Introduction
Tobacco smoking kills more than 8 million people a year worldwide (World Health Organization, 2019), making it the third highest risk factor for attributable disability-adjusted life-years, surpassed only by child and maternal malnutrition and high systolic blood pressure (Murray et al., Reference Murray, Aravkin, Zheng, Abbafati, Abbas, Abbasi-Kangevari and Lim2020). Total global costs of tobacco smoking are estimated around USD 1432 billion, which represents 1.8% of global annual GDP (Goodchild, Nargis, & Tursan d'Espaignet, Reference Goodchild, Nargis and Tursan d'Espaignet2018). In the US, tobacco is the leading cause of preventable death and disability (Center for Disease Control and Prevention, 2021), with direct and indirect medical costs exceeding USD 300 billion (US Department of Health Human Services, 2020; Xu, Bishop, Kennedy, Simpson, & Pechacek, Reference Xu, Bishop, Kennedy, Simpson and Pechacek2015).
The excess premature mortality associated with smoking can be mitigated by smoking cessation (Doll, Peto, Boreham, & Sutherland, Reference Doll, Peto, Boreham and Sutherland2004), but treatments are underutilized, and most major clinical trials have excluded those with serious mental illness (SMI) (Cather, Pachas, Cieslak, & Evins, Reference Cather, Pachas, Cieslak and Evins2017). Smoking rates have declined significantly over time among individuals without mental illness, but changed only slightly among those with mental illness (Cook et al., Reference Cook, Wayne, Kafali, Liu, Shu and Flores2014). More than half of individuals with schizophrenia spectrum disorders smoke cigarettes, a rate three times that of the general population (Sagud, Mihaljevic Peles, & Pivac, Reference Sagud, Mihaljevic Peles and Pivac2019; Šagud et al., Reference Šagud, Vuksan-Ćusa, Jakšić, Mihaljević-Peleš, Rojnić Kuzman and Pivac2018). Smoking prevalence is also particularly high among those with bipolar disorder (BD), who are two to three times more likely to have started smoking and are less likely to initiate and/or maintain smoking abstinence than individuals without psychiatric disorders (Heffner, Strawn, DelBello, Strakowski, & Anthenelli, Reference Heffner, Strawn, DelBello, Strakowski and Anthenelli2011). Life expectancy for individuals with SMI is reduced compared with the general population (Plana-Ripoll et al., Reference Plana-Ripoll, Pedersen, Agerbo, Holtz, Erlangsen, Canudas-Romo and Laursen2019), with tobacco accounting for a substantial portion of total deaths among patients with schizophrenia or BD (Callaghan et al., Reference Callaghan, Veldhuizen, Jeysingh, Orlan, Graham, Kakouris and Gatley2014).
Smoking cessation interventions would seem to be fundamental for improving life expectancy and quality of life in people with SMI (Doll et al., Reference Doll, Peto, Boreham and Sutherland2004; World Health Organization, 2018). However, mental health services are often unable to provide sufficient support for smoking cessation to patients (Bailey et al., Reference Bailey, Bartlem, Wiggers, Wye, Stockings, Hodder and Bowman2019), and staff attitudes about lack of usefulness of these strategies hinder the achievement of tobacco abstinence (Malone, Harrison, & Daker-White, Reference Malone, Harrison and Daker-White2018). Furthermore, individuals with SMI appear to have specific mediators/moderators of smoking behavior and barriers to smoking cessation compared to the general population (Lum, Skelton, Wynne, & Bonevski, Reference Lum, Skelton, Wynne and Bonevski2018; Moran, Betts, Ongur, & Janes, Reference Moran, Betts, Ongur and Janes2018; Slyepchenko, Brunoni, McIntyre, Quevedo, & Carvalho, Reference Slyepchenko, Brunoni, McIntyre, Quevedo and Carvalho2016). Smoking cessation self-efficacy, or confidence in one's ability to quit, is a known predictor of cessation outcomes with lower cessation self-efficacy predicting higher rates of relapse (Clyde, Pipe, Els, Reid, & Tulloch, Reference Clyde, Pipe, Els, Reid and Tulloch2017). Smokers with psychotic disorders such as schizophrenia commonly report low cessation self-efficacy, along with low motivation to quit due to a desire to better cope with symptoms, particularly cognitive impairment and negative symptoms, as well as to reduce side effects of antipsychotic treatment (Kumari & Postma, Reference Kumari and Postma2005; Lum et al., Reference Lum, Skelton, Wynne and Bonevski2018; Thornton et al., Reference Thornton, Baker, Lewin, Kay-Lambkin, Kavanagh, Richmond and Johnson2012). When schizophrenia patients attempt to quit, some reports find exacerbation of executive functioning deficits, which may contribute to high rates of smoking relapse (Tidey & Miller, Reference Tidey and Miller2015). In the case of BD, a substantial portion of patients report that they smoke to manage symptoms of their mental illness, and smoking is associated with greater severity of mood symptoms, comorbid psychiatric and addictive disorders, and suicidality (Tidey & Miller, Reference Tidey and Miller2015). Experimental evidence supports that nicotine can indeed influence symptoms in SMI (Featherstone & Siegel, Reference Featherstone and Siegel2015; Koukouli et al., Reference Koukouli, Rooy, Tziotis, Sailor, O'Neill, Levenga and Maskos2017; Kumari & Postma, Reference Kumari and Postma2005; Novak et al., Reference Novak, Zai, Mirkhani, Shaikh, Vincent, Meltzer and Le Foll2010; Sabe, Zhao, & Kaiser, Reference Sabe, Zhao and Kaiser2020), as well as cognitive performance in schizophrenia, BD and major depressive disorder (MDD) (Barr et al., Reference Barr, Culhane, Jubelt, Mufti, Dyer, Weiss and Evins2008; Caldirola et al., Reference Caldirola, Daccò, Grassi, Citterio, Menotti, Cavedini and Perna2013; Hong et al., Reference Hong, Schroeder, Ross, Buchholz, Salmeron, Wonodi and Stein2011; Jubelt et al., Reference Jubelt, Barr, Goff, Logvinenko, Weiss and Evins2008; Levin, Wilson, Rose, & McEvoy, Reference Levin, Wilson, Rose and McEvoy1996; Smith, Singh, Infante, Khandat, & Kloos, Reference Smith, Singh, Infante, Khandat and Kloos2002; Smucny, Wylie, Kronberg, Legget, & Tregellas, Reference Smucny, Wylie, Kronberg, Legget and Tregellas2017). Other studies, however, limit the neurocognitive effects of nicotine only to schizophrenia (Morisano, Wing, Sacco, Arenovich, & George, Reference Morisano, Wing, Sacco, Arenovich and George2013).
