Evaluating the PolarUs App in People With Bipolar Disorder

Background: BD is characterized by extremes of mood. While almost everyone experiences variations in mood state (i.e., periods of happiness, times when feeling low), people with BD experience more extremity in their moods, ranging from periods of clinical depression through to mild elation ('hypomania') or extreme elation and/or irritability ('mania'). Over half a million Canadians and over 5 million Americans live with BD. Importantly, BD is eminently treatable and people with BD can experience good health and quality of life (QoL). QoL is increasingly recognised as a key outcome of self-management interventions for BD, requiring tailored interventions beyond pharmacology alone. Self-management strategies (SMSs) are recommended in all major treatment guidelines as playing a key role in improving the health and QoL of people living with BD. DMHIs, particularly mobile apps, are one compelling means to improve access to self-management information around the clock, in a self-guided format. DMHIs can be useful in resource-limited environments where in-person care may be limited or fragmented, such as for those living in rural or remote areas. However, current publicly available apps for BD do not address the full spectrum of QoL domains, many are of dubious quality, lack an evidence-base for their efficacy, and may concerningly expose individuals to dangerous advice or lack basic privacy standards. Two challenges with DMHI development include: (1) a dearth of user involvement in app development to account for the desires and goals of end-users; (2) lack of culturally and linguistically appropriate interventions, despite evidence that cultural adaptation of existing interventions can be a cost-effective and scalable way of delivering interventions. Our multidisciplinary research network, the Collaborative RESearch Team to study psychosocial issues in BD (CREST.BD) specializes in research into psychological and social treatments for BD and knowledge exchange. We recently pilot tested a new DMHI co-designed with people with BD, the PolarUs mobile app (alpha version - iOS), to support self-management and QoL measurement in 170 people across North America (Canada and the United States). Quantitative results suggested that use of PolarUs was linked to improved QoL and mood (depression and mania) scores over time. Qualitative findings from interviews with a subsample of the participants (n=25) demonstrated feasibility. The PolarUs app functions as a self-guided, self-management aid to support people with BD in accessing and applying evidence-informed self-management strategies. The app is organized according to the Quality of Life in Bipolar Disorder (QoL.BD) scale, the first and only BD-specific QoL measure. The QoL.BD has been used across diverse settings internationally, with demonstrated sensitivity to treatment effects. It has been translated into over 30 languages, with psychometric validation demonstrated across several linguistic and cultural contexts. PolarUs contains psychoeducational content across 14 QoL domains and 100 evidence-informed self-management strategies. Over 500 resources have been curated from international evidence-informed sources and developed in-house to support self-management. Integrating feedback from the pilot study, we have now culturally and linguistically adapted and advanced the PolarUs app in English, French, Mandarin and Spanish to increase accessibility to diverse people living with BD in Canada and the United States. Cultural and linguistic adaptation processes were facilitated through co-design methods used with three Cultural and Language Advisory Groups (CLAGs), while general updates to the app were facilitated through CREST.BD's Community Advisory Group (CAG) and PolarUs User Group (PUG). The PolarUs app functions as a self-guided, self-management aid to support individuals in accessing and applying evidence-informed SMSs in their daily lives, and is not designed to provide diagnosis, treatment, or clinical decision-making. Objectives: This study will assess whether the PolarUs app, a low-cost and scalable DMHI that targets knowledge about self-management, significantly improves QoL and mood outcomes in people with BD through an RCT in Canada and the United States. The primary objective of this study is to determine the efficacy of the PolarUs app (deployed in English, French, Chinese, and Spanish) in an RCT in adults diagnosed with BD in Canada and the United States. Our specific hypothesis is that for participants with BD given access to the PolarUs app for a 12-week period, there will be significant improvements in QoL (measured by the Brief QoL.BD scale) during this period compared to those in a 12-week control period without access to the app (assessment-only control condition). Qualitative methods involving semi-structured interviews with a subsample of the RCT participants will be used to explore engagement with the PolarUs app using an embedded mixed methods approach. Study Population: Study participants will be North American residents (N=480) aged ≥ 18 years with a DSM-5 diagnosis of BD-I, BD-II or BD not otherwise specified (NOS) as