To the best of our knowledge, presented here is the first UK feasibility study with an embedded qualitative process evaluation exploring smoking cessation in the context of smokers accessing homeless centres, and the first known trial with adults experiencing homelessness on the use of ECs for cessation. This report presents data on participant recruitment, retention, engagement, service capacity, perceived value and preliminary efficacy and cost-effectiveness. The initial results are promising, and with careful consideration and future PPI and stakeholder (homeless charity) input, a future trial may be feasible with some changes to the structure and delivery of the intervention. We discuss the facilitators and barriers in more detail later in this section, including recommended changes needed to progress to a main trial.
In relation to our first objective, recruitment, of all those who were eligible, just over half of those invited agreed to take part and completed the baseline assessment. Only one person withdrew from the study. Our process evaluation has provided useful information that has helped us to understand how to increase participant trust and therefore willingness to take part in a future trial. One stand-out point relates to participant vulnerability and nervousness in taking part in research and sharing personal information. The longer we stayed at the centres – especially centres 1 and 2 – the more likely service users were to engage and to want to take part; this represents growing trust and a ‘snowball effect’, as other service users were seen to be taking part and speaking to the research team, although this also had the unintended consequence of ‘unblinding’ centre allocation. In terms of future work, this highlights the need for the research team to introduce the study to service users as well as staff before baseline assessment. Building trust and offering the opportunity to be seen and be known should, as per our qualitative interviews, reduce any concerns around confidentiality and misperceptions about the nature of the study.
In addition to this, the type of centres taking part in this trial and in a future full cRCT is important. Day centres, defined as those that offer services within usual working hours, were the more successful in terms of recruitment than residential units. Specifically, the two day centres (i.e. centres 1 and 2) accounted for 77.5% of the total sample. The research team and the participants themselves observed that, in the context of these services, a study taking place or a new service being offered was not particularly out of the ordinary. Both of these day centres already offer a range of services and support; thus, our study was within the realm of what was being offered. Furthermore, as noted by our participants when discussing the differences between centres, the day centres offered access to potential participants who, by virtue of their presence at the centres, were already engaged with the service. At the residential units, there was notably less activity and a significantly lower footfall of potential participants. Contrary to expectations, recruitment and follow-ups were not easier at residential units than at day centres. Residents who were ‘at home’ would often not respond to door-knocking or would reply that they did not want to be disturbed, whether or not they had scheduled appointments and received reminder messages. Offering the service in residential units at times felt more intrusive, and this needs to be considered for future studies. It should also be noted that there were fewer smokers in the residential units, and more people were at work. This point is important for considering targeting the most vulnerable smokers; it should not be assumed that, because these adults are facing homelessness, the levels of need will be the same across the population.
Centre 1 (EC) was the most successful in terms of recruitment and retention. There were very few limitations to this centre, but it was observed by the researcher that, although its scale was positive for recruitment, it was also often busy and there could be disturbances. Disturbances and commotion were common in both day centres. For example, at centre 2 (UC), it was common for events outside the staff/researchers’ control (e.g. arguments, drug poisoning, other urgent needs) to take over, making it difficult, or at times impossible, to recruit, collect data and elicit staff support for the study. Despite this, with flexible researchers and a good connection between the research team and the centre staff, these problems were overcome, and by extending the recruitment period to 6 weeks rather than 4 we were able to recruit 22 people. Follow-up was difficult at centre 2 not only for the aforementioned reasons but also because many of the participants were not able to come back for follow-up; many had moved on or were not visiting the centre during the follow-up period. The lower follow-up rate at centre 2 than at centre 1 may also reflect baseline differences between the groups and the fact that centre 2 was within a busy national capital city centre with a highly transient population.
The two residential units had the lowest recruitment rates, despite extending our recruitment period. Centre 3 (UC), despite being known as a ‘centre’, comprised three geographically separate sites and residents did not congregate in the communal spaces much when our researchers were present. Even during periods of good weather, very few people sat in the gardens; they often left the site or stayed in their rooms, and in some cases were active at night and slept during the day. A number of potential participants were out in the daytime, attending day centres, work, part-time education or health appointments, or volunteering. This is in comparison with the bustling environments and continual footfall of the day centres. In addition to this, with centre 3 being located at three different locations, the opportunities to ‘fill’ unattended appointments or serendipitously ‘catch’ one person while waiting for another were fewer than at centres 1 and 2, where the scale of daily interactions and activities was greater. Likewise, there were far fewer interactions between residents, so there was less effect from word of mouth, even related to the opportunity to receive an incentive gift card.
