Results

Baseline data

Baseline data are presented in Table 3.

The mean age of the sample was 42.66 years, and 65% of the participants were male. Participants were primarily white (76.3%) and heterosexual (85%).

Mental illness, health and substance use comorbidities

Seventy-four per cent of the sample self-reported a long-standing illness, 38.8% had been previously admitted to hospital owing to mental illness, 55% scored 10 or over on the GAD-7 indicating the presence of generalised anxiety disorder, and 60% of participants scored 10 or over on the PHQ-9 indicating the presence of major depression.

In relation to alcohol use, 33.8% of participants had an AUDIT score over 8, suggesting that they were drinking at harmful or hazardous levels. SDS scores were also high, indicating a high prevalence of substance dependence.

Half the participants reported having previously spent time in prison. A significant proportion (38.8%) of the participants, women and men, reported being a victim of domestic violence.

The arms also differed significantly on a number of baseline variables, namely GAD-7, SDS, cannabis joints per day, time spent in prison, physical illness and motivation to quit, with the UC arm scoring higher in all cases but having lower motivation to quit (see Table 3).

Education, employment and housing status

Overall, 37.6% were educated to A level/Highers (or equivalent) or above. Employment status varied: 2.5% reported being in current paid employment and 97.5% reported recourse to benefits (social security/benefits). Just over half of the sample (60%) were currently housed in supported accommodation or in a hostel, and 8.8% reported sleeping rough.

Cigarette dependence and smoking behaviour

The mean number of cigarettes smoked per day (including roll-ups) was 20 [standard deviation (SD) 15.33 cigarettes]; this is equal to a pack of cigarettes per day in the UK. The mean expired CO breath level was 20.29 parts per million (p.p.m.) (SD 10.04 p.p.m). The FTCD score was 5.51 (SD 2.47). In relation to smoking practices that increase the risk of respiratory infection, 55% reported that they shared cigarettes (24% reported doing this daily), and 43% reported that they had smoked discarded cigarettes (6% reported doing this daily).

Furthermore, MTSS varied considerably; although only 6% reported that they did not want to stop smoking, the large majority expressed a desire to quit smoking in the near future.

Differences between the intervention arms

The proportion of participants who had previously spent time in prison or who had a long-standing illness, disability or infirmity was significantly higher in the UC arm than in the EC arm. UC arm participants also scored higher on anxiety and substance dependence, smoked more cannabis joints per day and were less motivated to quit smoking. Given these differences and the differences between centres (see Description of settings), we conducted sensitivity analyses (one-way ANOVA with post hoc tests for continuous variables and Fisher’s exact tests for categorical variables; not per protocol) to explore differences between the four centres regarding variables in which there was a significant difference between arms. Cannabis smoking and substance dependence scores were significantly higher at the day centre in Edinburgh (centre 2; UC) than at all other centres (all p-values for cannabis smoking were < 0.01 and all p-values for SDS were < 0.05), which did not differ from each other. Similarly, anxiety (GAD-7) scores were also highest at centre 2, and these differed significantly from centre 1 (day centre, Northampton, EC) and centre 4 (residential centre, London; EC) (p < 0.05) but not from centre 3 (residential centre, London, UC). In addition, MTSS was significantly higher at centre 1 than at all other centres (p < 0.01), which did not differ from each other. A higher proportion of participants from centre 2 had spent time in prison (73% compared with 36%, 50% and 56% in centres 1, 3 and 4, respectively). The difference between centre 1 and centre 2 was statistically significant (p < 0.01). A higher proportion of participants from centres 2 (91%) and 3 (90%) (both UC) reported a long-standing illness than participants from centres 1 (67%) and 4 (44%). Significant differences were observed between centres 2 and 4 and 3 and 4 (p < 0.05).

Numbers analysed

Appendix 1 presents the CONSORT flow diagram showing the numbers for each intervention arm.

Objective 1: assess willingness of smokers to participate in the feasibility study to estimate recruitment rates and inform a future trial

Outcome: record the number of smokers asked to take part and the number of smokers who consent

Across the four centres, 177 participants were initially invited to take part. Of these, 24 were not eligible (16 in the EC arm and eight in the UC arm). Of the remaining 153 eligible individuals (90 in the EC arm and 63 in the UC arm), 80 consented to take part in the study: 48 (56%) in the EC arm and 32 (50.5%) in the UC arm. Although we did not reach the recruitment goal of 120 that we set ourselves (based on our preliminary scoping work), we did recruit these 80 participants during the originally specified 5-month period. Recruitment also differed between centres; the two day centres were most successful in terms of recruitment (39 consented at centre 1 and 22 at centre 2), together accounting for 77.5% of the total sample, and there was a waiting list of participants who could not be recruited at centre 1 as the researcher had to move on to the next centre. The residential units had fewer eligible individuals, including fewer smokers, and potential participants were less available because of work or other appointments and less interested in the study. Furthermore, there was less opportunity to recruit in the residential centres because of centre space and room availability, and issues around disturbing residents (see Discussion for more information).

Summary of learning points for a future trial

Ensure sufficient researchers are employed on the project to deal with recruitment and limit sites to day centres only.

Objective 2: assess participant retention in the intervention and control arms

Outcome: record (1) how many participants complete assessment measures in each arm at each time point, and (2) how many participants are still using electronic cigarettes in the intervention arm

The CONSORT flow diagram presents the participant numbers at recruitment, and follow-up data by intervention (Appendix 1). In the EC cluster, retention rates were 81%, 69% and 73% at 4, 12 and 24 weeks, respectively. In the UC cluster, retention rates were 66%, 53% and 38% at 4, 12 and 24 weeks, respectively.

There was a lower rate of attendance in the UC intervention at the 12- and 24-week follow-ups. The most common reason for not following up was that participants were no longer attending the services; although exact reasons were not formally documented by the research team, these reasons were highlighted informally by conversations with centre staff.

This difference in retention between arms did not appear to be due to the UC condition per se but, rather, due to baseline differences that might influence the ability to attend follow-ups. The UC arm was associated with a higher incidence of criminal background, illness, disability, substance abuse, anxiety and lower motivation to quit (see Table 3). This was particularly the case for centre 2 (day centre in Edinburgh; see Differences between the intervention arms), which was also associated with a far lower 24-week retention rate (26%) than centre 3 (residential centre in London; 67%).

We asked all participants in the EC intervention if they were still in possession of the EC that we had provided (although they did not have to present it at the appointment), and whether or not they were still using the EC. Assuming that all those who did not attend follow-up sessions did not still have, or were not still using, the EC, at 24 weeks 46% (22/48) still had the EC that we had provided and 56% (27/48) were still using it. Table 4 presents the data based on the number of participants who attended and answered the question at each follow-up. The greatest fall in use was between the 12-week and the 24-week follow-ups; however, 63% of those asked self-reported that they were still in possession of the device. Similarly, in relation to still using an EC, the greatest fall was between the 12-week and the 24-week follow-ups; however, of those in attendance at 24 weeks, 79% reported that they were still using an EC, either the device that was provided at baseline or a different one.