assessed by a structured diagnostic interview. Study Design: This study is a 12-week, 2 arm parallel, single blind RCT that will compare the effects of the PolarUs app (English, French, Chinese, Spanish versions) to an assessment-only control condition. Interested participants will be invited to a baseline assessment conducted via Zoom teleconferencing platform. The Diagnostic Interview for Anxiety, Mood, and OCD and Related Neuropsychiatric Disorders (DIAMOND) will be administered by trained research assistants to confirm diagnostic eligibility (i.e., a DSM-IV-TR diagnosis of BD-I, BD-II, or BD-NOS and an assessment of psychosis and suicidal ideation when screening scores indicate moderate or high risk) and identify any comorbid diagnoses. This structured baseline interview will also be used to collect self-report scales and participants will receive instructions on downloading the PolarUs app (one of the English, French, Chinese, Spanish versions, based on participant choice) to their device (iOS or Android). Ineligible participants will be provided with a list of mental health support resources for the 12-week waiting period. Participants will be centrally randomized in a 1:1 ratio (intervention: control) using computer-generated random numbers stratified by country, sex, and type of BD (specifically BD Type I/NOS and BD Type II). Concealment of allocation will be achieved using a secure online allocation system via MINIMPY; treatment allocation is only revealed after the unique participant number and demographics data (country, sex, type of BD) are entered. Due to the nature of the intervention, full blinding of participants will not be possible. To reduce the risk of bias in outcome assessment, research assistants administering diagnostic scales will be blinded to treatment allocation. Raters will be replaced if they inadvertently become unblinded. The intervention condition will receive access to the PolarUs app (one of the English, French, Chinese, Spanish versions, based on participant choice) on their device (iOS or Android). Content in the PolarUs app is organized according to 14 life areas (with the Physical domain further divided into 4 additional sub-domains) assessed by the QoL.BD. After completing a QoL.BD assessment at baseline, users select up to three QoL life areas that they would like to focus on, after which they can select up to three relevant self-management strategies to implement. Resources to support users with implementing these strategies (e.g., webinars, articles, worksheets, videos) are provided through the app. To support engagement and provide users with feedback, app users are encouraged to self-monitor their QoL, mood, and sleep daily. Users receive encouragement while using the app in various forms, including customizable daily reminder notifications and daily affirmations. The control condition will receive assessment-only (i.e., monthly completion of self-report outcome measures). Participants in the control condition will receive access to the full PolarUs app (English, French, Chinese, Spanish version) after the 12-week intervention period. Outcome measures will be collected at baseline and the end of the 12-week period and, where appropriate, monthly during the study period. The primary outcome measure (Brief QoL.BD), and all other self-report measures will be administered through Qualtrics, the University of British Columbia (UBC)'s secure online survey platform, which is hosted on Canadian servers. The Central Research Coordinator will monitor survey completion and contact participants as needed to support timely completion. App engagement metrics (e.g., frequency of app use, retention, attrition, dwell time on pages, strategies/resources chosen and accessed) will be collected in real time through Google Analytics. Self-report in-app measures such as daily mood, sleep, and wellbeing will also be collected and stored through Google Firebase's Cloud Firestore database. Data will be stored on Google Cloud servers located within Canada. Data will be regularly downloaded, encrypted and backed up onto UBC's secure Microsoft OneDrive cloud database hosted on Canadian servers. Outcomes: Our primary outcome measure is the Brief QoL.BD scale at 12 weeks. The Brief QoL.BD scale is contained in the full QoL.BD. The full QoL.BD assesses 12 core (Physical, Sleep, Mood, Cognition, Leisure, Social, Spirituality, Finance, Household, Self-Esteem, Independence, and Identity) and two optional (Work and Study) life areas across 56 items. The Brief QoL.BD contains 12 items representing the core life areas (minus Work and Study). During initial field testing, both versions had excellent internal reliability (Cronbach's α \> 0.8); the Brief QoL.BD demonstrated a higher sensitivity to change in clinician-rated symptoms of depression compared to non-BD specific (generic) QoL measures. Cronbach's α+0.88 was found for the Brief QoL.BD in our PolarUs pilot study with α = 0.966 for the Full QoL.BD. The QoL.BD scale has been widely used in clinical trials internationally, with demonstrated sensitivity to treatment effects and construct validity through associations with symptoms of mania and depression, generic QoL instruments, and