Similarly, centre 4 (EC) had four separate sites at three geographically separate locations. Generally, the residents we met who smoked at this centre had consistently complex needs (e.g. substance use and severe mental health comorbidities and side effects from prescription medications). There were few structured communal interaction events and, if these did exist, take-up by residents was low. Participation in activities such as coffee mornings was moderate, but such activities were more often attended by people who were already non-smokers. Newly hired staff already had full workloads with inductions, compulsory training and patterns of shadowing colleagues so were not able to prioritise supporting the research as much as had been anticipated at the planning stages. At the women-only accommodation (one site making up centre 4), there were no private meeting rooms for data collection or meetings, so all data collection was conducted in the multipurpose room. This was challenging at times because of the lack of privacy, as the garden and residents’ units were accessed through the multipurpose room. These are wider considerations that would need to be taken into account for a future trial by planning with the centres how the study needs to be run and what the conditions for a ‘good assessment’ look like.
However, despite some expected and unexpected challenges, the fact that we were able to recruit 80 participants across four centres in a 5-month period, despite the limited pool of participants at the residential centres, is encouraging. We are especially encouraged given that our participants presented with a high prevalence of mental illness and illicit substance use.
Retention was also assessed for a future trial. The overall 24-week total retention rate was 59%; this compares well with other studies that have recruited smokers from homeless services.4 Indeed, here we add to the growing evidence base that, despite many competing needs, the majority of our participants were willing to come and speak to the research team. As reflected in our qualitative process evaluation, the financial incentive was instrumental in securing a sufficient number of people in the trial, and, moving forward, the incentive system should remain in place. Financial incentives have been used widely across smoking cessation studies in this environmental context and have also been shown to be a direct measure of increased participation.4
The results show that 24-week retention rates were higher in the EC arm than in the UC arm. There are several explanations for this. Firstly, the EC arm received weekly sessions to provide e-liquid for the first 4 weeks. This differential level of contact may have encouraged retention in this arm. Nevertheless, this is unlikely to be a key factor given that 4-week retention rates did not differ dramatically between arms (81% vs. 66%). Secondly, baseline characteristics indicate that our UC arm presented with greater mental health and substance use comorbidities, in terms of both incidence and severity. This may be a consequence of our non-randomised design; however, the differences across arms appeared to be mainly due to higher levels of anxiety, and especially substance dependence and cannabis use at centre 2 in Edinburgh, which were significantly higher than at any of the other centres (which did not differ from each other). This is likely to affect attendance because of competing health and social needs, and indeed this centre was associated with far lower 24-week follow-ups than the other three. This sample (at centre 2) was also especially transient in nature, and it was difficult to keep in touch with the participants by 24 weeks (some of whom had been hospitalised or incarcerated). Nevertheless, it is encouraging that individuals with such high-level needs were willing to take part, and it provides proof of concept of the wide-reaching engagement of this study. Future research with this population would need to work closely with PPI representatives to explore how participants could be retained in research trials.
However, for those participants who did return, 6% achieved sustained CO-verified abstinence rate at 24 weeks (ITT) in the EC arm, compared with 0% in the UC arm. Cessation rates are low and our sample size is small but these figures are in line with those reported in other studies conducted in similar settings.16 However, to the best of our knowledge, our study is the first study to also monitor continuous abstinence at 24 weeks’ follow-up among this population; the general trend has been to capture 24-hour or 7-day point prevalence at 6 months.4 Given that we have measured continuous abstinence, our results compare favourably with those of two similar trials using NRT: 4% reported at 26 weeks by Segan et al.,36 and 5.6% and 9.3% 7-day point prevalence reported at 26 weeks in a larger sample of 430 smokers receiving, respectively, NRT or motivational interviewing and NRT at a homeless centre.16
Importantly, we were able to collect the data from our participants that would be needed for a full cRCT and include a full economic evaluation (although see later in this section regarding contact time with staff in the UC arm). Amendments can be made to the questionnaires that allow for less extensive intrusion of people’s background and lived experience but at the same time still allow the relevant data on health and demographics to be collected.
Completion rates of HRQoL and general health-care and social services questionnaires suggest that the use of these measures was feasible while indicating potential areas for the refinement of data collection. More detailed data collection will be included for services used frequently by this population, whereas services with little use can be omitted from the final data collection pro forma.
