Before outlining the trial design in detail, we wish to be transparent about the adjustment during the trial. Recruitment began in March 2023. However, by April 2024, enrolment was much lower than expected (N = 57 April 2024; we expected over 100 participants). In response, we implemented two key adjustments: 1) two GR centres and two hospitals were added, and 2) recruitment was extended by nine months. The following sections outline the original plan and the adjustments made.

The study is a partially cluster randomised stepped-wedge trial. The intervention was developed and tested following the steps of the first two phases of the MRC framework.²⁴ This study contained the third phase of the MRC framework, following the steps of the evaluation phase).^25,26

Table 1 provides an overview of the trial design. Initially, the clusters contained the nine GR centres. Every cluster began with care, as usual, and data were collected for the control group. Every three months, a cluster was initiated for the OTHER intervention, and all participating occupational therapists in the cluster delivered the OTHER intervention to new participants after stroke (see Table 1). Data was collected during the intervention period. Each cluster continued the OTHER intervention till the end of the trial. However, two hospital settings and two GR centres were added after one year of inclusion into the trial. They were formed based on geographical proximity, with Clusters 1, 3 and 6 comprising two Geriatric rehabilitation Centres (GR Centres), Cluster 7 and 8 containing the hospital setting, and the remaining clusters containing one GR Centre (see Table 1).

The participating GR centres were Vivium zorggroep (Naarden), GRZPLUS organisation, which is a collaboration with GR centres Omring and Zorgcirkel, Stichting Cicero Zorggroep (Brunssum), Stichting Tante-Louise (Bergen op zoom), Sevagram (Heerlen) and ZZG Herstelcentrum voor revalidatie en zorg (Groesbeek).

After one year of inclusion, the inclusion rate of six persons post-stroke per cluster every three months was not reached. We therefore adjusted the design by adding two additional clusters: one cluster containing two GR centres, treatment Centre SZR (Tiel) and AxionContinu (Utrecht), and two clusters each containing one hospital, OLVG (Amsterdam), Amsterdam Universitair Medisch Centrum and Canisius-Wilhelmina hospital (Nijmegen). We expected two to three participants per cluster every three months, starting one year after the inclusion date. Table 1 presents the designs and their corresponding clusters.

Baseline data were collected within seven days after inclusion of a participant into the trial, 4, 13, and 26 weeks after discharge from clinical rehabilitation or from the hospital. Figure 1 shows the study flow for participants.

The participants in this study were post-stroke persons admitted to one of the nine participating GR centres who were starting inpatient rehabilitation to return home (see Table 1), or those discharged from the participating hospitals (see Table 1). The persons post-stroke were eligible if: 1) they could walk at least a few steps with or without a walking device, 2) were 60 years or older, 3) had a score of at least 16 on the Montreal Cognitive Assessment (MoCA), and 4) had an indication for follow-up home-based GR. Persons post-stroke were ineligible if: 1) they were terminally ill, or 2) they could not understand verbal information (usually due to severe aphasia).

Written informed consent is obtained from all participants — including persons post-stroke, partners/caregivers, and participating occupational therapists — prior to their inclusion or participation in any study activity. Participants receive written and oral information and have a reflection period of up to five days before deciding whether to participate.

The process evaluation was conducted using mixed methods, guided by the MRC framework for process evaluation.²⁴ Quantitative measurement was used for 1) treatment fidelity (intervention dose, protocol process adherence) by filling in a logbook after every coaching session by OTHER; 2) level of treatment self-reported enactment of participants; 3) participants’ satisfaction with OTHER; and 4) Implementation of the intervention (training of occupational therapists, manual for occupational therapists and practical application). The professionals recorded the intervention dose in the case notes, and data on protocol process adherence were extracted from a structured list within them.

Qualitative data generation: individual semi-structured interviews with a randomly invited number of persons post-stroke; if the partner or informal caregiver was involved during the OTHER-intervention, they were asked to join the interview (N = 15). Focus groups were used to evaluate OTHER from the professionals’ perspectives on delivering OTHER (N = 8–10). The process evaluation will provide insight into the implementation of OTHER during ongoing rehabilitation at home in this trial, as well as into how persons post-stroke and professionals reflect on the application of OTHER in daily life.

