions [5–7] (i.e. neither personalized normative feedback nor motivational enhancement was provided to the control group), or the recruitment or treatment procedure required either contact beyond the internet or person-to-person communication [8–10]. To date, no trials have compared brief intervention alone to brief plus intensive self-help in an online setting [1–4]. This paper reports on a pragmatic randomized controlled trial comparing two versions of the online intervention ‘Balance’ [15]: the brief intervention-only versus the full version comprising both the brief and intensive self-help. We hypothesized improved treatment outcome in the intensive self-help condition compared to brief intervention only, 2 and 6 months after baseline.
METHODS Participants and procedure Participants were recruited by banner advertisements in online newspapers. When clicking on the advertisement, participants were routed to a website with information on the study. They were informed that participants would be assigned randomly to groups receiving different online tools that all started with screening and feedback on alcohol habits. They had the opportunity to explore the screening and feedback before deciding to participate in the study. They were informed that providing an e-mail and telephone number during the screening implied consent to participate. Participants were informed that by consenting they would receive additional follow-up, including surveys after 2 and 6 months. There was no reimbursement for participation. From the information page the participants were routed to the online screening procedure. After the personalized normative feedback, participants were asked to provide their e-mail address and telephone number if they wanted further information and follow-up, and consented to participate in the study. Trial participants thus signed up in the same way that people in a non-research setting would sign up for the treatment. To be eligible for inclusion into the study the participant had to be an at-risk drinker aged 18 years or older; complete the baseline assessment with no missing items; and provide a valid e-mail address and Norwegian mobile telephone number. At-risk drinking was defined as a score of 3 or higher on The Fast Alcohol Screening Test (FAST) [16]. A computerized automatic simple randomization procedure was performed throughout the recruitment period (from April to November 2011), assigning 244 participants either to the intensive or the brief self-help condition. The randomization took place immediately after each participant had provided an e-mail address. An e-mail was sent to participants © 2013 Society for the Study of Addiction
219
randomized to the intensive self-help condition on the consecutive day with a link and an invitation to the first follow-up session, whereas participants randomized to the brief self-help condition were routed to the booklet. Data were collected using online questionnaires at baseline and at 2 and 6 months post-baseline times. At follow-up points, an e-mail with a link to the questionnaire was sent to participants. Two e-mail reminders were sent to non-responders at both points. At the 6-month follow-up, participants not responding to e-mails were contacted by telephone. The telephone interviewer was blinded to allocation. Participants in the experimental condition logged on to the online sessions with a unique username and password, allowing for a continuous, unobtrusive and reliable measure of program exposure [17,18]. Apart from the telephone interview, there was no person-to-person interaction between participant and experimenters.
Interventions In a previous paper [15], we described the treatment rationale of ‘Balance’, an online intervention that combines the brief and intensive self-help formats. It starts with a screening and feedback session. The purpose is to provide personalized normative feedback; support an informed choice of whether to change or not; and recruit individuals to the intensive self-help intervention. The feedback compared the reported drinking habits to the recommended gender-matched low-risk drinking guidelines. Additionally, they were compared with national gender-matched averages; for example, ‘the average Norwegian male consumes more than 5 drinks less than once a month’, and ‘two thirds of Norwegian males report they have never experienced a blackout’. At the end of the feedback session, users identified as engaging in at-risk drinking are recommended to sign up for the intensive self-help program. The intensive self-help comprises 62 online sessions that are released one by one in a predetermined sequence during 6 months. The control group received an e-booklet, issued by the Norwegian Directorate of Health that covers general information about alcohol and potential risks and harms of drinking. Neither the screening session nor the booklet contains advice on how to achieve a change in drinking behaviour. The rationale for delivering the booklet was not to improve the brief intervention, but to provide something that appeared as a plausible follow-up to the control participants. This was necessary because the participants, prior to the screening, were told that they would receive more alcohol-related information and follow-up after they signed up for the study. Addiction, 109, 218–226
220
Håvar Brendryen et al.
