The major psychoses (ie, schizophrenia spectrum and bipolar I disorder) are recognized as a leading cause of disability worldwide [1] and are associated with poverty [2], premature mortality [3], impaired cognitive function [4], loss of education and employment [5], and increased global economic burden [6]. Published practice guidelines outlining the usual treatment for psychosis identify antipsychotic medication as a frontline treatment for positive symptom management and adjunct psychosocial interventions such as psychoeducation, family support, vocational interventions, and cognitive behavioral skills–based therapies [7-11]. However, despite concerted treatment efforts, disengagement from clinical services remains to be a significant barrier to recovery for this population as evidenced by high rates of nonadherence (25%-50%) [12], nonattendance (40%) [13], and service dropout (30%) [14].

Clinical engagement is a combination of health-oriented attitudes (ie, beliefs about the cause of illness and intentions of the care team) and behaviors (ie, appointment attendance, adherence to treatment plans, and the level of participation in a clinical relationship) [15,16]. Therefore, disengagement occurs when attitudinal factors cause patients to stop performing treatment behaviors such as taking medications and attending appointments [16]. Disengaged individuals often exhibit greater functional impairment and higher symptom burden over time [17,18], subsequently leading to higher rates of relapse and rehospitalization [12,13]. Cycles of relapse and rehospitalization create a unique conundrum for this population as evidence suggests that with each period of disengagement and relapse, positive symptoms become more severe and more difficult to treat [19]. This group is also less likely to fully recover or hold employment [18]. Moreover, studies show that early and sustained engagement in treatment can mitigate negative outcomes of psychosis such as significantly lowering all-cause mortality [20], reducing the risk of relapse [7-11] and ameliorating functional impairment [5,9]. As such, understanding and improving engagement in this population is a priority.

Barriers to clinical engagement are rooted in access to treatment including geographic, systemic, and financial obstacles. Psychosocial services are often located in urbanized areas, serving specific catchment areas [21] and limiting access for rural or remote residents [15,21-23]. Moreover, services are often only available in a business model (ie, Monday-Friday, 9 AM-5 PM) [24-26], with clinicians often treating double the recommended caseloads, further restricting appointment availability [26] because of funding constraints [24-26]. These challenges are often exacerbated by personal factors such as poverty, limited transportation, substance use, trauma history, and beliefs about mental health [15].

Given the financial and staff limitations, there has been a call for more cost-effective, flexible modes of engagement, such as SMS text messaging. SMS text messaging is a globally used mechanism for communication across both wealthy and poor countries [27], and cellular phone ownership and its use in populations with a psychiatric illness is high (72%-97%) [28-31]. SMS text messaging offers an easy, user-friendly platform to extend opportunities for engagement beyond clinical boundaries. SMS text messaging also does not require access to a smartphone, wireless internet, or cellular data, making it a more accessible platform for low-income populations, including those with psychosis. These characteristics may allow SMS text messaging to provide support remotely, and in real time, to boost engagement as shown in other areas of health care such as HIV [32], diabetes [33-35], coronary heart disease [36], obesity [37], and substance use disorders [38,39].

This systematic review aimed to examine studies investigating SMS text messaging strategies to improve clinical engagement in individuals with psychosis. To date, reviews have focused on the subdomains of clinical engagement such as clinic attendance or medication adherence; however, no such review or single study has examined engagement holistically nor have reviews focused on the use of SMS text messaging within populations with psychosis. This is an important knowledge gap given the challenging nature of engagement in the treatment of psychosis, the resulting considerable outcome disparities, and recent trends toward technologically aided health care. As such, even a modest improvement in clinical engagement could lead to better cost-effectiveness of psychosis treatments [40].

Studies were included regardless of the methodological framework or quality, patient setting (ie, inpatient, outpatient, or community care), age of the population, or stage of illness to provide an inclusive examination of clinical engagement. Studies with a range of quality ratings were included to fully examine the state of the literature on this topic. Included interventions used SMS text messaging as a delivery platform, whereas interventions with mixed or non–SMS text messaging technology (eg, mobile apps) were excluded. This review focused on treatment-seeking individuals with psychosis (ie, the portion of the population either beginning or receiving treatment) to isolate the effect of SMS text messaging on clinical engagement. Although it is true that disengagement is a challenge for those who have yet to be in contact with the mental health care system, attempting to study this population would shift focus from treatment engagement to treatment outreach.

