Background
Despite gradual decline in smoking prevalence in many high-income countries, a substantial proportion of the adult population still smokes [
1]. Evidence-based interventions such as pharmacotherapy, and individual or group behavioural smoking cessation counselling (SCC) can increase individual success rates, but most smokers attempt to quit without assistance [
2]. The success rate for unassisted quit attempts is only around 5% after 1 year among Dutch smokers, and for individuals with a lower socio-economic status (SES) this percentage is even smaller [
3].
Smoking is more prevalent and persistent among lower-SES compared with higher-SES groups, and contributes greatly to SES-based health inequities [
4‐
8]. SES can be defined in several ways [
9]. In tobacco research, SES is often defined by educational level, which has been found to be an important indicator of risk of smoking independent of occupational class and income [
9]. Compared to their higher-SES counterparts, lower-SES smokers typically are heavier smokers, have more difficulty to quit successfully, and often receive less social support for smoking cessation [
10,
11]. In addition, lower-SES smokers are more often hindered in quitting successfully by SES-related problems (e.g., financial debts, housing/family problems), but can be helped to quit by a personalized and proactive approach [
12‐
14]. Most interventions that intend to help smokers quit smoking are more effective among higher-SES than lower-SES smokers [
8,
15]. A review showed that, overall, individual-level interventions compared with no support can help lower-SES smokers quit. However, it did not matter for effectiveness whether individual-level interventions were targeted specifically at lower-SES smokers [
15], possibly because the community and population level are not sufficiently included. One of the few interventions published that targeted lower-SES smokers is the UK-based eHealth intervention StopAdvisor [
16‐
18]. This website-delivered intervention was based on the PRIME theory of motivation, the incorporation of a range of behaviour change techniques and experience in designing web-based interventions for behaviour change, and extensive user-testing with lower-SES smokers [
17,
19]. A randomized controlled trial among 4613 daily smokers showed that, among lower but not higher-SES smokers, StopAdvisor was used more often and resulted in significantly higher abstinence rates than an information-only website [
16].
eHealth interventions can help smokers quit more effectively, especially if interventions are personalized, interactive, and include text messages [
20,
21]. From a public health perspective, eHealth is promising for reducing smoking prevalence among people who do not (want to) use pharmacotherapy nor receive traditional SCC, given its low thresholds for use. However, eHealth interventions often suffer from low adherence (high drop-out) rates, and are used less often by lower-SES than higher-SES individuals [
22,
23]. Blended care, an integration of eHealth and face-to-face treatment, may be the most promising method for helping individual smokers quit [
24]. As part of blended care, eHealth can add valuable enhancements to face-to-face behavioural interventions, including availability and accessibility in the smoker’s own environment, tailoring to users’ needs, low costs, and easy scalability [
22,
24].
The implementation of eHealth interventions often fails, with shortage of financial resources being an important barrier [
25]. The complexity of the implementation process of eHealth interventions is often underestimated [
26]. This results in, for example, interventions with low acceptability and feasibility, organizations that are not ready for the intervention, or potential users having personal or professional reasons not to use the intervention. Implementation of blended interventions for smoking cessation likely depends on healthcare professional factors associated with implementation of SCC more generally, such as limited self-efficacy, knowledge or skills, or unbeneficial beliefs—e.g., that smokers are themselves responsible for smoking, or that smokers are not motivated to quit [
27‐
31]. Various theoretical models have been advanced to understand implementation processes. The Consolidated Framework for Implementation Research (CFIR), a widely used synthesis of nineteen frameworks and theories on implementation, proposes five interacting domains that determine implementation success: intervention characteristics (e.g., intervention quality), the outer setting (e.g., external policies) and inner setting in which the intervention is implemented (e.g., implementation climate), characteristics of individuals implementing the intervention (e.g., self-efficacy), and the implementation process (e.g., planning) [
32,
33]. Acceptance and use of eHealth interventions can be explained using the Unified Theory of Acceptance and Use of Technology (UTAUT) [
34]. UTAUT proposes that performance and effort expectancy (i.e., gains and ease associated with using the technology, respectively) and social influence determine intention, and both intention and facilitating conditions determine actual use. These relations are moderated by gender, age, experience and voluntariness of use.