The components of the usual care group, the intervention group starting in the GR centre, and the intervention group of people admitted from the hospital to home care are described below.

Inpatient GR begins with an intake, which may be conducted by the geriatric centre’s elderly care physician alone or as part of a multidisciplinary team. After the intake, assessments are conducted, followed by treatment from the multidisciplinary team. Most teams consist of an elderly care physician (a physician specially trained in the medical care of frail or older patients), a nurse, a physical therapist, and an occupational therapist. Other allied health professionals, such as dieticians, social workers or psychologists, are consulted on demand. The elderly care physician coordinates the participant’s multidisciplinary care and treatment plan.

After assessments and observations, a multidisciplinary care and treatment plan is designed at the first multidisciplinary meeting and discussed with the person post-stroke and their primary caregiver(s). All persons post-stroke follow a personalised, multidisciplinary rehabilitation program, tailored to their goals and preferences. In this multidisciplinary meeting, a preliminary indication for home-based GR is made. Home-based GR depends on the person’s post-stroke needs and on how care is organised in the person’s area. Home-based GR can include allied health professionals, such as occupational therapists, physical therapists, and dietitians, as well as nurses, social workers, or psychologists. Occupational therapy is sometimes involved in home-based GR, but not for all persons post-stroke and not for all organisations. Persons post-stroke are discharged home if they can safely function at home, either independently or with (in) formal care assistance.

When persons post-stroke were included for OTHER, they received the OTHER intervention after a short period of inpatient rehabilitation in a GR centre or after discharge from the hospital to their homes. OTHER is an intervention that combines activity monitoring, coaching at the person’s home, and hybrid coaching. The intervention is based on the Self-Determination Theory,^27,28 on principles of solution-focused brief therapy^29–32 and occupational performance coaching.^33,34 Self-Determination Theory highlights the importance of autonomy, competence, and relatedness, fostering intrinsic motivation by enabling participants to set meaningful goals and exercise control over their rehabilitation process.^27,28 Solution-Focused Brief Therapy adopts a strengths-based perspective that prioritises solutions over problems, encouraging individuals to recognise progress and build on their existing capacities.^29–32 Occupational Performance Coaching is a person-centred approach that integrates emotional support, structured goal-setting, and knowledge sharing to enhance individuals’ performance, self-efficacy, and problem-solving skills in performing meaningful daily activities.^33,34 Collectively, these theoretical frameworks ensure that the intervention is person-centred, empowering participants to take an active role in their rehabilitation and sustain engagement in everyday life. Collaborating with the person post-stroke is fundamental throughout the intervention and constitutes a core component of Occupational Performance Coaching.^29–32 The activity monitoring utilises a validated digital platform, and coaching is conducted either face-to-face or online via videoconferencing.

The intervention for persons post-stroke in GR was intended to start before discharge (at least three weeks before discharge) and with a follow-up at home during 12 weeks. Persons post-stroke who were discharged from the hospital started as soon as possible with the intervention when they were at home and for 12 weeks.

An occupational therapist who delivered the OTHER intervention coached persons post-stroke during rehabilitation post-discharge at home to improve daily functioning. During coaching by the OTs in OTHER, the person post-stroke will be empowered in self-management and in improving health-related quality of life.

The activity monitoring intervention tool (Hipper https://hippertx.nl) includes a wearable physical activity monitor (PAM AM300) (https://pamcoach.com) and a gateway (Raspberry Pi). The PAM consists of a 3-dimensional accelerometer, is worn on the hip, and measures 1) the amount of all daily activities in minutes per day and 2) the acceleration of body movements. The measured movements are expressed in a PAM score, representing the energy ratio expended during physical activity compared to resting energy. The PAM communicates with the gateway placed in the person’s home via a Bluetooth adaptor (WR300-E). Via a web application, users can view visualisations of their data on a tablet, phone, or computer.

Once it is clear that a post-stroke person is going home, they will start wearing a PAM during inpatient rehabilitation throughout the day (from waking up until they go to sleep at night). Through this introduction to activity monitoring, the person learns, with the help of the occupational therapist or a nurse, to wear the sensor every day. The therapist and person post-stroke monitor activities via a secure web application. The therapist uses the data as feedback to coach the person once a week during a coaching session. The components of the intervention are described from the GR centre ( Table 2) and the hospital ( Table 3).

