Introduction
Along with poor mental health, people with severe mental illness (SMI), such as bipolar disorder, schizophrenia, and other psychotic disorders, show elevated risks of engaging in adverse health behaviours (Carney, Cotter, Bradshaw, Firth, & Yung, Reference Carney, Cotter, Bradshaw, Firth and Yung2016; Firth et al., Reference Firth, Siddiqi, Koyanagi, Siskind, Rosenbaum, Galletly and Carvalho2019). For example, in comparison with the general population people with SMI are more likely to smoke cigarettes (Prochaska, Das, & Young-Wolff, Reference Prochaska, Das and Young-Wolff2017), are less physically active, and have higher daily calorie and sodium intake (Teasdale et al., Reference Teasdale, Ward, Samaras, Firth, Stubbs, Tripodi and Burrows2019; Vancampfort et al., Reference Vancampfort, Firth, Schuch, Rosenbaum, Mugisha, Hallgren and De Hert2017). This may be partly attributable to the psychotropic medications used to treat SMI, as antipsychotics have been found to increase appetite, delay satiety signalling, and cause sedation (Mazereel, Detraux, Vancampfort, Van Winkel, & De Hert, Reference Mazereel, Detraux, Vancampfort, Van Winkel and De Hert2020). Finding novel ways to promote healthy lifestyles in SMI is crucial for reducing morbidity and mortality (Firth et al., Reference Firth, Siddiqi, Koyanagi, Siskind, Rosenbaum, Galletly and Carvalho2019), with increasing evidence to suggest this could also improve mental health outcomes (Firth et al., Reference Firth, Solmi, Wootton, Vancampfort, Schuch, Hoare and Jackson2020; Pape, Adriaanse, Kol, van Straten, & van Meijel, Reference Pape, Adriaanse, Kol, van Straten and van Meijel2022).
Health behaviour change (HBC) interventions include a broad range of psychological techniques, and target modifiable health behaviours such as diet, physical activity, smoking, sleep, substance or alcohol use, and medication adherence. Traditional face-to-face HBC, while ideal in many respects, interventions are resource intensive (Bennett & Glasgow, Reference Bennett and Glasgow2009) and can be impacted by the capability and capacity of the person delivering the intervention. Interest in online HBC (web-based and smartphone) has grown rapidly in popularity (Arigo et al., Reference Arigo, Jake-Schoffman, Wolin, Beckjord, Hekler and Pagoto2019), given their potential to improve access to HBC for people with SMI, without relying on costly face-to-face interventions (Young et al., Reference Young, Cohen, Goldberg, Hellemann, Kreyenbuhl, Niv and Whelan2017). Previously, there have been concerns that people with SMI may experience socio-economic barriers – such as unstable housing, low income, and unemployment – which may limit their access to the internet and online interventions (Borzekowski et al., Reference Borzekowski, Leith, Medoff, Potts, Dixon, Balis and Himelhoch2009). Encouragingly however, smartphone and internet use is increasing among those with SMI (Firth et al., Reference Firth, Cotter, Torous, Bucci, Firth and Yung2016; Thomas, Foley, Lindblom, & Lee, Reference Thomas, Foley, Lindblom and Lee2017; Trefflich, Kalckreuth, Mergl, & Rummel-Kluge, Reference Trefflich, Kalckreuth, Mergl and Rummel-Kluge2015).
While previous reviews have focused on the feasibility and acceptability of digital interventions generally for symptom management and relapse prevention in SMI (Naslund, Marsch, McHugo, & Bartels, Reference Naslund, Marsch, McHugo and Bartels2015b), there is still limited understanding of how digital HBC could work to improve outcomes specifically in this population. Therefore, this review aimed to systematically identify and evaluate the current evidence around the feasibility, acceptability, and effectiveness of digital HBC for not only physical health, but also broader behavioural and psychological well-being outcomes, in people with SMI.
Specifically, this review addressed the following research questions: (i) are digital approaches towards delivering HBC feasible and acceptable for people with SMI?; (ii) can digital HBC for people with SMI change health-related behaviour?; (iii) can digital HBC for people with SMI improve health outcomes?; and (iv) what specific intervention components and strategies influence user engagement with digital HBC interventions in people with SMI?
Methods
The Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) checklist for reporting systematic reviews (Moher, Liberati, Tetzlaff, Altman, & PRISMA Group, Reference Moher, Liberati, Tetzlaff and Altman2009) guided this review, which was pre-registered on the online review protocol database, PROSPERO (CRD42021261267).
Search strategy
A systematic literature search was conducted in January 2022 using the following databases: Cochrane Central Register of Controlled Trials; Health Technology Assessment; AMED (Allied and Complementary Medicine); APA PsycInfo; Embase; and MEDLINE®, using the following keyword search algorithm: [psychosis OR psychotic OR schizophr* OR severe mental OR serious mental OR bipolar] AND [Behaviour change OR Behavior change OR behavioural change OR behavioral change OR Lifestyle OR Healthy Living OR Health Behaviour OR Health Behavior OR physical activity OR exercise OR smok* OR tobacco OR sexual health OR Sleep OR Alcoho* OR diet* OR Sedentary OR substance abuse OR weightloss OR weight loss OR obes*] AND [online or web-based or app-based or Internet or e-Health or mhealth or smartphone or mobile phone or iphone or android or wearable or digital].
