This pilot randomized controlled trial was performed at one of the acute general psychiatry units of a large academic psychiatric hospital, the Centre for Addiction and Mental Health (CAMH), which provides secondary and tertiary inpatient care in Toronto, Canada. The study was registered on ClinicalTrials.gov (identifier: NCT04841603). In addition to the clinician researchers, a peer researcher (ie, a researcher with lived experience of mental illness) was involved in recruitment of patients and in the collection and analyses of qualitative data as per best practices for patient involvement in mental health research [21]. This report was prepared according to the CONSORT (Consolidated Standards of Reporting Trials) statement (Multimedia Appendix 1).

All patients admitted to the unit between April 5 and May 5, 2021, were screened by the admitting inpatient team for eligibility and referred to the research team if they assented to hear more about the study. Inclusion criteria were age between 18 and 65 years, fluency in English, a Dynamic Appraisal for Situational Aggression (DASA) score < 3 at the time of referral (which indicates lower acute risk of violence/aggression [22]), and capacity to consent to participation as assessed by the treating team. Exclusion criteria were diagnoses of moderate or severe intellectual disability, learning disability, or neurocognitive disorder, as these participants may have had difficulty navigating the app.

The study was performed in accordance with the Ethical Principles of Psychologists and Code of Conduct as set out by the British Association for Behavioural and Cognitive Psychotherapies and British Psychological Society, and with the Declaration of Helsinki. All participants provided written informed consent after being provided verbal and written information about the study and before the initiation of any study procedures. The CAMH Research Ethics Review Board approved the protocol, all supplementary documents, and the informed consent form (#116/2020-01). Participants could be withdrawn from the study at their request, the treatment team’s request, or if they experienced significant worsening of symptoms as determined by the research team or treating physicians. Study data were deidentified and anonymized, and participant records were stored in a secure locked drawer in a locked room. Participants were not compensated for their participation in the study. After the trial period, participants in the control group were provided access to the MindShift CBT app.

After provision of informed consent and confirmation of eligibility, participants were randomized without stratification, using the open-source randomizer randomizer.org, to MindShift+treatment as usual (TAU) or TAU. By design, neither the research team nor participants were blinded to group assignment upon commencement of the study.

After being randomized, participants were involved in the study for 6 days (ie, participants whose length of stay was longer than 6 days were involved for only 6 days). The 6-day study period started on each Tuesday of 4 consecutive weeks. On the first day of the study, each participant met individually with a research team member uninvolved with clinical care on the unit. Participants in both conditions were provided with a tablet to complete the baseline questionnaires (see below). Participants randomized to the intervention (ie, MindShift+TAU) completed an individual introductory session to the app, comprising a review of several sections: the “Home” section, including CBT-based tools such as a fillable thought record and exposure ladder; the “Learn” section containing basic information about anxiety and CBT; the “Quick Relief” section containing meditation exercises; and the “Goals” section, which allows users to set and review progress toward specific and measurable goals. Patients were asked to open the app and use their preferred features for at least 10 minutes and to use the “check in” function daily. This feature requests that users: (1) rate their current mood on a 1-10 Likert scale calibrated with qualitative labels and emoticons; (2) type a response to an open question stating “What’s going on? Describe what’s going on in your life right now and/or what’s on your mind”; and (3) indicate active symptoms of anxiety selected from a checklist. Due to concerns about confidential patient information being shared with others on the unit inadvertently, the tablets used were programmed to require login to the MindShift app each time they were opened. On the third day, each participant was offered a 15-minute session with a clinician or peer member of the research team to address any questions about the study. For participants randomized to the intervention, the use of the app and its core features were also reviewed during these sessions. On the sixth and final day, all participants completed poststudy questionnaires and a debriefing session. Participants in the intervention condition also completed a 30-minute semistructured qualitative interview with a peer or clinician researcher.

All tablets were stored in the nursing station when not in use by patients or when requiring charging. The tablet given to the participants randomized to the control (TAU) condition did not include the MindShift app but had otherwise identical apps installed and accessible, including a video-streaming app and internet browser. TAU on the inpatient unit comprised daily assessments by a psychiatrist, meetings with a social worker as determined by the clinical team, 24/7 access to nursing staff, and pharmacotherapy management. No group psychosocial activities were conducted as part of TAU during the study period due to COVID-19 pandemic–related restrictions. After completing the exit questionnaires on the sixth day of the study, control group participants were offered access to tablets with the MindShift app.