Table 1 Overview of the online sessions across program phase. Program structure and phases Brief intervention Single session with screening (i.e. FAST) and personalized normative feedback Active behaviour change phase Fifty-six sessions: one new session available daily for 8 weeks. Daily logging of alcohol consumption and self-efficacy. Supportive mobile telephone text messages available on demand Follow-up phase Six sessions: one session per week for 4 weeks (four sessions), then once every fourth week for the remaining period (two sessions); supportive mobile telephone text messages available on demand For each interactive online-session an e-mail reminder was sent to the subjects. Within the 6 months of follow-up, the Balance intervention comprises a total of 62 follow-up sessions (not counting the screening session). FAST = Fast Alcohol Screening Test.
The intensive self-help program was delivered to the treatment group through multiple interactive sessions, reminder e-mails and mobile telephone text messages (Table 1). The central concept of Balance is to support continued self-regulation throughout the behaviour change process [15,19]. There are four key aspects of the program; the first is focus on goal-setting and tracking of alcohol consumption on a day-to-day basis. The second is on relapse prevention; for example, when clients report drinking more than their pre-set target, they receive personalized content aimed at preventing a full-blown relapse [20]. The third is emotion regulation [21], where content and assignments from positive psychology and from cognitive behavioural therapy are used [22–24]. Finally, the intervention covers alcohol education (i.e. the same topics as in the booklet provided to the control group). Balance uses tunnelled information architecture [25]. The program withholds and gradually releases sessions in a predetermined sequence. This facilitates learning moments that are short and many rather than few and lengthy, and are in accordance with the need for continuous self-regulation. The awareness of one’s own attempt to self-regulate and change behaviour will be stimulated by the distributed and frequent contact points, which represent a tacit way of telling the clients that behaviour change is a process that calls for sustained effort [26–28]. In a recent trial, participants randomized to a freedom of choice condition were more satisfied but learned less compared to participants in a tunnel design condition [29], demonstrating that restricting the freedom of choice can improve learning in e-health interventions. Although restricting choice, a © 2013 Society for the Study of Addiction
tunnel design does not dictate passive users, as it is well suited to foster interactive dialogues. Hence, the sessions include interactive tasks, cognitive behavioural assignments and quizzes. As part of a lapse prevention system, participants are given the option of scheduling a supportive mobile telephone text message tailored to their reported need on that particular day (motivation, mood, or self-efficacy). An average session consists typically of 1000 words, split between 10 and 15 screens. Each session takes from 3 to 10 minutes to complete, depending on the clients’ depth of processing and speed of reading. Completing the 62 follow-up sessions thus requires up to 10 hours of reading. Unless actively withdrawing, the participant will receive an e-mail reminder each time a new session is released. There is a danger that the many e-mails create annoyance among participants with low motivation to complete the program. The treatment rationale of the intervention is described thoroughly in a separate paper [15]. The Norwegian version of the program is currently publicly available [30].
