Chapter 12Summary and conclusions

Publication Details

This study has successfully tested the feasibility of all aspects of a behaviour change intervention study which used text messages to deliver a brief alcohol intervention to disadvantaged men in early mid-life. Because the target group has been little studied and the design and delivery of the intervention posed interesting challenges, the study provided the opportunity to make a series of important theoretical and methodological advances. The use of extensive process measures shed light on all aspects of study design and conduct including the recruitment strategy and the design, delivery and impact of the intervention. The results of the study have been discussed fully in the main chapters of the report.

This chapter presents the main findings from each chapter, assesses the extent to which the aims of the feasibility study have been met, and concludes with modifications to be made to the proposed full trial. It should be noted that the term feasibility study is used here. In fact this study contains elements of both a feasibility study and a pilot study. The research could be described as feasibility because it assesses whether or not the full trial could be done (e.g. availability of participants, willingness to be randomised, follow-up rates). However, it also fits with the definition of pilot study: a version of the main study run in miniature. The term feasibility has been chosen because it captures the exploratory and developmental spirit of the study.

Main findings

Focus group analysis

  1. Explanations for current drinking patterns and their change over time appear rooted in shifts in three interacting conceptual areas: private purpose (the individual's personal reasons for drinking), social roles (expectations, duties and sanctions on drinking) and concrete experience (of the adverse effects of excess drinking).
  2. The set of social expectations stem from recognition of the person's wider social roles and responsibilities (employee, husband/partner, parent) and abilities (self-discipline/control, ability to tolerate alcohol, judge limits and resist social pressure).
  3. Drinking motives had significantly changed from when the men were younger. This stemmed from a shift in the private purpose and social roles facilitated by alcohol use. It was accompanied by a shift from ‘freedom to’ enjoy the pleasures of alcohol to ‘freedom from’ life stresses through the use of alcohol.
  4. Consumption was influenced by obligations to a range of competing social roles including parent, partner and employee that had not been extant when younger. In addition, recognitions of these changes produced what might be termed a ‘mature drinker’ role with associated expectations.
  5. The mature drinker was viewed as someone who can tolerate greater amounts of alcohol without becoming drunk, has greater self-discipline, is more likely to be able to resist social pressures to drink, and drinks as a means of reward, relaxation and release from life pressures.
  6. Knowledge of the negative experiences of alcohol was detailed and widespread. However, reported changes in behaviour appeared rooted in concrete personal or vicarious experience of these issues rather than more abstract risk.
  7. Peers and social networks influenced drinking in this group in complex and subtle ways. Although some men felt that their friends were more accepting of their decisions to slow their drinking down, other men still felt pressure from their peers to consume large amounts of alcohol. In the second context, the consumption of alcohol signified membership of the group and gave the men a sense of belonging.
  8. Drinking refusal self-efficacy largely depended on whether the group was predominantly defined by its drinking behaviour or another organising factor. A brief intervention might therefore highlight that the men have a choice about the different social networks they are accessing, and that this choice will affect the amount of alcohol they consume and ultimately their health and well-being.
  9. There was often a marked discrepancy between an individual's drinking habits and the intended aims of drinking. Thus, brief interventions could highlight the mismatch between being drunk, and success in socialising, working, ‘doing good turns’ and feeling good about oneself.

Recruitment

  • 10. The intended target of 60 men from disadvantaged areas was easily exceeded using two recruitment strategies: through GP lists and by RDS. The use of evidence-based techniques may have assisted the recruitment.
  • 11. Obtaining ethical consent from individuals recruited through GP posed a challenge because there was no face-to-face contact. This was overcome by obtaining consent by a confirmatory text message.
  • 12. Recruitment by RDS required prior fieldwork to establish suitable venues for recruitment. The study found that participants could be recruited if approached in familiar settings. For the full trial, venues such as post offices, training centres for the unemployed, pharmacies and communal areas in high-rise buildings could be used.
  • 13. Flexible recruitment strategies involving multiple attempts at contact at different times of the day and days of the week are essential.
  • 14. Some potential respondents were concerned that the initial approaches might be part of a scam to steal money. This was more marked when first approach telephone calls came from an institutional 0845 telephone number. Use of a dedicated study mobile telephone number overcame this difficulty.