The primary outcomes of this trial were indicators of feasibility and acceptability related to the use of the MindShift CBT app in an inpatient setting. Feasibility was assessed with the rates of consent, study completion, and withdrawal, and by completeness of data in pre- and poststudy questionnaires. Acceptability was assessed quantitatively with the Client Satisfaction Questionnaire (CSQ-8), a self-reported scale with 8 items that describes satisfaction with a health service [23]; participants in the intervention condition also completed a user-experience questionnaire for app-based interventions (Multimedia Appendix 2). Given the feasibility-sized sample, changes in symptoms from baseline to the end of the study were classified as secondary quantitative outcomes. The Generalized Anxiety Disorder (GAD-7) scale was used to assess anxiety [24], the 9-item Patient Health Questionnaire (PHQ-9) was used for depressive symptoms [25], and Kessler Psychological Distress Scale (K10) was used as a global assessment of psychological function [26].

To assess qualitative data, a thematic analysis was applied to transcribed records of the semistructured qualitative interviews using the 15-point checklist of criteria for good thematic analysis published by Braun and Clarke [27]. Transcripts were reviewed and coded independently by the lead clinical researcher and peer researcher, and themes arising from the codes were identified independently. The researchers then met to establish a consensus regarding themes, and items were recorded in alignment with the themes identified by consensus.

One of the main goals of this feasibility study was to obtain data that could be used to calculate sample sizes for a future larger, confirmatory trial. Prior research has suggested that a sample size of 12 in each arm suffices for a pilot feasibility study [28]. Given the patient flow on the inpatient unit on which the study was conducted, we planned to conduct the study over 4 weeks with the expectation that this would be long enough to enable randomization of 24 participants.

Statistical analysis was completed using SPSS 28 software. Descriptive statistics, including mean, SD, and range, were calculated for the entire study sample and for each group. The baseline characteristics were compared between the two groups using Kruskal-Wallis tests.

Figure 1 presents the flow of referrals and participation. During the study period (April 5 to May 5, 2021), 33 patients were referred to the study, 5 of whom were discharged or transferred prior to being invited to participate. Of the 28 patients invited to participate, 4 declined, resulting in a consent rate of 24/28 (86%). Of the 24 consented participants, 1 withdrew and 3 were transferred or discharged before being randomized, resulting in a randomization rate of 20/24 (75%). Of the 20 randomized participants, 11 (55%) completed the study and provided pre- and poststudy data that were analyzable. Of the 9 participants who did not complete the study, 2 withdrew (1 in each condition) and 7 were withdrawn because of being discharged or transferred prior to study completion (3 in the MindShift+TAU condition and 4 in the TAU condition).

Table 1 summarizes the demographic and baseline clinical characteristics of the participants in the two study conditions. Overall, intervention and control condition participants did not show significant differences in baseline characteristics, except for a higher proportion of women in the control condition. Of note, 65% of the participants were admitted with a primary diagnosis of psychotic disorder, and none of the patients who participated was admitted for treatment of a primary anxiety disorder. Overall, 35% of the patients were involuntarily admitted under the Mental Health Act. The mean length of stay of the 20 participants was 20 (SD 4.4, range 3-21) days from the time of admission to the first of either discharge, study completion, withdrawal from the study, or transfer to another unit. Table 1 also presents the clinical outcomes of the 11 participants who completed the study and for whom pre- and poststudy clinical measures were available. The differences in GAD-7, PHQ-9, or K10 scores were not interpretable given the small sample size.

The scores on the CSQ-8 questionnaire (possible range 8-32, higher scores indicate higher satisfaction) completed by the 5 participants in the MindShift+TAU condition (mean 20.2, SD 3.4; range 17-25) indicated overall positive satisfaction with the intervention. Similarly, the scores on the user-experience questionnaire (possible range 0-110, higher scores indicate better usability), available for 4 participants (mean 80.5, SD 27.4; range 46-108), reflected moderate to high user engagement and satisfaction.

Four of five participants in the MindShift+TAU condition completed exit interviews, providing information that was grouped into five major themes: feasibility of the intervention in the inpatient setting, usability and patterns of app use, acceptability of the features of the app, a desire to provide feedback and to be seen as active participants during inpatient admissions (which we characterized as agency), and impact of the app on symptoms (see Textbox 1). The benefit of having access to the tablets and app on the inpatient unit was a recurring theme. For instance, one participant commented, “It helped a lot…It gave me access to a tablet and the programs on it, so indirectly I had access to the programs…and do your work as well, so I think it was beneficial.” Users found the app simple to navigate. Frequency of use varied from multiple times per day to just a few times over the week. The “chill zone” feature of the app was the most commented upon and received consistently positive feedback. The participants spontaneously provided suggestions for improvement or additional features, and talked about their participation in research and how their feedback could impact future programming. For instance, one participant commented, “I’ve been asked to participate in the research. This means it’s not just inpatient and that’s it...you could put input in it, you feel that you’re worth something.” There were mixed comments about the impact the app had on symptoms, with some believing use was beneficial, while one participant commented that the app may trigger negative thoughts: “I didn’t like to read too much into it because the info could scare me. It’s like phobias, ‘I’m scared I’m scared!’” Two participants described being confused about the purpose of the pre- and poststudy questionnaires, with one participant conflating the prestudy questionnaires with the CBT app itself.

