Studies were selected for inclusion in the review if they met the following criteria:

Studies were excluded if SMI was caused by dementia or brain injury, they were not available in English, they investigated telecare or social care technology such as remote sensors for falls, the intervention focused on carers or health care professionals rather than service users, participants were not randomized, technology was not the primary focus of the intervention, and participants had major depressive disorder or other mood disorders without psychosis.

Electronic searches using the following databases were conducted from inception to March 2016: MEDLINE, EMBASE, PsycINFO, Cochrane Central Register of Controlled Trials, Allied and Complementary Medicine Database, Health Technology Assessment, CINAHL Plus with Full Text via EBSCOhost, and NHS Economic Evaluations Database by searching “all fields” using the following search terms: TELEHEALTH, TELE*, TECHNOLOGY, E?HEALTH, M?HEALTH, ONLINE, WEB*, INTERNET, COMPUTER, MOBILE, APP, VIRTUAL CONSULTATION, PHONE, POCKET PC, IPHONE, SHORT MESSAGE SERVICE, SMS, TEXT MESSAG*, WIRELESS, SMARTPHONE, REAL-TIME, ELECTRONIC DIAR* and INTERVENTION, PROGRAM*, THERAPY, SUPPORT, EDUCATION, TRAINING and SEVERE MENTAL ILLNESS, MENTAL DISORDER, PSYCHOTIC DISORDER, MOOD, AFFECTIVE DISORDER, PERSONALITY DISORDER, BIPOLAR DISORDER, SCHIZOPHRENIA, DEPRESSION, SCHIZOAFFECTIVE DISORDER and RANDOMI?ED, TRIAL, CLINICAL TRIAL, COMPARATIVE STUDY, SYSTEMATIC REVIEW, META-ANALYSIS, REVIEW, CROSSOVER PROCEDURE, DOUBLE BLIND*, and SINGLE BLIND*.

The first author checked titles, and all 3 authors reviewed abstracts to exclude any irrelevant articles. Full texts of the remaining articles were obtained; all 3 authors screened these. Any disagreements were discussed within the research team to reach consensus. References were searched for additional papers.

The first author extracted data using a standardized form developed by the Cochrane Collaboration [15], which included information on the following: study characteristics (including aim and design), participants (including population description, inclusion and exclusion criteria, and baseline imbalances), features of the intervention and comparison groups (including description, timing, and providers), outcome measures, statistical methods used in analysis, results, and conclusions.

The Cochrane Collaboration tool to assess the risk of bias in RCTs [16] was used. The Cochrane risk of bias tool includes a list of potential sources of bias in clinical trials in 7 main areas: random sequence generation, allocation concealment, blinding of participants and personnel, blinding of outcome assessment, incomplete outcome data, selective reporting, and other bias [16]. The rating scale for each area of bias ranges from “low risk of bias” to “high risk of bias” and an option of “unclear risk of bias” for studies that do not provide enough details to be able to make a clear judgment [16]. Two of the review authors (SLW, KM) independently judged each article for risk of bias, discussing disagreements until a consensus was reached.

Electronic searches identified 13,907 unique articles, and reference searches of key articles identified a further 17 potentially eligible articles. The inclusion criteria were met by 31 articles, which have been presented in a Preferred Reporting Items for Systematic Reviews and Meta-Analyses [17] diagram (Figure 1). A meta-analysis was not performed due to the heterogeneity of the interventions, including their methods of delivery and the outcome measures of the included studies. The characteristics of the final 31 articles can be seen in Multimedia Appendix 1.