Pharmacologic treatments for smoking cessation in SMI include bupropion and varenicline, as well as nicotine replacement therapies when necessary (Peckham, Brabyn, Cook, Tew, & Gilbody, Reference Peckham, Brabyn, Cook, Tew and Gilbody2017). However, quit rates in SMI are low and relapse rates are higher than in the general population (Brunette et al., Reference Brunette, Mueser, Babbin, Meyer-Kalos, Rosenheck, Correll and Penn2018b; Peckham et al., Reference Peckham, Brabyn, Cook, Tew and Gilbody2017). While psychosocial treatments like cognitive behavioral therapy, motivational interviewing, and contingency reinforcement have demonstrated short-term efficacy in some trials, other studies have shown no effects (Baker et al., Reference Baker, Bucci, Lewin, Kay-Lambkin, Constable and Carr2006; Cather et al., Reference Cather, Pachas, Cieslak and Evins2017; Gelkopf et al., Reference Gelkopf, Noam, Rudinski, Lerner, Behrbalk, Bleich and Melamed2012; Gold et al., Reference Gold, Otto, Deckersbach, Sylvia, Nierenberg and Kinrys2018; Lum et al., Reference Lum, Skelton, Wynne and Bonevski2018; Williams et al., Reference Williams, Steinberg, Zimmermann, Gandhi, Stipelman, Budsock and Ziedonis2010). With the advance of digital technologies and eHealth, several smoking cessation digital interventions have been developed for use in non-psychiatric population, although evidence regarding their efficacy shows mixed results (Bricker, Watson, Mull, Sullivan, & Heffner, Reference Bricker, Watson, Mull, Sullivan and Heffner2020; Chulasai, Chinwong, Chinwong, Hall, & Vientong, Reference Chulasai, Chinwong, Chinwong, Hall and Vientong2021; Garrison et al., Reference Garrison, Pal, O'Malley, Pittman, Gueorguieva, Rojiani and Brewer2020; Pbert et al., Reference Pbert, Druker, Crawford, Frisard, Trivedi, Osganian and Brewer2020; Peek et al., Reference Peek, Hay, Hughes, Kostellar, Kumar, Bhikoo and Marshall2021). Although digital interventions would seem to be particularly relevant for patients with SMI (Sawyer, Hassan, Guinart, Martinez Agulleiro, & Firth, Reference Sawyer, Hassan, Guinart, Martinez Agulleiro and Firth2022), most major clinical trials have excluded such patients (Cather et al., Reference Cather, Pachas, Cieslak and Evins2017), and most interventions have not been assessed in properly powered clinical trials (Haskins, Lesperance, Gibbons, & Boudreaux, Reference Haskins, Lesperance, Gibbons and Boudreaux2017), have not included behavior change techniques that have been shown to be effective in smoking cessation interventions (Ubhi et al., Reference Ubhi, Michie, Kotz, van Schayck, Selladurai and West2016) or have not been customized to match users' personal characteristics (Hoeppner et al., Reference Hoeppner, Hoeppner, Seaboyer, Schick, Wu, Bergman and Kelly2016). A recent narrative article has explored the feasibility of smoking cessation apps in people with schizophrenia (Sawyer et al., Reference Sawyer, Hassan, Guinart, Martinez Agulleiro and Firth2022), and indicated that these interventions might represent a realistic alternative for the treatment of nicotine use in patients with SMI. However, due to its narrative nature, information on effectiveness, usability, and other relevant outcomes was scarce and not systematically assessed.
Thus, the purpose of this review was to systematically describe and evaluate the characteristics of digital interventions for smoking cessation validated in patients with SMI, namely schizophrenia-spectrum disorder (SSD), BD, and MDD. We aimed to describe and evaluate (1) effectiveness, (2) acceptability and adherence, (3) usability, and (4) safety of these interventions.
Methods
This systematic review was conducted according to Preferred Reporting Items for Systematic Reviews and Meta-Analysis (PRISMA) guidelines (Moher, Liberati, Tetzlaff, Altman, & Group, Reference Moher, Liberati, Tetzlaff, Altman and Group2009). The review protocol for this review was registered in PROSPERO (CRD42021262481).