functioning. The QoL.BD has been validated for use in multiple languages, most notably for this study in French, Chinese, and Spanish. Our key secondary outcomes are symptoms of depression and mania, which will be assessed using the Personal Health Questionnaire Depression Scale (PHQ-8) and the Internal State Scale (V2), respectively. Both will be administered at baseline and then monthly. Other secondary outcomes will be assessed using validated self-report measures of generic QoL (WHOQOL-BREF), self-compassion via the Self-Compassion Scale-Short Form (SCS-SF), self-efficacy via the Self-Efficacy for Managing Chronic Disease (SEMCD) Scale, healthcare utilization and app engagement. These secondary outcome measures will similarly be administered at baseline and then monthly. App engagement will be captured through app usage patterns, tracked automatically and continuously through metrics such as login frequency, session duration, and features used; and subjectively reported via the User Engagement Scale-Short Form (UES-SF), administered monthly from 1 month. Treatment factors such as medication use and healthcare visits will be assessed using the Matryoshka Service Needs Profile (SNP) at 12 weeks, allowing us to control for whether participants are in pharmacological or psychosocial treatment (if indicated). Sample size and power calculation: In our previous pilot study, the mean improvement in the primary outcome measure Brief QoL.BD over 12 weeks was 1.74, with a baseline standard deviation of 8.15 and a correlation of 0.5 between baseline and 12 weeks. This is compared to a wait listed control with an expected improvement of zero along with the same baseline standard deviation and correlation in our power analysis. A G\*Power 3.1.9.7 analysis for this 2-armed parallel design with 80% power and a 5% 2-sided significance level suggests a total sample size of 346 participants for a test of statistical superiority. However, given that only 72% of participants completed their week 12 QoL.BD assessment in our pilot study, the sample needs to be adjusted to contain 480 participants. Statistical analysis: Descriptive analyses for the outcome and demographic data will be presented and drop-out over time appraised using graphical analyses. Randomness of the allocation will be tested for significant demographic and clinical associations using independent samples mean and distribution tests for continuous variables and crosstab tests for categorical variables. If any such associations are found, these variables will be controlled for in the ensuing analyses. Similarly, any significant predictors of dropout found from binary logistic regression analyses will be controlled for (e.g., whether the participant is engaged in pharmacological or psychosocial treatment). A Linear Mixed Model analysis, controlling for the variables described above, will be used for all outcome analyses, treating the intercept and time effects as random effects. Based on results from our pilot study, compound symmetry will be assumed for the repeated measures correlation structure. Correlations between the random intercept and time effects will be used to determine to what extent baseline measures affect changes over time. Sub-group moderation effects (e.g., sex and gender, country, treatment status, of BD type, language group) will be tested by incorporating fixed time interaction effects in the above models. A Linear Mixed Model analysis is an Intention-to-Treat analysis that assumes that all missing data is missing at random (MAR). The analyses will be repeated using Multiple Imputation, again assuming MAR. In addition, Pattern Mixture Models and a Delta-adjusted Tipping Point Analysis will be used to analyze the data allowing for the missing data to be NOT missing at random (NMAR). Core project objectives are to create an app inclusive of diverse people with BD including those belonging to diverse genders and linguistic communities, among others: Regarding sex and gender considerations, the analysis aims to capture the intersections of sex, gender and sexual orientation alongside other identities and experiences, to avoid reinforcing gender normativity. Data on transgender and gender diverse individuals in BD is limited, but one Canadian study reported that LGBTQ2S+ people with BD face unique barriers relating to problems of accessing support and creating social networks. Key outcome variables will be analyzed and reported disaggregated by sex, with sex examined as a potential moderator of QoL outcomes using a Mixed Model analysis while allowing for the interaction of Time\*Sex as a Fixed Effect. Regarding considerations pertaining to diverse linguistic communities, the analysis aims to account for the influence of language and cultural context on participant experiences, engagement, and outcomes, while avoiding assumptions of linguistic or cultural homogeneity. Key outcome variables will be analyzed and reported by language group where sample sizes permit, with language examined as a potential moderator of QoL outcomes using a Mixed Model analysis while allowing for the interaction of Time\*Language as a Fixed Effect. Exploratory analyses will also consider potential