Clinical engagement is a broad term describing patients’ participation in treatment and is typically measured using its behavioral outcomes such as appointment attendance (eg, number or percentage of appointments attended, cancelled, or rescheduled), medication adherence (eg, pill counts, clinician or self-estimates, pharmacy prescriptions, or blood plasma levels), and service dropout (eg, complete disengagement from care). However, the consideration of behavioral outcomes alone does not provide a holistic examination of clinical engagement as it excludes the attitudinal aspects of this complex and nuanced concept. To this end, we also included studies with outcomes measuring attitudes toward medication, the level of involvement or participation in treatment plans, and therapeutic rapport.

Feasibility studies were also included to assess 2 outcomes: the level of patient engagement with the SMS text messaging intervention itself, and the degree of practicality of SMS text messaging interventions and their ease of implementation. Outcomes of feasibility trials include user feedback on ease of use, satisfaction, likelihood of future use, and overall intervention design. Outcomes related to practicality and implementation include associated risk and economic, technical, and legal considerations. These are important outcomes to consider when assessing the efficacy of interventions because both directly affect the interventions’ clinical utility. An intervention of any kind is only as successful as its ability to engage its users. In this case, engagement in the SMS text messaging interventions aims to improve engagement in the treatment of psychosis. Thus, this review examined both aspects of engagement to comprehensively evaluate the efficacy and clinical utility of SMS text messaging interventions to improve clinical engagement in populations with psychosis.

Web-based databases cataloging meta-analyses, systematic reviews, and protocols were searched to ensure the originality of this protocol (ie, Cochrane and PROSPERO); however, there were no results on existing reviews on the use of SMS text messaging in psychosis. A unique protocol was, then, created following the Preferred Reporting Items for Systematic reviews and Meta-Analyses guidelines and in consult with a librarian at the University of Toronto. The resulting protocol was registered with PROSPERO (reference number: CRD42018091962), an international protocol registry for prospective reviews in health care managed by the National Institute of Health Research.

In all, 3 core databases powered by OVID were used: PsycINFO, MEDLINE (Medical Literature Analysis and Retrieval System Online), and EMBASE, using the following search strategy: schizophrenia spectrum.mp. OR psychotic disorder.mp. OR exp psychosis/ OR exp bipolar disorder/AND (sms or short messag* service* or texting or text messag*).mp. Please note that .mp. is a mapping command that allows you to search using a word or phrase across titles, keywords, and abstracts using databases powered by OVID.

CINAHL and Google Scholar were used as peripheral databases to ensure studies were not missed. Search terms and keywords related to psychosis and SMS text messaging that were used to search gray literature included the following: (1) SMS OR Short Message Service OR SMS-Survey OR Texting OR Text Message OR SMS Based System OR SMS Reminder OR Text Message Reminder and (2) Psychosis OR Schizophrenia OR Bipolar OR Schizoaffective OR Schizophrenia Spectrum OR Psychotic Disorders OR First-Episode Psychosis OR Early-Episode Psychosis. Additionally, reference lists of identified studies were hand searched.

The search criteria were limited to studies published in the English language, restricting publication dates from January 2000 to March 2019 as there were no SMS text messaging intervention studies published before the year 2000.

Studies were independently reviewed by title, abstract, or full-text using a Web-based blinded review platform named Covidence, which allows the reviewers to log-in independently to conduct their review and tracks conflicts to be resolved. Conflicts were resolved first by the 2 primary reviewers, authors JD and JC, and then by a third reviewer, GF, if necessary. Only full-length empirical studies were reviewed; conference abstracts and commentaries were excluded given their limited descriptions of methodologies and inability to be adequately appraised.

We extracted sample sizes, and demographic characteristics including age and gender, and diagnoses. Methods were also extracted, including the study design, intervention type, primary and secondary outcome variables, and statistical plan. To better understand the significance of the results, effect sizes were also extracted from studies with randomized controlled trial (RCT) and quasi-experimental designs and converted into a comparable measure of effect size, Cohen d. Finally, results reported in each study were extracted, including primary and secondary results (including P values, if available) as well as any other findings of interest to the topic of the review.