Baseline characteristics

  • 15. The men in this study were socially disadvantaged whereas those recruited to previous studies on alcohol brief interventions were mostly well educated and in employment.
  • 16. Baseline interviews were successfully conducted by telephone. Data on alcohol consumption over a 30-day period, including total consumption and frequency of binge drinking, heavy binge drinking and alcohol-free days were obtained.
  • 17. Binge drinking was common, with 57% of men consuming ≥ 8 units of alcohol in a session at least three times per week. In addition, 25% of men drank at least 20 units in a session.
  • 18. Most men were drinking at a hazardous level (84% had a positive test score on the FAST screening questionnaire).
  • 19. Randomisation was successful and the intervention and control groups were balanced at baseline.
  • 20. Participants recruited by RDS were less well educated and more likely to be unemployed. These men may be marginalised in society to such an extent that they have become a hidden group.
  • 21. Men recruited through RDS were much more likely to be drinking hazardously. These men also frequently suffered episodes of memory loss following drinking sessions. Recruiting solely through GP would miss high-risk drinkers.
  • 22. Knowledge of the conventional definition of binge drinking was low, with most men guessing the level was much higher. There is an opportunity to incorporate education on binge drinking into the intervention.
  • 23. RSOD was common, with episodes of heavy binge drinking interspersed with many alcohol-free days. This pattern of drinking has not been previously described in this age group. This type of drinking could explain why disadvantaged men are at such high risk of alcohol-related harms, even although the total alcohol consumed is not high.
  • 24. Heavy binge drinking was common despite a high level of awareness of the harms of binge drinking. This discrepancy, using the techniques of MI, provides a lever for behaviour change.

Intervention design

  • 25. The content of the intervention text messages was derived from seven types of literature: alcohol brief interventions, text message interventions, two psychological models (the TPB and the TTM), MI, communication theory and reviews of specific behaviour change techniques.
  • 26. The literature reviews identified several issues to be resolved including style and tone of the texts, the specific behaviour change strategies to be included, and the sequence and frequency of the messages.
  • 27. A combination of SMS texts and MMS messages with images were used to deliver the intervention. The images provided colour and variety and were used to complement the text messages.
  • 28. Despite the restriction of SMS texts to a maximum of 160 characters, the components of behaviour change strategies were successfully presented in the messages.
  • 29. The user group representatives provided insight into drinking motives and behaviour which identified opportunities for behaviour change. They also helped ensure that the interventions were acceptable and the outcomes were relevant and measurable.
  • 30. Messages were tailored to the target group and the individual by embedding them in the language and the drinking culture of disadvantaged young men. The messages used popular texting terms and abbreviations. The participant's first or preferred name was used, particularly in the early text messages.
  • 31. The text messages emphasised the credibility of the source (University of Dundee), used the informal style and abbreviations of everyday text messages and made frequent use of humour.
  • 32. A variety of techniques were employed to increase message effectiveness: tailoring of messages; use of gain-framed texts; pairing of messages; inclusion of questions to promote interactivity.
  • 33. Images were designed to develop and extend the narrative and to make concrete the harms of alcohol.
  • 34. The messages were constructed to take advantage of the conventional pattern of heavy weekend drinking. Messages were designed to tap into three windows of opportunity: before weekend drinking, after a heavy drinking episode and mid-week sobriety.
  • 35. The initial set of messages were tested in focus groups and improvements made. They were then mapped against reviews of behaviour change techniques to ensure that relevant techniques were not overlooked.
  • 36. The control group received a similar number of texts, but these were intended to have no influence on drinking behaviour. The control texts addressed four health behaviours: diet, physical activity, sexual health and mental health.

Intervention delivery

  • 37. Delivery of the messages was computerised, and followed the same schedule for all men. Messages began on the Monday morning following recruitment and continued for 28 days. This enabled messages to be tailored to day of the week.
  • 38. The computerised system could identify receipt of SMS text messages and could collect responses from the participants. This enabled messages to be terminated for individuals who wished to withdraw from the study. In the event no-one asked to leave.
  • 39. The use of a computerised system for the automated delivery of the text messages means that large numbers of men could be targeted at low cost.

Retention

  • 40. Through the use of multiple evidence-based strategies embedded throughout the study, a retention rate of 96% was achieved with a difficult-to-retain group.
  • 41. Over 90% of men thought that participation in the study was informative and worthwhile. The text messages were very popular and many men would have liked more text messages.
  • 42. Three features of the messages – anticipation of the content of the next message, the varied time of sending of the messages and the frequent use of humour – may have contributed to the popularity of the study.