Self-reported SSS use showed that few participants in the UC arm sought SSSs for quitting and they did not ask for help from GP services after 4 weeks. In the EC arm, patients did not seek any SSSs during the study period and instead used GP services when they were no longer offered help by their keyworkers in the centre.
Overall, feedback from participants about the study was positive; as noted in Harms, there were very few unintended consequences, and as noted by the qualitative process evaluation the experience of being offered an EC and use after assistance was well received. Encouragingly, the majority of participants reported that they still had and were still using, though not exclusively, the EC at the 24-week follow-up. This is an important issue for breaking the stereotype that this may be a group of adults who will not be able to keep an EC in their possession. There were extra signs of encouragement in the form of participants purchasing their own products, which can be taken as a personal investment in vaping. Around one-third of participants reported having purchased their own EC; this may also explain why the number of participants reporting having used an EC was greater than the number of participants originally supplied with one.
The use of an EC was also commonly reported in the UC arm; around two-thirds of participants reported that they had used one, and one-third had purchased an EC at some point during the study period. However, we did not explore the extent or the frequency of this use. The use of an EC is not uncommon in smokers experiencing homelessness,14,37 and use and experimentation are common among the general population of smokers in the UK.38 As has been reported elsewhere, cluster designs cannot fully mitigate the occurrence of EC use in UC, and, in fact, EC use is beginning to be supported or actively encouraged in SSSs under ‘usual care’. Although there was some EC use in our UC arm, no participants experienced the EC intervention ‘as delivered’ in the EC arm (i.e. a free EC starter kit and e-liquid offered at homeless centres).
Staff capacity to support the study was generally good, and staff across three of our four centres reported few issues in the process evaluation interviews. The additional responsibilities placed on staff in the EC arm, including initial device set-up, administration of e-liquid and general participant support, were met well by the staff in our study. The high value that staff and management placed on the EC intervention probably helped to better engage staff in these processes. Staff training on ECs and researcher support were also important in building staff belief in ECs as a harm reduction tool and developing their confidence for EC intervention delivery. This helped staff overcome initial nervousness at supporting participants in using their EC. As the study progressed, staff concerns surrounding ECs were alleviated; for example, they reported receiving few problems regarding batteries or charging devices and they were able to manage participants’ e-liquid supply well during the intervention phase. Although some of our participants would have welcomed a greater variety of e-liquid, staff were content with the quantity of different e-liquids on offer, which needs to strike the right balance between participant choice and staff ability to administer.
Daily challenges for service delivery, particularly those experienced at the busier drop-in centres, highlight that service providers may need further support by providing efficient and effective processes for facilitating the study alongside their usual work. Although daily challenges were similar for the drop-in centres (centres 1 and 2), centre 1 was better able to support the study by having better resources to draw on, including staffing and space. Lack of adequate staffing and space presented some challenges for the study at centre 2. Although staff at centre 1 could cope with and support the high level of interest in the study from clients, staff at centre 2 were not able to facilitate participant signposting to SSSs or follow up with participants weekly to check their engagement with SSSs, and where they were able to we could not collect data on this. Busy workloads and client demands primarily accounted for reduced capacity for staff to support the study at centre 2, and alternative ways of capturing contact time to fully cost UC would need to be considered if moving to a full trial. However, it is also possible that a lower value placed on the UC intervention may also account for lower staff engagement with these aspects of these study. Service providers could be further supported by suggesting systems for managing referrals, appointments, signposting and follow-ups, taking into account the differences at each site. Training for staff in the UC arm could also encourage them to see the benefits of UC support.
The qualitative interviews found that perceived value of the study and EC intervention was high among participants. Given participants’ reported desire to change their smoking behaviour, irrespective of whether their primary goal was cessation or harm reduction, the chance to engage in smoking-related discussions was welcomed. The fact that this opportunity could be accessed at a place where participants resided or regularly attended and was incentivised through provision of the EC starter kit and/or Love2Shop gift cards was an added bonus for participants, facilitating their engagement despite low expectations for behaviour change. Those participants in the EC arm who took part in an interview were particularly enthusiastic about the EC intervention, with strong belief in the value of ECs as a harm reduction tool. Social dynamics, especially at centre 1, where some individuals with social standing became vaping advocates, contributed greatly to this belief. This contrasted with the lower perceived value of the UC approach, which largely stemmed from previous negative experiences of NRT, and probably accounts for the poorer engagement with local SSSs. The fact that many participants in the UC arm expressed interest in ECs as a result of the study may also provide an explanation for the level of contamination across study arms.