Searches were restricted to publication in English language in peer-reviewed journals and all articles were included regardless of publication date. Reference and citation list searches were also conducted to search for additional studies, alongside a basic search of Google Scholar and the Journal of Medical Internet Research (JMIR).
Eligibility criteria
English language articles were included. Randomised controlled trials (RCTs), non-RCTs, pilot studies, feasibility studies, quasi-experimental studies, and qualitative studies examining the feasibility, acceptability, or effectiveness and other outcomes of a digital HBC, delivered online via computer smartphone apps, social media and/or ‘wearable’ formats, among people with SMI were eligible.
For the purpose of this review, ‘SMI populations’ included any groups of individuals (of any age) diagnosed and/or receiving treatment for bipolar or psychotic disorders. Studies of non-entirely SMI samples were included, if either (i) where data pertaining to the SMI sub-sample were reported separately, or, (ii) where the overall sample contained over two-thirds of individuals diagnosed/treated for SMI.
Studies that reported changes in health behaviours relevant to physical health and overall well-being (such as smoking, substance use, sleeping, diet, physical activity, and sexual behaviours) as primary or secondary outcomes were included. Studies will be eligible for inclusion if they deliver a behavioural change intervention fully online, or where the digital technology forms a well-defined and central part of a multi-component intervention. Interventions in which the HBC aspect only made up a tangential or minor part of the intervention, or where relevant technological aspects were limited to text messages, emails or phone calls, were excluded.
Study selection process
Initial screening of titles and abstracts was conducted by one reviewer (C.S.). Two reviewers (C.S. and G.M.), who were blind to each other's review, independently reviewed all full-text articles meeting the inclusion criteria (interrater agreement: 83%). A third reviewer (J.F.) resolved any disagreements between the two reviewers.
Data extraction
Data were independently extracted by two reviewers (C.S. and G.M.), using a pre-determined data extraction form specifically designed for this review. The extraction form collected the following data: (i) study information (sample size, mean age of participants, diagnostic information, and study design); (ii) intervention features (intervention platform, app/programme name, trial/feasibility details, regularity of instructed use, digital intervention summary, any additional intervention components, and details of the control condition); and (iii) effects on behaviour and health outcomes (changes in behaviour, changes in physical and/or mental health before and after interventions).
Given there are no established standards for assessing feasibility, acceptability, and usability (Greenhalgh et al., Reference Greenhalgh, Wherton, Papoutsi, Lynch, Hughes, Hinder and Shaw2017; Jacob, Sezgin, Sanchez-Vazquez, & Ivory, Reference Jacob, Sezgin, Sanchez-Vazquez and Ivory2022), measures were chosen from validated scales [e.g. the System Usability Scale (SUS) (Hyzy et al., Reference Hyzy, Bond, Mulvenna, Bai, Dix, Leigh and Hunt2022)] and in line with prior research (Balaskas, Doherty, Schueller, & Cox, Reference Balaskas, Doherty, Schueller and Cox2021). Feasibility measures included recruitment rates, attrition to study, reasons for refusal or ineligibility, and adherence to intervention. Acceptability measures included usage data (e.g. duration of use, modules completed, etc.) and user perspectives from interviews and quantitative questionnaires (e.g. regarding relevance of content and readability). Usability measures included the SUS, task scores, and interview comments regarding design, layout, and/or other aspects of the user interface. Finally, behavioural outcomes included intervention effect sizes and/or changes in target behaviour, depression, and/or anxiety.
Data synthesis
Due to variations in design, intervention approaches, and primary outcome measures it was not appropriate to conduct a meta-analysis and therefore quantitative findings were synthesised narratively (Liberati et al., Reference Liberati, Altman, Tetzlaff, Mulrow, Gøtzsche, Ioannidis and Moher2009). For qualitative studies, themes were identified and the detailed analytical narratives reported within the text were summarised, following the principles of thematic synthesis (Thomas & Harden, Reference Thomas and Harden2008). Mixed methods studies contributed separately to both types of synthesis. The data in each section were different and were therefore not double counted.
Results
Overall, 2196 results were identified from the databases. Figure 1 presents a PRISMA flowchart of study selection procedures. After removing duplicates (N = 262), 1934 titles and abstracts were screened. Following initial screening, 1827 articles were removed, leaving 107 full texts to be reviewed. A further 76 were then excluded, leaving 31 papers. Five additional articles were found through separate literature searches of reference lists and Google Scholar, yielding 36 papers in total.
Study characteristics
Tables 1–3 summarise the characteristics of selected studies, including descriptions of participants and interventions. There were 14 pilot RCTs, 14 non-RCTs, three qualitative studies, and five mixed-method studies. The majority (N = 35) of studies were conducted in the United States, with only one study conducted in the UK, Australia, Portugal, Netherlands, and Canada. Fourteen of the studies consisted of mainly White participants. Sample sizes ranged from 5 to 2570 participants. Schizophrenia was the most common diagnosis in smoking and physical activity studies, whereas bipolar disorder was the most common diagnosis for ‘other behaviours’ (Tables 1–3).
BD, bipolar disorder; CBT, cognitive behavioural therapy; FTDN, Fagerström test for nicotine dependence; LTAS, Lets Talk About Smoking; LTQ, Learn To Quit; MDD, major depressive disorder; MH, mental health; NA, not applicable; NCI, National Cancer Institute; NR, not reported; NRT, nicotine replacement therapy; NS, not significant; PNTS, prefer not to say; PNOS, psychosis not otherwise specified; PPA, point prevalence abstinence; QG, quit guide; RCT, randomised controlled trial; RMD, recurrent major depression; s.d., standard deviation; SMI, serious mental illness; SZ, schizophrenia; SZ-AFF, schizoaffective disorder; SUS, System Usability Scale; TAU, treatment-as-usual; mcm, smartphone-based application contingency management, iCOMMIT, multi-component mobile-enhanced treatment for smoking cessation.