Summary of learning points for a future trial

Larger numbers, more clusters and official randomisation should reduce baseline differences between arms. Randomisation with stratification by region should also ensure that variables that predict dropout are equally distributed between arms, thus reducing the difference in retention across arms. In the event of a future trial, it would be important to consider ways in which retention could be maximised, including other ways of maintaining more regular contact with participants, for example increasing text and verbal contact between appointments.29

Objective 3: examine the perceived value of the intervention, facilitators of and barriers to engagement, and influence of local context

Outcome: qualitative interviews with 4-week completers and non-completers, quitters and smokers (N = 24, approximately six per site) between weeks 4 and 8 across both arms

This section presents the findings of 22 in-depth qualitative interviews with a subsample of study participants. Given the difficulty in contacting most of those participants who had not completed their 4-week follow-up, and with the majority of study participants smoking at week 4, it was not possible to conduct interviews with a good range of completers and non-completers, quitters and smokers as intended. However, there is a good spread of qualitative participants across the four homeless centres and both study arms (see Appendix 3 for the qualitative sample characteristics). Nineteen interviewees (86.4%) had completed their 4-week follow-up and three (13.6%) had not completed this follow-up. At the 4-week follow-up, 18 of the 22 interview participants (81.8%) were smoking, one was abstinent and three did not provide smoking information.

We explored participants’ experiences of the study in line with research objective 3. The findings present six topic summary themes: barriers to and facilitators of the study and intervention engagement; experiences of the EC intervention; experiences of UC; perceptions of study processes; unintended consequences; and perceived value. Local differences, if they exist, are highlighted in the text. The section ends with a summary of key learning points for a future trial.

Barriers to study and intervention engagement

Participants’ personal barriers

Interview narratives highlighted the psychological and emotional vulnerability of many of our participants. Some explained that their mental state, attention difficulties, anxiety around social interaction and difficulties with appointments made engaging in treatment programmes challenging:

It is partly my mental health . . . I do panic about talking to people . . . It’s very hard sometimes, just to sit and explain yourself.

Several displayed mistrust in authorities and research:

I thought ‘oh this definitely a government initiative. They’re going to run a test on the homeless . . . maybe they’ve got a dodgy batch of [e-liquid] and they just want to see if it takes anyone out before they put them up for sale.

Although some reported being comfortable providing personal data as part of the study, this was a major concern for others, with particular anxiety around whether or not data would be shared:

. . . if I thought my information was being shared, then I wouldn’t take part.

Participants also lacked confidence in their ability to stop smoking, and many reported that their ideal outcome would be to cut down first, rather than stop straight away: ‘if you can cut down, then you can go that further field’ (Participant 09 C02, UC). This was due to their reported reliance on smoking to deal with stress, isolation or boredom. Traditional stop smoking approaches, particularly NRT, which was reported to have interacted unpleasantly with substance use or mental health in previous quit attempts, were unappealing to participants. This presented a significant barrier to engagement with local SSSs for those in the UC arm.

Cannabis smoking

For participants regularly smoking cannabis mixed with tobacco, most reported wanting to reduce or stop their cannabis consumption, although this presented concerns and was highlighted as especially difficult given the associated pleasure: ‘I feel relaxed [smoking cannabis when] I’ve been wound up all day’ (Participant 11, C02, UC). Some noted how cannabis use had drawn them back to smoking in previous quit attempts:

I’ve quit smoking a couple of times and the thing that always brought me back was smoking weed with the tobacco.

Two participants, previously dependent on heroin, described reliance on cannabis as a protective factor in abstaining from other substances:

I’m an ex-addict, I used to inject heroin etcetera, now I stopped all that and started smoking cannabis and that stopped me taking any other drug.

Participants experiencing mental health distress explained that cannabis helped them to cope and were anxious that stopping cannabis use would exacerbate their symptoms:

I suffer from a lot of anxiety and depression, so it’s just heightened it. It made it worse, trying to stop.

Facilitators of the study and intervention engagement

Opportunity: right time and place

All our interviewees reported a desire to stop or reduce their smoking, expressing concerns about finances or health. Although most had tried to quit previously, negative experiences with NRT and mainstream routes of support [e.g. general practitioner (GP) appointments] meant that these approaches were unattractive. This made opting in to an incentivised, on-site smoking treatment programme appealing. Many illustrated their receptivity to engaging with support at a place they already attended. Bringing smoking cessation aids and services to participants in their own environments was therefore a key facilitator of engagement:

I just thought if you were coming here and Tuesday is my day off, I’d definitely benefit by taking part . . . So why not?

So I’m glad that this has come along because I don’t think, if this hadn’t have happened, I would have bothered to give up. I had no plans of giving up, until that day, none.

Free electronic cigarette starter kit and gift cards

The incentive of receiving a free EC starter kit and/or Love2Shop follow-up gift cards played an important role in study engagement. Although the gift cards received for attending follow-up appointments were valued by all participants, they were especially important for facilitating study recruitment for the UC arm:

My main incentive – I can’t tell lies – is the £15 voucher aye. Why not? If you can come to the centre and make £15 pounds, well then that is fine. Happy days.

In the EC arm, the offer of a starter kit capitalised on participants’ desire to quit and perceived lack of existing cessation options. Despite some participants’ lack of initial enthusiasm for vaping, participants said it prevailed as worth trying, given the availability and offer of the device and e-liquid, especially at a time of financial hardship:

[I] didn’t have to go out and buy it you see, you know, ‘here it is, just try it’.

. . . the thought of maybe it might work and there’s nothing to be lost, there’s only something to be gained, you don’t have to pay for the device . . .

Providing free starter kits also overcame start-up cost barriers for those who had expressed previous interest in vaping:

I have wanted to do it before but thought it was going to be really costly to start up. I haven’t got a spare 30 quid to lay down, ever.

Social context

Social dynamics facilitated EC intervention engagement. Some participants described becoming strong study and vaping advocates. EC intervention buy-in, particularly from individuals with social status among peers, helped raise awareness and interest. Word-of-mouth communication quickly relayed vaping benefits to potential participants:

Everyone I see who smokes I tell them you’ve got to try it, and I really try and talk them into it. But I’m worried I’m becoming one of those [laughs] people that gives up smoking and just like is a pest. But I want to pester them because it’s – I know it’s good.

. . . now four or five key people have done it and no one, there was no ‘oh are you vaping?’

Positive peer influence was especially prevalent at the day drop-in centre (centre 1), with high demand for the intervention and the creation of a vaping community, characterised by peer advice on equipment and technique, and social vaping norms:

I’ll go out, still socially . . . six or seven of us are vaping away . . . there’s no stigma attached to it anymore. You can just vape happily.

. . . they encourage you to stop smoking, you encourage them. So it’s a bit of like support in a way. Moral support to try and cut down.

A vaping community was not as evident in the residential centre (centre 4). Generally, in these residential units, the service users we met tended not to mix as much and reported more solitary activity, often staying in their rooms:

I don’t [notice others vaping] . . . I don’t pay any attention to them. I just come in and go out.