Measures at baseline Total weekly alcohol consumption was defined as the sum of drinks reported for each of the previous 7 days. The scale for each day ranged from zero to 10, giving a weekly consumption range from zero to 70. In Norway, a standard alcohol unit is equivalent to 12 g of pure alcohol. The Fast Alcohol Screening Test (FAST) is a brief version of The Alcohol Use Disorder Identification Test (AUDIT) [16]. The four items of FAST was administered at baseline. The first item of FAST, which measures frequency of binge episodes during the last year, was modified to reflect the Norwegian official guidelines for low-risk drinking. A binge drinking episode was defined as more than four or more than six drinks in one session for females and males, respectively. FAST also includes items that assess how often during the last year responders have experienced memory problems, failed to do what is normally expected from them due to alcohol consumption, and if anyone has been concerned about their drinking. FAST scores range from zero to 16, each item score ranges from zero to 4 and a score of 3 or higher suggests at-risk drinking. Gender and age were the only socio-demographic characteristics that were assessed. A broader baseline assessment was not included due to practical and methodological issues. By preserving the screening procedure in the way it was already used in practice, ecological validity would be higher compared to a more complex screening assessment. Addiction, 109, 218–226
RCT of an online alcohol intervention
Assessed for eligibility (n = 276)
221
Excluded (n = 32):
Enrolment
Not at-risk drinker (n = 20) No valid e-mail (n = 6) No valid mobile phone number (n = 2) < 18 years (n = 4)
Analyses
Follow up
Allocation
Randomised (n = 244)
Figure 1 Flowchart
Allocated to brief plus intensive self-help (n = 125)
Allocated to brief self-help (n = 119)
Lost to follow-up 2 months (n = 65) 6 months (n = 52)
Lost to follow-up 2 months (n = 28) 6 months (n = 25)
Intent-to-treat analyses (n = 125)
Intent-to-treat analyses (n = 119)
Complete case analyses at 2 months (n = 60)
Complete case analyses at 2 months (n = 91)
Complete case analyses at 6 months (n = 73)
Complete case analyses at 6 months (n = 94)
Measures at 2- and 6-month follow-up
RESULTS
Total weekly alcohol consumption was the primary and only outcome reported herein. This was measured in the same way at the 2- and 6-month follow-up as at baseline.
There were 276 unique registrations, 32 of which did not fulfil the inclusion criteria (Fig. 1). This left 244 participants to be randomized into two experimental conditions. Table 2 shows participant characteristics at baseline. The response rate of the web surveys fell from the 2- to the 6-month follow-up. With additional telephone-based follow-up at the 6-month point, the total response rate increased (Table 3). Significantly more participants in the control group responded to the survey both at 2 months (χ2 = 19.8, P < 0.001) and 6 months (χ2 = 11.0, P < 0.001). There were no significant differences between non-responders and responders on any baseline variables at any of the follow-up assessments for the sample as a whole, or for experimental conditions separately. Table 4 shows the results of the series of regression analyses comparing alcohol consumption per week between the control and treatment groups. While the results at two months was inconclusive, the final model for the main comparison at 6 months showed that intervention affected alcohol consumption significantly [B = 2.96; 95% confidence interval (CI) = 0.02–5.90; P = 0.049]. Specifically, participants in the intensive
Analyses All the analyses were based on the standard alpha level of 0.05 (two-tailed). Assuming an effect size of Cohen’s d = 0.35, a sample size of 260 was necessary to achieve an 80% chance (power) of detecting an effect at the P < 0.05 level of significance. Linear regression analyses were used to compare outcomes across conditions for each of the two follow-up points. The 6-month follow-up is considered the prime outcome. The primary comparisons applied the intent-to-treat principle, in which all missing values were substituted with baseline values. Complete case analyses were performed as secondary comparisons. Three linear regression models were performed. The first model included experimental condition as the only predictor; the second model included baseline weekly alcohol consumption as a covariate; and the third model included all the baseline variables (i.e. baseline alcohol consumption, FAST, age and gender). © 2013 Society for the Study of Addiction
Addiction, 109, 218–226
222
Håvar Brendryen et al.
Table 2 Baseline characteristics and drinking outcome at 2 and 6 months. Treatment n = 125
Control n = 119
Baseline Female Age (years) FAST Drinks/week Two months Drinks/week, complete case Drinks/week, intent-to-treat Six months Drinks/week, complete case Drinks/week, intent-to-treat
n (%) 38 (30) Mean ± SD 39 ± 14 6.3 ± 3.0 19.8 ± 14.0
37 ± 13 6.2 ± 2.8 19.4 ± 12.8
14.9 ± 15.6 16.3 ± 13.5
17.3 ± 13.0 18.1 ± 13.4
12.7 ± 12.0 15.4 ± 13.6
17.3 ± 16.0 17.9 ± 15.7
43 (36)
Fast Alcohol Screening Test (FAST) ranges from 0 to 16; a score greater than 2 indicates at-risk drinking. SD = standard deviation.
