Focus group discussions (FGDs).

Five FGDs involving twenty-nine community-dwelling chronic stroke survivors were conducted (between September and December 2022) in Kura town of Northern Nigeria, with the aim of identifying the common daily activities that stroke survivors find difficult to accomplish. Kura town is located about 27km from Kano city and its inhabitants are predominantly Hausa farmers engaged in rice farming.

FGD: participants and sampling strategy.

In order to generate diverse information with a range of experiences in performing daily activities after stroke, participants were purposefully selected based on gender blend, age, and from diverse locations within the community. Local community contacts and some identified stroke survivors within the community helped with participant recruitment. All stroke survivors who expressed interest and met the following criteria were recruited into the study: aged 18 and above; diagnosed with ischemic or haemorrhagic stroke; stroke onset of more than six months; speak and understand the Hausa language; capable of giving informed consent.

Participants were excluded if they had: cerebrovascular events due to malignancy or head trauma; limited comprehension (receptive and/or expressive aphasia) as indicated by having a score of < 80 on the Comprehensive Aphasia Test [15]; been diagnosed with any other neurological/mental disorder.

FGDs: procedure.

FGD sessions were scheduled with those who gave consent and met the inclusion criteria. Four of the focus groups had 6 participants while one focus group had 5 participants. The discussions took place in a convenient classroom in one of the schools within Kura town. Using a semi-structured interview guide (S1) which was developed by two members of the research team (IUL and RI), the participants’ experiences in performing daily activities after stroke were explored.

Discussions were conducted in Hausa language and audio-recorded. Maximum duration of the FGDs was 1h 40 min. The FGDs were facilitated by IUL and field notes were taken by RI. The concepts of data saturation and collection of rich and thick data [16] were used to end the data collection.

Developing the initial task training items.

This was done in two steps; first, the results of the focus group discussions were considered by three physiotherapists (who were part of the research team) with more than 15 years’ experience in stroke rehabilitation. All the tasks that the stroke survivors reported as being difficult were identified. Secondly, the three physiotherapists in collaboration with a clinical kinesiologist (who obtained Master’s degree in kinesiology, also, having more than 15 years working experience in a rehabilitation centre) isolated and grouped similar tasks based on the pattern of movement and the body part (i.e. upper extremities, lower extremities, and the trunk) involved. The training requirements in each group were further categorized based on the activities. This process resulted in the generation of a list of potential tasks (S2) to include in the intervention model which was used in the Delphi exercise.

Selection of panellists.

Using a mixture of purposive and stratified sampling techniques, 30 stroke rehabilitation experts (comprising 10 from Nigeria, 10 from other parts of Africa and 10 from other parts of the world) from diverse professional groups involved in stroke rehabilitation, with at least 15 years of experience in stroke management, were contacted through email for their consent (S3) to participate in the Delphi process. The selection criteria considered for the Delphi panellists included healthcare professionals such as physiotherapists, occupational therapists, neurologists, and rehabilitation specialists who work directly with stroke survivors, as well as researchers who have published studies on stroke rehabilitation or self-rehabilitation strategies. We emphasized the importance of experts with hands-on experience in stroke rehabilitation, particularly those focusing on task-specific strategies, and researchers who have made significant contributions to the field. Additionally, geographical diversity was prioritized to ensure representation from various regions, accounting for cultural and systemic variations in rehabilitation practices. Finally, it was essential that experts were currently active in their fields, as this increases the likelihood that they remain informed about the latest research and practices.

Potential panellists were identified through their published work in the field of stroke rehabilitation and some were suggested by the main study team. These potential panellists were also asked to suggest others who they considered suitable to participate (snowball sampling), and these were also invited, provided they met the inclusion criteria. Our aim was to achieve a panel of 11 to 30 panellists, which is a range of sample size considered effective and reliable for the Delphi technique [18,19].

To determine the expertise of potential panellists, we reviewed their academic publications, presentations, and involvement in research projects related to stroke rehabilitation. We also sought recommendations from trusted colleagues or institutions to identify individuals recognized for their expertise in this area. In cases where a potential panellist’s expertise was difficult to ascertain through these methods, we directly inquired via email about their personal experience, views on rehabilitation, and familiarity with task-specific strategies.