BD, bipolar disorder; BMI, body mass index; C, control; CBT, cognitive behavioural therapy; CVD, cardiovascular disorder; I, intervention; lbs, pounds; MCBT, mindfulness cognitive behavioural therapy; MH, mental health; mHealth, mobile health; PA, physical activity; RCT, randomised controlled trial; s.d., standard deviation; SMART, specific measurable attainable realistic and timely; SMI, severe mental illness; SZ, schizophrenia; SZ-AFF, schizoaffective treatment-as-usual (TAU); 1×/week, once a week; 2×/week, twice a week.
BD, bipolar disorder; e-platform, electronic platform; FU, follow-up; HI, high intensity; IQR, inter-quartile range; LI, low intensity; MDD, major depressive disorder; NS, non-significant; OMI, other mental illness; s.e.m., standard error of mean; SUD, substance use disorder; SZ, schizophrenia; SZ-AFF, schizoaffective disorder; TAU, treatment-as-usual; TAU + TES, treatment-as-usual + therapeutic education system.
The HBC used included those delivered entirely digitally and those using ‘multi-component’ approaches (i.e. digital and in-person aspects). Tables 4–6 present study outcomes and interventions. Nineteen studies focused on smoking as the primary behavioural outcome (Tables 1 and 4). Fourteen studies focused on physical activity, weight loss, and cardio-metabolic health (Tables 2 and 5) and three papers focused on ‘other behaviours’ (Table 6), specifically sleep (Taylor, Bradley, & Cella, Reference Taylor, Bradley and Cella2022), substance use (Hammond, Antoine, Stitzer, & Strain, Reference Hammond, Antoine, Stitzer and Strain2020), and invoking changes in the perceived benefit of changing health behaviours, rather than the behaviour itself (Melamed et al., Reference Melamed, Voineskos, Vojtila, Ashfaq, Veldhuizen, Dragonetti and Selby2022). No HBC targeted sexual health.
AEs, adverse events; CPD, cigarettes per day; LTAS, Lets Talk About Smoking; LTQ, learn to quit; MH, mental health; NCI, National Cancer Institute; NS, NOT SIGNIFICANT; PNTS, prefer not to say; PPA, point prevalence abstinence; QG, quit guide; RCT, randomised controlled trial; SAEs, serious adverse events; s.d., standard deviation; SMI, severe mental illness; SUS, System Usability Scale; TAU, treatment-as-usual.
BMI, body mass index; C, control; CBT, cognitive behavioural therapy; CVD, cardiovascular disorder; I, intervention; lbs, pounds; MCBT, mindfulness cognitive behavioural therapy; MH, mental health; mHealth, mobile health; PA, physical activity; RCT, randomised controlled trial; s.d., standard deviation; SMART, specific measurable attainable realistic and timely; SMI, severe mental illness; SUS, system usability scale; TAU, treatment-as-usual; 1×/week, once a week; 2×/week, twice a week.
e-platform, electronic platform; FU, follow-up; HI, high intensity; IQR, inter-quartile range; LI, low intensity; OMI, other mental illness; NR, not reported; NS, not significant; s.e.m., standard error of mean; SUD, substance use disorder; TAU, treatment-as-usual; TAU + TES, treatment-as-usual + therapeutic education system.
Overview of included studies
Results of the included studies were synthesised in their respective classes of health behaviours, namely: (i) smoking; (ii) physical activity, weight loss, and cardio-metabolic health; and (iii) other behaviours. For each HBC class, feasibility, acceptability, usability, and impacts on behaviour/outcomes were summarised.
Smoking
Eight studies delivered HBC through smartphone apps and nine used web-based interventions. The remaining two were multi-component interventions (Table 1).
Feasibility. Recruitment rates ranged from 13% to 91% across studies (Table 4). Overall, attrition was low, ranging from 0% to 23%; common reasons for dropout included hospitalisation and loss of interest. Adherence to digital interventions was generally high, ranging from 43% to 100% (Table 4).
Acceptability. Apps developed for those with an SMI had greater engagement when compared with apps for the general population (Browne, Halverson, & Vilardaga, Reference Browne, Halverson and Vilardaga2021; Vilardaga et al., Reference Vilardaga, Rizo, Ries, Kientz, Ziedonis, Hernandez and McClernon2019, Reference Vilardaga, Rizo, Palenski, Mannelli, Oliver and Mcclernon2020). One website, Lets Talk About Smoking (LTAS), was developed specifically for individuals with SMI, providing interactive tailored smoking cessation advice. This scored more highly on patient satisfaction when compared to users of static National Cancer Institute (NCI) patient education handout, which was developed for the general population (Brunette et al., Reference Brunette, Ferron, Robinson, Coletti, Geiger, Devitt and Greene2018).
Participants reported links between symptom severity and smoking, and saw benefits in tracking smoking alongside their mental health (Klein, Lawn, Tsourtos, & van Agteren, Reference Klein, Lawn, Tsourtos and van Agteren2019). Real-time support, such as a person or distraction task, was deemed essential to help with cravings (Klein et al., Reference Klein, Lawn, Tsourtos and van Agteren2019).