The coaching takes place once a week during the rehabilitation period. Physical coaching sessions at home alternate with online coaching. These are based on the five coaching steps ( Figure 2). The web application provides the therapist and the post-stroke person with access to the collected data via a personal login. The issues and priorities related to daily activities that are relevant and essential to the person are the point of departure for the coaching sessions. The sensor data can be used to discuss current activity levels across daily or weekly periods.

The focus of the coaching is to help persons post-stroke consider possibilities and take small steps to increase their daily activities.

The five coaching steps are:

The basis of this coaching is creating a partnership to work together on an equal footing. According to the theory, this basis is necessary at every step of coaching to ensure effectiveness and optimally support self-management. This basis is a principle of occupational performance coaching³³ and self-determination theory,^27,35 with step 1 also being based on solution-focused brief therapy.^29–31

Step 1: Reflection on the desired situation: daily activities that are important for persons post-stroke, the person’s overall activity during the day, and how this affects their well-being.

Step 2: Goal setting based on the desired situation: the occupational therapist supports the person post-stroke in setting goals that reflect daily activities and participation that are important and meaningful for the person post-stroke.

Step 3: Plan daily activities: by translating goals and daily activities into small, achievable steps. Also, a coping plan is made to address barriers and facilitators and act on these. Working on small steps is based on solution-focused brief therapy.^29–31

Step 4: Doing activities: The person post-stroke will perform daily activities to see what works and take small steps without the occupational therapist. The occupational therapist will support the person; daily activities will be discussed and evaluated. The person can do the activities alone or with support from the occupational therapist if necessary.

Step 5: Evaluating daily activities: This step evaluates the performance of daily activities, self-management in daily life, the person’s activity level in general throughout the day, and appointments for the next (small) steps.

The sensor data reports are a starting point for the conversation about the daily patterns and activities essential to practice, and for making new, realistic plans for following activities based on these objective reports and reflections on the person’s experiences. The daily and weekly reports can also be used to evaluate rehabilitation progress.

In addition to coaching at participants’ homes, online coaching sessions will support post-stroke persons. Occupational therapists will use the software available at the GR organisation to book and start a videoconference. The Hipper system was available for persons post-stroke during this trial.

Every three months, the occupational therapists in a cluster that was about to switch to the OTHER intervention received the OTHER training. After completing this training and new inclusion, they started treating persons post-stroke with the OTHER intervention. The OTHER training consists of a two-day course followed by a one-day refresher. During the course, all occupational therapists received information about the study, the procedures, and the six-step coaching, as well as guidance on how to utilise the sensor to instruct and coach persons post-stroke (both face-to-face and via videoconference). In preparation for the course, professionals received a manual with this information and were trained on the technical aspects of the activity sensor and its use with the web application. Lastly, the professionals received training on the trial’s procedures and participant recruitment.

After the feasibility study³⁶ adjustments were made to the occupational therapist’s training to explain the concept of ‘OTHER’ more clearly for participants, to practice using sensor data and coaching techniques more effectively, and to plan consultations after the training.

Primary outcome

The primary outcome measure is ‘daily performance’, 4 weeks, 12 weeks and 24 weeks after the start of OTHER compared to baseline functioning, measured with the performance scale of the Canadian Occupational Performance Measure (COPM).³⁷ The COPM is a person-centred, occupation-focused outcome measure for detecting changes in perceived daily performance over time. The COPM results in a performance score (COPM-p) and a satisfaction score (COPM-s). Research has shown that the COPM exhibits excellent test-retest reliability and measures changes in the performance of daily activities.³⁷ Through a semi-structured interview, persons will prioritise up to five daily activities deemed most important and rate each on a 10-point scale regarding perceived performance (COPM-p) (1 = unable to do at all and 10 = able to do exceptionally well). The mean COPM-p will be obtained by dividing the ratings by the number of prioritised activities. Change in scores can be calculated after a reassessment interval to measure the change in the perception of daily performance. A 1.3-point difference between pre- and post-measurement indicates a minimally clinically important difference.^37–39 A trained research assistant completed the COPM interview and scored the results in this study.

