Effectiveness of a Customized Digital Platform to Increase Coordination of Care and Uptake of Evidence-based Practices

Research questions and sites The research questions guiding this study are: 1) What are the potential barriers and facilitators in the implementation of the BRILLIANT platform at the local teams' level?; 2) What is the impact of the BRILLIANT platform compared to usual cardiovascular and neurological care on the primary outcome, intensity of rehabilitation (direct minutes of rehabilitation per week); 3) Does the platform have an impact on secondary outcomes: time to be admitted to rehabilitation (inpatient or outpatient), health-related quality of life, care experience, and if found to be effective, costs?; 4) What are the implementation factors that explain potential impact (e.g., relevance, acceptability)? An integrated knowledge translation (iKT) approach will be used by involving all relevant partners throughout the research process. The partners are stakeholders within four health regions with two to four programs each in various rehabilitation stages for cardiovascular and neurological care in Quebec. In four health regions and 8 cardiovascular and neurological programs, this implementation study follows two phases aiming answering the 4 research questions.

Study phases and procedures Phase I is for measuring and addressing the potential barriers and facilitators related to the implementation of the BRILLIANT platform and prepare teams for implementation. With each program investigators will review the workflow (clinical processes and the material and human resources available) for cardiovascular and neurological care pathways, and use the Business Process Model and Notation (BPMN) method and analyze them to maximize successful implementation during the trial. The purpose of this first phase is to determine what the BRILLIANT platform will look like. Through focus groups with 8-10 participants per program comprised of all health professions, coordinators, managers, and patient/family participants with experience with different phases of the rehabilitation process. In the focus groups, investigators will review how the platform will work, validate certain tasks, and ensure ease of use so that investigators may adjust as needed. Identifying barriers and facilitators to implementing the platform during the focus groups will be guided by the Consolidated Framework for Implementation Research (CFIR). Focus group sessions will be recorded and transcribed, and further analysed to feed back information on barriers and facilitators during subsequent design workshops to also inform adaptation and implementation of the platform.

The second phase is to evaluate the impact of the BRILLIANT platform on primary (intensity of rehabilitation, i.e. direct minutes of rehabilitation per week), and secondary outcomes (time to be admitted to inpatient or outpatient rehabilitation, health-related quality of life, care experience, and costs if the platform is found to have an impact) and evaluating the factors that impact implementation. The control condition will be usual care, which consists of an interdisciplinary team of physiotherapy, occupational therapy, nursing, medicine, psychology, speech therapy, social work, and nutrition. Clinicians mostly work on paper evaluating patient needs (mostly by checklist) using static standardized forms (e.g., Functional Independence Measure, FIM, at admission) without reminders, scoring and interactive functionalities. The choice of standardized assessment is left to the discretion of each clinician. Since the pandemic Microsoft teams is used to share scanned documents between clinicians and remote program meetings. During the control period clinicians will have access to static forms on the BRILLIANT Platform.

Intervention details The intervention will be the use of the BRILLIANT Platform. The platform provides five Modules aimed at increasing coordination of care and direct rehabilitation time (Table 1): 1) Standardized measures: provided consistently with scoring and reminders to complete assessments, across all stages of recovery and care settings to guide and evaluate treatment effectiveness of clinical outcomes (e.g. FIM, Mayo Portland Adaptability Inventory - MPAI), performance based outcomes (e.g. gait speed), patient reported outcome measures (e.g. pain level during walking); 2) Communication Module: clinicians can communicate together and with patient/family to clarify patient-related information (e.g. safety concerns, intervention plan progression) and update colleagues for time sensitive information, synchronously (scheduled tele-videoconference) and asynchronously; 3) Shared intervention goals and plan: personalized and dynamic care trajectory and intervention plan guided by standardized assessments, reminders, and evidence-based interventions recommendations to direct patients and caregivers to the right services and interventions (e.g. treadmill walking practice) at the right time based on clinical and social profile and best practices;1,2 4) Self-management support: To provide patients and families with self-management dashboards (e.g. physical activity tracking and reminders) and interactive online intervention plan (goals, action plan, monitoring progress) to actively engage the patient and family in setting goals with their care team. Access to evidence-based patient education and resource modules, tailored to stage of recovery (e.g. Education on how to handle arm or shoulder); 5) Quality Improvement Dashboards to provide feedback and reports allowing observation of clinical/administrative/clinical performance indicators (based on MSSS indicators, Quadruple Aim framework) for continuous program evaluation.

Each health region will be provided with cloud-based access to the platform as agreed in collaboration with their IT units. Each site will have support from a knowledge broker from the research team to promote the platform's purpose, features and benefits and build understanding, and engagement of the local programs. Pre-implementation, there will be: 1) a 2-hour training session (respecting COVID restrictions) in each site and an online training video that participants can access anytime; 2) a platform guidebook explaining each functionality, that can be accessed anytime.