Limitations
We did not meet the original recruitment target we set ourselves of 120. However, this was a pragmatically chosen target based on our early scoping work and there was potential for further recruitment from some sites if our schedule had permitted it. A second limitation was that we were not able to randomise clusters to condition as planned, and this may have led to the differential retention rates, retention being lower in the UC arm. Our arms also differed on some important baseline characteristics, which may have contributed to the lower retention rates at UC sites. This was particularly the case for centre 2, in Edinburgh, which differed from the English sites on many of these baseline variables. A fully powered randomised cluster design stratified by region in a main trial should mitigate baseline differences across groups. Blinding was not possible for the assessment of outcome measures, which may have led to assessment bias or reduced effort in attempting to contact some participants for follow-up. Furthermore, staff working at the centres were not blinded to the intervention arms, and this could potentially have introduced some bias, for example staff members’ beliefs regarding who would be interested and most able to take part, not least their beliefs regarding who would attend and quit. Finally, we were unable to collect information in the UC arm on staff contact time with research participants.
The qualitative process evaluation has some additional limitations. Owing to the small subsample of trial participants and sample of centre staff, the qualitative findings may not be representative of all our trial participants or staff involved in supporting the study. As in the previous paragraph, it is possible that those who agreed to take part in an interview may have held stronger views about the EC intervention or the UC intervention; however, many of our interviewees in the UC arm had not engaged with UC support, which was reflective of the wider trial sample. Given this, the qualitative process was not able to fully detect any potential bias across intervention arms. We had difficulty recruiting 4-week non-completers to take part in an interview, primarily as a result of challenges in maintaining contact rather than a reluctance to take part. Although interviewing a greater number of non-completers may have identified further barriers to engagement, interview narratives of the small number of non-completers who did take part were not substantially different from those of the majority of the sample.
Generalisability
Our data derive from three geographically distinct areas of the UK and from a variety of different centres. The demographics of our data are similar to those of the wider UK population experiencing homelessness (see www.homeless.org.uk/facts/understanding-homelessness/impact-of-homelessness; accessed 16 March 2021), although the percentage in current employment was lower and the percentage experiencing long-term physical health problems and using drugs was higher than the percentage reported among the UK homeless population by www.homeless.org. Based on these facts, and although centre staff recruited those who were known and actively engaging with the services, it does not appear that our sample was any less disadvantaged than the UK homeless population as a whole. The sample is small relative to the numbers of people engaging with the homeless sector, but we see no reason why the results and methods here cannot be generalised to other homeless service settings as a potential area for future work. Furthermore, it is important to stress that our sample included those with both current and historical substance use and mental health comorbidities, meaning that our participants presented with the complex needs that one expects in these services. It is encouraging that we are able to reach participants from a wide range of backgrounds with differing levels of needs. Central to DHSC’s vision of reducing smoking prevalence rates is directly tackling smoking where tobacco-related health inequalities are most felt by driving smoking rates down among groups with a high incidence of social disadvantage. This work provides a baseline for that vision. This feasibility study provides proof of concept that, even among adults with complex and competing needs, there is a desire to take part in cessation research and attempts to quit are good; our participants used and held on to their EC. Encouragingly, overall, the intervention was well accepted and provides a baseline for future work. Specifically, findings suggest that a fully powered cRCT to explore efficacy and cost-effectiveness may be feasible, but, to fulfil this, recruitment across multiple sites must be considered against service provision type and plans for retention need to be explored to increase follow-up rates. A larger trial is important to establish a stronger evidence base for smoking among this population and to work towards the shared vision of reducing smoking prevalence among adults experiencing homelessness.
Research recommendations
Objective 1: recruitment – assess willingness of smokers to participate
The residential units had fewer eligible individuals, and potential participants were less available and less interested in taking part in research.
Recommendation
Any future study should ensure that a sufficient number of researchers is employed on the project to deal with recruitment and explore limiting sites to day centres only. This is especially important given the transient nature of the population; staff may need to be responsive to the availability of the participants.
Objective 2: assess retention and engagement in the electronic cigarette and usual-care arms
Retention rates were moderate to good and were comparable to those seen in other studies on deprived groups; however, one site’s poor retention rates negatively skewed retention rates.