The 31 articles reported 29 trials. A total of 17 trials focused on people with schizophrenia, 9 on people with schizophrenia or schizoaffective disorder or psychotic disorder, and 3 on people with bipolar disorder. A range of telehealth devices were used with varying aims. Computers were used to improve cognitive functioning and disease-specific knowledge; websites aimed to improve psychosocial functioning; hand-held devices were used to improve communication with medical staff; telephones were used to improve medication adherence and disease-specific symptoms; virtual reality (VR) aimed to improve social and work-related functioning; and electronic medication dispensers aimed to improve adherence. The specific outcome measures included medication adherence (4 studies), social functioning (including work behavior; 5 studies), health care utilization (4 studies), neurocognitive functioning (16 studies), knowledge about medication or SMI (2 studies), self-esteem (3 studies), self-efficacy (1 study), quality of life (4 studies), mood (4 studies), insight into condition (1 study), perceptions of deprivation of liberty (1 study), satisfaction with life or treatment (5 studies), and illness perceptions (1 study). The 29 included trials recruited a total of 4338 participants. Sample sizes ranged from 29 to 507. Details about each intervention can be seen in Multimedia Appendix 2.

Results for the risk of bias assessment can be found in Multimedia Appendix 3. Of note, only 5 studies were rated as high quality. One study demonstrated a particularly high and, at times, unclear risk of bias for the majority of sources of bias [18]. Several studies did not provide enough details to be able to make a clear judgment that they had not introduced bias into their findings.

Results for each of the 31 studies can be found in Multimedia Appendix 4. A description of the findings is presented in the following sections.

The aims of this review were (1) to identify which, and how, telehealth interventions have been trialed for people with SMI and (2) to synthesize the evidence in relation to the effectiveness or efficacy of these interventions. This review identified 31 articles describing 29 trials, including a total of 4338 participants with schizophrenia, schizoaffective disorder, psychotic disorder, and bipolar disorder. The studies in the included articles evaluated the use of computers for cognitive rehabilitation, patient education, consultations with key workers, and interactive Web-based CBT–based self-management interventions; the use of VR to simulate work and social situations and to deliver cognitive training; the use of telephones to prompt medication use and to report SMI symptoms to health care teams; and the use of remote sensors to monitor medication use.

This review found evidence for using some types of technology to support the management of SMI while finding that not all technology is effective, depending on the outcome of interest. Interventions containing telephone support from the medical team, including phone calls and short message service text messaging prompts about medication, were consistently found to be effective in improving medication adherence while also reducing the severity of mania symptoms and reducing inpatient days. CACR was found to be effective in improving cognitive outcomes in schizophrenia in most, but not all, studies. A Web-based CBT-based self-management intervention for bipolar disorder was found to be effective at improving quality of life and social functioning. The use of VR was found to be effective in improving work-related behavior, conversational skills, assertiveness, and cognitive functioning, although face-to-face social skills training was found to be more effective in improving nonverbal social skills than VR-delivered training. Computers appeared not to have a superior benefit for delivering patient education over traditional nurse-delivered education, and in fact, participants preferred the nurse-delivered method. This finding is perhaps reflective of the nature of interactions with health care staff, which can provide a personal and individualized approach to supporting people with SMI, for example, by offering opportunities like discussing diagnoses and modeling nonverbal social cues. Hand-held devices to support health care consultations were found to be effective in improving quality of life and satisfaction with care, but not in improving positive or negative schizophrenia symptoms. Telemonitoring was found to be effective in improving medication adherence and also led to fewer medical and emergency visits, including inpatient days, while simultaneously improving schizophrenia symptoms.

Notably, 5 studies demonstrated a low risk of bias for 6 out of the 7 sources of bias listed in the Cochrane risk of bias tool [28,39,41,43,44], and 6 studies demonstrated a low risk of bias for 5 out of the 7 sources [23,37,40,45,49,50]. One study demonstrated a particularly high and, at times, unclear risk of bias for the majority of sources of bias [18]. Several studies did not provide enough details to be able to make a clear judgment that they had not introduced bias into their findings, particularly in reference to whether participants were assigned to groups using an adequate randomization method and whether allocation had been truly concealed. The findings of this review should, therefore, be viewed in light of the potential bias introduced into the findings of some of the studies included.