Literature search
We conducted a search of the electronic databases PubMed/MEDLINE, EMBASE, CINAHL, Web of Science, Cochrane Tobacco Addiction Group Specialized Register, and PsycINFO from database inception until 31 March 2022. Reference lists of eligible studies were screened, but no other sources (including unpublished studies) were sought. No restrictions on language were applied. Published, full-text peer-reviewed articles reporting the effectiveness, acceptability, adherence, usability and/or safety of digital interventions for smoking cessation in patients with SSD, BD, and/or MDD were included. Articles were excluded if: (1) the study population was composed of patients with a primary diagnosis different from SSD, BD and/or MDD; (2) the intervention was exclusively non-digital; (3) the intervention was not for smoking cessation; and (4) the paper was an abstract, systematic review, protocol, or poster communication. The search string was designed to include any study reporting digital interventions for smoking cessation in patients with SMI, as follows: (digital OR app OR web OR ehealth OR mhealth OR smartphone) AND ((smoking OR tobacco) AND (cessation OR abstinence)) AND (severe mental illness OR serious mental illness OR SMI OR psychosis OR schizophrenia OR schizoaffective OR bipolar OR mania OR major depressive disorder OR major depression OR depression).
Study selection and data extraction
After initial search and removal of duplicates, each study was independently screened for eligibility by two reviewers (BP, LMA). If a study was deemed relevant by at least one reviewer, or in case of doubt, the full-text article was retrieved and examined in depth for eligibility. Differences in search and selection results were discussed and when consensus was not reached, a third reviewer (DG) was consulted.
Data extraction was performed independently by two reviewers (BP, LMA) using a template developed for this systematic review including the following characteristics and data: (1) article design and characteristics: first author, year, country, study design, primary and secondary outcomes, comparators; (2) characteristics of the digital intervention: name, type (smartphone app, web), device (smartphone, cellphone, laptop, computer), cost, theoretical approach, duration; (3) sample characteristics: eligibility criteria, sample size, diagnosis, age, sex, race, illness duration, setting (inpatient and/or outpatient), symptom severity, pattern of tobacco use and severity of nicotine dependence; and (4) results of the intervention: effectiveness, acceptability, adherence, usability, and safety.
Data on effectiveness of the intervention were extracted according to study-defined criteria as any smoking reduction and/or abstinence/cessation, measured either as self-report, structured/semi-structured interviews, or expired carbon monoxide (CO). Acceptability, adherence and usability were extracted using the quantitative instruments reported in the study. Additionally, qualitative results on acceptability and usability were extracted and summarized from available articles. Safety of the intervention was extracted if the publication reported the onset of adverse events (AEs) during the trial.
Study quality was evaluated using the National Heart, Lung, and Blood Institute (NHLBI) Study Quality Assessment Tools for Controlled Interventions and Before-After (Pre-Post) Studies With No Control Group (National Institutes of Health, 2021). As most digital interventions were not accessible online to be fully evaluated with the Adapted Mobile App Rating Scale (A-MARS) (Stoyanov et al., Reference Stoyanov, Hides, Kavanagh, Zelenko, Tjondronegoro and Mani2015), we used author-reported information whenever possible (see online Supplementary Table S3). For the two interventions that were available (QuitGuide and quitStart), ratings reflect the assessment made independently and agreed upon by the investigators (BP, LMA) while using the apps.
Results
The search retrieved 2214 articles. After duplicates were removed, 1419 unique studies were screened and 1388 were excluded. A full-text review of 31 articles was conducted, and finally 13 articles were included in this review. The inter-rater agreement was 98.5% [95% CI (97.6–99.0)]. The study selection process is presented as a PRISMA flow diagram in Fig. 1. A detailed summary of the characteristics of the included articles can be found in Table 1.
Figure 1. PRISMA flowchart of the review process.
Table 1. Characteristics of the included studies
AA, African American; BD, bipolar disorder; CM, contingency management; FTDN, Fagerström Test for Nicotine Dependence; MDD, major depressive disorder; NA, not applicable; NHLBI SQAT, National Heart, Lung, and Blood Institute Study Quality Assessment Tools; NR; not reported; PNOS, psychosis not otherwise specified; RCT, randomized controlled trial; SMI, serious mental illness; SZ, schizophrenia; SZ-AFF, schizoaffective disorder.
The 13 included articles covered 9 different digital interventions. All included articles were published after 2011 and pertain to five distinct research groups. The number of articles studying each digital intervention ranged from 1 to 3. Six articles used controlled trial methods to study the digital intervention, while the rest used quasi-experimental and prospective designs. All samples were of outpatients with sample size ranging from n = 5 (Vilardaga et al., Reference Vilardaga, Rizo, Kientz, McDonell, Ries and Sobel2016) to n = 162 (Brunette et al., Reference Brunette, Ferron, McGurk, Williams, Harrington, Devitt and Xie2020). SSD was the most common diagnosis, followed by BD and MDD (Table 1).
General characteristics of the included digital interventions
Nine unique digital interventions for smoking cessation were identified in this systematic review. At the time this review was conducted, only two (QuitGuide and quitStart) were commercially available and readily accessible via App Store and Google Play, all of them in English with no alternative language option. The remaining seven interventions were not accessible for evaluation and data were obtained from associated publications or directly from the authors whenever possible (See Table 2 for detailed description of the interventions). Briefly, six interventions were delivered as smartphone-based apps, and three were computer-based. The theoretical foundations of the digital interventions range from acceptance and commitment therapy (ACT) (Hayes, Strosahl, & Wilson, Reference Hayes, Strosahl and Wilson2011) to theory of planned behavior (Ajzen, Reference Ajzen1991), mindfulness (Quaglia, Brown, Lindsay, Creswell, & Goodman, Reference Quaglia, Brown, Lindsay, Creswell and Goodman2015), and contingency management (CM) (Petry, Reference Petry2011). The specific interventions were as follows:
(1) iCOMMIT is a behavioral therapy-based smartphone-based app, which includes mobile CM (mCM). Additionally, the full intervention includes five cognitive-behavioral counseling sessions with a therapist. For the digital part of the intervention, participants are asked to self-record a video of themselves blowing into a handheld CO monitor to confirm smoking abstinence and upload it onto the secure website linked to the app. The website then analyses the CO monitor reading in the video and translates that information to deliver financial reinforcement through the app (Medenblik et al., Reference Medenblik, Mann, Beaver, Dedert, Wilson, Calhoun and Beckham2020).