Table 3 Number of responders in treatment (n = 125) and control group (n = 119) at specified data collections. Data collection
Treatment
Control
Two months (web) Six months (web) Six months (telephone) Six months (web+telephone)
60 (48.0%) 44 (35.2%) 29 (23.2%) 73 (58.4%)
91 (76.5%) 78 (65.5%) 16 (13.4%) 94 (79.0%)
self-help group drank approximately three drinks fewer compared to participants in the brief self-help group. This effect corresponds to a Cohen’s d of 0.20, which is considered a small effect size. The mean and standard deviations for drinking outcome across condition at 2 and 6 months are reported in Table 2. Of the 244 people in the study, 109 reported drinking 10 or more drinks on any day during the three assessment weeks, suggesting a ceiling effect. Treatment adherence As randomization took place at the end of the screening session, all randomized participants completed the normative feedback session. For technical reasons, we have no data for the control group’s use of the e-booklet. Of the 125 participants allocated to the active treatment condition, half (n = 63) dropped out, completing fewer than three sessions. Of the 63 ‘early dropouts’, 38 did not even log on to the first follow-up session. Of the 62 ‘stayers’ (≥3 sessions), 31 completed 20 sessions or more and 10 individuals stayed for the full 6 months of the program (62 sessions). In other words, treatment dose was distributed unequally. Moreover, post-hoc analyses showed a substan© 2013 Society for the Study of Addiction
tial overlap between being an early treatment dropout and not responding to the questionnaires: at the 2-month follow-up, the measurement attrition rates were 75 versus 29% for the early dropouts and the stayers, respectively. At the 6-month follow-up, the measurement attrition rates were 56 versus 27% for the early dropouts and stayers, respectively. The response rates among those completing three sessions or more was similar to that observed in the control group. Treatment adherence and its relation to baseline characteristics and treatment outcome will be analysed in detail in a subsequent paper.
DISCUSSION In this study we compared two types of fully automated online self-help interventions: a traditional brief intervention, which included personalized normative feedback, and an intensive self-help program that also included the brief component. The results partially supported the hypothesis that adding an intensive self-help program to an online brief intervention would increase its effectiveness; that is, an effect was apparent, but statistical significance depended upon analytical strategy, follow-up time and covariates. At the 2-month follow-up the evidence was inconclusive, indicating reduced drinking in the intensive group when examining the complete case analysis, but no difference when using an intent-to-treat approach. By the 6-month follow-up, the intensive group showed improved outcomes with both analytical approaches. None the less, the intent-to-treat analyses were significant only when controlling for baseline characteristics, and the effect size was small. However, the intervention had no apparent negative effects on alcohol consumption, indicating intervention safety regardless of follow-up time, analysis and control for covariates. Due to limitations on time and budget, we were not able to recruit as many participants as we planned to, meaning that this trial can be slightly underpowered to detect true effects. Nevertheless, the primary outcome comparison of this trial, using intentto-treat analysis with baseline carried forward imputation 6 months after baseline, showed that participants in the intensive self-help condition had reduced their weekly consumption by three drinks more than the people who received the brief intervention only. Currently, the most common format for online alcohol programs is brief intervention [1–4]. Brief interventions are effective, but effect sizes tend to be small [3]. In a previous trial [14], two additional sessions of screening and feedback did not enhance the effect of brief intervention. In the current trial, instead of giving more of the same, intensive self-help was provided after a screening and feedback session. The current trial adds to this literature by providing preliminary evidence in support of this Addiction, 109, 218–226
© 2013 Society for the Study of Addiction
5.22 4.51 0.78
4.68 0.76 0.85 0.17 2.61
−1.56
−0.40 0.60
−0.25 0.52
−0.26 −0.02 −2.79
2.06 0.69
2.22 0.64
0.29 0.08 −0.09
95% CI
1.83
B
0.29 0.10 0.95
0.078