Participants’ experiences of the electronic cigarette intervention

Participants in the EC arm provided detailed descriptions of their initial expectations, the information and instructions they received, the EC and e-liquids, and their experiences with vaping.

Initial expectations

Most participants had not tried vaping before, reporting little knowledge of ECs. Many described their initial scepticism and low expectation that ECs would help them to reduce or stop smoking. It was common for participants to think the intervention would not be suited to them. For others, engagement in the study meant going against their negative view on vaping:

I was anti-vape.

I did not think for a minute it would fit for me, and work with me at all. I had no hope . . . hugely low expectations . . . I didn’t give myself a hope in hell.

Electronic cigarette information and instructions

Most participants reported satisfaction with the information provided. Instructions were considered thorough and participants particularly valued the practical demonstration, which included advice on device set-up, maintenance and use. Some said the practical help leaflets were useful to refer back to:

They basically showed me how to use the vape and how to change the liquid . . . how to charge it . . . change my coil . . . what to do when I need to clean the vape out. So it was just a really big help because I didn’t really understand anything about vaping and vape pens before.

Electronic cigarette device

Perceptions of the device were mixed. Some found it easy to use and liked its compact size, light weight and long-lasting battery. Although there were several reports of batteries overheating, this issue was a suspected manufacturer fault. Others reported problems with durability, reporting that the device broke easily when dropped (a common occurrence), resulting in cracks and leak of e-liquid:

I mean I’ve dropped mine several times . . . I’ve sort of repaired it . . . it’s still leaking juice.

With no money readily available to replace devices or replacement parts, some had used tape to repair their devices. One highlighted the importance of accessible and affordable replacement parts:

If you are going to use a vape on a study with vulnerable and homeless people, I would suggest you pick a vape that the parts are more accessible. Because, most of these things you have to go online to get them. And most of these people don’t have a bank account, let alone be able to go online.

E-liquid

Finding a preferred e-liquid was important for continued use and participants had differing preferences. Perceptions of the e-liquid provided were generally positive. Many interviewees said that they were satisfied with the variety of flavours and amount and nicotine content of the e-liquid. Experimentation was important for participants to find what suited them best. The flavour options were not liked by all, however, and several suggested the study should provide a greater variety. Some described mixing flavours, exchanging flavours with others and purchasing different flavours:

I think [the variety] covered all bases . . . I was telling everyone three quarters blueberry or half blueberry and half polar bear and seems to have gone down rather well.

Some also commented that the available strengths were too high, leading to initial adverse effects such as coughing and feeling sick, which could put them off continued use:

I shouldn’t have started on 18 [mg] straight away . . . I think I went full out too strong . . . I didn’t give myself a chance.

. . . the 12 [mg]s and 18 [mg]s are very high. A lot of people get head rushes and things . . . I can hit it about twice before nearly passing out. I have to mix mine down.

Vaping experience

Many interviewees noted that their vaping frequency increased over time, with the reported realisation that vaping was helping with smoking reduction and was more useful than NRT. Many were pleasantly surprised with these changes and intended to continue:

It surprised me . . . I know 65% [smoking reduction] isn’t massive but I’ve been smoking 30 years. It’s a dramatic amount in the short space of time . . .

Well at first . . . I thought maybe it’s just going to be a couple of days thing and then just put it on the side but no I use it every single day . . . it’s helped.

For some, vaping easily fitted in with existing routines and habits; it was easy to vape and more convenient than rolling a cigarette, especially when outdoors. One noted that vaping helped them to relax. Others found it more difficult to replace certain cigarettes with vaping, for example the first one in the morning or after dinner, and some described quickly choosing to smoke if their EC was unavailable or they were experiencing stress:

I probably smoke about 35% cigarettes and 65% vape . . . It changed for a little bit where I was doing about 50/50, but I was having a well stressful time . . . if I have a stressful day like I know I’m going to have today, I will have about five [cigarettes].

Participants’ usual-care experiences

There was little evidence of engagement with UC among our interviewees. Although some said that they appreciated the information provided in the help-quit leaflet, many acknowledged that they had not looked at it since the baseline appointment: ‘I’m sorry to say that I didn’t actually read through it’ (Participant 12, C02, UC).

Only three of our interviewees made contact with local cessation support. When accessing support, participants emphasised that making contact with, or contact being facilitated by, someone they trust was important. At centre 2, participants with a methadone prescription reported good relationships with the local pharmacist. Although one participant obtained a nicotine inhalator this way, another did not obtain support as he was reluctant to speak with anyone else:

At centre 3, two of our interviewees attended the local SSS accompanied by their keyworker. Without this keyworker support, one participant noted: ‘I wouldn’t have gone on my own’ (Participant 15, C03, UC). Participants who had tried the UC approach were ambivalent about their experiences. The participant who used an inhalator complained of a sore head and nausea, had not found the nicotine delivery satisfying, and had not returned for further advice or support. At centre 3, participants received a combination of patches and lozenges; however, they were critical of the advice provided and lack of availability of appropriate products:

. . . they didn’t have the appropriate strength to my taste and they gave me a weaker one. But I’ve tried it maybe 4 or 5 days . . . I didn’t quit smoking completely, but I reduced it like notably.

One UC interviewee reported greater success by combining NRT with an EC. It was common for UC participants to be interested in ECs or to have purchased and used one during the study. Cross-contamination was a feature of the study, as one participant explained:

[Vaping] is helping me – buying my own one . . . [Researcher] told me that the other places were doing that. I thought ‘well, I’ll try my own one’.

Perceptions of study processes

Understanding of the study

Most of our interviewees reported awareness of which study arm they were allocated to prior to the baseline meeting. Some also noted a lack of detailed initial information about the study and its aims. This ties in with staff reports of how they communicated the study (see objective 4). There were two notable exceptions whereby EC arm participants likened the baseline assessment to a test, as they perceived that the offer of an EC would be ‘based on the answers’ (Participant 03, C01, EC).

Pragmatic cluster design

Participants’ accounts gave some indication of the appropriateness of the pragmatic cluster study design. In centre 1, some participants perceived resentment from those not able to be recruited because the sample target had been reached as they observed others receiving a free EC. This, combined with interviewees saying they would have felt ‘scorned’ (Participant 02, C01, EC) or ‘like I’d just wasted my time’ (Participant 21, C04, EC) if they had not been allocated an EC after completing the baseline questionnaires, suggests disharmony should both study arms have operated in one centre.

Case report form

Some of our interviewees reported issues with the length of the baseline CRF, describing how sitting through a long appointment was difficult: ‘it’s not as easy for me to just sit, I fidget a lot’ (Participant 03, C01, EC). Others observed that questions alluding to substance use and mental health appeared to be misplaced in a smoking study: ‘why do you want to know about drug history or drinking?’ (Participant 18, C04, EC). Some noted emotional difficulty in responding to more sensitive questions:

It was difficult . . . it just brings back memories of things that – I tried to commit suicide three times . . . it did stress me a bit, but [researcher] was a nice man and he got me through it and we sat and sorted it out.