(2) Learn To Quit is a smartphone-based app based on ACT and specifically developed for patients with SMI following a user-centered design (Vilardaga et al., Reference Vilardaga, Rizo, Zeng, Kientz, Ries, Otis and Hernandez2018). Notably, it is the only included intervention who was developed in co-production with patients with SMI. It focuses on three processes of change: creating awareness of experienced smoking urges, openness to that experience, and committing to quitting smoking through specific values. Learn To Quit uses simple screens and large buttons, gamification and prioritizing visual engagement among other features to minimize usability challenges in people with SMI. It also implements a tracking option for daily ecological momentary assessments (EMA) of mood, number of cigarettes smoked, tobacco craving, and nicotine replacement therapy (NRT) use (Vilardaga et al., Reference Vilardaga, Rizo, Ries, Kientz, Ziedonis, Hernandez and McClernon2019, Reference Vilardaga, Rizo, Palenski, Mannelli, Oliver and McClernon2020).
(3) Let's Talk About Quitting Smoking is a web-based intervention focused on Theory of Planned Behavior. It comprises 12 modules aimed to address attitudes, social norms, and perceived behavioral control for smoking and cessation treatment. It includes 48 interactive sessions that encourage the user to enlist reasons for smoking and track smoking related activities, practice coping skills, or learn about NRT and other medications, among others. The interface was designed to minimize reliance on cognitive functions to improve usability in patients with SMI (Brunette et al., Reference Brunette, Ferron, Geiger, Guarino, Pratt, Lord and Adachi-Mejia2019).
(4) Let's Talk About Smoking is web-based motivational decision support system designed to increase motivation to quit. This program can be completed over a single sitting within 30–60 min. It includes a video program host who guides the participants through three modules to assist decision making about cessation treatment options. The program was tailored for people with cognitive deficits to ensure ease of use (Brunette et al., Reference Brunette, Ferron, McGurk, Williams, Harrington, Devitt and Xie2020).
(5) mSMART MIND is a smartphone-assisted intervention based on mindfulness and CM. The intervention encourages mindfulness practices, incorporates EMA and uses a progressive mCM payment schedule through videos with a handheld CO monitor showing expired CO levels which are uploaded until 14 days after the set quit date (Minami et al., Reference Minami, Brinkman, Nahvi, Arnsten, Rivera-Mindt, Wetter and Brown2018).
(6) QuitGuide, QuitPal, and quitStart are smartphone-based apps developed by the US National Cancer Institute based on different clinical practice guidelines for smoking cessation. QuitGuide is aimed toward the adult population, the app monitors behaviors and mood associated with quitting smoking, enhances motivation to maintain abstinence and helps patients setting a quit date. The intervention includes psychoeducation, tracking smoking habits and tips for quitting. QuitPal allows setting a quit date, daily tracking of smoking and mood, and allows users to log smoke-free days if the individual has been smoke-free for an entire day. The app allows for customized reminders, saving goals, and provides graphs of the progress (Vilardaga et al., Reference Vilardaga, Rizo, Kientz, McDonell, Ries and Sobel2016). quitStart was designed to target the teenage population. It encourages the user to set a quit date, provides education on smoking cessation strategies with interactive swipable cards, and allows for daily tracking of information. Additionally, it offers distraction games to aid with smoking craving. The app automatically sends check-in notifications to ask the user for the number of cigarettes smoked since the last check-in.
(7) WebQuit Plus is a web-based intervention based on ACT, built on the WebQuit Plus program (Bricker, Mull, McClure, Watson, & Heffner, Reference Bricker, Mull, McClure, Watson and Heffner2018). This program has four sequential components to help the user by making a quit plan, developing awareness and coping skills of triggers, and supporting long-term abstinence by engaging personal values and self-compassion. WebQuit Plus adds to the ACT program exercises to tackle specific challenges to smoking cessation, testimonials, and one- and two-way messaging to promote adherence to NRT and assistance with mood-specific triggers. It also uses tracking data to display money and minutes saved via reducing or quitting smoking on the main dashboard (Heffner et al., Reference Heffner, Kelly, Waxmonsky, Mattocks, Serfozo, Bricker and Ostacher2020).
Table 2. Effectiveness, acceptability, adherence, usability and safety of smoking cessation digital interventions
AE, adverse events; EOT, end-of-treatment; NR, not reported; PPA, point-prevalence abstinence; SAE, severe adverse events; SUS, System Usability Scale [mean (s.d.)].
a Results are reported for the whole group, not just people with severe mental illness (SMI).
b Results are reported jointly for QuitGuide and quitStart.