This study evaluated StopCoach, a mobile phone delivered eHealth intervention (app) targeted at lower-SES smokers based on StopAdvisor, in a real-world setting. We aimed to implement StopCoach in blended care settings within five municipalities in The Netherlands. The Netherlands were chosen as several organizations in the field indicated that they were in need of a smoking cessation intervention for lower-SES smokers [
35]. The project team cooperated with five local project leaders who engaged healthcare professionals to implement the intervention among smokers intending to quit. Smokers could use the app regardless of their SES. We conducted individual semi-structured interviews with project leaders, healthcare professionals, and participating smokers and examined quantitative data generated by the app. The following research questions were addressed:
(1)
What is the experience of project leaders and healthcare professionals implementing StopCoach, which barriers and facilitators emerge from the implementation process, and what is the experience of smokers using StopCoach?
(2)
Is app adherence to StopCoach related to participant (i.e., municipality), smoking (e.g., number of cigarettes per day) and initial app usage characteristics (e.g., enabling notifications)?
(3)
What percentage of app users undertake a quit attempt?
(4)
How do users evaluate their progress in quitting?
(5)
Are quit attempts related to participant, smoking and initial app usage characteristics?
Methods
Design and participants
This real-world study was performed in five municipalities involved in the proof-of-concept implementation project of StopCoach in The Netherlands in 2019–2020. Municipalities were recruited by Pharos. Pharos supports 155 municipalities that participate in the “Healthy in the City” program by the Dutch Ministry of Health, Welfare and Sport in reducing local health disparities. Any interested municipality could take part. Participating municipalities (i.e., Goeree-Overflakkee, Hulst, Roermond, Stadskanaal, and Weststellingwerf) were located in various regions in The Netherlands with different degrees of urbanisation, with numbers of inhabitants ranging from around 26,000–58,000 in 2019. Each municipality appointed their own project leader, who selected a setting in which to implement the app, and involved local healthcare professionals that could offer the app to smokers as part of a SCC program. Project leaders were instructed to take local health and social services into account, such that healthcare professionals could refer smokers for other problems (e.g., housing, financial debts) if necessary. In three municipalities, the implementation project was embedded in ongoing projects that focused on reducing smoking rates among pregnant women in the region, or promoting health among either the municipality inhabitants or the regional population which contained a relatively large share of lower-SES individuals. StopCoach was a stand-alone project in the two other municipalities, where it was implemented in midwifery practices and existing SCC programs serving the broader population of smokers. Smokers could use the app regardless of their SES.
Qualitative data was collected through interviews with five project leaders, seven healthcare professionals and ten participating smokers between November 2019 and February 2020. We aimed to include mostly lower-SES smokers. One project leader also participated as a healthcare professional and was interviewed twice (project leader 1/healthcare professional 4). Healthcare professionals were recruited through project leaders, and smokers were contacted by one of the interviewers after they had indicated that they were open to being interviewed on an information sheet about the project.
Quantitative data was retrieved from all app users (
N = 235) who indicated that they resided in one of the five municipalities involved, and who used the app during the period when the project was running (i.e., May 2019 through June 2020) regardless of how long they had used the application (see Additional file
1 for descriptive statistics). Ethical approval was provided by Trimbos Institute’s Ethical Committee (2548425).
Intervention
StopCoach comprised an 8-week stepwise program to support smokers in the initial phase of their quit attempt. The app was based on StopAdvisor [
16,
17], and fully complied with Dutch clinical guidelines for smoking cessation and tobacco dependence [
36,
37]. We ensured that all behaviour change techniques used in StopAdvisor were incorporated in the Dutch StopCoach app. We involved lower SES smokers and healthcare professionals in several design choices, such as the appearance of the virtual coach, which was an important feature of the app. The virtual coach provided practical and motivational support throughout the process. The application was tested on ease of understanding and accessibility in think-aloud sessions with five lower-SES smokers also reported low literacy, and modified based on their feedback (see Additional file
2 for more details). The app could be downloaded for free.
Procedure
Qualitative data were collected using in-depth semi-structured individual telephonic interviews (see Additional file
3 for interview protocols). We sent potential participants information about the study’s aims and procedures, processing of personal data and protection of privacy by e-mail. We contacted potential participants 1 week later to discuss any questions regarding the project, and scheduled an interview. Oral informed consent was recorded in a separate audio file. Interviews focused on evaluation of the app in all groups, complemented by experiences during the implementation process and support by Pharos for project leaders; experiences with SCC and the app for healthcare professionals; and smoking history and behaviour, and experiences with blended care for participants. Two interviewers (EH and another Pharos employee) specialized in lower-SES groups conducted most of the interviews. JK, at the time Master student in Medicine, interviewed four smokers. Duration of the interviews was 30 min on average, ranging from 15 to 46 min. Interviews were recorded and transcribed verbatim. The interview recording of healthcare professional 6 was aborted after 12 min due to technical malfunction. All interview participants received a €15 gift voucher.