Follow-up appointments

Although some interviewees reported no difficulty in attending weekly or the 4-week follow-up appointment, this was problematic for some, who attributed missed appointments to issues such as other commitments, difficulty with remembering, hospital admission and being distracted by other life events, such as those requiring police involvement. Having no credit on their mobile phone to rearrange, having their phone stolen and sleep patterns were reasons given for non-completion by our small number of interviewees who had not completed their 4-week follow-up.

Unintended consequences

Use of electronic cigarettes to vape other substances

There was minimal evidence from our interview participants that the EC provided through the study was used to consume other substances. Only one interviewee said that they had vaped cannabis e-liquid, which they had made, though they reported using this in a device previously purchased, not obtained through the study.

It was, however, common for interviewees across both arms to report anecdotes of other people, not connected to the study, vaping cannabis e-liquid [‘. . . yeah, people smoke weed in it . . .’ (Participant 01, C01, EC)], cannabidiol (CBD) oil or tetrahydrocannabinol (THC) [‘. . . of course I’ve heard about it because I’m in [that] lifestyle . . .’ (Participant 09, C02, UC)]. In addition, one participant at a UC centre, where no EC had been provided as part of the study, reported to have heard of somebody who had ‘cooked up heroin and put it in’ (Participant 11, C02, UC). In response to interviewers’ enquiries, some participants with a history of drug use speculated about the possibility of using EC to vape other substances:

[I] suppose you could put heroin in an e-cigarette and it would do something. Not so with crack or coke. Don’t think there are many drugs you can put in there. Liquid ones, yes.

There’s a definite similarity to smoking crack in its method and the sucking in . . . Yes the thought did pop into my head . . .

Love2Shop gift cards

None of our participant interviews provided evidence that the gift cards received for attending follow-up appointments had been traded for other substances or cash (although there were a small number of informal reports of this occurring from trial participants who did not take part in a qualitative interview). Rather, the qualitative data highlighted that, in most cases, participants had used their gift card to buy essentials, most commonly food: ‘I really need[ed] some food in. I had nothing. I was probably eating about twice a week at that point’ (Participant 05, C01, EC). Some described the gift card as an opportunity to save up or buy something for a family member: ‘I can get my child school shoes now’ (Participant 10, C02, UC). One interviewee said that they had purchased alcohol with the gift card.

Perceived value of the study and intervention

All participants in the EC arm placed great value on the intervention and the opportunity to try vaping. Although most had not made a complete switch from smoking to vaping, they described a strong belief in the EC as a quitting aid, and valued their increased vaping and tobacco harm reduction knowledge. Many described feeling healthier and better about themselves since starting vaping:

I look in the mirror sometimes and see that I’m progressing, I feel a bit healthier. I’m not always having a cigarette.

I feel very proud of myself for the first time in ages . . . and I’m glad that it’s had a healthy impact on my life . . .

It does give you confidence . . . it’s made me think, ‘I can do something’. Cos my confidence has been knocked a lot. It’s good.

Some believed that the EC intervention could be wide-reaching in its ability to help disadvantaged smokers, saying that it could ‘help a lot of people’ (Participant 04, C01, EC) and ‘potentially save lives’ (Participant 02, C01, EC):

You’re giving people a chance because you’re giving them the equipment so, you know, they’re halfway there and then it’s up to you . . . that’s a good thing. People, like myself, that would have never thought of buying one.

Among the UC arm, some participants said that they valued how the study made them consider their smoking and appreciated the information on where to access support: ‘. . . even just participating, just even the thought that there is help out there to stop smoking is good enough’ (Participant 08, C02, UC). Most, however, believed that the EC would have been a more useful and effective approach:

It could have been useful to have been in the half of the study that had electronic cigarettes given to them.

Overall, however, UC participants said that they were pleased to have taken part. Across both arms, participants valued how the study had enabled them to have their ‘voice heard’ (Participant 17, C03, UC) in relation to smoking. Some said that they liked the thought that the study findings may help other homeless smokers in the future:

I’ve enjoyed it. It makes me feel important to be part of something that is important.

Summary of learning points for a future trial

Participants’ anxieties presented a significant barrier to study engagement. A sensitive and flexible research approach, characterised by participant/researcher trust, can help to engage participants.

Although cannabis smoking is a barrier to smoking cessation more generally, a future trial would need to consider how best to manage and monitor cannabis use so that study objectives are not undermined.

Bringing smoking cessation aids and services to participants in their own environments was a key facilitator of engagement. Gift cards also incentivised participants to engage.

Despite low expectations, participants were willing to engage with the EC intervention. Clear, practical instructions, a durable device with easily accessible parts, and the option of a lower e-liquid nicotine strength may help facilitate use.

UC support was less attractive for participants. Previous negative experiences with NRT and anxieties about engaging with others were barriers to engagement.

Participants’ accounts suggest that a cluster design is appropriate in homeless settings. Participants may have benefited from clearer information on the study at the recruitment stage and a shorter CRF. A future study should carefully consider the questions included in the CRF to minimise the burden on participants and reduce potential areas of sensitivity.

There was minimal evidence of unintended consequences, namely use of the EC provided to vape other substances and trading of gift cards; however, a future study should continue to monitor this closely. Nevertheless, device breakages were reported (see Harms).

Perceived value of the study and intervention was high among all participants. Participants’ desire to change their smoking behaviour and willingness to engage in the study presents a significant opportunity to engage homeless smokers in smoking discussions and treatment programmes.

Local context had implications for the study on two issues. First, social dynamics at the busy drop-in centre facilitated intervention engagement and vaping social norms. Future EC interventions might therefore benefit from capitalising on peer influence to create a supportive vaping community. Second, centre links and relationships, that is with keyworkers and the local pharmacy, were a key feature of whether or not participants in the UC arm engaged with support. Studies should consider the impact of these relationships on UC uptake.

Objective 4: assess service providers’ capacity to support the study and the type of information and training required

Outcome measure: qualitative interviews with keyworkers and front-line staff (n = 12, approximately three per site) across both arms

This section presents the findings of 12 in-depth qualitative interviews with keyworker and support staff involved in facilitating and supporting the study. This included assisting with recruitment and appointments, setting EC arm participants up with the device, for example showing them how to use the device and helping them choose an e-liquid, providing EC arm participants with weekly EC support and further e-liquid, and providing UC arm participants with brief advice on quitting and information about and signposting to the local SSS. The sample is spread across the four centres (see Appendix 3). We explored their experiences of the study in line with research objective 4. The findings present five key issues: staff assumptions; staff training; communicating the study; staff capacity to support the study; and perceived value of the study. The section ends with a summary of key learning points for a future trial.

Staff assumptions about smokers experiencing homelessness

Most staff, across centres and study arms, perceived that stopping smoking was not a concern or priority for clients given other issues, including mental health, substance use and chaotic lifestyles: ‘[smoking is] the less of all the evils’ (Staff 01, C01, EC). Other support needs were prioritised, as staff believed that smoking provided clients with comfort and enjoyment. These assumptions underpinned the lack of discussion about smoking between staff and clients: ‘it’s not something I would really discuss with them’ (Staff 07, C02, UC). Staff had few links to the local SSSs, little knowledge of smoking cessation and few established processes to support clients:

They have bigger things on their plates to deal with. I’m not going to try take away the one thing that they might enjoy. I’m already working with bigger challenges than that.