Secondary outcome

Healthcare costs will be valued using Dutch standard costs and prices derived from www.medicijnkosten.nl. Informal care (i.e. care by family and friends) and unpaid productivity losses (i.e. costs associated with reduced productivity levels related to unpaid activities, such as volunteer work) will be valued using a recommended Dutch shadow price. Absenteeism will be valued according to the Friction Cost Approach and using gender-specific price weights. Presenteeism will be valued using gender-specific price weights as well.⁵⁰

Besides the outcomes mentioned above, at baseline, data was collected on:

The process evaluation was conducted using a mixed-methods approach, and the process evaluation of complex interventions from the Medical Research Council guidance will be used.²⁵ Quantitative measurement was used for 1) treatment fidelity (intervention dose, protocol process adherence) by occupational therapists filling in a logbook after every intervention by OTHER; 2) level of treatment enactment of persons post stroke from individual interviews and questionnaire during measurements at home; 3) participants’ satisfaction with OTHER by a questionnaire during the measurements at home (one month, three months and six months after discharge); and 4) Implementation of the intervention (training of OT, manual for OT’s and practical application). Details on intervention dose were recorded by the occupational therapists in the logbook, and data for protocol process adherence by the professionals were extracted from a structured list in the logbook (see Table 4).

Qualitative data generation: individual semi-structured interviews with a randomly invited number of persons post-stroke; if a partner or caregiver was involved during the OTHER-intervention, they will be asked to join the interview (N = 15). Topics of the interview included satisfaction with the intervention, content, delivery, results of rehabilitation on daily functioning and participation in joint goal setting, tailoring of the rehabilitation, as well as barriers and facilitators. A focus group was conducted to evaluate OTHER from professionals’ perspectives on delivering OTHER, the mechanism of impact, contextual factors influencing the intervention, barriers and facilitators of implementing OTHER during this trial (N = 8–10).

Table 4 gives a detailed overview of outcome measures at each time point.

Trained research assistants conducted all outcome assessments for GR centre: at baseline, 4 weeks home (T1), 12 weeks at home (T2) and a follow-up 26 weeks at home (T3). Measurements for the effect are taken at baseline in GR centre, and subsequent measurements are conducted at home. For the cost questionnaire, a second phone call appointment was made to complete the questions. The feasibility study made clear that the measurements are exhausting for the participants. For this reason, measurements were split up. Data were obtained from the electronic patient record (EPR) through standardised questionnaires and assessments. All data were entered into the CASTOR database in accordance with good clinical practice guidelines, using an identification code for each participant.

The initial sample size calculation at the beginning of the recruitment yielded a sample size of 180 based on the following assumptions:

A new sample size calculation was performed one year after the start of recruitment because the number of participants included was much lower than we had aimed for. Using the exact same assumptions as above and adding an additional assumption of an average cluster size of 2 or 3 participants in the new GR centres, the sample size went up to 114 (137 assuming 20% loss to follow-up) for the care as usuals group and 171 (205 assuming 20% loss to follow-up) for the OTHER group respectively, with corresponding powers of 0.7843 and 0.9093, respectively. For more information about the STATA command used to calculate sample size (see S1).

Figure 1 shows the flow of clusters and participants post-stroke through the trial. We deemed obtaining equal amounts of patient outcome information for each treatment condition as more important than unrestrictedly randomising the five clusters for two reasons. First, randomisation of so few units does not guarantee baseline comparability. Second, and related to the first reason, statistical adjustment of treatment effect estimates will be needed, which is more effective if both treatment groups have comparable sizes. Thus, assignment of clusters to the five initial sequences was performed four weeks before the start of data collection by GtR, a methodologist, not involved in the day-to-day logistics of care delivery in three steps: (i) the five clusters were ranked in order of the numbers of patients we expected them to recruit (based on prior administrative information obtained from the centres); (ii) the two largest clusters were manually put into sequences 1 and 5, respectively; the intermediate-sized cluster was manually put into sequence 3; (iii) we randomly allocated the remaining two (smallest) clusters into sequence 2 and 4 using the sample command in STATA 18.