Quantitative data were obtained from the app. During the installation process of the app, participants were asked to actively agree with the app’s privacy statement. The statement was drafted in understandable Dutch language, as lower literacy is more common among the lower-SES target group. If they agreed, they were asked again in a pop-up window whether they were sure. The privacy statement explained which data were collected, purposes of data collection (e.g., scientific research), and how users could have their data removed (i.e., by sending an e-mail). This procedure suffices the GDPR and Dutch privacy law requirements for consent.
Measures
Measures had no missing values, unless indicated otherwise.
Participant and smoking characteristics
The app asked participants about the municipality participants resided in (self-report with five options, and ‘other municipality’), number of cigarettes smoked per day, smoking within 30 min after waking up (yes/no), previous quit attempts ever (yes/no; no criteria specified for duration), and whether they currently had a professional coach for SCC (yes/no). For cigarettes per day, two values over 80 were recoded into 80 in order to correct for inaccurate data entry [
38]. Participants could indicate in the preparation step which reason(s) to quit they had. Participants were not obliged to answer this question (n = 18, 8% missing values). If there was any activity recorded in the preparation step, we also calculated number of reasons to quit reported. Demographic variables including SES were not asked, to prevent putting off app users in the onboarding process.
App usage
The app registered whether participants had enabled push-notifications, and had chosen the male/female virtual coach. We calculated duration of app usage (in days) by subtracting the date of last activity from the date of initial activity in the app, and assessed in which step participants showed their last activity within the app, in how many of the steps they had shown activity, and how many activities were registered in the app in total.
Analysis
Qualitative analysis
To answer research question 2, initial qualitative data analysis was performed by JK and KO, and supervised by EM who also performed the final, integrative analytic steps. Analysis took place according to the principles of the framework approach and cross-case analysis [
39,
40]. Separate coding trees were constructed for project leaders, healthcare professionals and participants, based on two randomly selected transcripts from each group. We combined inductive and deductive analysis, such that theoretical frameworks guided the analysis, but at the same time we were open to novel themes within the theoretical domains that emerged during the analytic process. The CFIR and UTAUT were used as theoretical frameworks [
32‐
34]. The CFIR facilitated our understanding of factors involved in the implementation process, and the UTAUT was used in addition to CFIR’s intervention domain in order to gain in-depth insight into performance expectancy (i.e., gains attained from using a technology) and effort expectancy (i.e., ease associated with using a technology) with regard to StopCoach. Data were coded and processed by JK and KO using Atlas.TI. Whilst coding, the coding trees were improved when necessary. We ensured reliability of the analysis by having five (three healthcare professionals, two participants) randomly selected transcripts coded independently by JK and KO, after which the coded transcripts were discussed and discrepancies were solved. One coded project leader transcript was reviewed in detail by EM and interpretations were discussed. Relevant quotes were subsequently brought together per code, and compared among interview participants within each group to create a cross-case analysis. We made sure that our analysis was grounded in the data by verifying our interpretations with the overall data from the interviews, and by removing answers to overly suggestive questions. Themes emerging for the within-group analysis were merged in an overall synthesis of findings. Illustrative quotes presented in the Results section were abbreviated for length and clarity.
Quantitative analyses
For research question 2, a dichotomous adherence variable was constructed, with the non-adherent category including participants who only did the preparation phase, and adherent participants showing activity after the preparation phase. We performed separate univariable logistic regression analyses, with adherence as the dependent variable and participant, smoking and initial app usage characteristic variables (e.g., enabling push-notifications or not) as independent variables. Dummy variables were created for municipality, with the municipality that most app users resided in (Roermond) as the reference category. Independent variables associated with adherence at p < 0.05 were then included in a multivariable logistics regression analysis. We performed a sensitivity analysis without the reasons to quit variables, as these variables had missing values and models were using cases with complete data for variables included in the respective model. Descriptive statistics were estimated to answer research questions 3 and 4. For research question 5, we fit univariable and multivariable models as for research question 2, with quit attempt as the dependent variable.