Process evaluation of the delivery and impact of the intervention

  • 43. Novel methods of process evaluation were employed in this study. These shed light on the fidelity of delivery of the intervention and on the impact of the components of the intervention. The novel methods complement and strengthen conventional process measures.
  • 44. The fidelity of delivery of the text messages can be successfully measured through the computer system which managed the sending of the messages. Overall, 95% of SMS text messages were delivered to the participants' telephones.
  • 45. The computer system also captured responses to text messages. Responses to text messages were common, particularly for texts which asked questions. However, many participants responded to other texts.
  • 46. Eighty-eight per cent of men in the intervention group responded to one or more text questions. Over half of the men replied to at least seven of the nine texts which asked questions. Frequency of responding did not vary by age, social deprivation or level of hazardous drinking, suggesting this approach is suitable for different types of drinkers.
  • 47. There was no evidence that the frequency of responding decreased over the duration of the intervention, suggesting that, once enlisted, engagement is maintained.
  • 48. The questions gave the men the opportunity to think about the components of the behaviour change strategy. The questions thus form an important element of the behaviour change strategy.
  • 49. Analysis of the content of the responses showed that the men gave considered and personal responses to the questions, indicating a high level of engagement with the texts.
  • 50. Analysis of specific questions identified which of the components of the causal chain to behaviour change was achieving the intended response. The effective components included: subjective norms, control beliefs, perceived behavioural control and behavioural intentions.
  • 51. The process evaluation provides strong evidence that the proposed full trial will be a technical success.
  • 52. Men in the control group also showed a high readiness to respond to text questions. These questions concerned other health behaviours: sexual health, mental health, physical activity and diet. This indicates that the text message questions can be successfully used in studies targeting other health behaviours.
  • 53. Given the highly informative nature of the text question responses, they should be a feature of feasibility studies of interventions delivered electronically.
  • 54. This type of process measure is unobtrusive and gives real-time feedback on the components of the intervention.
  • 55. As text message questions have the potential to increase the effectiveness of the intervention, they should be used for each of the key components of the behaviour change strategy.
  • 56. Two text messages were found to be inadequate; one was ambiguous and the other (a loss-framed message) received a poor response. These messages will be changed for the full trial.
  • 57. Almost all (95%) of the men told someone else about the study and 40% discussed it with more than five people. This could increase the impact of the intervention on social norms (gaining support from family and friends). It could also facilitate national roll-out of the intervention, promoting uptake in the general population. Through this mechanism the intervention could act to promote culture change by encouraging discussion of the benefits of more moderate drinking.
  • 58. The extent to which study participants discussed the text messages with friends and family members has an unexpected consequence for RDS. As this technique aims to recruit several individuals from close social networks, contamination of the control group could occur. Abandoning the community outreach recruitment strategy would seem unwise, since recruiting only through GP would exclude many high-risk drinkers. The solution is to use an alternative strategy, time–space sampling. This technique recruits individuals from multiple venues, but avoids recruiting groups of individuals from the same social network.

Outcome assessment

  • 59. The primary and secondary outcomes proposed for the full trial were successfully measured on all subjects who were followed up.
  • 60. More men in the intervention group reduced their frequency of binge drinking, and heavy binge drinking, than in the control group. There was also a greater increase in the number of moderate drinking days in the intervention group.
  • 61. More men in the intervention group increased intention to reduce their alcohol consumption, as measured by the Readiness to Change Questionnaire and a question specifically about intentions. These changes were accompanied by a fall in the number of men who thought the benefits of their drinking outweighed its harms.
  • 62. The success of the intervention most likely reflects four key design features.
    1. The use of a delivery medium, the mobile telephone, which is an attractive, valued means of communication by the target group. It gave the men control over when they read the messages and gave them time to reflect on them and respond to them if they chose (most men did respond).
    2. The consistent use of a friendly supportive style in which behaviour change strategies were interspersed with novelty and humour. The messages were of sufficient popularity to be shared widely with friends and family.
    3. The use of questions to encourage extended engagement with the behaviour change strategies. These questions elicited carefully thought-out, highly personal responses.
    4. The phrasing of text messages in the language of the target group, based on the focus groups conducted prior to and during intervention development.
  • 63. There was a high awareness of both the harms of alcohol and the benefits of moderate drinking in both intervention and control groups. This indicates that these items should not, as was proposed in the protocol, be used as secondary outcome measures.
  • 64. The feasibility study identified that drinking refusal self-efficacy was not influenced by the intervention. For the full trial the intervention will be improved by giving increased emphasis on self-efficacy.