Prior assumptions meant that some staff made judgements about which clients, if any, would engage in the study: ‘I thought no one would be interested’ (Staff 09, C03, UC). Other staff expected EC arm participants to sell their devices. These assumptions, however, were not borne out by their study experiences:

I’m amazed that [devices] have not been sold as far as I know . . . It was not at all how I expected it to go. It was nice as well because it taught me something – so you don’t know everything about these guys. There are still things you can learn.

Staff training

Prior to study commencement, the research team delivered staff training on smoking prevalence and harm, ECs, local SSSs and supporting the study. The training was well received by those who participated in the session and was described as a key facilitator of their engagement with, and belief in, the study; staff felt ‘better equipped’ (Staff 10, C03, EC) to have more confident discussions with clients on smoking, local SSSs and vaping. For many, the training facilitated a change in attitude regarding the importance of stopping smoking for their clients, as well as greater motivation to discuss smoking. Some also noted a new understanding of tobacco harm reduction, making a comparison with drug harm reduction:

Doing that training . . . it made sense because we work with a lot of people with COPD [chronic obstructive pulmonary disease] and actually smoking wasn’t a priority at all. So the whole study actually has made me rethink . . . it’s the same discussion, part of harm reduction as we have about safer injecting techniques.

Our staff sample described having limited prior EC experience and initial ambivalence towards ECs. Post training, all reported increased knowledge of and interest in ECs as a harm reduction and smoking cessation aid. Staff in the EC arms appreciated the practical, ‘hands-on’ EC training, researcher support, and clear, simple handouts. Some said that this training assisted them in intervention delivery and also helped alleviate initial nervousness at providing the starter kit and instructions:

The handout sheets, I think those were the best things. It actually showed you – if you forget, this is what [to do].

The timing of the training was important. Some staff expressed concerns about remembering information if there was a gap between training delivery and study commencement. Some suggested that additional advice on local vape shops and vaping policies would be useful. Aware of some participants’ anxiety around providing data, some suggested that a better overview of the CRF be given, to give staff greater understanding of the types of questions and measures used in the study. Some staff noted that more training around facilitating recruitment, approaching participants and what information should be given to potential participants would be useful in communicating the study (see the following section):

Going back to the information stage, the recruitment stage. Just having a bit more information available for the staff to be able to use as a bit of a prompt. Or as a consistent communication style for all of them. I think that would be useful.

Communicating the study

How staff communicated the study to participants was one of two deviations from the feasibility study protocol. As per the protocol, all participants were to receive the same study information at recruitment and complete baseline assessments before being told about their allocation to the EC arm or UC. For practical reasons, staff knew whether their centre was allocated to the EC arm or the UC arm. When facilitating recruitment, our sample of staff in the EC arm reported routinely introducing the study as an EC or ‘vape study’, in which participants would receive a free EC if they signed up:

I just basically told them, ‘hey listen, I know you are smoking, are you interested in vaping, they are doing a smoking study, they are probably giving you free vapes if you are wanting to do that’.

I did hear some conversations where it was really quite low level, where it was a vape [study] and you are going to get the vape.

Conversely, our sample of staff in the UC arm said that they focused their description of the study on smoking, quitting and local support, and did not suggest that participants would receive an EC:

I’m not sure actually I introduced it really well . . . Just first asking them if they were a smoker . . . if they ever think of stopping and if so, that they could meet with [researcher] . . . you could help them and give them advice on how to stop . . . I don’t think I even mentioned the e-cigarette. I completely forgot about the e-cigarette, because we don’t get it.

Staff capacity to support the study

Electronic cigarette condition

Few staff capacity issues to support the intervention were described by EC arm staff. There was some initial trial and error with setting up efficient processes, but once resolved facilitating the intervention was described as straightforward. High demand for the intervention at centre 1 led the manager to take control of managing referrals and baseline appointments, by organising an appointment diary consisting of a main and ‘backup’ list: ‘[Manager] was able to manage it really well from there . . . just one central point’ (Staff 03, C01, EC). This same model was less successful at the smaller, residential units in centre 4, where there were fewer alternative candidates and less motivation to be disrupted from their ordinary activity.

Staff across centres highlighted that weekly follow-ups, during which staff provided participants with e-liquid and vaping support, could be challenging as some participants lost their appointment cards, did not show up at agreed times or could not be contacted: ‘it did happen quite a bit – people missing appointments, turning up late, turning up early’ (Staff 03, C01, EC). In centre 1, implementing a daily vaping ‘surgery’ overcame some of these difficulties; however, staff said that it could still be challenging if several participants came at once or did not attend on their set day. Some noted the importance of flexibility if participants approached keyworkers at different times: ‘if someone came in the afternoon and I’m free I would issue them the liquids then’ (Staff 01, C01, EC). One staff member also noted the importance of maintaining consistency for the follow-up sessions:

So what we did was set up a smoking surgery between 10 and 11 every day. So people could come along, talk to other people that were vaping, and also get help if they needed it . . . also get their liquids and coils if it was their day to get them.

Mainly, staff reported responding to participant queries relating to coils: ‘it was just about changing the coil . . . that was the main thing’ (Staff 11, C04, EC). Some noted requests to help with broken devices and suggested that staff having access to spare parts would be useful. Some also noted that they had been asked to open e-liquid bottles because some participants had poor dexterity. Most staff reported receiving few problems with batteries or charging devices. For managing participants’ e-liquid supply, some noted that the variety on offer through the study was a good amount, any more may have overwhelmed participants and staff:

If you give people too much choice, it just becomes a nightmare. So having that restricted number I think was absolutely fine.

Usual care condition

There were some challenges for staff supporting the UC arm around recruitment and weekly follow-ups. Although some described facilitating recruitment as straightforward, others found this challenging. At centre 2, the day drop-in centre, staff reported that, once the small number of regular clients had agreed to take part, they had difficulty engaging those who were more transient. Although the Love2Shop follow-up gift cards had been especially important for facilitating UC recruitment, some expressed discomfort in their reliance on the gift cards as an incentive:

It was almost like we were encouraging people to do it because they would get a voucher out of it. But it just felt a bit uncomfortable if I’m honest . . .

In addition, some reported negative feedback from participants about the length of the baseline CRF and the types of questions, and expressed concern about recruiting some clients to the study: ‘knowing that actually . . . it was going to be a challenge for them to answer all those questions’ (Staff 06, C02, UC).

Staff signposting UC services and maintaining weekly follow-up contact was the second main deviation from the study protocol. At centre 3, keyworker staff reported that this process worked well. Two staff members had escorted their participating clients to local SSSs and had attempted to quit alongside their clients:

I escort clients to the stop smoking van, where they can get nicotine patches, all sorts of paraphernalia to help them stop smoking . . . [I’m] still going with them to the van. Every Tuesday . . . we all go for a coffee afterwards as well. So we’ve turned it into a real social event actually.