Discussion
This real-world study evaluated StopCoach, a smartphone application intended to support lower-SES smokers in quitting smoking, in an implementation project in five Dutch municipalities with 235 participating smokers. In addition, we investigated app adherence and quit attempts among app users. Qualitative results, based on individual interviews with local project leaders, healthcare providers and participating smokers suggest that the implementation process is primarily subject to factors related to the intervention itself and the setting in which the intervention is implemented. We found that the implementation was facilitated by project leaders’ and healthcare professionals’ positive attitudes towards the app, which seemed to result largely from positive effort and performance expectancies. That is, they perceived the app as easy to use and useful specifically for the lower-SES target group. This was reflected in participants’ accounts as well, for example, participants appreciated the practical tips and social support provided by the virtual coach, which felt genuine despite their awareness that the coach was virtual. Notably, although the study was not designed to detect SES differences in evaluation of the app, we found that these evaluations were similar among lower, middle, and higher SES smokers. Given that lower-SES smokers typically are less supported in quitting by their social environment than higher-SES smokers [
10,
11], eHealth interventions that use messages from a virtual coach to support participants are promising. Supporting this, quantitative results showed that people who had disabled push-notifications, which contained the messages from the coach, were more likely to drop out.
Overall, 33% of app users reported abstinence at some point, indicating that they had undertaken a quit attempt while using the app. Exploratory analyses indicated that participants who were more adherent to the app (i.e., longer use, and more activity in the app) were more likely to attempt to quit, suggesting that the app facilitated their smoking cessation process. It also seemed that participants who lapsed or relapsed were more likely to stop using the app, given that most participants who kept using the app reported that they had not smoked and that quitting went quite well. The current study design does not allow for assessing whether adherence led to quit attempts or vice versa. The app was perceived as less useful for dealing with lapse or relapse, as indicated by some smokers who had stopped using the app after having resumed smoking. In contrast to StopAdvisor, StopCoach allowed smokers to continue using the app after they had smoked, but the motivational messages, information about lapses, and the possibility to restart the app after a lapse do not seem sufficient to ensure that smokers who had smoked kept using the app. StopCoach was also felt to be less useful for facilitating long-term abstinence, as several participants and healthcare professionals found the app’s 8-weeks duration too short. A main barrier to implementation that emerged from the analysis of ‘Intervention’ factors were difficulties installing the app encountered by smokers with limited digital literacy. This occurred despite elaborate testing with low literate individuals and the general finding that the app was experienced as accessible and understandable, and suggests that developing inclusive eHealth interventions remains is challenging [
24].
With regard to ‘Setting’ factors, we found that the implementation was facilitated by good compatibility between the app and existing SCC programs, which allowed for blended care [
24]. In addition, project leaders, healthcare professionals and participants reported positive attitudes toward blended care for smoking cessation. Blended care was perceived as promising for lower-SES smokers in particular, as adhering to an app requires a pro-active attitude from the smoker, which could be complicated by other problems that lower-SES smokers may experience [
12,
13]. Quantitative results also showed that adherence to StopCoach was better among smokers who used the app alongside regular SCC, and that those who were supported by a professional SCC coach were more likely to attempt to quit. At the same time, the implementation of the app in a blended setting appeared hindered by healthcare professionals’ insufficient preparation and familiarity with the app, and possibly perceptions that integrating the app with their existing SCC program was not part of their role [
26]. However, since even the limited integration of the app with regular SCC programs observed in the current study was associated with better app adherence and quit attempts, blended care seems a promising route to dealing with the problem of attrition that is common in eHealth interventions for smoking cessation and eHealth more generally [
22], as well as to improving smoking cessation outcomes compared to stand-alone eHealth. In order for blended care to succeed, current results suggest that healthcare professionals need to be sufficiently prepared and feel responsible for integrating eHealth into their treatments [
26]. Furthermore, blended care is likely to be facilitated if the intervention has specific functionalities for the healthcare professional, such as a mode in which they can quickly walk through the app when preparing blended care, and a monitoring function that allows them to keep track of their patients’ progress. Notably, although blended care seems the most promising route to support people in smoking cessation, a substantial group of smokers in the larger population prefers to quit smoking without formal assistance [
41]. Arguably, using a stand-alone eHealth intervention is more beneficial than quitting without any support, and stand-alone eHealth interventions can make important contributions to public health. A stepped care model can be used, such that smokers who fail to quit smoking with a stand-alone eHealth intervention are stimulated to seek professional help. Importantly, matched care—in which smokers are directly referred to treatment that optimally meets their needs—is most appropriate for smokers for whom quitting is urgent (e.g., pregnant smokers, or smokers with smoking-related disease) as well as for smokers experiencing complicating factors (e.g., psychiatric disorders, socioeconomic problems) that they urgently need help with [
36].