Effectiveness
The definition of effectiveness varied widely among the different articles. The most frequently used definition was point-prevalent abstinence (PPA), either at 7 or 30 days, measured either as self-report or confirmed with expired CO. Abstinence rates ranged between 9 and 40%, with varying comparison groups. The highest abstinence rates were found for mSMART MIND, with biochemically verified 7-day PPA abstinence rates of 40% at 2 weeks after end of treatment (EOT), and 16% at 3 months post-EOT (Minami et al., Reference Minami, Nahvi, Arnsten, Brinkman, Rivera-Mindt, Wetter and Brown2022). Similar rates were found in iCOMMIT, with 7-day self-reported PPA at EOT of 38–40% for the successive cohort design (Wilson et al., Reference Wilson, Thompson, Currence, Thomas, Dedert, Kirby and Beckham2019), but 9.5–19.0% for the pilot RCT trial (Medenblik et al., Reference Medenblik, Mann, Beaver, Dedert, Wilson, Calhoun and Beckham2020). Other definitions of effectiveness used included reduction of cigarettes smoked per day (CPD), engagement in smoking cessation behavior, and smoking cessation knowledge. Vilardaga et al. (Reference Vilardaga, Rizo, Palenski, Mannelli, Oliver and McClernon2020) found that Learn to Quit led to a significant greater reduction in CPD than the comparator (QuitGuide), and iCOMMIT was found to reduce CPD, although significant reductions were only reported for patients with BD. Regarding engagement in smoking cessation behavior (e.g.: seek counsel, start smoking cessation medication, enroll in support groups), the interventions did not seem to outperform their comparators. In this regard, Brunette et al. (Reference Brunette, Ferron, McHugo, Davis, Devitt, Wilkness and Drake2011); Brunette et al. (Reference Brunette, Ferron, Robinson, Coletti, Geiger, Devitt and McHugo2018a) and Brunette et al. (Reference Brunette, Ferron, McGurk, Williams, Harrington, Devitt and Xie2020) reported inconsistent evidence for the effect of Let's Talk About Smoking, and Heffner et al. (Reference Heffner, Kelly, Waxmonsky, Mattocks, Serfozo, Bricker and Ostacher2020) and Vilardaga et al. (Reference Vilardaga, Rizo, Palenski, Mannelli, Oliver and McClernon2020) did not find differences in NRT use between groups for WebQuit Plus and Learn to Quit, respectively. Finally, iCOMMIT was not found to improve smoking cessation knowledge (Medenblik et al., Reference Medenblik, Mann, Beaver, Dedert, Wilson, Calhoun and Beckham2020). Detailed information regarding effectiveness can be found in Table 2.
Acceptability and adherence
In the nascent field of digital health, the definitions of acceptability and adherence are oftentimes blurred and overlapped. For the purpose of this review, we report acceptability as the willingness and satisfaction to use a digital intervention (Nadal, Sas, & Doherty, Reference Nadal, Sas and Doherty2020), and adherence as the degree to which patients are able to use the digital intervention as indicated by the research protocol.
Acceptability was reported only for three interventions (Let's Talk About Quitting Smoking, Let's Talk About Smoking, and WebQuit Plus). Across studies, acceptability was evaluated with author-developed scales rather than standardized measures. For Let's Talk About Quitting Smoking (Brunette et al., Reference Brunette, Ferron, Geiger, Guarino, Pratt, Lord and Adachi-Mejia2019), 87% of the sessions were rated ⩾3 on a 4-point Likert scale, and 85% of participants gave a positive response when interrogated about the intervention. Similarly, Let's Talk About Smoking was found to be easier to understand and to have a higher satisfaction rate than its comparator (Brunette et al., Reference Brunette, Ferron, Robinson, Coletti, Geiger, Devitt and McHugo2018a). For mSMART MIND, satisfaction was measured on a 5-point Likert scale, and the intervention was found ‘enjoyable’ (author-reported score: 4.6/5) and ‘helpful’ (author-reported score: 4.7–4.9/5) (Minami et al., Reference Minami, Brinkman, Nahvi, Arnsten, Rivera-Mindt, Wetter and Brown2018, Reference Minami, Nahvi, Arnsten, Brinkman, Rivera-Mindt, Wetter and Brown2022). WebQuit Plus outperformed its comparator (Smokefree.gov) in satisfaction and perceived usefulness, but not in other items (such as usefulness of the plans for quitting or the support forum, site organization or accessibility) (Heffner et al., Reference Heffner, Kelly, Waxmonsky, Mattocks, Serfozo, Bricker and Ostacher2020).
Additionally, three articles (Gowarty, Aschbrenner, & Brunette, Reference Gowarty, Aschbrenner and Brunette2021; Vilardaga et al., Reference Vilardaga, Rizo, Kientz, McDonell, Ries and Sobel2016, Reference Vilardaga, Rizo, Ries, Kientz, Ziedonis, Hernandez and McClernon2019) used semi-structured qualitative interviews to measure acceptability. Patients showed positive attitudes and high levels of acceptability for QuitGuide and quitSTART, emphasizing the motivational content (Gowarty et al., Reference Gowarty, Aschbrenner and Brunette2021). QuitPal was found to be excessively focused on quitting smoking, and that a shift toward reducing tobacco use would be preferable (Vilardaga et al., Reference Vilardaga, Rizo, Kientz, McDonell, Ries and Sobel2016). Finally, patients found Learn to Quit to be more engaging than QuitGuide, pointing out the use of gamification to boost engagement (Vilardaga et al., Reference Vilardaga, Rizo, Ries, Kientz, Ziedonis, Hernandez and McClernon2019). In general, the most remarkable features of the interventions were cigarette tracking, aids to avoid craving and monetary incentives (Gowarty et al., Reference Gowarty, Aschbrenner and Brunette2021; Vilardaga et al., Reference Vilardaga, Rizo, Kientz, McDonell, Ries and Sobel2016, Reference Vilardaga, Rizo, Ries, Kientz, Ziedonis, Hernandez and McClernon2019).