differences in intervention engagement and retention across linguistic groups to inform the cultural and linguistic appropriateness of the intervention. Exploratory analyses of sub-groups by gender, ethnicity, and socioeconomic-related factors will be conducted; however, the trial is not powered for efficacy in these sub-groups. Health Economic Analyses: We will conduct health economic analyses alongside the RCT. Health economic analyses in this study are exploratory in nature. A healthcare payer perspective will be applied to create estimates of the incremental cost for an additional unit of QoL. Estimates will be summarized using an incremental cost-effectiveness ratio and an incremental net benefit statistic. Healthcare service use data (i.e., healthcare provider visits, inpatient stays, emergency department visits, and psychotropic medication) will be collected using the Matryoshka SNP (based on the Client Service Receipt Inventory), which has successfully been used in evaluations of mental health programs. Matryoshka SNP can also be adapted for different contexts. Unit prices will be based on Canadian or US estimates, as relevant. We will use the EQ-5D-5L for health-related QoL as it is the preferred applied preference-based measure for generating quality-adjusted life years (QALYs). The primary objective of the EQ-5D-5L is to examine the cost-effectiveness of the intervention by calculating the incremental cost-effectiveness ratio (ICER) to help determine whether the intervention provides good value for money. Estimates will be summarized using the ICER and the incremental net benefit (INB) statistic. Estimates uncertainty will be set using 95% confidence intervals and cost-effectiveness acceptability curves. Net benefit regression will be conducted to confirm estimation robustness and uncertainty of findings. Additionally, the net benefit regression approach will assess whether there are differences among sociodemographic and clinical subgroups in terms of the intervention's cost-effectiveness. We will also compare effectiveness outcomes for these different groups. The full list of measures for the RCT and health economic analyses includes: * Brief QoL.BD - primary outcome measure (extracted from the Full QoL.BD below) * Full QoL.BD * Personal Health Questionnaire Depression Scale (PHQ-8) * Internal State Scale (V2) * The Brief World Health Organization Quality of Life Scale (WHOQOL-BREF) * Self-Compassion Scale-Short Form (SCS-SF) * "Manage Disease in General" subscale of Stanford's Chronic Disease - Self-Efficacy Scale (SEMCD) * User Engagement Scale-short form (UES-SF) * Matryoshka Service Needs Profile (SNP) (based on the Client Service Receipt Inventory) * 5-level EQ-5D (EQ-5D-5L) Trial Steering and Data and Safety Monitoring Committees: A trial steering committee has been established. The steering committee will meet monthly to update progress, review recruitment, and ensure adherence to the study protocol. An independent Data Safety and Monitoring Committee (a psychiatrist, biostatistician, and patient advisor who are otherwise not involved in the study) has also been established. This committee will review recruitment data, monitor adverse event reports and participant distress, and ensure these are managed in accordance with operating procedures as outlined by UBC. Qualitative Methods: Post intervention completion, semi-structured individual interviews will be conducted via Zoom teleconferencing with a purposively selected subsample of participants (approx. n=30) invited to participate based on app engagement (high and low engagers), baseline QoL.BD scores, country, primary language (i.e., the language they chose to use the app in), and sex and gender. Interviews will assess app acceptability and engagement throughout the intervention period. Interviews will also help to provide clarity on the quantitative outcomes by examining participants' use of self-management strategies and subjective impacts. Analysis will involve reflexive thematic analysis to analyze and identify common themes for patterns and meaning in the data. Transcripts will be familiarized through repeated reading and analyzed using open coding to identify meaningful units of text. An initial coding framework will be developed using inductive approaches. Codes will then be organized into themes, reviewed for coherence, and clearly defined. Reflexivity will involve explicit consideration of how analytic assumptions are shaped through in-depth engagement with the data. Supported by detailed documentation and interview notes, this process will enhance rigour and contribute to an audit trail of analytic decisions and study procedures to support transparency and reproducibility. All analyses will be conducted using NVivo (version 15; QSR International). Participants that request to be withdrawn from the study will be asked to complete an exit survey. They will be informed that completing the exit survey is entirely voluntary, and that they may withdraw from the study without completing it. If participants express significant distress or suicidal ideation during study procedures, the research team will follow a predefined risk management and distress protocol.