An important barrier emerging from the analysis of ‘Setting’ factors was the difficulty experienced by project leaders and healthcare professionals to engage colleagues, organizations, the municipality and smokers in the project. This resulted, at least partially, from suboptimal communication at various levels. As a consequence, much of the work had to be done by a small number of people, which threatens sustainability of the implementation [
26,
42]. Barriers to providing SCC more generally also played a role, such that the implementation of StopCoach was hindered by HCP’s perceived resistance of patients to discuss smoking and smoking cessation, and insufficient referral networks both for SCC and other problems that lower-SES smokers may have [
28,
29]. A randomized controlled trial showed that a functional referral system for services in social domains (e.g., for employment, food, literacy) can facilitate successful quitting, provided that smokers use the referral, underscoring the importance of improving these structures if they are not in place [
12].
This study has limitations. First, selection bias likely plays a role in the interview sample for healthcare professionals and participants. Most project leaders stated that it was difficult to enthuse healthcare professionals for working with StopCoach for a longer period of time, but the healthcare professionals included in this study were positive about the project. Relatedly, one municipality was not represented by healthcare professionals, as the implementation process lagged behind and healthcare professionals were not yet involved when the interviews were conducted. Likewise, participating smokers were recruited from three municipalities only. App data furthermore showed that most people using the app did not have a professional helping them quit, whereas all but one of the interview participants participated in SCC and were advised to use the app by their healthcare professional. However, all project leaders were interviewed and provided their views on the implementation process, and we used both qualitative and quantitative data, which reduces risks associated with selection bias and helps provide a representative answer to the research questions. Second, although we assessed a number of smoking characteristics during the installation process of the app, we had to be selective in which variables to measure to prevent putting off app users. As such, the exact proportion of lower-SES smokers in the quantitative sample is unknown as SES was not asked (the same holds for age, gender etcetera, but key smoking characteristics were assessed). This also means that we were unable to assess associations between SES and both adherence and quit attempts. With regard to the interviews, a small majority of participating smokers (6/10) had a lower-SES, but the sample also contained a number of middle and higher SES participants, reflecting the real-world nature of this study. Although the study was not designed to study SES-based differences in evaluation of the app, we did not observe these either among our ten participant interviewees. Pending further research, this suggests that StopCoach may be useful for smokers regardless of SES, despite the fact that it was developed for the lower-SES target group. This corresponds with research showing that higher-SES individuals, like people with lower-SES, prefer simple language [
43]. Finally, low app adherence prevented thorough analysis of the smoking cessation process. As noted, low adherence is a well-known problem for eHealth interventions [
22]. The current study adds to the literature by offering explanations for low adherence and providing directions for reducing attrition from eHealth interventions for smoking cessation in the future. The real-world setting used in the current study allows for high ecological validity of these results [
22].
Conclusions
This real-world study demonstrated that it is possible to develop an accessible and supportive smoking cessation app for lower-SES smokers. Results based on interviews with project leaders, healthcare professionals and smokers suggest that future eHealth interventions can be made useful for lower-SES smokers by providing practical guidance using understandable language and visual material, including feedback on progress made, and incorporating a virtual coach that sends motivational messages which address the user by name. In addition, such interventions ideally last longer than 8 weeks, and provide adequate and tailored support for smokers who lapsed or relapsed. The implementation of an eHealth smoking cessation intervention in local settings proves to be a complex process, in which intervention characteristics and setting characteristics seem to play a key role. Given that the implementation of many eHealth interventions fails, the current study provides important insight into factors that can facilitate and obstruct successful implementation. Blended care appears to be both challenging and promising, as it requires effort on behalf of the healthcare professionals, as well as smokers willing to participate in both regular SCC and an eHealth intervention, but at the same time can increase adherence to the app and facilitate quit attempts. We believe that, for individual smokers, blended care combines the best of both face-to-face and digital coaching in helping smokers quit smoking successfully.
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.