Adherence was reported for two interventions (iCOMMIT, and mSMART MIND). The intervention iCOMMIT was reported to have an adherence rate of 40–63% over the 4-week period of the trial (Wilson et al., Reference Wilson, Thompson, Currence, Thomas, Dedert, Kirby and Beckham2019). Adherence, considered as percentage of videos uploaded as part of the mCM intervention, was reported to be 66.2–68.0% of the total requested videos (Minami et al., Reference Minami, Brinkman, Nahvi, Arnsten, Rivera-Mindt, Wetter and Brown2018, Reference Minami, Nahvi, Arnsten, Brinkman, Rivera-Mindt, Wetter and Brown2022). Relevant information regarding acceptability and adherence of the reviewed digital interventions can be found in Table 2.
Usability
Five articles applied the System Usability Scale (SUS) to measure usability of four different digital interventions (Gowarty et al., Reference Gowarty, Aschbrenner and Brunette2021; Vilardaga et al., Reference Vilardaga, Rizo, Kientz, McDonell, Ries and Sobel2016, Reference Vilardaga, Rizo, Palenski, Mannelli, Oliver and McClernon2020, Reference Vilardaga, Rizo, Ries, Kientz, Ziedonis, Hernandez and McClernon2019). This 10-item scale provides a reliable measure of usability, with scores higher than 68 considered ‘above average’ (Brooke, Reference Brooke1996). The highest score was found for Learn to Quit (85.2 ± 15.5), while the lowest was reported for QuitPal (65.5 ± 18.6). In addition, Vilardaga et al., (Reference Vilardaga, Rizo, Ries, Kientz, Ziedonis, Hernandez and McClernon2019) reported statistically significant higher SUS scores for Learn to Quit than its comparator (QuitGuide); and in a different article (Vilardaga et al., Reference Vilardaga, Rizo, Kientz, McDonell, Ries and Sobel2016), they found that patients with SMI found QuitPal challenging to use due to its layer structure. Heffner et al. (Reference Heffner, Kelly, Waxmonsky, Mattocks, Serfozo, Bricker and Ostacher2020), found no differences between intervention and control groups for number of logins, number of unique login days or total number of days in use. Noteworthy, it was reported that none of the participants accessed the section where most of the content targeted to smokers with BD was located.
Additionally, three articles (Vilardaga et al., Reference Vilardaga, Rizo, Kientz, McDonell, Ries and Sobel2016, Reference Vilardaga, Rizo, Ries, Kientz, Ziedonis, Hernandez and McClernon2019; Wilson et al., Reference Wilson, Thompson, Currence, Thomas, Dedert, Kirby and Beckham2019) used semi-structured qualitative interviews to measure usability. Using a think-aloud approach, users of QuitPal found obstacles to use several components of the app (e.g.: enter information, pull up the keypad, navigate through several layers), which ultimately required direct guidance (Vilardaga et al., Reference Vilardaga, Rizo, Kientz, McDonell, Ries and Sobel2016). Learn to Quit was found easier to use than QuitGuide (Vilardaga et al., Reference Vilardaga, Rizo, Ries, Kientz, Ziedonis, Hernandez and McClernon2019). Finally, Wilson et al. (Reference Wilson, Thompson, Currence, Thomas, Dedert, Kirby and Beckham2019) report that initiating CO readings was the main hurdle while using iCOMMIT. Relevant information regarding usability of the reviewed digital interventions can be found in Table 2.
Safety
AEs were reported only for five digital interventions (Learn to Quit, Let's Talk About Quitting Smoking, Let's Talk About Smoking, mSMART MIND, and WebQuit Plus) in six articles (Brunette et al., Reference Brunette, Ferron, Geiger, Guarino, Pratt, Lord and Adachi-Mejia2019, Reference Brunette, Ferron, McGurk, Williams, Harrington, Devitt and Xie2020; Heffner et al., Reference Heffner, Kelly, Waxmonsky, Mattocks, Serfozo, Bricker and Ostacher2020; Minami et al., Reference Minami, Brinkman, Nahvi, Arnsten, Rivera-Mindt, Wetter and Brown2018; Vilardaga et al., Reference Vilardaga, Rizo, Palenski, Mannelli, Oliver and McClernon2020, Reference Vilardaga, Rizo, Ries, Kientz, Ziedonis, Hernandez and McClernon2019), although standardized assessments were not used. Most of the interventions were found to be safe, and no AEs were reported for Let's Talk About Smoking (Brunette et al., Reference Brunette, Ferron, McGurk, Williams, Harrington, Devitt and Xie2020) and mSMART MIND (Minami et al., Reference Minami, Brinkman, Nahvi, Arnsten, Rivera-Mindt, Wetter and Brown2018).
For Learn to Quit, Vilardaga et al. (Vilardaga et al., Reference Vilardaga, Rizo, Ries, Kientz, Ziedonis, Hernandez and McClernon2019, Reference Vilardaga, Rizo, Palenski, Mannelli, Oliver and McClernon2020) found no differences for ‘related’ or ‘possibly related’ AEs between intervention and control groups, with 87% of these AEs being linked to NRT use. Additionally, PANSS general psychopathology scores remained stable (pre- v. post-intervention: 40 v. 32, no p values reported), and although one participant was hospitalized due to an unrelated psychiatric event, no serious AEs (SAEs) were found. One participant complained of increased paranoia with Let's Talk About Quitting Smoking (Brunette et al., Reference Brunette, Ferron, Geiger, Guarino, Pratt, Lord and Adachi-Mejia2019). During the WebQuit Plus trial, Heffner et al. (Reference Heffner, Kelly, Waxmonsky, Mattocks, Serfozo, Bricker and Ostacher2020) reported 18 AEs, of which nine were classified as SAEs. Five unique patients (two on the WebQuit Plus arm, and three on the control arm) reported seven SAEs that were characterized as psychiatric, but only one of these was deemed to have a potential relationship to the intervention. The remaining AEs were considered to be related to nicotine withdrawal symptoms or adverse effects of NRT use. In addition, depression and mania scores remained stable throughout the trial. Finally, Wilson et al. (Reference Wilson, Thompson, Currence, Thomas, Dedert, Kirby and Beckham2019) did not report any medication-related AEs with iCOMMIT, although no information was available for the digital component of this intervention. An overview of the safety of the interventions can be found on Table 2.