In contrast, at centre 2 our interviewees reported some capacity issues to complete these tasks because of lack of staff, busy workloads and client demands. Some described dealing with challenges requiring immediate attention, rather than resources to focus on stronger client relationships as a factor. Finding private space for researchers to conduct baseline or follow-up appointments could also be a challenge, with rooms often purposed for other support services:

I have found it, personally, really difficult to do the follow-up . . . You are always on the go and basically I forget . . . It seems feasible . . . and then the reality is it hasn’t been . . . maybe because here we don’t keywork as such.

Staff’s perceived value of the study

Similar to our participants, staff also placed high value on the study. Across arms, staff appreciated the opportunity to become involved in supporting clients’ quit attempts: ‘it’s really important that we are involved in stuff like this’ (Staff 06, C02, UC). Many highlighted that the study had resulted in positive shifts in practice, increasing smoking-related discussions:

I can’t think of any other time we had a conversation about smoking with any other client.

It’s an excellent thing, just being part of this study that is potentially going to inform practice and policies and how we deal with smoking amongst the homeless, it’s just immense. The study is going to change the way we do things.

Although staff at the UC centres were positive about the study and pleased to have taken part [‘I feel privileged that we were included in this’ (Staff 03, C03, UC)], some expressed less enthusiasm for UC cessation support and disappointment that they had been unable to provide clients with ECs: ‘I think getting the e-cigs . . . I think that would be very helpful for people’ (Staff 05, C02, UC).

Staff in the EC arm were particularly enthusiastic about the intervention. Many noted positive impacts on their clients’ well-being. They valued being able to offer clients a tangible, practical tool, rather than just offering them stop smoking information. Many reported that the intervention had facilitated client trust and interactions and boosted staff morale:

It has brought them to the centre every day, which is good. Building a bit of trust, which is always a winner.

I think this is a lot more immediate [than other cessation support]. With the attention deficit some of our clients have got, to be given something tangible – to be given something, not just information, something practical they can use.

Summary of learning points for a future trial

Staff training was key to changing staff misperceptions about smokers experiencing homelessness and engaging staff in the study. Recommendations to further improve staff training include consideration of the timing of training in relation to study commencement; advice on local vape shops and policies; a more comprehensive overview of the CRF measures and questions; and greater training on facilitating recruitment.

For a cluster study design, consideration should be given to how best to communicate the study to participants, and whether or not it is practically feasible for participants across arms to be provided with the same information at recruitment.

Staff capacity to support the intervention was good at the EC centres and in one UC centre, suggesting that service providers are well placed to offer stop smoking interventions.

Local differences made some study processes more challenging. In one UC centre, lack of adequate staffing numbers, a busy environment, and staff having to support people with acute challenges requiring immediate attention, outside their allocated one-to-one keyworker hours, highlighted capacity issues in signposting to UC and following up participants. Service providers could be further supported by suggesting systems for managing referrals, appointments, signposting and follow-ups, taking into account differences at each site.

Objective 5: assess the potential efficacy of supplying free electronic cigarette starter kits

Outcome measure: measure breath carbon monoxide levels, self-reported quit rates/cigarette consumption at each follow-up time point

Table 5 presents the cessation and smoking reduction outcomes. The CO-validated sustained abstinence rate at 24 weeks was 3 out of 35 (8.57%) for the EC arm and 0 out of 12 (0%) for the UC arm. Assuming that all those with missing follow-up data were smoking, the 24-week sustained abstinence rate was 6.25% (3/48) in the EC arm and 0% (0/32) in the UC arm.

Seven-day point prevalence rates at 24 weeks were the same as sustained abstinence rates.

Of those who returned for 24-week follow-up (not ITT), the percentage of participants who reported > 50% reduction in CPD from baseline was 43% in the EC arm and 25% in the UC arm. The percentage of those participants with > 50% reduction in expired CO was a little lower and similar between arms (EC = 20%; UC = 25%).

Intracluster correlation coefficient

The ICC was calculated by the Fleiss–Cuzick method31 and was estimated to be 0.0157. The ICC is the ‘proportion of the total variance in the outcome attributable to variance between clusters’.31 Although the estimated ICC is relatively small, neglecting to factor this into the sample size calculation for a full cRCT would result in reduced statistical power.

Summary of learning points for a future trial

Based on the feasibility average recruitment of 16 participants per month, it would take 24 months to recruit 384 participants. A main trial would therefore need more (at least double the number of) researchers recruiting and collecting data simultaneously.

The estimation of the ICC used the Fleiss–Cuzick method and was not adjusted for individual-level baseline measures. In the full trial, the analysis is likely to be carried out by mixed-effects models, in which adjustment for prognostic covariates (e.g. baseline nicotine dependence, number of years smoking) is possible. This adjustment for covariates is likely to reduce cluster-level correlation in outcomes; therefore, an estimated ICC of 0.01 is used for the sample size calculation for the full trial.

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

Outcome measure: record participant scoring on EQ-5D-3L and utilisation of primary and secondary health-care services using a self-report service-use questionnaire at each time point

The missing information on EQ-5D-3L was mostly due to absence at follow-up visits. Most of those who attended completed the measure. At baseline, 46 out of 48 participants in the EC arm and 31 out of 32 participants in the UC arm completed EQ-5D-3L. At 4 weeks, all participants in both arms who were followed up completed it. At 12 weeks, 33 out of 34 attending participants in the EC arm and all 18 attending participants in the UC arm completed it. At 24 weeks, 34 out of 35 attending participants in the EC arm and all 12 in the UC arm completed EQ-5D-3L.

Among the observed cases, no one in either arm reported severe problems with mobility at any time point. The domain with the highest percentage of severe problems was anxiety/depression at baseline, 4 weeks and 12 weeks in both arms, followed by pain/discomfort. At 24 weeks, the percentage was similar in the two domains in each arm. The proportion of participants who reported no problem was consistently higher in the EC arm than in the UC arm across all domains and all time points. This led to a consistently higher mean utility score in the EC arm among those who completed the measure (Table 6). Among those participants who completed EQ-5D-3L at all time points, the mean QALY gain was 0.195 (SD 0.097) in the UC arm (n = 11) and 0.315 (SD 0.120) in the EC arm (n = 26), over 24 weeks.

There were three sections concerning services use: SSS use at follow-ups, smoking cessation help from primary/community health-care professionals, and general health-care services. Overall, most of those participants who attended the follow-up visits completed these sections, with only individual participants having an item missing in a few questions.

Attendance at SSS sessions was very low in the UC arm. At 4 weeks, four participants out of 21 contacted a SSS. One attended two sessions, three attended one and one did not answer. At 12 weeks, two participants out of 18 contacted a SSS; one attended one session and the other attended two. At 24 weeks, 2 out of 12 participants contacted a SSS; one attended one session and the other attended two. Using the resource impact template for smoking cessation service,34 the unit cost per session was estimated at £12.82. In the UC arm, the mean cost of SSS sessions was £3.21 (SD £9.18) at 4 weeks (n = 20), £2.14 (SD £6.60) at 12 weeks (n = 18) and £3.21 (SD £7.97) at 24 weeks (n = 12). In the EC arm, there was no SSS session cost incurred as none of the participants reported attending any session.