As studies were included regardless of the study design, 3 checklists from the Joanna Briggs toolkit were used: RCTs, quasi-experimental designs, and qualitative studies [41]. This toolkit is an open-access, widely used appraisal tool for systematic reviews accessible through the citation provided. Using the checklist that corresponded to the study type, studies received 1 point per criterion met. Points were tallied and converted to a percentage based on the total number of criteria outlaid on the checklist used for easy comparison across study designs. However, results were compared and contrasted based on the matching outcome variable and study design.

As outlined in Table 1, studies were conducted in a range of countries, although most were Western, developed countries from Europe and North America, with 1 study each from Nigeria, China, and India. They were published from 2010 to 2018, with steady publication rates throughout. Most studies included patients with a diagnosis of schizophrenia spectrum disorders, aged 18 years and older, with 1 study focusing on a younger population, aged 18 to 29 years. Sample sizes ranged from 17 to 1139 with some studies based on the same sample but reporting on different outcomes (eg, studies by Ben-Zeev [42] and Aschbrenner et al [43]; and studies by Kauppi et al [44], Kannisto et al [45], and Välimäki et al [46]). RCTs tended to examine the effects of SMS text messaging reminders on medication adherence and/or clinic attendance paired with secondary outcomes related to symptom severity and functioning. Qualitative studies examined the feasibility and acceptability of SMS text messaging interventions.

Studies included in this review were of a reasonable quality (range 31%-85%); however, no study met all of the criteria laid out in the Joanna Briggs frameworks. Individual study ratings are included in Table 2. RCTs received a range of scores from 31% to 85%, with sample sizes between 28 and 1119 participants. Common unmet criteria in RCTs were blinding expectations, as none of the included trials were double-blinded and many failed to use a concealed assignment. Additionally, some trials were underpowered for their results. In the quasi-experimental study, there was no control group used, causing multiple criteria to be unmet. Qualitative designs tended to satisfy a higher proportion of the quality criteria, receiving scores from 40% to 80% with sample sizes ranging from 17 to 562. These studies often lacked information around experimenter bias and sampling methods and an adequate representation of the population. Of note, feasibility studies do not fit smoothly within the qualitative study category and, therefore, appraisal was limited in this domain. Despite these methodological limitations, these studies provide adequate assessments of and insights into a variety of outcomes related to SMS text messaging, clinical engagement, and the consideration of SMS text messaging implementations.

The quality of outcomes within the included studies was fair, although heterogeneous. The included studies showed inconsistent methods of measurement across medication adherence, often relying on self-report, and the calculation of attendance rates was unclear. Other considerations included small sample sizes as 4 of the experimental studies had samples under 100, and only 1 trial had insufficient power to detect a significant difference [50]. The resulting effect sizes were also variable, ranging from moderate to high. Additionally, the intervention design was inconsistent from the type of SMS text messaging (ie, one-way vs two-way messaging) to the frequency, length, and content of the messages themselves. Only 2 of the included studies were designed to include a follow-up period, both examining participants’ adherence up to 3 months postintervention [47,49]. Taken together, this amount of variance makes it difficult to assess the overall success of the included interventions; however, no clear patterns emerge to suggest that sample size, message type, or effect size varies in response to any methodological factors in these trials. Individual study intervention details are outlined in Table 2 and procedures are outlined in Table 3.

A summary of the main results has been provided in Multimedia Appendix 1.

Using SMS text messaging for medication reminders appears to have a significant positive effect on both attitudes about medication [47] and medication adherence [47,50,54], with the potential to have lasting effects [49]. However, there are many caveats to these findings. First, findings, although positive and significant, had moderate effect sizes based on small samples, with some trials reporting nonsignificant changes or significant changes for only the subgroups of the study sample. Second, most studies did not investigate effects after the termination of the intervention, highlighting a need for additional longitudinal follow-up studies.

It is also important to acknowledge that medication reminders only address involuntary nonadherence (ie, motivation, forgetfulness, and understanding) and do not directly address voluntary nonadherence (ie, lack of willingness) [57] except potentially indirectly through participants' attitude toward medication. As a cautionary note, this should be considered when evaluating medication reminders as different causes of nonadherence may require different approaches. Furthermore, results indicate SMS text messaging reminders to be most efficacious for patients in a midrange of symptom severity and functioning [47,52,54].