Usability. People with SMI viewed easy navigation and engaging content and design as preferable, or even essential, for digital HBC (Brunette et al., Reference Brunette, Ferron, McHugo, Davis, Devitt, Wilkness and Drake2011; Klein et al., Reference Klein, Lawn, Tsourtos and van Agteren2019; Vilardaga et al., Reference Vilardaga, Rizo, Kientz, McDonell, Ries and Sobel2016, Reference Vilardaga, Rizo, Zeng, Kientz, Ries, Otis and Hernandez2018). Issues were reported with readability, difficulty using support chatrooms, and navigation for certain websites or apps, particularly if pages had multiple functions. Difficulties simultaneously filming and uploading carbon monoxide readings were also reported (Wilson et al., Reference Wilson, Thompson, Currence, Thomas, Dedert, Kirby and Beckham2019). In Brunette et al.'s (Reference Brunette, Ferron, Devitt, Geiger, Martin, Pratt and McHugo2012) study, four websites developed for the general population were difficult to use among people with SMI who had less experience using computers. Promisingly all participants learnt to use the website developed specifically for SMI populations, regardless of experience, with minimal training (one to three sessions) (Brunette, Ferron, Gottlieb, Devitt, & Rotondi, Reference Brunette, Ferron, Gottlieb, Devitt and Rotondi2016).
Several apps developed for the general population (QuitGuide, quitSTART, QuitPal) scored below acceptable standards on the ‘System Usability Scale’ in some studies. In Vilardaga et al.'s (Reference Vilardaga, Rizo, Ries, Kientz, Ziedonis, Hernandez and McClernon2019) study, both QuitGuide and the ‘Learn to Quit’ (LTQ) app – developed for people with SMI – met usability cut-offs as rated by two (all) participants. It is worth noting that when detecting usability problems, studies using samples as small as five can be deemed acceptable (Lewis, Reference Lewis1994).
Behaviour and health outcomes. All 13 studies, which evaluated digital HBCs' impact on smoking behaviours, found self-reported smoking reductions (Table 4). Five studies confirmed smoking abstinence through biochemical verification (Table 4). Unpromisingly, in one intervention, which took a multi-component approach, self-reported 7-day point prevalence abstinence decreased from 38–40% (from both cohorts) to 9.4% in the pilot RCT (Medenblik et al., Reference Medenblik, Mann, Beaver, Dedert, Wilson, Calhoun and Beckham2020; Wilson et al., Reference Wilson, Thompson, Currence, Thomas, Dedert, Kirby and Beckham2019).
Notably, Brunette et al. (Reference Brunette, Ferron, Robinson, Coletti, Geiger, Devitt and Greene2018) found, after 3 months, greater percentage of LTAS participants had biologically verified abstinence, compared to the NCI education group. Additionally another app developed for people with SMI (LTQ) was more effective in promoting smoking cessation, with those assigned to the QuitGuide app (developed for the general population) making more quit attempts and subsequently more relapses (Browne et al., Reference Browne, Halverson and Vilardaga2021; Vilardaga et al., Reference Vilardaga, Rizo, Palenski, Mannelli, Oliver and Mcclernon2020).
Only two studies measured mental health outcomes. Heffner et al. (Reference Heffner, Kelly, Waxmonsky, Mattocks, Serfozo, Bricker and Ostacher2020) reported a potential improvement in depression and mania scores, while Vilardaga et al. (Reference Vilardaga, Rizo, Palenski, Mannelli, Oliver and Mcclernon2020) demonstrated a reduction in negative symptoms and a small non-significant reduction in depression, anxiety, and symptom severity across both groups.
Physical activity, weight loss, and cardio-metabolic health
One study delivered HBC through a smartphone app (WellWave), two studies used a web-based intervention and one study used an app with an associated wearable device. Seven studies used a multi-component approach (Table 2). Three studies compared web-based interventions with in-person interventions (Muralidharan et al., Reference Muralidharan, Niv, Brown, Olmos-Ochoa, Fang, Cohen and Young2018; Olmos-Ochoa et al., Reference Olmos-Ochoa, Niv, Hellemann, Cohen, Oberman, Goldberg and Young2019; Young et al., Reference Young, Cohen, Goldberg, Hellemann, Kreyenbuhl, Niv and Whelan2017).
Feasibility. Recruitment rates varied across studies, with studies recruiting from veteran centres reporting lower rates (19% participated and 58% were ineligible) than studies recruiting from mental healthcare services (42–45% participated and 28% were ineligible; Table 5). The highest recruitment rates were observed in studies conducted through outpatient clinics with subsequent participation not involving additional in-person sessions (Campos et al., Reference Campos, Mesquita, Marques, Trigueiro, Orvalho and Rocha2015; Looijmans, Jörg, Bruggeman, Schoevers, & Corpeleijn, Reference Looijmans, Jörg, Bruggeman, Schoevers and Corpeleijn2019).
Where reported, overall retention was high, with 75–90% of participant completing follow-up measures at the final time point, which ranged from 1 to 6 months (Table 5). Retention rates dropped after 12 months, to around 33% (Aschbrenner et al., Reference Aschbrenner, Naslund, Gorin, Mueser, Browne, Wolfe and Bartels2021; Looijmans et al., Reference Looijmans, Jörg, Bruggeman, Schoevers and Corpeleijn2019). Reasons for dropout included health concerns, hospitalisation, and competing time commitment.