Comparability at the level of clusters and participants immediately after the assignment of the sequences will be assessed to detect potentially important imbalances in prognostic indicators. Descriptive data will be used to determine any time trends of persons’ post-stroke characteristics at recruitment, since patient selection bias is a threat in cluster trials that cannot be blinded for allocation. Except for one secondary outcome, all outcomes will be analysed using mixed linear regression analysis on measurement times 4, 13 and 26 weeks (with baseline values as a covariate). The crude (unadjusted) treatment effect and its effect after adjustment for a set of relevant confounders will be estimated. Additional decisions on relevant confounders (variables that appear unbalanced at baseline and are not in the confounder set) will be made using the 10 per cent change-in-estimate criterion, which works well in linear regression (as opposed to its use in logistic regression).⁵³ Participating centres will be represented as dummy variables in all analyses. Models will include a random intercept for participants. Random slopes will be fitted to assess the improvement of the model fit. Using information from regularly updated and semi-standardised occupational therapists’ logbooks and repeated measurements of the main time-varying confounders, we will analyse the treatment effect (if any) as a function of treatment fidelity and, if possible, participants’ adherence using structural mean models.^54–56 Baseline comparability of participants will be assessed. Baseline assessments will be summarised using interquartile ranges (for continuous and ordinal variables) or percentages (for proportions).

The main analysis focuses on a comparison between the intervention group OTHER and the care as usual concerning the primary outcome, the perceived daily function scores, with scores at T1, T3 and T6 as dependent variables and the baseline score as a covariate. The differences between the two groups will be evaluated using a mixed linear regression analysis. Two-sided 95% confidence intervals will be calculated. An intention-to-treat analysis will be conducted. The most recent versions of SPSS, STATA or R-CRAN will be used to analyse the data.

The treatment effects on the secondary outcomes will be estimated using the same multi-level approach. Survival data (institutionalisation and mortality) will be analysed using flexible parametric survival models.^57,58 Restricted mean survival models will be used as a sensitivity analysis where appropriate.^59–61 Two-sided 95% confidence intervals will be calculated. Following the above-mentioned approach, the ordinal outcome (ICECAP-O) will be analysed using mixed-effects ordinal logistic regression analysis.

Cost-effectiveness analyses will be performed according to the intention-to-treat principle. In the main analysis, missing cost and effect data will be imputed using Multivariate Imputation by Chained Equations (MICE).⁶² Rubin’s rules will be used to pool the results from the different multiple imputed datasets. Multilevel analyses will be performed to estimate cost and effect differences between the OTHER-intervention and usual care, while adjusting for confounders if necessary. In doing so, the stepped wedge design of the trial will be considered. Incremental Cost-Effectiveness Ratios (ICERs) will be calculated by dividing the differences in costs between groups by the differences in the primary outcome (i.e. COMP-p) and QALYs. Bias-corrected and accelerated bootstrapping with 5000 replications will be used to estimate 95% confidence intervals around the cost differences and statistical uncertainty surrounding the ICERs. Uncertainty surrounding the ICERs will be graphically presented on cost-effectiveness planes. Cost-effectiveness acceptability curves will be estimated, showing the probability that OTHER-intervention is cost-effective compared to usual care for a range of different ceiling ratios, thereby showing decision uncertainty. Sensitivity analyses will be done to assess the robustness of the results (e.g. complete-case analysis, per-protocol analysis).

Process evaluation will be performed based on a mixed-methods design. Methods for qualitative analysis follow the constant comparative method. The individual and focus group will be transcribed verbatim and imported into MAXQDA. The researcher will read the data to get familiarised. Two researchers will perform the analysis and will apply open coding techniques derived from constant comparison methods. Open data coding was conducted by the researcher and a part of the transcript was also coded by the second researcher followed by comparison and discussion of the codes in order to reach consensus on the coding procedure and content. Both researchers will identify potential categories among initial codes, and the potential categories will be discussed by members of the research group and further grouped into final, main themes related to the research question. Quantitative data of the process analysis will be descriptive.

Dissemination

Study results will be published in peer-reviewed journals, reported in line with the literature Consolidated Standards of Reporting Trials guidelines for stepped wedge cluster randomised trials.⁶³

Study status

Recruitment started in April 2023. We completed recruitment at the end of July 2025. Data collection continued until the end of January 2026. We included 116 participant, 60 care as usual and 56 OTHER. At the time of submission of this protocol, data analysis is ongoing, and a primary results manuscript is in preparation.