The study that did not find SMS text messaging reminders to have a significant impact on medication adherence used hospital admission rates as the primary outcome [46]. Despite reductions in hospital admissions being one goal of treatment, decreases in hospitalization rates as an indicator of improved medication adherence may be too far removed from the adherence process and may not be an appropriate indicator of the efficacy of SMS text messaging reminders.

The results shown here are similar to those across other areas of health care such that many show positive trends toward improved treatment adherence, with mixed results with regard to significance. For example, some studies find a positive nonsignificant trend toward improved medication adherence in postsurgery populations [58] and individuals with diabetes [33]. Other studies report strong findings in areas such as HIV antiretroviral treatment [32] and general health populations [59]. In contrast, some studies have found no effect of SMS text messaging reminders on populations such as patients with tuberculosis [60].

Some of the aforementioned studies [33,58] as well as many in this review used one-way messaging, which may explain some of the nonsignificant findings. A meta-analysis examining one-way versus two-way SMS text messages as medication reminders found that two-way messaging can increase oral medication adherence in a breadth of medical conditions by 23%, whereas one-way messaging shows little or no effect [61]. In this review, 6 interventions used one-way SMS text messaging reminders, yet only 2 studies found nonsignificant changes in either attendance or medication adherence; thus, one-way reminders may still be effective in psychosis populations.

Regarding clinic attendance, SMS text messaging reminders were also found to increase the likelihood of attending appointments [51,52]. The target population in the study by Thomas et al [51] was patients who had not yet had contact with psychosis services, yet the use of SMS text messaging reminders increased the likelihood of clinic attendance two-fold at initial appointments. Similar findings are reported for initial attendance at treatment start in substance use literature [62]. This is notable considering the high rates of disengagement and dropout for patients in early stages of accessing care [14] and, therefore, represent a population at elevated risk of not receiving treatment.

Findings in this review replicate previous positive findings on attendance within various health conditions and settings, including primary care, mental health care, and dental care, where reminders improve attendance and decrease the probability of missed appointments [63-65]. Taken together, the positive effects of SMS text messaging on clinic attendance may have the potential for annual national cost-savings in the millions [66], and large health insurance providers have begun to consider reimbursement for use of digital aids [31].

Therapeutic rapport may also be enriched by the addition of SMS text messaging as patients in 1 included study reported improved therapeutic rapport with SMS text messaging interventionists compared with their clinical teams [42]. This may be due to factors relating to perceived availability of the mobile interventionist compared with the clinical team [42], as the mobile interventionist is able to reply to questions and problems in real time and communicate with patients daily. This is an important finding given that therapeutic rapport is important for sustained engagement [14,15].

Feasibility outcomes were assessed by both RCTs and qualitative studies, providing important insights for future implementation. All studies examining user outcomes found SMS text messaging to be feasible, safe, and acceptable, gaining the majority endorsement in each sample [42,44,54-56]. These findings help assuage popular concerns around the feasibility of SMS text messaging interventions as cell phone ownership [55,56] and response rates were high [42,54], most were familiar with SMS text messaging or found it easy to learn and use [49,52,56], and only 1 patient reported increased paranoia around usage of mobile phones [56].

In comparison with other mobile strategies such as mobile apps, SMS text messaging studies reported slightly higher response/usage rates (83%-87%) than mobile app studies (69%-86%) in this population [29], although attrition rates were similar [29]. In a direct comparison of SMS text messaging and mobile apps for daily symptom monitoring, authors noted that the set up used for apps is a more involved process and may affect participation rate; however, the user platform in apps was preferred over SMS text messaging by participants [67].

A clear advantage of SMS text messaging over mobile apps is cell phone versus smartphone ownership, and concerns around the exclusion of economically disadvantaged populations, such as those reliant on government financial support. Cell phone ownership steadily decreases with income and is significantly lower among individuals with severe mental illness (SMI). Ownership among earners of an annual income under US $30,000 per year is 84% and 81% for those with psychosis, yet ownership among the lowest earners drops to 50% and 35%, respectively [29]. Cell phone ownership among SMI populations has been reported at 93% in the United States, yet only 50% to 60% owned smartphones [68,69], and 67.9% in Israel [70]. In developing and newly developed countries such as India, however, smartphone ownership is still only present in a minority (22.7%) of mental health populations [71]. Importantly, trends in smartphone ownership, although still below cell phone ownership rates as of 2018, have been steadily increasing for years in the general population [72] as well as populations with psychosis [29] and will likely become more ubiquitous as younger populations grow in this digital age.