Levels of adherence were generally high, particularly with digital components of the studies (Table 5). In one study of 32 participants, the in-person exercise sessions achieved only a 28% attendance rate, while 100% and 76% used the provided Fitbit and private Facebook group, respectively (Aschbrenner, Naslund, Shevenell, Kinney, & Bartels, Reference Aschbrenner, Naslund, Shevenell, Kinney and Bartels2016a). Similar findings were reported in Aschbrenner et al.'s (Reference Aschbrenner, Naslund, Gorin, Mueser, Browne, Wolfe and Bartels2021) study, with 70% of PeerFit participants attending at least one in-person exercise session, while 97% of BEAT participants attended at least one online coaching session (Aschbrenner et al., Reference Aschbrenner, Naslund, Gorin, Mueser, Browne, Wolfe and Bartels2021).
Acceptability. Usage of digital interventions was also generally high (Table 5); in Muralidharan et al. (Reference Muralidharan, Niv, Brown, Olmos-Ochoa, Fang, Cohen and Young2018), Olmos-Ochoa et al. (Reference Olmos-Ochoa, Niv, Hellemann, Cohen, Oberman, Goldberg and Young2019), and Young et al. (Reference Young, Cohen, Goldberg, Hellemann, Kreyenbuhl, Niv and Whelan2017), more modules were completed by those in the digital v. the in-person arm.
Feedback indicated that peer interaction, particularly interacting with peer coaches and learning about others experiences, seemed to be a popular component of interventions among patients. Conversely, the main barriers to use were physical limitations and pain and, when attending in-person sessions, time constraints and travel burden (Aschbrenner et al., Reference Aschbrenner, Naslund, Barre, Mueser, Kinney and Bartels2015; Muralidharan et al., Reference Muralidharan, Niv, Brown, Olmos-Ochoa, Fang, Cohen and Young2018; Olmos-Ochoa et al., Reference Olmos-Ochoa, Niv, Hellemann, Cohen, Oberman, Goldberg and Young2019). Some participants attending in-person sessions also found it difficult to engage with new people (Olmos-Ochoa et al., Reference Olmos-Ochoa, Niv, Hellemann, Cohen, Oberman, Goldberg and Young2019; Young et al., Reference Young, Cohen, Goldberg, Hellemann, Kreyenbuhl, Niv and Whelan2017). Other less commonly reported barriers included concerns about their environment and safety (for in-person interventions), financial barriers, control over food preparation, and lack of support from others (Olmos-Ochoa et al., Reference Olmos-Ochoa, Niv, Hellemann, Cohen, Oberman, Goldberg and Young2019). Concerning wearables, participants found them helpful for setting goals, motivation, and useful for self-monitoring (Aschbrenner et al., Reference Aschbrenner, Naslund, Barre, Mueser, Kinney and Bartels2015; Naslund, Aschbrenner, Barre, & Bartels, Reference Naslund, Aschbrenner, Barre and Bartels2015a; Naslund, Aschbrenner, & Bartels, Reference Naslund, Aschbrenner and Bartels2016). Some participants did experience frustration due to forgetting to wear pedometers (Young et al., Reference Young, Cohen, Goldberg, Hellemann, Kreyenbuhl, Niv and Whelan2017) and the cost of wearables was identified as a barrier in one study (Naslund et al., Reference Naslund, Aschbrenner, Barre and Bartels2015a, Reference Naslund, Marsch, McHugo and Bartels2015b).
Usability. None of the studies measured usability using formally validated scales, complicating evaluation. Some participants did comment that they found wearables easy to use (Naslund et al., Reference Naslund, Aschbrenner, Barre and Bartels2015a, Reference Naslund, Marsch, McHugo and Bartels2015b). Notably some participants did experience technical issues when using equipment (Young et al., Reference Young, Cohen, Goldberg, Hellemann, Kreyenbuhl, Niv and Whelan2017) or logging into digital interfaces for the first time. Peer coaches were noted as helpful in combatting such issues. Of note, in an intervention that involved participants playing physically active video games, 69% of participants completed the intervention using Kinect, although 85% reported would not have done so without technical support (Campos et al., Reference Campos, Mesquita, Marques, Trigueiro, Orvalho and Rocha2015). Thus suggesting that without support this is not acceptable for those with an SMI and technical support would be required for real-world implementation in mental health settings (Campos et al., Reference Campos, Mesquita, Marques, Trigueiro, Orvalho and Rocha2015).
Behaviour and health outcomes. Nine studies assessed the impact of the digital interventions on physical activity and/or weight loss (Table 5). Five studies showed at least some promising results, with two studies in particular reporting participants lost at least 5% of their body weight and clinically significant reductions in cardiovascular risk (⩾5% weight loss or improved fitness) (Aschbrenner et al., Reference Aschbrenner, Naslund, Shevenell, Kinney and Bartels2016a, Reference Aschbrenner, Naslund, Gorin, Mueser, Browne, Wolfe and Bartels2021). Additionally Aschbrenner et al. (Reference Aschbrenner, Naslund, Shevenell, Kinney and Bartels2016a, Reference Aschbrenner, Naslund, Shevenell, Mueser and Bartels2016b) reported 17% of participants showed clinical significant improvements in cardiovascular fitness. Two interventions lead to increases in physical activity (Macias et al., Reference Macias, Panch, Hicks, Scolnick, Weene, Öngür and Cohen2015; Muralidharan et al., Reference Muralidharan, Niv, Brown, Olmos-Ochoa, Fang, Cohen and Young2018). Only one study, which used a web tool designed to help patients set goals, monitor their progress, and receive feedback via a mental health nurse (Looijmans et al., Reference Looijmans, Jörg, Bruggeman, Schoevers and Corpeleijn2019) found no significant reductions in body mass index (BMI) or waist circumference at 6/12-month follow-ups.