Quality assessment
Overall, quality of the studies was ‘fair’ to ‘good’ as assessed with the NHLBI Study Quality Assessment Tools, although with limitations inherent to the experimental or pilot nature of most of the studies. Inter-rater agreement for quality assessment was 90.0% [95% CI (55.5–99.7)]. Studies showed high adherence to protocol and methods, with few drop-outs from the trials. However, blinding and sample size were found the main concerns when rating quality (see online Supplementary Tables S1 and S2 for detailed NHLBI assessments).
Digital interventions were evaluated using specific items of the A-MARS. In this regard, quitStart showed the greatest scores, mostly due to its highly user-friendly interface, with high ratings for engagement and functionality aspects. By contrast, iCOMMIT was deemed the lowest since the digital component of the intervention was mostly limited to mCM. Overall, the digital interventions showed good results in items of ‘interest’, ‘ease of use’ and ‘goals’, with lower ratings for addressing ‘multiple health issues/symptoms’ and ‘real-time tracking’. See online Supplementary Table S3 for detailed A-MARS ratings.
Discussion
To our knowledge, this is the first systematic review of digital interventions for smoking cessation in patients with SMI. We found nine different digital interventions that were assessed in patients with SSD, BD, and MDD. First, the highest abstinence rates (measured as 7-day PPA) were found with mSMART MIND and iCOMMIT, two smartphone-based apps developed on the basis of mCM. In this regard, mSMART MIND seemed to be especially beneficial for patients with BD, showing a significant improvement in abstinence rates when compared with enhanced standard treatment without mCM (aOR = 8.12, 95% CI 1.42–46.6, p = 0.019). Additionally, mSMART MIND showed the greatest CPD reduction (79.4% at 2 weeks after EOT, decreasing to 61.4% at 3 months), followed by iCOMMIT and Learn to Quit. Results on motivation enhancement and improvement of smoking cessation knowledge do not support the use of the reviewed digital interventions for these purposes. Participants reported high acceptability rates for Let's Talk About Quitting Smoking, Let's Talk About Smoking, iCOMMIT, mSMART MIND and WebQuit Plus, and in the case of Let's Talk About Smoking and WebQuit Plus, acceptability measures outperformed those of the comparators. However, adherence rates to the interventions were low (40–68%). The highest usability score was found for Learn to Quit (SUS = 83.1–85.2%), which also was reported to be significantly better, in terms of usability, than QuitGuide. This seems in accordance with its user-centered design since the early phases of development, which allowed to tailor the intervention to patients with SMI. Results on qualitative measures were found to be similar to quantitative approaches, stressing the strength of using multiple methodologies when studying digital interventions. Finally, safety was reported for only five interventions, but broadly, the interventions were safe to use, and no psychopathological decompensations were reported while carrying out the trials. No AEs were reported for Let's Talk About Smoking or mSMART MIND, and in sum, only WebQuit Plus reported one psychiatric SAE that could potentially be related to the intervention.
In general, these findings suggest that digital interventions represent a suitable and promising alternative to promote smoking cessation in patients with SMI, although larger RCTs are clearly needed. Overall results are in line with similar reviews of digital interventions for smoking cessation in patients without SMI. Regmi, Kassim, Ahmad, and Tuah (Reference Regmi, Kassim, Ahmad and Tuah2017) reported an improvement in quitting rates among smokers, although it was lower than in our findings (13–24%). This might be explained by the fact that in their review, most of the subjects were recruited through social media signs and advertisement. This could bias the recruitment: taking on less motivated subjects than the clinical populations included in the articles of our review may lessen quitting rates. In a 2019 update of their review, the Cochrane Tobacco Addiction Group found discouraging results for smoking cessation smartphone apps, reporting no evidence of benefit of these digital interventions for smoking cessation (Whittaker et al., Reference Whittaker, McRobbie, Bullen, Rodgers, Gu and Dobson2019). Again, recruitment methods (online, through app stores) and the use of non-clinical populations might explain the discrepancies with our findings. The quality of the studies conducted on non-SMI population was deemed below average by the recent reviews on the matter (Chu et al., Reference Chu, Matheny, Escobar-Viera, Wessel, Notier and Davis2021; Regmi et al., Reference Regmi, Kassim, Ahmad and Tuah2017; Whittaker et al., Reference Whittaker, McRobbie, Bullen, Rodgers, Gu and Dobson2019), and authors agreed on the need for more uniformity when conducting trials in this field. Using less stringent criteria due to the exploratory purpose of most of the included studies in our review, we found that the quality of the studies was fair to good, although more robust methodology is needed in order to be able to generalize these results and provide specific recommendations.
In the addiction field more broadly, some reviews have also found optimistic results on the use of digital medicine approaches for the treatment of methamphetamine (Rubenis, Baker, & Arunogiri, Reference Rubenis, Baker and Arunogiri2021) and alcohol (Colbert, Thornton, & Richmond, Reference Colbert, Thornton and Richmond2020) use disorders, albeit a nascent field, and methodological quality remains a significant barrier. Comparable results in terms of efficacy and acceptability were found in other areas of healthcare, such as cardiac rehabilitation (Wongvibulsin et al., Reference Wongvibulsin, Habeos, Huynh, Xun, Shan, Porosnicu Rodriguez and Martin2021) or maternal health and pregnancy (DeNicola et al., Reference DeNicola, Grossman, Marko, Sonalkar, Butler Tobah, Ganju and Lowery2020; Overdijkink et al., Reference Overdijkink, Velu, Rosman, van Beukering, Kok and Steegers-Theunissen2018), indicating the meaningful opportunity for digital medicine while highlighting the need for larger and methodologically comparable studies.