At 4 weeks, 3 out of 21 participants in the UC arm used other medications supplied by a SSS, including inhaler, patches and lozenges, with a mean cost of £5.68 (SD £14.30). At 12 weeks, 2 out of 18 used patches and lozenges in the UC arm, with a mean cost of £5.43 (SD £16.01). At 24 weeks, 1 out of 12 in the UC arm used gums and patches, with a mean cost of £3.38 (SD £11.71). None of the attending participants in the EC arm reported using cessation aids supplied by a SSS. The majority of them reported using their EC device only. Among those in the EC arm who claimed that they were not using any aids supplied by a SSS or the research team, two reported using another EC device.

No one reported using varenicline or bupropion. No one purchased any NRT product. The purchase of an EC device or e-liquid was reported by a small group of participants in both arms (Table 7). Although the number of participants who reported purchasing an EC device or e-liquid remained low in the UC arm at all times, spending was higher than that in the EC arm. In the EC arm, an increasing number of participants reported purchasing an EC device/e-liquid, after the initial 4 weeks of the study.

There were very few participants receiving smoking cessation help from primary/community care professionals (GP, practice nurse, pharmacist, NHS helpline), and those who did did not access them frequently. Owing to the rare and low frequency of use, the costs of smoking cessation help from primary health-care professionals were low and, in some cases, null. Especially for the UC arm, from 12 weeks onwards, no one reported any use. The mean costs in the UC arm were £1.45 (SD £6.75) among 31 participants at baseline, and £5.81 (SD £11.83) among 21 participants at 4 weeks. In the EC arm, the mean costs were £3.66 (SD £14.41) among 47 participants at baseline, null cost among 39 participants at 4 weeks, £1.56 (SD £9.09) among 34 participants at 12 weeks, and £2.11 (SD £12.51) among 35 participants at 24 weeks. Overall, the majority of the costs were contributed to by GP advice on smoking cessation.

Very few participants reported getting NRT products, mostly on prescription or from a SSS. The participants who had got NRT products all came from the UC arm, with only one exception: two participants in the EC arm at 24 weeks also received NRT. In the UC arm, only one or two participants at each follow-up point reported getting NRT on prescription or from a SSS. At 12 weeks, one of those reported four packs of patches and four packs of lozenges in total, but five packs of each were on prescription or from a SSS, causing a discrepancy in quantity. The only participant who received NRT that was not on prescription or from a SSS was in the EC arm at 24 weeks. Overall, the costs of NRT on prescription or from a SSS were £1.72 (SD £9.59) at baseline (n = 31), £4.13 (SD £18.73) at 4 weeks (n = 21), £16.30 (SD £51.81) at 12 weeks (n = 18) and £11.46 (SD £39.69) at 24 weeks, in the UC arm. In the EC arm, no costs of NRT on prescription or from a SSS were incurred until 24 weeks (n = 35), when it was £0.89 (SD £5.26).

Among the general health-care and social services asked, some services were used by none or close to none. Among all participants who answered in this section, no one received any visit from a GP at any point in the study. In the UC arm, only one participant received one visit from a practice nurse. In the EC arm, at both 4 weeks and 12 weeks, one participant received one visit from a practice nurse, whereas at 24 weeks one participant received two. These were probably due to the homeless situation of the participants. No participant accessed a maternity service during the study period. Sex health clinic was not visited at all at baseline (i.e. the 4 weeks prior to the baseline appointment) or at 12 weeks. At 4 weeks, one participant in the UC arm reported having visited a sex health clinic twice, whereas two participants in the EC arm had done so, one once and one four times. At 24 weeks, one participant in each arm reported having visited a sex health clinic once. Except for one participant in the EC arm at 12 weeks and one in the UC arm at 24 weeks, no participants had visited an early intervention team. Both participants visited the team only once. No participant in the UC arm stayed in a detox or rehab unit. One participant in the EC arm reported both at baseline and at 4 weeks having stayed in a detox or rehab unit, and another participant reported having done so twice at the 12-week follow-up.

There was a low but constant use of emergency and hospital services in both arms. Although the majority of the responding participants did not report any use, a small group of participants visited accident and emergency (A&E) frequently or had hospital admissions (see Appendix 6). In the UC arm, 4–7 participants reported having visited A&E at different time points. The number of visits per participant ranged from one to nine. In the EC arm, 5–7 participants reported having visited A&E at different time points. The number of visits per participant ranged from one to six. Owing to the design of the questions, we were unable to tell the number of visits that led to admission. However, in the UC arm, at least five participants at baseline, two at 4 weeks, one at 12 weeks and one at 24 weeks were admitted following the A&E visit. In the EC arm, at least two participants at baseline, three at 4 weeks and one at 12 weeks were admitted following the A&E visit. The highest number of hospital outpatient appointments reported at baseline, for over the 4 weeks before baseline, was 10 in the EC arm. The longest hospital stay was 14 days, reported in the UC arm at baseline and 4 weeks, and in the EC arm at 24 weeks.

Most of the participants in both arms received prescriptions. Other more frequently used services included GP, practice nurse and adult mental health team. Overall, 10 participants in the UC arm and 14 in the EC arm accessed a drug and alcohol service at baseline. This number dropped to almost none afterwards in the UC arm, whereas it remained more or less stable in the EC arm. Similarly, a housing team was visited by nine participants in the UC arm and 12 in the EC arm at baseline. The number of participants in the UC arm who used this service then dropped to one at 4, 12 and 24 weeks, whereas in the EC arm service use ranged from 7 to 10 participants at different time points (see Appendix 7).

Owing to the ambiguous phrasing of the question regarding detox/rehab unit, the meaning of ‘visit’ was unclear, namely whether or not it indicates a period of stay. Consequently, two possible unit costs to quantify the number of use were identified: by day or by week. We could not estimate the unit cost of a simple visit to the facility. Despite this, the overall mean costs of the health-care service did not differ much because the use of a detox/rehab unit was very rare. We therefore reported the higher one of the two (by week) to present a conservative number.

The mean costs among observed cases were consistently higher in the UC arm than in the EC arm (Table 8). The large difference between arms at baseline and 4 weeks was mostly due to the much longer hospital stay in the UC arm. At 12 weeks, although the mean cost of hospital stay was still higher in the UC arm, the mean cost of drug/alcohol service and early intervention team in the EC arm offset part of the difference. At 24 weeks, the mean cost of hospital stay and drug/alcohol service in the EC arm was much higher, whereas the mean cost of A&E, early intervention team and adult mental health team in the UC arm was much higher. However, this comparison has to be taken with caution, as around 30–60% of the participants in the UC arm were lost to follow-up during the study.