The variety of comparison groups employed in the studies included in this review is also noteworthy. In the majority of articles (n=17), the authors reported that the comparison group received standard care or treatment as usual. This may have differed across trials. For example, in the context of inpatient care, this included medication and attendance at routine therapy groups [33], but in the context of community care, this might have included medication, physician visits, and support from available community centers [49]. However, for 9 of the studies, the control group received a comparative intervention, consisting of health care professional-delivered education about their condition [18,29], text-delivered education [40], therapist-delivered skills training or psychosocial interventions [28,35,42], or computer-delivered activity, such as computer games or education about their condition [21,25,36].

This review supports the findings of previous reviews evaluating cognitive training in schizophrenia [51,52], although evidence suggests that cognitive training is effective regardless of whether it is delivered via computer technologies or noncomputerized psychological interventions [51]. Substantial variations in the intensity and duration of the interventions that used software to support neurocognitive training may have impacted the results of these studies. If training was particularly intensive, for example, half a day or delivered over several months, it is possible that this might have overburdened participants with SMI, leading to a lack of significant findings, or reduced participants’ attention abilities. A recent meta-analysis suggests that as little as 5-15 hours of cognitive remediation could be sufficient for improving cognitive outcomes in people with schizophrenia [52], suggesting that more intensive cognitive training may not be necessary and that future interventions should be designed with this in mind.

CACR was most commonly assessed offline using personal computers and CD-ROM software, with little change in the software used over time. With the rise in more sophisticated technologies, including VR, which in this review was found to improve social and neurocognitive outcomes in people with SMI, perhaps we can expect to see VR interventions delivered more widely to support SMI outcomes in the coming years. VR has the potential to provide usable and safe, ecologically valid assistance in the management of SMI, as already found in general health care [53]. In particular, our findings support a recent systematic review of VR for people with SMI, which was found to be more interesting training than control conditions [54].

We identified surprisingly few studies that had evaluated Web-based interventions to support individuals with SMI. Educational websites were not found to be effective for improving knowledge about medication, adherence, self-esteem, insights into SMI, or deprivation of liberty over traditional nurse-delivered education. However, when websites provided more interactive elements than traditional education, such as peer support or CBT-based self-management techniques, improvements in quality of life, mood, and social adaptation were seen. Further research on the psychosocial benefits of Web-based interventions for people with SMI is suggested, particularly given the small number of studies to date and the growing evidence for such interventions in common mental health disorders, including anxiety and depression [55,56]. Web-based interventions will, however, need to take the specific needs of people with SMI into account to improve accessibility. For example, service users often report that common website design guidelines produce websites that are confusing for them to use, particularly in the presence of cognitive deficits [57].

The finding that telephones were effective for promoting adherence to medication use and attendance at health care appointments was consistent with a recent systematic review [58] of mobile phone-based technologies for supporting general health care and another review for the use of remote technology in SMI [58]. Further evidence suggests that the use of telephones has a beneficial impact on adherence over interventions that do not use technology, for example, psychoeducation for patients who are nonadherent due to forgetfulness [59]. While telephone-based interventions may not consistently improve clinical factors associated with SMI, the findings of the studies included in this review give early promise to the use of telephones in supporting adults with SMI to manage their medication. With advances in the function of telephones, including smartphone apps, this provides further opportunities for supporting people with SMI in the future. One might have expected more interventions, however, as Bakker et al [60] recently emphasized, mental health apps have not to date utilized the designs made available by physical health and social networking apps nor have the hundreds of apps available been tested using formal experimental methods.

The present review also found that the use of telephones improved patients’ attitudes toward using their treatment as well as their quality of life. Leach and Christensen [61] suggest that telephones are acceptable and cost-effective uses of technology to support health care due to their accessibility and convenience. A large-scale survey of users of mental health services in the United States in 2013 found that 72% users reported owning a mobile device, and both users and nonusers expressed an interest in future services being offered through mobile devices [62], suggesting that accessing a mobile device may not pose a barrier to people with SMI.