To minimize potential selection bias, a broad selection criterion was employed to include a diverse array of experts from various disciplines, backgrounds, and geographical locations. A stratified sampling technique based on geographical areas (countries and regions) was also utilized to ensure representation from different perspectives. Panellists were asked to disclose any potential conflicts of interest that could influence their views, enabling us to consider excluding individuals with significant biases. Furthermore, we ensured that confidentiality and anonymity were maintained throughout the Delphi process to encourage honest feedback and minimize the influence of dominant personalities. A copy of the set of tasks, including an instruction note (S4) on how to rate the items, was sent to each of those who met the inclusion criteria and gave their consent to participate. All correspondence with the panellists was done via e-mail.

Defining consensus.

Consensus on a topic or an issue can be regarded as the general agreement among a group of people that are well informed (experts) about the topic or issue in question. In the literature, consensus was defined as the “gathering of individual evaluations around a median response with minimal divergence” [20,21]. While experts can reach 100% agreement on only a few issues [22], consensuses are determined by a certain percentage of experts who agree to an issue in a Delphi survey, and this should be defined and stated before the conduct of the survey [23].

In this study, the level of consensus regarding the relevance of task training was measured at the conclusion of both the first and second rounds of assessment. The final level of consensus on the overall suitability of the intervention model was evaluated at the end of the third round. To quantify the consensus on the relevance of each specific task training, we employed a systematic approach. For each task, we calculated the proportion of experts who rated it as 3 or 4 on a Likert scale, where 1 represented “not relevant” and 4 signified “highly relevant.” This calculation involved dividing the number of experts who rated a task training as 3 or higher by the total number of experts participating in the evaluation.

A consensus was deemed achieved if a task training received a score of 0.78 or higher [24], indicating that a significant majority of the panellists recognized its relevance. Conversely, any task training that scored below this threshold ( < 0.78) was either modified for further consideration in the subsequent round or removed from the evaluation process entirely, ensuring that only the most pertinent tasks were advanced.

For assessing the overall suitability of the intervention model, we determined the level of consensus through a similar methodology. Specifically, we computed the proportion of items rated as 3 or 4 by the panellists involved, following the guidelines established by Waltz and Bausell [25]. A score of 0.8 or above was accepted as indicative of moderate to high consensus, based on existing procedural evidence from the literature [22,26–28]. This threshold helped to ensure that the intervention model reflects a broad agreement among experts, ultimately enhancing the reliability and applicability of our findings in real-world rehabilitation settings.

Delphi round 1.

In this round, the panellists rated each item of the model in terms of its relevance to the underlying construct. The item ratings were on a 4-point scale to avoid a neutral and ambivalent midpoint [24] and to produce stable findings in Delphi studies [19]. The rating interpretation was as follows: 1 =  not relevant, 2 =  somewhat relevant, 3 =  quite relevant and 4 =  highly relevant. After all the panellists had finished their ratings, the package was retrieved and computation to determine consensus on each item’s relevance was done. Items that were not rated by at least two-thirds of the panellists were not included in the computation. These were regarded as items on which no consensus was reached and were included in the next round.

For each task item, options for the panellists to provide text comments in support of their rating, reason for abstaining from rating or to suggest changes to the task were available. The panellists were given a period of three weeks to respond, and reminders were sent on a weekly basis.

Delphi round 2.

In this round, the written opinions of all the panellists were summarized and shared among them. Experts were invited to consider their scores for the remaining task trainings that did not make the required item validity score and were not removed, based on group responses in relation to the overall responses received. They suggested what they deemed appropriate between their rating and the average opinion of other respondents. Additionally, further explanation of some culture-based tasks was made available to the panellists, particularly, those who were not familiar with the culture of the intervention target population. Panellists were also asked to give written opinions (S5) on the overall structure and inclusiveness of the included tasks. The item rating and validity score computation was as in round 1.

Prior to the commencement of the third round, an in-person meeting of the study team was held, purposely to decide on some tasks which are peculiar to the culture of the study area and that some panellists were not familiar with, and as a result had abstained from rating. The study team also finalized the structure and inclusiveness of the task training items in the model by considering the comments from the panellists in round 2.