A strength of this review is that we took a comprehensive perspective to engagement and provided an inclusive review of clinical engagement in populations with psychosis [14]. However, an even more comprehensive examination of technologically aided engagement could have been created by including other platforms that use mobile devices, such as mobile apps, as a helpful comparison group. This would help further elucidate patient preferences and effectiveness to ascertain information on which platform is best suited for implementation in the treatment of psychosis. Another limitation of this review may be the design itself, as a review is limited to data extracted from the published articles of the included studies, without attempts to contact corresponding authors of the included studies to obtain further information. In addition, this review was designed and registered as a qualitative review rather than a meta-analysis, although we acknowledge that a meta-analysis may have provided some more quantitative insights into the use of SMS text messaging to boost engagement in the treatment of psychosis; the meta-analysis itself would be limited given the small number of published studies and the significant heterogeneity in methods, primary outcomes, and durations of interventions..

Limited conclusions can be drawn from included studies owing to the scarcity of effectiveness trials and novelty of digital strategies in the treatment of psychosis. Additionally, many current trials are pilot studies that are time-limited with small sample sizes using convenience sampling and, in some cases, with underpowered analyses. Another limitation is that no study examined impacts on clinician workflow or cost-effectiveness or privacy considerations beyond educating participants on privacy risks. Furthermore, there are no standard measurements for clinical engagement or any of its domains, making results too heterogeneous to be accurately compared.

The central theme drawn from the presented studies is that individuals with psychosis prefer a person-centered approach, with a personal feel, and that this approach tends to be more effective. Suggestions for future trials using SMS text messaging include the involvement of participants in co-design, employing two-way messaging to engage participants in a conversation and allowing for personal choice and autonomy within the intervention, namely personal choice over content, timing, and frequency. This is especially important given the considerable heterogeneity in psychosis spectrum disorders and the flexibility of SMS text messaging to afford opportunities for individualized treatment approaches. Moreover, digital tools have the potential to increase patient autonomy and independence by extending care outside the clinic and putting it literally in patients’ hands.

Given the heterogeneity of clinical presentation and treatment, it should also be considered that not all patients with psychosis are prescribed medication, so standalone medication reminders would not be universally helpful. Other subgroups requiring additional planning include individuals with physical disabilities, low literacy levels, severe cognitive impairments, developmental delays, experiencing homelessness, or share a cell phone with others [42]. Another key area of consideration is safety protocols. Despite a study’s report that several urgent issues had to be escalated to the primary care team resulting in expedited prescription fills, home visits, and safety assessments [43], none of the included studies reference precautions or procedures relating to safety. Such issues underscore the importance of developing safety protocols for digital tools.

Finally, due to the potential increase in patient autonomy and independence, SMS text messaging–augmented care could lead to a decreased dependence on clinical staff, allowing clinicians more flexibility to manage caseloads and potentially leading to positive financial and economic outcomes while simultaneously allowing for better, more personalized care. Future studies should, therefore, report on the cost-effectiveness of the interventions as well as impact on workflow to assess how readily SMS text messaging can be implemented.

In sum, this review demonstrates the potential of SMS text messaging to be used as an adjunct platform to support clinical treatment as a means of improved engagement. These findings show that SMS text messaging is extremely well tolerated, safe, and accepted among individuals with psychosis, testifying to its potential to not only improve engagement in care but also to extend care beyond the clinic. To this end, SMS text messaging stands to offer a pragmatic solution to boost clinical engagement and provide an alternative avenue to access treatment. To this end, a more thorough examination of practical considerations, consistent measures for engagement factors, and rigorous examinations of its outcomes (eg, medication blood plasma levels or electronic medication caps compared with self-reporting) are required to adequately assess its efficacy, cost-effectiveness, and readiness for implementation.