A pragmatic randomised trial of two counselling models at the Swedish national alcohol helpline

A pragmatic randomised controlled trial was initiated in 2015 at the SAH. The Ethical Review Board of Stockholm, Sweden approved the study (DNR 2014/1732–31/5), and the corresponding protocol was registered in the ISRCTN registry (ID: ISRCTN13160878). The analytical approach to hypothesis testing has been changed [26] from that reported in the trial register from non-inferiority to superiority due to a lower recruitment rate then what was expected at inception.

Participants were clients from the general population seeking help at the SAH for at least hazardous alcohol use. Counsellors assessed the eligibility of clients at their first contact, before informing about the study. Eligible clients who expressed interest in participating were contacted by telephone within a week by trained interviewers not involved in the counselling at the helpline. The interviewers performed the following sequence of tasks: 1) obtained formal consent, 2) conducted the baseline interview, and 3) opened a sequentially numbered, sealed envelope containing the results of the randomization algorithm and communicated the group allocation to the participant. Clients were allocated with 1:1 ratio to the two groups. The research coordinator prepared the envelopes containing the computer-generated allocation sequence. Detailed information on the randomisation and recruitment process can be found in the study protocol [26]. The enrolment period was from May 27, 2015, to December 15, 2017, while the 6 months follow-up was completed between December 5, 2015 and June 20, 2018.

First-time callers or callers with a washout period of at least 1 year since last SAH contact, that were adults (≥18 years) who spoke Swedish and sought help for at least hazardous alcohol use were eligible to participate in the trial. Hazardous alcohol use was identified by the Alcohol Use Disorders Identification Test (AUDIT), using a cut-off of 6 and 8 points or more for women and men respectively [27]. Clients were excluded if the counsellor made the assessment that a caller required referral to treatment for severe alcohol problems or if the client reported the concurrent use of illicit drugs, or the suffering from severe psychiatric conditions or other acute health problems that required medical attention. Further, counsellors refrained from informing clients about the study if the overall assessment at the first call indicated that clients were not able at that time to understand the conditions for participation in the study.

Data was collected by structured interviews at baseline and at 6 months follow-up. At baseline, the interview covered questions of demographics (sex, age, education, employment, and living arrangements), and social support. Additional information included co-morbidity, indicators of general health, and help-seeking for alcohol-related problems (past 6 months). To screen for major depressive episodes (MDE, past 2 weeks) and generalized anxiety disorder (GAD, past 6 months), the measured mental-health problems, two sections of the M.I.N.I. (Mini-International Neuropsychiatric Interview M.I.N.I.) were used [28, 29]. Indicators of general health status were: sick-leave during the past 6 months, and self-assessed health, the latter measured by “How would you rate your overall health status?” in which responses were collapsed into three alternatives: ‘Very good to Good’, ‘Fair’, ‘Bad to Very bad’) [30]. Motivation to change was measured at baseline as it could be a predictor of the outcome [31, 32]. This was measured by the Readiness to change ruler ranging from 0 to 10, where 10 defines very high degree of readiness [31]. At baseline, AUDIT was primarily completed at the first call, since this scale is usually administered at the SAH as an essential tool for the diagnosis of the severity of alcohol problems. In some cases when AUDIT was not completed at the first call, it was administered in the subsequent baseline interview. Data retrieved from the client record at the helpline included AUDIT score at baseline and number and length of calls. A call was considered a counselling session if it lasted at least 5 min. AUDIT was re-administrated at the 6-month follow-up interview. Interviewers who performed the follow-up interviews were blinded as to the participant’s experimental group.

Problematic alcohol use was measured by the Alcohol Use Disorders Identification Test (AUDIT) [27]. AUDIT is a validated instrument, sensitive to problematic alcohol use in the lower end of the spectrum [33, 34]. Risk levels were based on cut-offs described in the Swedish AUDIT manual: ‘low-risk use’ (score 0–5 women; 0–7 men), ‘hazardous use’ (score 6–13 women; 8–15 men), ‘harmful use’ (score 14–17 women; 16–19 men), and ‘probable dependence’ (score ≥ 18 women; ≥20 men) [27]. Further, the AUDIT-C was used to assess frequency and quantity of drinking. This measure includes the first three questions of the AUDIT instrument and has a maximum score of 12.

The primary outcome was defined as any downward shift in AUDIT risk level at 6 month follow-up, i.e. a ‘downward change in risk level compared to baseline’. An upward shift or no change in risk level was defined as ‘no downward change in risk level’. Secondary outcome measures were: 1) change to low risk level according to the AUDIT cut-offs, 2) mean change in the total AUDIT score, and 3) mean change in AUDIT-C at 6-month follow-up, this latter representing a measure of current alcohol use.

The analysis was carried on as a modified intention-to-treat analyses (ITT), i.e. participants with outcome information at 6-month follow-up were analysed according to their randomly allocated counselling model, irrespective of the counselling received. To assess the impact of attrition a sensitivity analysis was performed assuming that AUDIT risk level among those lost to follow-up did not change or was worsened as compared to their baseline score; and by the last observation carried forward (LOCF) scenario, i.e. assuming that participants lost to follow-up did not change their baseline AUDIT score.

The retention in this study was reasonably high compared to previous studies in the field [16, 45], with no differential loss to follow-up between the two counselling groups. The similar retention between experimental groups, besides reassuring about selection bias also suggests that both counselling models were acceptable to many of the SAH clients. The study protocol was implemented as planned, and the delivery of the brief, structured intervention was ascertained by regular meetings between the employees at SAH and the study team. Further, the personnel conducting the interviews were not part of the SAH and blinded to the client’s group assignment. The random allocation was concealed.

Limitations of this study include the low recruitment rate, as only 46% of first-time callers were assessed as eligible. This was due to a “service-based” screening, which left great freedom of decision to each counsellor at the SAH. In fact, counsellors assessed the eligibility of each caller not only based on the pre-defined criteria, but also taking into account the complexity of the problem presented by the individual at the moment of the call. For instance, clients who were too emotionally disturbed at the time of the call, or who reported serious social consequences of their drinking calling for immediate actions, such as involvement of minors were judged unreceptive to questions about the study, therefore discarded as not eligible. Of the eligible and only 39% participated. Besides resulting in a small sample size, this selection certainly limits the generalizability of the results. However, age and alcohol use profile of the participants in this study were comparable to those of the overall SAH clients [26].

Some contamination of the experimental groups may have occurred since the counsellors interacting with the clients were the same in both groups. However, the average number and length of the sessions were substantially higher in the usual care group, confirming that the brief, structured intervention was less labour intensive, as expected. Additionally, the documentation of the content in the proactive call did not indicate major deviations from the study protocol. As in most trials of behavioural modification the endpoints in this study rely on self-reported data, thus are potentially prone to information bias. However, the follow-up assessment was conducted by interviewers not involved in the counselling, which should at least limit the risk of socially desirable reports.

The main outcome measure AUDIT has traditionally been scored as a three-factor screening instrument, however a review of studies support a two factor model; patterns of intake/consumption (item 1–3) and alcohol problems (item 4–9) [46]. Therefore a change in the AUDIT score may not necessarily depend on changes in alcohol consumption, which is instead captured by the AUDIT-C subscale. This was included as an additional sensitivity analysis, yielding results in line with those obtained from the complete AUDIT.

As this was a trial enrolling callers who sought help for alcohol use at a well-established service, a no-treatment control condition was considered unethical. Therefore, the conclusions of this study apply to the model of the counselling, and not to the effectiveness of the helpline in comparison with other counselling settings or a no-treatment control group.