Papers investigating a digital intervention called ‘WebMOVE’ (Muralidharan et al., Reference Muralidharan, Niv, Brown, Olmos-Ochoa, Fang, Cohen and Young2018; Young et al., Reference Young, Cohen, Goldberg, Hellemann, Kreyenbuhl, Niv and Whelan2017) reported more weight loss than the in-person comparator intervention (MOVE-SMI). Both provided pedometers and access to peer coaches and comprised of the same educational content, differing only in delivery mode. However, in another study, both individual mHealth coaching and in-person HBC were similarly effective; with both groups achieving clinically significant weight loss and reduction in cardiovascular risk at 6 and 12 months (Aschbrenner et al., Reference Aschbrenner, Naslund, Gorin, Mueser, Browne, Wolfe and Bartels2021).
Only one study looked at the effect of digital HBC on mental health (Campos et al., Reference Campos, Mesquita, Marques, Trigueiro, Orvalho and Rocha2015) and found slight, non-significant improvements in these domains.
Other health behaviours
Two studies delivered HBC through web-based interventions, with one promoting the treatment for substance use disorder (SUD) (Hammond et al., Reference Hammond, Antoine, Stitzer and Strain2020) and the other changing attitudes towards health behaviours as a route to behavioural changes (Melamed et al., Reference Melamed, Voineskos, Vojtila, Ashfaq, Veldhuizen, Dragonetti and Selby2022). One study targeted sleep, which used an app (Taylor et al., Reference Taylor, Bradley and Cella2022).
Feasibility, acceptability, effectiveness, and outcomes of other interventions. Across the three studies, recruitment appeared to be challenging, with issues around screening and ineligibility (Table 6). Retention for the primary endpoint of theses interventions was excellent, ranging from 93% to 97%, though longer term follow-up (24 weeks) dropped to 40% in one study (Melamed et al., Reference Melamed, Voineskos, Vojtila, Ashfaq, Veldhuizen, Dragonetti and Selby2022).
Adherence to interventions varied across the three studies (58–100%). Adherence was highest for the sleep intervention. Overall participants had positive experiences, but many felt the 6-week intervention was not long enough and needed more variety of content and games (Taylor et al., Reference Taylor, Bradley and Cella2022).
The app-based sleep intervention had a large effect on behaviour (sleep) and a small-to-medium effect on mental health (Taylor et al., Reference Taylor, Bradley and Cella2022). The attitude-focused intervention led to positive changes in individual attitudes but did not ultimately change behaviours (Melamed et al., Reference Melamed, Voineskos, Vojtila, Ashfaq, Veldhuizen, Dragonetti and Selby2022).
The SUD intervention was rated highly across several measures, including acceptability (Hammond et al., Reference Hammond, Antoine, Stitzer and Strain2020). At the end of the web intervention period, similar rates of participants had enrolled in SUD treatment, at 30 days post discharge, as that observed under treatment-as-usual conditions.
Discussion
This paper reviewed 36 studies and systematically identified 29 digital HBC (with overlap of components for some of the physical activity interventions) for people with SMI. Feasibility, acceptability, and outcomes of interventions were evaluated and intervention components and strategies which were preferred by people with SMI were identified. Overall, 70% of the studies established support for the acceptability and/or feasibility of digital behavioural change interventions. However, themes around the need for human support for both digital literacy/navigation and engagement were common across all clinical targets. Given the pilot nature of studies and the heterogeneous outcomes, it is not possible to determine an effect size estimate, but current evidence shows that these interventions do have the potential to change health behaviours.