Post-trial assessment

  • 65. The main reasons for taking part in the study were curiosity about the study, interest in the topic of alcohol and feeling valued.
  • 66. In the recruitment process, more emphasis could be placed on the importance of each individual to the research study. The potential to help others could also be stressed.
  • 67. Some men expressed initial concerns about participation, particularly being classed as an alcoholic and uncertainty about what the study would involve.
  • 68. The majority of men thought that the financial incentives promoted retention within the study.
  • 69. By the end of the study almost all the men said they had found the experience sufficiently worthwhile that they would recommend participation to their friends.
  • 70. The friendly style and the varied content, particularly the use of humour, were highly valued by the participants.
  • 71. Text messages with images were thought to have had more impact than text messages alone. Making images more realistic could increase their relevance to the participants.
  • 72. Men in the intervention group thought that the study increased intention to change drinking, through concerns about personal health and their families.
  • 73. Some men in the control group reported being very surprised by the statement in the baseline questionnaire that consumption of four pints constituted binge drinking. They may also have been influenced by baseline questions on the harms of alcohol and the benefits of moderated drinking. These questions should be omitted for the full trial.
  • 74. The control messages may also have influenced drinking intentions. This could be because the control messages were comparable to the intervention package in the duration and number of texts, the use of questions to elicit responses, the provision of health information and the frequent use of humour. Thus, for the full trial the number of control texts should be reduced, as should the amount of health information. This would make the control group messages more comparable to those used in conventional alcohol brief interventions.

Implications for other research studies

  • 75. Questions embedded in text messages provide a simple method of process evaluation for trials of behaviour change interventions delivered by electronic media. The questions should tap into key components of the behaviour change strategy, to assess the extent of engagement with each component. Rather than being intrusive, this method of process evaluation may promote engagement with the intervention and thus increase its effectiveness.
  • 76. Follow-up rates are likely to be higher if the intervention is inherently enjoyable. Behaviour change interventions should address the enjoyability of their intervention. The challenge is to make sure the intervention is not boring or ‘preachy’.
  • 77. The intervention and delivery method could easily be rolled out for use with other social groups and other adverse health behaviours. The responses to the control group text messages showed a high willingness to engage in improving mental health, sexual health, physical activity and diet.

Does the feasibility study indicate that a full trial should be conducted?

The feasibility study sought to answer three questions to determine whether or not a full trial of the intervention should be conducted.

  1. What are the best ways to recruit and retain disadvantaged men in a study aimed at reducing the frequency of heavy drinking?
  2. What is the type of content and timing of the delivery of a series of text messages and images that is most likely to engage young to middle-aged men?
  3. Is the intervention likely to be an acceptable way to influence the frequency of heavy drinking?

Recruitment and retention

Recruitment was successful. It is recommended that GP recruitment together with a community outreach strategy be used in the full trial. RDS, by recruiting several individuals from close social networks, could lead to contamination of the control group. An alternative strategy, time–space sampling, could be used. Retention, at 96%, was very high and the existing strategies to promote retention should be maintained in the full trial.

Content and timing of messages

The messages were derived from an extensive review of several relevant fields of research. Components of behaviour change theories and techniques were successfully conveyed in short text messages and images. The content and timing of the messages was carefully planned to fit with the pattern of drinking of the target group. The high level of engagement with the messages indicates that the content was warmly received. The spontaneous ‘thank-you’ text messages at the end of the study support this conclusion. The post-trial evaluation showed that the frequency and timing of the messages was not only acceptable but also popular.

Influence on heavy drinking

The study was not powered to detect changes in drinking. However, the frequency of binge drinking reduced slightly more in the intervention group. In addition, there was a greater improvement in intentions to reduce drinking in the intervention group. These findings, together with the high level of engagement with the intervention, are compatible with the intervention being effective.

Summary

In summary, this study has successfully tested the feasibility of all aspects of a behaviour change intervention study which used text messages to deliver a brief alcohol intervention to disadvantaged men in early mid-life. The use of extensive process measures has shed light on many aspects of study design and conduct including the recruitment strategy and the design, delivery and impact of the intervention. This study design has demonstrated the potential to recruit, engage and modify drinking behaviour among disadvantaged young to middle-aged men. We recommend that a full trial of the intervention be carried out to assess the effectiveness of the intervention. We further recommend that a cost-effectiveness study be incorporated into the trial to determine whether or not the approach is truly a low-cost method of reaching large numbers of disadvantaged men.

