Aha BOOST Arm-hand BOOST Therapy to Enhance Recovery After Stroke: Clinical, Health Economic and Process Evaluation

Data collection and transfer:

• Data will be collected by the clinical sites and by the external researchers from the KU Leuven, VUB and Jessa Hospital. Source documents will be used to collect the data, these can be on paper or digital document.
• Different kinds of data will be collected on paper, for example results of the clinical tests, questionnaires and diaries. A standardized case report form (CRF) will be completed during data collection, containing researchers notes, remarks concerning data quality, contextual information, deviations from the protocol, etc. These CRFs will be kept on paper, in the same folder as the research data that are collected on paper and will become part of the participant's source documentation. These forms will be digitalized by the researchers in the e-CRFs via REDCap. After digitalizing the data, these paper forms will be stored in a locked cabinet in the Department of Rehabilitation Sciences, Building The Nayer, of the KU Leuven. During data collection temporary storage is possible at building K on the health campus in Jette, of the VUB. Only authorized personnel will have access to this locked storage rooms as they will need to be granted access by the PI.
• The data collected by the clinical sites will consists of a combination of paper forms and digital documents. A ManGO platform (hosted by KU Leuven IT department) will be used for collecting these different pseudonymized data forms. The different paper forms and data received via ManGO will be processed to REDCap by the external researchers. After digitalizing the data, the paper forms will be stored in a locked cabinet in the Department of Rehabilitation Sciences, Building The Nayer, of the KU Leuven. During data collection temporary storage is possible at building K on the health campus in Jette, of the VUB. Only authorized personnel will have access to this locked storage rooms as they will need to be granted access by the PI. The recorded video's will be transferred via ZIVVER, after which they will processed by the external researchers and deleted both on the computer of the external researcher and in the clinical sites.
• Any participant records or datasets that are transferred to the Sponsor or any partners of the Sponsor will contain the study-specific participant identifier only; participant names or any information which would make the participant identifiable will not be transferred. All pseudonymized data relating to the Investigation must be transmitted in a secure manner to the Sponsor or any partners of the Sponsor via the ManGO platform or Zivver. Interim reports and a final report at the group level (without individual patient data) will be provided to the funder (FWO).

Power calculation:

• A power calculation was done based on the primary research question (evaluating the short-term clinical effect of Aha BOOST intervention versus the control intervention on arm-hand capacity). This give the following results: In the proof of concept RCT, an average difference of 7.8 points in ARAT between both groups, in change from pre- to post intervention was found. Furthermore, the standard deviation of change from pre- to post intervention was estimated to be 11.7 points for the ARAT scale. With 74 patients, the comparison between experimental and control group for change from pre- to post intervention will have 80% power to detect an average difference of at least 7.8 points at the 5% two-sided level of significance, assuming a standard deviation of at most 11.7 points.
• To account for an expected dropout of at most 8%, 80 patients will be recruited.

Statistical analysis:

1. Clinical data: Changes in ARAT-scores from baseline between patients in the experimental and control group will be compared. Change from baseline will be estimated in both groups based on a multivariate ANOVA model for the original ARAT scores, with time and treatment as main effects and with time by treatment interaction. As a further sensitivity analysis, the analysis will be repeated correcting for patient characteristics such as age, time post stroke, and cognitive impairments. Secondary outcome measures: Multivariate ANOVA and sensitivity analyses, as described above. Subgroup analyses: pre-specified subgroup analyses will be undertaken to explore the effects of the interventions in different types of stroke survivors, such as experiencing cognitive impairments, using appropriate caution about multiplicity in the interpretation of these results. If there appear to be different effects in different subgroups, this will be investigated using interaction or trend tests rather than the statistical significance of the result for the individual subgroup.
2. Health economic data: The economic evaluation will incorporate costs and health gains during the trial and follow-up period, adopting healthcare and societal perspectives. On the cost side, the evaluation will focus on the direct medical and non-medical costs, and on indirect costs depending on the perspective. The direct medical costs encompass all costs for treatment and follow-up from the healthcare perspective and all out-of-pocket contributions by the participant. Direct non-medical costs include transport costs, and home care help, whereas indirect costs include productivity loss (e.g. the number of days away from work). Productivity losses due to informal care will be documented and valued using the human capital approach. The effects are expressed in utilities, derived from the national values of the EQ-5D-5L. The cost-effectiveness of the intervention will be expressed in incremental cost per QALY (quality-adjusted life years), using a decision-analytic model based on the trial data with a time horizon of one year and on a lifetime horizon. The incremental cost per QALY will be calculated as a ratio of (Cost Experimental-Cost Control) / (Outcome Experimental-Outcome Control). The robustness of the results will be analyzed by probabilistic sensitivity analyses on the cost as well as on the outcome. Tornadodiagrams will be used to measure impact of individual components in healthcare utilization. Probabilistic sensitivity analyses by bootstrapping with replacement will be employed to test the robustness of the results, utilizing MS Excel, using a minimum of 1000 iterations to obtain 2.5% and 97.5% percentiles of the incremental cost-effectiveness ratio (ICER) distribution. In a second phase, the health economic model will be further extended from a one-year time horizon to a life-time horizon based on literature. A Markov-model is developed defining the health states in chronic stroke. Estimations on resource use and utilities for each health state are derived from international peer-reviewed literature. Special attention is paid to the transferability of the data to the Flemish healthcare context. This will done by validation checks within the research consortium and advisory board. Similar to the one-year model, probabilistic sensitivity analyses will be used to account for uncertainty around the input parameters
3. Process evaluation data: Qualitative data will be analyzed using content analysis (a deductive analysis based on the defined indicators) and thematic analysis (an inductive analysis where themes will be generated from the data by the researcher(s) using a generic qualitative approach). The aim is to interview +/- 16 interviewees to achieve meaning saturation per respondent group (patients, expert therapists and treating physicians). Approximately 48 interviews will be conducted, although the final numbers will be determined by saturation. Interviews will be audio recorded, transcribed at verbatim and analyzed in NVIVO. Different types of data-triangulation, by using different methods or data-sources to gain a more in depth understanding, will be applied: (1) methodological triangulation e.g. using a combination of interviews and observations; (2) data triangulation e.g. patients, Aha BOOST therapists and treating physicians; (3) theoretical triangulation e.g. involving researchers of different disciplines to look to the data from different perspectives; (4) investigator triangulation e.g. involving different researchers in data analysis.