This study provides informative data on the feasibility of studying an app-based intervention on an acute psychiatric inpatient unit. A survey of health care provider attitudes toward digital interventions in this setting raised important questions about feasibility, including implementation and suitability for acutely ill patients [29]. Our study supports that the use of digital psychotherapeutic interventions on the inpatient unit requires careful planning and considerations, but could have benefits for an acutely unwell, transdiagnostic population. The high consent rate (24/28, 86%) among inpatients referred to the study demonstrates interest in and willingness to participate in such an intervention. However, the relatively low completion rate (11/20, 55%) emphasizes the challenges of conducting research in acute inpatient settings, where rapid patient turnover can impact the participation and retention of participants. Our goal of obtaining data to calculate sample sizes for a larger confirmatory trial was not achieved because of limitations in our study design and time allotted for data collection, which will inform the design of future trials moving forward. We gathered pre- and poststudy clinical measures of anxiety, depressive symptoms, and psychological distress, but could not accurately interpret these data owing to the small sample size.

While the limited sample size also restricts our ability to draw broad conclusions from the thematic analysis, the quotes provide rich context to reflect patient experiences of the intervention. An unexpected finding was that participants felt empowered by contributing to research and were eager to offer feedback. Voluntary participation in inpatient research projects may stand in psychological contrast to the disempowerment experienced by at least one-third of our participants who were admitted involuntarily. Some participants indicated an intent to download the app on their personal devices after completion of the study and planned to continue to use it after discharge from the hospital. If included in discharge planning, such an intervention could serve as a “transitional object” between inpatient and outpatient settings [30].

A core feature of our study was the collaboration between clinician researchers and a peer researcher for recruitment, data collection, and analysis. However, collaborative efforts must occur earlier in the research process to guide research question development and study design [21]. Compiling research teams of diverse stakeholders, including people with lived experience of mental illness, will be key to designing patient-centered research objectives in the historically coercive inpatient setting.

Of the 20 randomized participants who withdrew from the study, 7 (35%) had to be withdrawn due to being transferred to another more specialized unit or being discharged. Due to limited research personnel, the study operated on a fixed schedule with a 1-week intervention that started on each Tuesday of 4 consecutive weeks, which contributed to low completion rates. A future study would need to use a rolling-entry design in which enrolled participants would be randomized and start the intervention on the day they provide consent. Although the focus of our study was transdiagnostic use of the MindShift app for anxiety symptoms, no participants in the intervention condition had a primary anxiety disorder, which may have impacted the acceptability of the app that is tailored to people with a primary anxiety disorder diagnosis. The requirement to log on to the app at each use presented an additional barrier compared to typical use on personal devices, where users are only required to log in once upon first use of the app and then remain logged in. Additionally, there were unanticipated logistical challenges, which can be generalized to researching other inpatient technology-based interventions, including requiring a system for tablet sign-out, technical troubleshooting, and a need for policies on the inpatient unit about acceptable use of tablets. In addition to the challenges posed to research, these limitations are also relevant to implementation of app-based therapy for clinical use.

Despite the limitations of this study, our findings support the overall feasibility and acceptability of use of psychotherapeutic apps by inpatients admitted with a variety of psychiatric diagnoses. Specifically, the interactive features of guided meditation and mindfulness in the MindShift CBT app were well-received among our participants. An unintended but important finding was the empowerment described by inpatients from being included in a research study run collaboratively by peer and clinician researchers. Our feasibility trial adds to the limited literature on the use of digital psychosocial interventions on acute psychiatric wards. These interventions could be particularly useful in acute transdiagnostic inpatient settings where tablets or mobile devices are already available, or when traditional group-based psychosocial interventions are not practical (ie, when patients are on infection control precautions). However, it is unclear whether investing in technology-based psychosocial interventions for psychiatric inpatient units would be universally appropriate, or whether these are reasonable substitutes for existing psychosocial treatments. Potential future directions for research include exploring gaps in knowledge about the efficacy of this type of inpatient intervention; exploration of the use of other apps with psychotherapeutic features that are interactive, such as guided meditation and mindfulness; and head-to-head comparison with standard-of-care inpatient group psychotherapy interventions.