Modifications to the full randomised controlled trial

This feasibility study has identified several ways in which the design of the full trial could be improved. These cover the recruitment strategy, intervention, process measures and outcome measures. These revisions will improve the conduct of the stages of the trial and increase the effectiveness of the intervention. This section only mentions issues which arose from the feasibility study it does not attempt to provide a protocol for the full trial.

Recruitment strategies

  1. The evidence-based recruitment techniques should be retained for the full trial.
  2. In the recruitment process, more emphasis could be placed on the importance of each individual to the research study. The potential to help others could also be stressed.
  3. Given the differences in the types of drinkers recruited by the GP and community outreach strategies, it is recommended that both GP and community outreach recruitment strategies be used. With each method recruiting 50% of the intended sample, coverage of all types of drinkers will be achieved.
  4. As RDS could give rise to contamination of the control group, an alternative outreach strategy, time–space sampling, should be employed. This recruits from areas which the target group is known to frequent and stresses the need for flexibility of setting, time of day and day of the week for recruitment. It requires initial fieldwork to identify a wide range of venues and suitable times for recruitment.

Randomisation

  • 5. As the randomisation method used in the feasibility study was successful and maintained blindness of the research team which collected and analysed baseline and outcome data, it should be used again.
  • 6. Randomisation should be stratified by recruitment method because of the difference in the types of men recruited by the two recruitment strategies. This will enable the analysis to determine whether or not the impact of the intervention differs between the two types of men.

Baseline questionnaire

  • 7. The question which gives information on the definition of binge drinking should be deleted as it could lead to contamination of the control group. This question will still be asked at follow-up.
  • 8. Fewer questions should be asked about the harms of alcohol, the benefits of moderated drinking and intentions about future drinking to ensure that the baseline questionnaire does not act as part of the intervention. This will be important for national roll-out, which would not involve the administration of a baseline questionnaire.

The intervention

  • 9. The use of the novel method of process evaluation identified a text message which was misunderstood and another which elicited a poor response. These will be modified.
  • 10. Information on the conventional level of binge drinking (≥ 8 units in a session) should be added to the intervention. The post-trial evaluation showed that the men were surprised to learn that as little as four pints constituted binge drinking.
  • 11. The analysis of the follow-up data revealed that there was insufficient improvement in self-efficacy to refuse a drink. Additional texts should be created to address this.
  • 12. Negative experience of alcohol should be presented in a more concrete manner to increase impact.
  • 13. More attention should be given to the discrepancy between drinking aims (relaxing, having fun and socialising) and the consequences of heavy drinking.
  • 14. The intervention should be extended to incorporate the HAPA, which seeks commitment to action and also promotes maintenance. Thus, new texts will be created to promote goal setting, to gain commitment to action, to overcome barriers to action and to promote maintenance of the reduced frequency of binge drinking.
  • 15. The planned modifications will increase the length of the intervention. Feedback from the participants indicated that additional text messages would be welcomed. Because of the affection that the intervention generated in many participants, it would seem sensible to taper the last few messages over several weeks.
  • 16. The control messages should be fewer in number and give less information about health. This will ensure that the control messages do not prompt a re-evaluation of drinking behaviour. This would make the control group messages more comparable to those used in conventional alcohol brief interventions.

Outcome assessment

  • 17. The primary outcome should remain the frequency of binge drinking. It is the most appropriate measure for men whose drinking pattern is episodes of very heavy drinking interspersed between periods of abstinence.
  • 18. The secondary outcome measures should be changed. Awareness of harms of drinking and benefits of moderated drinking should be omitted as levels of knowledge are high in the target group. Total consumption should be added as a secondary outcome. Goal setting and action planning could also be secondary outcome measures.

Process measures

  • 19. Additional process measure questions will need to be added to the intervention to check that the proposed new texts on goal setting and action have the desired impact.
  • 20. New process measures on self-efficacy should also be added.

The proposed revisions to the study protocol clearly demonstrate the value of conducting a feasibility study prior to a full trial. Although the broad structure of the study is unchanged, important changes will be made to all phases of the trial. These will improve internal and external validity and, crucially, increase the effectiveness of the intervention.