The research team members who are Hausa natives played two major roles. First, they served as cultural brokers by facilitating understanding between Western rehabilitation concepts and Hausa cultural context, and second, they provided insight into Hausa cultural values, customs and healthcare practices, and clarified some cultural issues. It was ensured that all decisions taken during the research team meeting were strictly based on the recommendations and responses from the panellists. All items that had the required score and did not have suggestions made about them by the panellists, were not debated in the meeting and panel consensus was deemed final.

Delphi round 3.

Round 3 involved content validation of the task training items in the model (including those that had been retained in rounds 1 and 2, and those modified or added during the in-person study team meeting. In this round, panellists were invited to reassess the content and rate the whole model using the criteria in round 1. All responses were collected individually.

Data analysis.

FGD: The audio recordings of the focus group discussions were transcribed verbatim and anonymized with codes. The Hausa language transcript was translated into English at the English Language department of Bayero University, Kano. Using the deductive open coding process, two members of the research team (IUL and RI) agreed on an initial flat coding frame, developed a set of predetermined codes based on the study’s objective, both read through the data line by line, and assigned excerpts to codes until themes were developed.

Delphi: Analysis of data was done using Microsoft Excel and IBM SPSS Statistics 23. Analyses involved calculating the proportions of agreement and Kendall’s coefficient of concordance (W). For the first and second rounds of the Delphi, for each task training, the number of experts who gave a rating of 3 or 4 was computed and divided by the total number of experts. A score of ≥  0.78 indicated that consensus had been reached to keep the particular task training [24].

In the third round of the Delphi study, the number of items given a score of 3 or 4, divided by the total number of items was calculated for each panellist (following the methodology outlined by Waltz and Bausell) [25], then the results were averaged and considered. A threshold of ≥  0.8 was deemed a consensus [26–28]. In each round, Kendall’s W was used to assess the level of agreement and statistical significance.

Personal care.

The major activities reported under this theme are categorized into feeding, dressing/undressing and bathing and washing. Many mention the difficulty of performing basic tasks, such as eating, dressing, bathing, and washing clothes, using only their unaffected side. For instance, one person has to eat with their left hand, despite disliking it, while another struggles to tie a wrapper and requires assistance from a family member. Several participants mention relying on a walking stick for mobility, needing help from loved ones for dressing and bathing, or performing tasks like washing clothes on their own.“… I have no option but to eat with the left hand... I don’t like it, but I have to do it.” (Female, 50 yrs, right-sided hemiparesis)

“It is always difficult to tie a wrapper around my waist, because I use one hand, so my daughter has to assist me in tying it.” (Female, 38 yrs, right-sided hemiparesis)

Religious activities.

The participants were predominantly Muslims and many of them reported difficulty in performing ablution (sequential washing of some body parts which is a pre-requisite for performing the Islamic daily prayers) and praying:

“In the aspect of performing religious rituals, I am weak; thank God that now I can bend down during prayers, but I can’t prostrate. Also I perform ablution slowly, the person that assists me is also not feeling well now. So somehow, I manage to do it, even though my hand is heavy.” (Male, 51 yrs, left-sided hemiparesis)

Social participation activities.

The participants describe various mobility challenges due to their hemiparesis, highlighting the limitations they face in daily activities. One person recounts falling off a motorcycle after attempting to use their weaker left leg, which led to frustration.

“… For example, when I want to stand up to walk, I do that from the right side of the body because it is the stronger side. Just not long ago, I fell down while trying to get off a motorcycle. I tried getting down from the left side and placed the left leg on the ground first, but it couldn’t carry me and gave way, so I fell down... it was frustrating.” (Male, 48 yrs, left-sided hemiparesis).

Another participant struggles with climbing, as stepping up with the affected leg causes them to become stuck, though stepping with the unaffected leg first helps. A female respondent discusses how her condition has made it impossible to continue her work, such as frying and selling groundnuts, leading to her dependence on others for basic needs.

“… You go out every day to work and earn a living for you and your family, but now I can’t. I rely totally on others to give me … I used to fry groundnuts and sell, but I can’t do that now. Because of the sickness, I am not able to cook.” (Female, 50 yrs, right-sided hemiparesis)