Based on ongoing implementation of new practices in these programs the investigators will use the following six evidence-based implementation strategies: 1) organize implementation meetings; 2) provide educational materials (e.g. videos) on platform use; 3) conduct ongoing training for new platform users; 4) audit and feedback of optimal use indicators to platform users; 5) engage health region digital health leads in the implementation process; 6) provide local assistance for rapid resolution of technical issues. The duration of follow up is 12 months to allow time for sites to practice and adopt the platform and increase likely impact on change in rehabilitation intensity and outcomes given the recovery and clinical process post ABI. Twelve months are necessary to account for differences across sites that stem from admission rates.

Randomization After a period of 2 months, when all programs have access to the BRILLIANT platform only with static forms (control), randomisation of the first 2 sites will start. Programs will enter the trial in four consecutive waves based on stratified randomisation controlling for health region and type of program (cardiovascular or neurological). This will help to minimise, in part, temporal trends due to region and program, if any exist. Each step will enroll 2 programs at a time (paired stepped-wedge), to allow the training and trouble-shooting period to take place in a limited number of sites at a time. The enrolment process will take 27 months, allowing for a minimum 12-month follow-up in the last two programs receiving the intervention. Programs will receive sufficient notice of the dates to cross over to the intervention.

Knowledge Translation Plan The four health regions, patient, and industry partners have outlined their commitment to this project. The investigators will have knowledge exchange sessions between health regions to share best practices throughout all phases of the project. The iKT and Living lab approach entails continuous co-creation and exchange throughout the study. For this study, the primary end-of-grant KT activities will consist of 1) a stakeholder-generated report integrating all our findings; and 2) a one-day symposium, where the investigators will present this report to the stakeholders including coordinators/managers decision-makers of health organizations and discuss implications and next steps for testing the platform more broadly across Canada and internationally with team's VITALISE network. In addition to traditional end-of-grant methods (e.g., conference presentations, open access publications), the investigators will also post regular updates and tailored executive summaries of findings on social media and on the CISSS and CIUSSS websites. This will be coupled with an end-of-project webinar to share results of the team's work. Training tools and sessions described above available online to help knowledge users build capacity for the use of the platform.

Data analysis Data will be analysed by type of stakeholder and program. The investigators will analyse the transcripts using content analysis techniques based on a deductive directed approach typically done in studies using the CFIR. While it will not be possible to sample for gender or sex in the composition of site implementation teams, the investigators will explore differences in the interview questions, analysis of qualitative data and modeling of quantitative data using participant self-identified gender. Sex and gender influence 1) behavior, 2) individual and organizational decision-making, 3) implementation and 4) may influence the platform design, how and why it is implemented, and use of the platform by clinicians and patient/families. As such the investigators will consider sex and gender in 1) development of knowledge tools for each site; 2) in development, usability testing, and interviews to evaluate facilitators and barriers and mechanisms for adoption and 3) in the usability testing in the design phase which will include 8 women and 8 men per stakeholder group (to explore gender/sex differences qualitatively).

Based on the one-year program indicators (2019-2020), the investigators have on average 569 patients with stroke and 393 with TBI, and total of 962 across the four regions and 8 programs. Data for all participants for the primary outcomes will be obtained from the SIPAD administrative database. It is anticipated a conservative estimate of 60% of patients (n=577) will consent and have complete data for research based on previous studies in the same settings and available to evaluate the secondary outcomes. In total, 28 managers and 60 clinicians across all programs were contacted. The investigators expect nearly 100% completion of questionnaires by clinicians as was found in a recent study in the programs.

Sample size:

The anticipated sample size is 8 programs (clusters) and a total of 962 participants for the primary outcome and 577 for the secondary outcomes. For the primary outcome, the study will have 80% power to detect standardized mean differences in direct rehabilitation time (primary outcome) of >=0.4 between control and implementation periods. This is based on a conservatively assumed intra-cluster correlation (ICC) of 0.20, and allowing for a coefficient of variation in cluster sizes of 0.42 (SD=50)/(mean=960/8=120),77,78 and taking the ICC and the design effect of 1.2 for cluster size variations ([1+(0.42^2+1)*120-1)*0.2] / [1+(120-1)*0.2] into consideration, the effective cluster size is n=100. This takes into account the stepped wedge design and the confounding effect of time (i.e. for a continuous outcome: time of direct rehabilitation; i.e. the investigators will be able to detect a change as small as 0.35 times the standard deviation of the outcome). The power calculations consider mixing both new patients and existing patients across trial steps (cluster autocorrelation=0.80). For the secondary outcomes, with a power of 90% the study will be able to detect a standardized effect size of >=0.50 (effective cluster size n=72/1.2 = 60 to account for heterogenous cluster sizes i.e. a total of 60 x 8 =480 observation points).

Apart from basic descriptive statistics, the analysis approach will be intention-to-treat using a generalize linear mixed effects (GLM) model. This model estimates the effect of order (intervention duration) on outcome accounting for clustered and correlated nature of the data (repeated assessments within nested levels of clusters: patients and sites which will be modeled as random effects). A log-link GLM will be employed to account for right-skewed time data (e.g. minutes per week and days of therapy). Estimated effects will be reported with 95% confidence intervals. For the secondary outcomes related to experience and health outcomes, a composite outcome approach will be used testing the effect of order on outcome independent of the source (experience) or measure (HRQL) used for that outcome. This approach increases power and reduces the need to consider multiple comparisons. As the effect of order (source, measure) may vary by time owing to secular effects, interaction terms with time will be included although not hypothesized to have a strong effect. Table 2 outlines the analysis plan for primary and secondary outcomes. The term for measures provides information on which measures are more contributing to the composite.