Across the studies reviewed, there was a relatively consistent result that digital interventions to change behaviours are both feasible and acceptable for use among people with SMI. This is an important finding, due to the large health disparity among this group and insufficient resources in mental healthcare settings to provide lifestyle interventions in mental healthcare settings (Firth et al., Reference Firth, Siddiqi, Koyanagi, Siskind, Rosenbaum, Galletly and Carvalho2019). Despite concerns about smartphone use as a main barrier to digital interventions in this population, the majority of participants with SMI reported digital interventions were easy to use and several studies even reported participants completed additional modules or sessions voluntarily (Aschbrenner, Naslund, Shevenell, Mueser, & Bartels, Reference Aschbrenner, Naslund, Shevenell, Mueser and Bartels2016b; Aschbrenner et al., Reference Aschbrenner, Naslund, Shevenell, Kinney and Bartels2016a; Brunette et al., Reference Brunette, Ferron, Gottlieb, Devitt and Rotondi2016, Reference Brunette, Ferron, Robinson, Coletti, Geiger, Devitt and Greene2018). It is important to note some participants did struggle with accessibility, internet access, and/or needed additional support, in particular for those with limited experience using technology (Campos et al., Reference Campos, Mesquita, Marques, Trigueiro, Orvalho and Rocha2015; Ferron et al., Reference Ferron, Brunette, McHugo, Devitt, Martin and Drake2011; Naslund et al., Reference Naslund, Aschbrenner and Bartels2016; Olmos-Ochoa et al., Reference Olmos-Ochoa, Niv, Hellemann, Cohen, Oberman, Goldberg and Young2019; Taylor et al., Reference Taylor, Bradley and Cella2022; Vilardaga et al., Reference Vilardaga, Rizo, Kientz, McDonell, Ries and Sobel2016, Reference Vilardaga, Rizo, Ries, Kientz, Ziedonis, Hernandez and McClernon2019; Young et al., Reference Young, Cohen, Goldberg, Hellemann, Kreyenbuhl, Niv and Whelan2017). Promisingly those with little experience using digital platforms could use them after assistance from peer coaches (Olmos-Ochoa et al., Reference Olmos-Ochoa, Niv, Hellemann, Cohen, Oberman, Goldberg and Young2019; Young et al., Reference Young, Cohen, Goldberg, Hellemann, Kreyenbuhl, Niv and Whelan2017) or training sessions (Brunette et al., Reference Brunette, Ferron, Gottlieb, Devitt and Rotondi2016). Therefore, to reduce the digital divide in future, it would be crucial to have human support available in mental healthcare settings to facilitate use, such as digital navigators (Sylvia et al., Reference Sylvia, Faulkner, Rakhilin, Amado, Gold, Albury and Nierenberg2021; Wisniewski, Gorrindo, Rauseo-Ricupero, Hilty, & Torous, Reference Wisniewski, Gorrindo, Rauseo-Ricupero, Hilty and Torous2020; Wisniewski & Torous, Reference Wisniewski and Torous2020).
It is important to mention participants recruited from veteran centres and inpatient settings had high rates of ineligible participants, which may limit the generalisability of the results to other patient populations and settings. Further, some of the digital interventions (such as Microsoft, iCOMMIT, and LTQ) are not publicly readily available to use. Two interventions financially compensated participants for ongoing engagement, which may not be sustainable in real-world healthcare services (Linardon & Fuller-Tyszkiewicz, Reference Linardon and Fuller-Tyszkiewicz2020).
Compared to in-person interventions, digital HBC had benefits such as greater adherence, lower resource intensity, and the potential for non-clinical staff to deliver them (Aschbrenner et al., Reference Aschbrenner, Naslund, Gorin, Mueser, Browne, Wolfe and Bartels2021). Further, the outcomes/changes in behaviour from digital interventions seemed similar to in-person interventions of the same content (Muralidharan et al., Reference Muralidharan, Niv, Brown, Olmos-Ochoa, Fang, Cohen and Young2018; Olmos-Ochoa et al., Reference Olmos-Ochoa, Niv, Hellemann, Cohen, Oberman, Goldberg and Young2019; Young et al., Reference Young, Cohen, Goldberg, Hellemann, Kreyenbuhl, Niv and Whelan2017). Such findings are promising given the lack of capacity in mental healthcare services for in-person HBC (Ayerbe et al., Reference Ayerbe, Forgnone, Foguet-Boreu, González, Addo and Ayis2018; Bailey et al., Reference Bailey, Bartlem, Wiggers, Wye, Stockings, Hodder and Dray2019). However, future work is required to compare the effectiveness of delivering an intervention digitally v. non-digitally to people with SMI.
Therefore, digital HBC are poised to play a crucial role in the near future. Digital interventions can also increase engagement and overcome socioeconomic and barrier issues reported by participants regarding the in-person elements of the multi-component interventions.
Peer/social support – offline and online – was perceived positively among many of the physical activity interventions (Aschbrenner et al., Reference Aschbrenner, Naslund, Barre, Mueser, Kinney and Bartels2015; Macias et al., Reference Macias, Panch, Hicks, Scolnick, Weene, Öngür and Cohen2015; Muralidharan et al., Reference Muralidharan, Niv, Brown, Olmos-Ochoa, Fang, Cohen and Young2018; Young et al., Reference Young, Cohen, Goldberg, Hellemann, Kreyenbuhl, Niv and Whelan2017) and, from the interviews, social support was a strongly desired element for smoking cessation apps (Gowarty, Aschbrenner, & Brunette, Reference Gowarty, Aschbrenner and Brunette2021a; Gowarty et al., Reference Gowarty, Longacre, Vilardaga, Kung, Gaughan-Maher and Brunette2021b; Klein et al., Reference Klein, Lawn, Tsourtos and van Agteren2019). Also, design features and content that made platforms more interactive, usable, and tailored to those with SMI enhanced engagement (Aschbrenner et al., Reference Aschbrenner, Naslund, Shevenell, Mueser and Bartels2016b; Browne et al., Reference Browne, Halverson and Vilardaga2021; Brunette et al., Reference Brunette, Ferron, Gottlieb, Devitt and Rotondi2016, Reference Brunette, Ferron, Robinson, Coletti, Geiger, Devitt and Greene2018, Reference Brunette, Ferron, McGurk, Williams, Harrington, Devitt and Xie2020; Klein et al., Reference Klein, Lawn, Tsourtos and van Agteren2019; Naslund et al., Reference Naslund, Aschbrenner, Barre and Bartels2015a, Reference Naslund, Marsch, McHugo and Bartels2015b; Taylor et al., Reference Taylor, Bradley and Cella2022; Vilardaga et al., Reference Vilardaga, Rizo, Palenski, Mannelli, Oliver and Mcclernon2020).