Recommendation
In designing a future trial, we advise, based on our findings, that a larger number of clusters with fewer people at each one and official randomisation should help to reduce residual confounding factors between the groups. Randomisation should also ensure that variables predicting dropout are equally distributed between groups, thus reducing the difference in retention across arms.
Objective 3: examine the perceived value of the intervention, facilitators of and barriers to engagement, and influence of local context
Perceived value of the intervention was high. Barriers were participants’ personal and psychological difficulties, especially in relation to answering the questionnaires, a sense of nervousness around research, cannabis use and staff assumptions about smoking. Facilitators were participants’ desire to change, lack of existing cessation support, a free EC and social dynamics.
Recommendation
Future studies should make time to introduce the research team to the participants and outline the project to the service users (all those who access the service, who may or may not take part in the study). Researchers should also identify opportunities to increase trust and build relationships before recruitment. We strongly recommend reconsidering questionnaire burden and intrusiveness, including shortening some questionnaires. Future work should also seek to understand what person-centred language would be most appropriate and acceptable with users of homeless services – there is currently a dearth of literature on this.
Specifically for intervention studies that include staff training, from our own experiences, we suggest adding an element to the staff training on fostering a community around the study, highlighting the benefits that this brings to motivation. Qualitative work should be designed to fully capture service user and staff views about their experiences of taking part in research studies, including not only barriers but also facilitators; this would complement our own findings and develop this evidence base.
Objective 4: assess service providers’ capacity to support the study and the type of information and training required
Overall, in the EC arm, staff were well equipped to support the intervention, with few capacity issues, although there was some evidence that staff misremembered key points from the EC training and facts around their use. Some challenges were highlighted in one UC centre, but this is reflective of the unique challenges of this centre and not the UC intervention.
Recommendation
If a full trial was successfully awarded, planning randomisation and a larger number of clusters should reduce any similar issues. Furthermore, fewer participants in each cluster will reduce the burden on recruitment and follow-up at each centre. A future trial should also explore topping up staff training with smaller, intermittent sessions, especially at EC centres.
Objective 5: assess the potential efficacy of supplying free electronic cigarette starter kits
The offer of a free EC starter kit was well accepted by staff and participants. Using ITT analysis, we found sustained 24-week abstinence rates of 0% in the UC arm and 6.25% in the EC arm, which provides preliminary evidence of potential efficacy; however, the absolute numbers were small (3/48 in the EC arm and 0/32 in the UC arm). The UC sample size was also lower than expected; the 0% abstinence rates may therefore be an underestimate of true cessation rates with UC.
Recommendation
To definitively measure efficacy, a future trial with smoking cessation rates as the primary outcome should include more clusters with a fully powered sample size, perhaps using a slightly higher abstinence rate with UC (e.g. 0.5–1%). To fully understand the mechanisms through which change occurs, we recommend an embedded process evaluation that seeks to explore fidelity of implementation and treatment context.
Objective 6: explore the feasibility of collecting data on contacts with health-care services among this population as an input to an economic evaluation in a full cluster randomised controlled trial
Almost all participants who were present at follow-up visits provided data on health-care service utilisation and HRQoL measures; however, as reflected in our interview data, some participants found this data collection intrusive or the questionnaires overly complicated.
Recommendation
Alongside experts with lived experience, any future trial should consider the structure and wording of the health utilisation questionnaire.
Objective 7: estimate the cost of providing the intervention and usual care
Although it was feasible to collect information relating to the cost of delivery in the EC arm, collecting data from staff about contact time was not always possible in the UC arm.
Recommendation
Alternative ways of capturing contact time (or supporting staff in capturing contact time) should be considered (e.g. improving reporting mechanisms and collecting data at a quieter time, such as the end of the follow-up data collection period at each centre) to fully cost UC.
Conclusion
This study focused on a group of people who are not well considered in traditional smoking cessation treatments, and, despite some issues with recruitment and retention, there is promising evidence that a full trial may be feasible with iterations to the main trial design and with further discussion with project partners. We observed early signs of acceptability and potential efficacy of offering free EC starter kits. Centre staff were a crucial element in helping people to take part in the study, training helped to support staff in regard to how to be involved in the study, and building trust between the researchers and the participants was also key. However, the study was challenging in some sites, and parts of the delivery of the study need to be optimised so that we can plan a trial that would be likely to progress fully through a clear stop–go criterion. In considering this, and planning to mitigate future issues, we will explore the findings from the process evaluation alongside the quantitative data, so as to help inform the design of a main trial.