The general form of the model will be Outcome = order of randomization + time of assessment + order*time. The order*time interaction estimates whether the effect on the outcome of order (more or less exposure to the BRILLANT platform) depends on time which may happen if secular trends affect the impact of the intervention. Investigator's hypothesis is that this is unlikely, simplifying the analysis. For the secondary, experience outcomes and the HRQL outcomes, a composite analysis will be done where the source or measure is considered another cluster. The general form of the model is Outcome construct = Source (Who) [or Measure (What)] + Order + Time + Source [Measure]* Time which estimates whether time influences the association between the measure and outcome.

Economic evaluation will be performed only if an effect of intervention is detected. Additional data collection required is minimal as majority of cost data is from DPS database. Costs will be reported separately as: (1) provider time per patient, in hours by provider type; (2) healthcare utilization, measured as number of ED visits, number of hospitalizations, and hospital length of stay in days, obtained from DSP databases. Although services outside the organization may be missed, Quebec's regionalized healthcare mergers ensure that most patient services within a geographic area are captured. Although investigators may miss service utilization outside of the organization, because of mergers in Quebec, healthcare organizations include most services that one would use in a geographic area and as such, this approach should provide complete data.

Implementation analyses is needed to evaluate mechanisms for impact on outcome. Guided by Proctor's Framework the investigators will evaluate: relevance, acceptability by interview, meeting transcripts and the Technology Acceptance Measure (e.g. intention to use, perceived usefulness, and perceived ease of use), change in practice, feasibility, fidelity, and implementation cost (time for training, equipment); Platform Analytics and logs related to activity duration (comparing dates (first and last activities completed within each functionality), proportion (number of activities completed and total number of activities in each functionality), and clinical notes and communications on the platform will be analyzed thematically and presented descriptively.

Feasibility of implementing the platform There are three conditions in place that support successful implementation of the platform and trial feasibility. The first is stakeholder partnerships: the four health regions that will participate in this study are clinical partners of investigators' research center (CRIR), and each region has a lead investigator and knowledge user. The health regions acknowledges that they accept to be randomised to start using the platform in 2024/2025. The second is that many researchers have their research programs within the participant sites and are members of a larger digital health initiative, BRILLIANT (https://www.brilliant-cfi.ca/). Three of four health regions are a partner of the BRILLIANT CFI infrastructure (in-kind and cash contribution). The third condition, in parallel, there is a provincial mandate to implement Canadian ABI guidelines with local recommendations including standardized assessments and criteria for programs along the continuum of care. The researchers and clinicians participating in this study in each of the four health regions actively participated in the development and adaptation of recent ABI clinical guidelines adopted by the MSSS in 2015.

The guidelines are informing the reorganization of healthcare that will allow for continuity of care services, and more harmonized transitions at different levels of the healthcare system. Currently, the team is working with each region to implement clinical measures and patient reported outcome measures within the BRILLIANT Platform as part of the clinical workflow. The authors are implementing and evaluating the use of the MPAI measure to ensure that the impacts of ABI on physical, cognitive, emotional, behavioral, and social health are considered for rehabilitation planning.

Discussion This stepped wedge cluster randomized trial design aims to evaluate the impact of the BRILLIANT platform on intensity of rehabilitation, time to admission, health-related quality of life, and care experience for patients with cardiovascular and neurological conditions in Quebec's context. Without this trial, digital health platforms will be implemented in the health regions without evidence of their effectiveness. The scarcity of evidence on the mechanisms by which interventions work plagues implementation science. The investigators' longstanding clinical partnerships, the BRILLIANT infrastructure and the research network, provide an unprecedented opportunity for rehabilitation to: 1) inform decision-making in cardiovascular and neurological care by building in evidence-based guidelines in the reminders and recommendations to clinicians and patients; 2) provide evidence on if, and how digital health can improve cardiovascular and neurological best practices related to patient-centered care processes, assessment and treatment, and patient outcomes; 3) monitor the implementation process across health regions; 4) provide detailed mapping of the variability in the way the platform is implemented in each health region and across the cardiovascular and neurological care continuums (hospital, in- and outpatient rehabilitation) and serve as an exemplar to rehabilitation implementation teams; 5) if found to be effective, the trial will also measure the associated cost of implementation and cost savings (e.g., saved coordination and travel time) relative to the derived benefit from using the platform; 6) the implementation evaluation will provide information to inform the transferability and evaluation of the BRILLIANT platform internationally; and 7) when aggregated at a health region level, data from the BRILLIANT platform will inform quality improvement cycles, as part of a learning health system to identify processes and interventions that improve patient outcomes and health.

This study will provide valuable evidence on the effectiveness and feasibility of the BRILLIANT platform in improving rehabilitation care for patients with cardiovascular and neurological conditions. By addressing the challenges and pursuing future directions, the research team can contribute to improving patient outcomes and healthcare delivery.