Recommendation for future revision of the questionnaire

The structure and wording of the health-utilisation questionnaire would need to be considered carefully if a full trial were to go ahead. Some of the questions about health-care utilisation were confusing for participants, repetitive or not applicable to everyone in the sample. Others (such as A&E admissions) were useful and endorsed by all participants, suggesting that a more granular follow-up question was needed.

Objective 7: estimate the cost of providing the intervention and usual care

Outcome measure: record all resources used in the delivery including staff costs, electronic cigarettes and other costs incurred. Staff will complete a pro forma to record contact time, non-contact time and other resources used in delivery

The training for the centre staff to deliver standard support in the UC arm and EC guidance in the EC arm was provided by the research team. For the convenience of costing, we divided the training into three components: core component, auxiliary component and extra component.

The core component of the training was the group training programme, which was delivered by the centre: two for the UC arm and two for the EC arm. For the UC arm, it was a 1.15-hour programme delivered by four trainers to one centre (seven staff) and three to the other (five staff). The help-to-quit leaflet used was printed in-house at a cost of £0.20 per leaflet. For the EC arm, the training was slightly longer, taking 3 hours (including individual coaching). Each member of staff who attended the training was provided with an EC device instruction pamphlet and an EC helpful tips pamphlet, which were both printed in-house at a cost of £0.20 per pamphlet. The training was similarly delivered by four trainers to one centre (12 staff) and three to the other (eight staff). The demonstration models of the EC were shown to the staff at the ratio of one unit per two trainees during the training, and six bottles of e-liquid (three flavours with two strengths each) were also shown. The EC starter kit was costed at £20 per kit and the e-liquid was £3 per bottle. One of the research team members who delivered the training was grade 9, three were grade 7 and three were grade 6. Their hourly costs, including superannuation and national insurance, were calculated based on the financial records of the project. The staff who received the training held various roles, resulting in varied pay grades. We used the costs information collected from centre 1 as an estimate for all those involved. The time cost of staff was £13.38 per hour, including superannuation and national insurance. The cost of staff time in delivering the EC training was £1040.93 in total and the cost in delivering the UC training was £246.65 in total. Overall, 20 staff attended EC training and 12 attended UC training. Their staff time costs were therefore £802.80 in the EC arm and £184.64 in the UC arm. The demonstration EC kit and e-liquid were reused for each training. The maximum number of staff for one training was 12. Therefore, the number of demonstration models was estimated to be six. Combining this with six bottles of e-liquid, the total costs of the EC and e-liquid were £138.00. The two pamphlets were printed for all 20 staff attending the EC training, costing £8.00 in total. The pamphlets for the UC training were printed for 12 staff, costing £2.40 in total. The costs of core component of the training were £433.70 in the UC arm and £1989.73 in the EC arm.

The auxiliary component of the training included travel, accommodation and meals for those delivering and receiving the training. These costs would vary depending on the arrangement of the training programme. In this study, we included the costs of these items if they were claimed in expenses. For travelling, only train journeys were included. Travelling within a city was not taken into account. For the train journeys, the journey duration was estimated using the National Rail website.35 It was then multiplied by the hourly costs of the personnel who took that journey. Centres 3 and 4 were located in the same city as most of the research team and, therefore, they did not require train travel; as a result, the travelling expense and time were not costed. The return train journey to centre 1 cost £29.30 in terms of ticket and 2 hours in terms of time. To centre 2, the return journey cost £118 (one-way £59) in terms of ticket and 9 hours in terms of time (4.5 hours one-way). Three team members took a return journey to centre 1, costing £205.83 in terms of travelling time and £87.90 in terms of train tickets. Two team members went to centre 2 and stayed in a hotel. Owing to personal arrangements, only one return ticket to centre 2 was claimed and the other was a one-way ticket. This led to £425.97 in terms of travelling time and £177.00 in terms of train tickets. Their stay cost £500 for accommodation and £40 for meals. The costs of the auxiliary component of the training were £293.73 in the EC arm and £1142.97 in the UC arm.

The extra component of the training was ECs given to staff who were smokers so that they themselves could become familiar with the device. This was not applicable to all staff trained; only four EC devices were given to the staff in one centre and two to the staff in the other centre during EC training. In total, six EC devices cost £120.

In total, the cost of the training for the UC arm was £1576.66 and the cost of the training for the EC arm was £2403.46. Allocating the costs evenly to participants in each arm, they amounted to £49.27 per participant in the UC arm and £50.07 per participant in the EC arm (Table 9).

The delivery of intervention in both arms was recorded in the keyworkers’ form at baseline (visit 0), 1 week post baseline (visit 1), 2 weeks post baseline (visit 2) and 3 weeks post baseline (visit 3). Contact time was costed by multiplying the staff hourly cost by the duration of the contact. The printing cost of the help-to-quit pamphlets for the UC arm and the EC device instruction pamphlets and the EC helpful tips pamphlets for the EC arm was recorded by the research team at £0.20 per pamphlet.

Owing to staff capacity, the data on delivery sessions were not collected in the UC arm. We therefore did not know the attendance of weekly sessions (visit 0–3) and if the pamphlet was given to the participants.

For the EC arm, one participant did not attend any session and did not receive the EC starter kit or e-liquid. The remaining 47 participants attended visit 0 and received an EC starter kit and five bottles of e-liquid. Excluding one participant with a missing response, 46 participants also received the EC ‘tips and tricks’ help sheet. The mean duration of session 0 (baseline) with homeless centre staff/keyworkers was 25 minutes (SD 16 minutes, range 7–80 minutes). Overall, 30 participants attended visit 1, of whom 29 received five bottles of e-liquid and one received two bottles. Except for a missing answer for one participant, the mean session duration of visit 1 was 8 minutes (SD 4 minutes, range 1–20 minutes). Among the 22 participants who attended visit 2, 20 received five bottles of e-liquid, one received four and one received 10. Except for a missing answer for one participant, the mean session duration of visit 2 was 6 minutes (SD 3 minutes, range 1–15 minutes). The number of participants who attended visit 3 was also 22, all of whom received five bottles of e-liquid. Duration was not recorded for two participants. Among the remaining 20 participants, the mean session duration was 5 minutes (SD 2 minutes, range 2–10 minutes).

The mean cost of the EC starter kit and e-liquid was £57.46 (SD £19.91) among the 48 participants in the EC arm. The mean cost of the pamphlet was £0.39 (SD £0.06) among 47 participants in the arm. The mean cost of the sessions was £7.51 (SD £4.60) among 43 participants. The cost of intervention delivery was therefore estimated for 43 participants whose data were complete in this part; the cost was £64.35 (SD £22.89) per participant. In total, the mean cost of the EC intervention, including training and delivery, was £114.42 (SD £22.89) among 43 participants in the EC arm (Table 10).

Summary of learning points for a future trial

Although it was feasible to collect information relating to the costs of delivery in the EC arm, collecting data from staff about contact time was not feasible in the UC arm. If staff are unable to complete this information each time they engage with a research participant, approaching them at a quieter time (e.g. at the end of the follow-up data collection period at each centre) may improve compliance.