The positive findings regarding the use of telemonitoring to improve global schizophrenia symptoms, medication adherence, and medical and emergency visits are consistent with the reviews that have found positive effects of telemonitoring for managing chronic health conditions, including heart failure [63] and respiratory conditions [64]. Furthermore, a recent review of the use of remote technology for SMI found this to be a feasible and acceptable method of health care delivery [65]. These are positive initial results for a form of telehealth that was found to be acceptable from both the service user and health care perspective, and future research evaluating the use of telemonitoring for SMI could have implications for the delivery of future services for people with SMI.

The finding that computer-mediated consultations led to improvements in quality of life and reduced unmet need in service users with SMI is promising for a novel use of technology, where there is currently a limited evidence base. One recent qualitative study reports that health care professionals perceive tablet computers to fill a need between smartphones and desktop computers and have some value in supporting consultations with patients [66]. Tablet computers offered support in structuring consultations, ensuring that patients’ priorities were discussed [39]; thus, patients might feel their needs have been better dealt with.

Few studies evaluated the use of remote medication telemonitoring, computer-mediated consultations, or Web-based self-management resources, suggesting these are novel technologies for managing SMI, although all were found to be effective for improving psychosocial and behavioral outcomes. We were surprised to find the limited use of these more advanced technologies in an SMI context given the rise in their use to support broader health care in recent years, for example, Web-based self-management programs for diabetes [67]. We suggest that future research should seek to establish the full potential benefits of these novel uses of telehealth for improving the management of SMI.

Of note is the varied quality of the studies included in the review, with only 5 studies rated as high quality. This suggests that high levels of bias were potentially introduced into the results of these studies. Of particular note is the small sample size of the majority of articles in this review, with 19 of the total 31 studies having samples of <100 participants. Furthermore, the type of comparison or control group employed might have had an influence on the effect size of the intervention evaluated, and the variety of comparison groups included in this review is noteworthy. The generalizability of the findings from many of these studies is, therefore, limited.

To our knowledge, this is the first systematic review of RCTs across a range of telehealth technology interventions delivered to people with SMI. The studies included in this review measured a plethora of outcomes using heterogeneous measures and a range of interventions, making meta-analysis of the results impossible. In addition, we excluded articles that had not been published in English. It is possible, therefore, that we excluded relevant foreign language articles from this review. One author performed the data extraction process; while this was not verified by a second author, all authors were involved in the screening process, and 2 authors independently completed the risk of bias assessment. Thus, several stages of the review process were validated by more than 1 author.

The lack of studies that formally evaluated the acceptability of telehealth interventions is noteworthy. Of those that did measure the opinions of their participants, telehealth was found to be acceptable, particularly in comparison with more traditional face-to-face methods of the delivery. However, this may depend on whether a face-to-face element to the intervention is also offered, as it has been suggested that this may be preferable over unguided interventions [68]. If telehealth is to be developed to support the care of people with SMI, it is important that acceptability of the interventions is considered as part of the evaluation and formally measured. From a health-commissioning perspective, it is also possible that telehealth delivery costs are higher than those of usual care [1]. Cost-effectiveness was not formally evaluated by the studies included in this review; thus, it is unclear whether this is also the case with interventions in SMI. Future studies should evaluate the costs of using technology to support the management of SMI.

This systematic review has identified a range of ways in which telehealth has been used to support the management of SMI and its symptoms. The studies found some strengths of cognitive remediation for schizophrenia, whether delivered through a personal computer and CD-ROM or VR. The use of telephone support from the medical team was consistently found to be effective for improving medication adherence and reducing the severity of symptoms and inpatient days. Few studies evaluated the use of remote medication telemonitoring, VR, Web-based self-management, and hand-held devices, suggesting that these are novel technologies for managing SMI, although all were found to be effective in improving some psychosocial and behavioral outcomes. Patient preferences should be assessed and accommodated, as some may prefer traditional methods of delivery with health care staff over computer-based methods. Given the poor quality of all but 5 of the included trials and that few studies have evaluated the acceptability and cost-effectiveness of using technology to support people with SMI, further studies are needed to establish the potential benefits to these areas.