The results from this review should be interpreted with caution. Due to the different methodologies and ways of reporting effectiveness, acceptability, usability and safety, we are unable to provide a definite recommendation on which specific digital intervention should be used in clinical practice. Additionally, the relatively small number of participants in each study hinders the possibility of performing an adequately powered meta-analysis of the results. While mSMART MIND and iCOMMIT seem to be the most effective interventions in terms of abstinence, they are not available for public use in the most common operating systems, which deems its immediate application impossible. It is also important to note that although this review identified a total of nine digital smoking cessations trialed in SMI, only two of these (QuitGuide and quitStart) are readily available online, which showed poorer results in terms of effectiveness and have not been assessed as thoroughly as mSMART MIND and iCOMMIT for acceptability and safety. The general inaccessibility of evidence-based digital interventions for SMI is a problem which extends far beyond smoking cessation, with a recent review finding that only ~15% of apps trialing in schizophrenia are currently available online (Kwon, Firth, Joshi, & Torous, Reference Kwon, Firth, Joshi and Torous2022). Clearly, further work is needed to improve the dissemination of effective digital approaches into the public sphere more generally.
In terms of effective strategies for smoking cessation in SMI, the results reported by both mSMART MIND and iCOMMIT appear to indicate that mCM could be the most effective approach. Nonetheless, future trials in this field should report standardized measures that allow a comprehensive comparison of effectiveness among different strategies, particularly because delivery of cash monetary reinforcement might be challenging in many settings (e.g.: public health systems, private practices), whereas a digitally delivered reinforcement may easily overcome such limitations. Moreover, results from the Learn to Quit and WebQuit Plus trials highlight the need for simplified interfaces to improve acceptability and usability in SMI. In this regard, the idiosyncrasies of patients with SSD, BD, and MDD must be considered, and research efforts targeted to each individual disorder must be developed. For instance, nicotine has been reported to affect neurocognitive performance in patients with schizophrenia, but not in patients with BD and MDD (Morisano et al., Reference Morisano, Wing, Sacco, Arenovich and George2013). Also, patients with SSD exhibit impaired cognitive performance (Green, Kern, Braff, & Mintz, Reference Green, Kern, Braff and Mintz2000; Mallet et al., Reference Mallet, Godin, Dansou, Mazer, Scognamiglio, Berna and Dubertret2022), which might impact their ability to use digital interventions. Considering this, patients with SSD might face harder cognitive challenges than patients with BD and MDD when quitting smoking, and thus interventions tailored to their ability might be necessary.
Another limitation when considering the results of this review is that all the digital interventions were studied in the US and in stable outpatients, which may limit the generizability of the results to other settings and patient populations. Given that most digital interventions were not readily available to be systematically evaluated with a standardized evaluation scale, results from our assessments should be cautiously considered. Further, reports of AEs in behavioral interventions were scarce and not systematically reported, in contrast with other types of clinical trials, which could lead to the misguided conclusion that non-pharmacological interventions are without risk. As described in previous literature, AEs are also present in both psychotherapeutic and behavioral interventions and should be systematically defined and reported to avoid harming in other patients (Linden & Schermuly-Haupt, Reference Linden and Schermuly-Haupt2014). Finally, concerns regarding cybersecurity should be monitored in the future to guarantee privacy and confidentiality to the users.
In conclusion, digital health appears to be a promising field to develop and further validate interventions for smoking cessation, which should also be tailored to patients with SMI. While some interventions have shown positive preliminary data, more research is needed to provide clinicians and patients with robust recommendations to treat tobacco use with digital interventions.
Supplementary material
The supplementary material for this article can be found at https://doi.org/10.1017/S003329172300123X.
Funding
Dr Martinez Agulleiro was supported by a fellowship funded by Fundacion Alicia Koplowitz (Madrid, Spain). Dr Guinart received funding from Instituto de Salud Carlos III (CM21/00033). Dr Firth was supported by a UK Research and Innovation Future Leaders Fellowship (MR/T021780/1).
Acknowledgements
The authors thank Uxia Gutierrez Couto for her help with article search and database screening.
Conflicts of interest
Dr Fonseca has been a consultant for and/or has received speaker honoraria from Otsuka, Lundbeck, Camurus, Servier, Angelini, Gilead, Abbvie and Esteve. Dr Guinart has been a consultant for and/or has received speaker honoraria from Otsuka America Pharmaceuticals, Janssen Pharmaceuticals, Lundbeck and Teva. Dr Kane has been a consultant to or has received honoraria from Alkermes, Allergan, Click Ther., Dainippon Sumitomo, H. Lundbeck, Indivior, Intracellular Therapies, Janssen Pharmaceutical, Johnson & Johnson, Karuna, LB Pharmaceuticals, Merck, Minerva, Nuerocrine, Newron, Novartis Pharmaceuticals, Otsuka, Roche, Saladax, Sunovion, and Teva; grant support from Otsuka, Lundbeck, Sunovion, and Janssen; shareholders from Vanguard Research Group, LB Pharmaceuticals Inc., North Shore Therapeutics, and HealthRhythms; and royalties from Up to Date. Dr Torrens has been a consultant to and/or has received speaker honoraria from Otsuka, Lundbeck, Camurus, Servier, Angelini, and Molteni. The rest of the authors declare no competing interests.
Open access