With regards to behavioural change techniques, it appears that setting goals and reviewing progress may not be enough to change behaviour for people with SMI. Setting diet and physical activity goals, behavioural monitoring, and receiving feedback from health professionals failed to reduce BMI or waist circumference at 6/12-month follow-ups in one study (Looijmans et al., Reference Looijmans, Jörg, Bruggeman, Schoevers and Corpeleijn2019). In contrast, interventions that involved exercise sessions, information about preparing healthy meals, wearables, provided rewards/trophies or had social support, led to weight loss for the majority of participants (Aschbrenner et al., Reference Aschbrenner, Naslund, Shevenell, Kinney and Bartels2016a, Reference Aschbrenner, Naslund, Shevenell, Mueser and Bartels2016b, Reference Aschbrenner, Naslund, Gorin, Mueser, Browne, Wolfe and Bartels2021; Muralidharan et al., Reference Muralidharan, Niv, Brown, Olmos-Ochoa, Fang, Cohen and Young2018; Naslund, Aschbrenner, Marsch, McHugo, & Bartels, Reference Naslund, Aschbrenner, Marsch, McHugo and Bartels2018; Young et al., Reference Young, Cohen, Goldberg, Hellemann, Kreyenbuhl, Niv and Whelan2017). Previous research has shown that demonstrating exercises at home yielded large impacts on physical activity in low-income groups (Bull et al., Reference Bull, McCleary, Li, Dombrowski, Dusseldorp and Johnston2018). Further research would be required to determine the feasibility and acceptability of digital home workouts in people with SMI.
People with SMI appear more amenable to HBC tailored to consider their needs. This review highlighted examples where digital HBC developed for those with SMI were found to have superior outcomes, including higher rates of smoking abstinence and/or greater reduction in cigarettes smoked (Browne et al., Reference Browne, Halverson and Vilardaga2021; Brunette et al., Reference Brunette, Ferron, Robinson, Coletti, Geiger, Devitt and Greene2018), fewer relapses (Vilardaga et al., Reference Vilardaga, Rizo, Palenski, Mannelli, Oliver and Mcclernon2020), and enhanced usability (Brunette et al., Reference Brunette, Ferron, McGurk, Williams, Harrington, Devitt and Xie2020). For example, with smoking interventions, tailoring could mean normalising relapses and integrating their mental health symptomology, while with physical activity interventions considering the physical limitations of people with SMI could be important (Aschbrenner et al., Reference Aschbrenner, Naslund, Barre, Mueser, Kinney and Bartels2015; Klein et al., Reference Klein, Lawn, Tsourtos and van Agteren2019; Muralidharan et al., Reference Muralidharan, Niv, Brown, Olmos-Ochoa, Fang, Cohen and Young2018; Olmos-Ochoa et al., Reference Olmos-Ochoa, Niv, Hellemann, Cohen, Oberman, Goldberg and Young2019).
Strengths and limitations
A strength of this study was the comprehensive nature of the methods, which applied a systematic approach and broad search terms to capturing digital HBC for people with SMI, including various study designs. Although only one reviewer was responsible for an initial screening at title and abstract stage, this was done only to remove the obviously ineligible articles swiftly (i.e. those in which no part of the title or abstract indicated relevance to this review). Any study with an indication of eligibility from title/abstract content was subject to full-text screening, conducted by two reviewers. Given the auxiliary search methods conducted alongside the main search, we are confident this review captures the relevant published literature on this nascent but growing topic. However, a key limitation is that due to the preliminary nature of most studies conducted so far (which were largely focused on feasibility, or pilot studies with small-sample sizes consisting of mostly, if not all, Caucasian participants), the research may be too nascent at present to draw any definitive conclusions on the effectiveness of digital approaches for health promotion in SMI. Additionally due to the short-term follow-up of most studies (<4 months), the degree of engagement with digital interventions over longer durations is unknown.
While this review was able to summarise the acceptability/feasibility of digital HBC from a range of different metrics in SMI samples, a further limitation is that many of the included studies were conducted in the United States, which may limit generalisability to other healthcare systems. Furthermore, most studies only recruited participants who had the ability and/or interest in using digital technologies, making it hard to determine the actual feasibility of such approaches, across the entire clinical populations of those treated for SMI (i.e. beyond those individuals who are eligible and willing to join the reviewed studies, to begin with).
Conclusions and future research
Current results suggest digital HBC overall are acceptable and useful for people with SMI, but some individuals may need extra support with technology. The effectiveness of these interventions has yet to be fully established. Nonetheless, there are many provisional findings of digital technologies can result in positive HBC among people with SMI, and even better engagement when compared with some in-person intervention components. To ensure accessibility and usability, the design process of digital interventions should aim to involve people with SMI throughout. Future research should also examine the cost-effectiveness and implementation of digital HBC for promoting health behaviours into real-world clinical settings and healthcare systems for people with SMI.
Financial support
This study was supported by a University of Manchester Presidential Fellowship (P123958) and a UK Research and Innovation Future Leaders Fellowship (MR/T021780/1).
Competing interest
J. F. has received honoraria/consultancy fees from Atheneum, Informa, Gillian Kenny Associates, Big Health, Nutritional Medicine Institute, ParachuteBH, Richmond Foundation, and Nirakara, independent of this work and Dr Guinart has been a consultant for and/or has received speaker honoraria from Otsuka America Pharmaceuticals, Janssen Pharmaceuticals, Lundbeck and Teva. All other authors have no competing interest.