Stroke is the second leading cause of death and a major cause of disability worldwide (Katan & Luft 2018). Disability because of stroke is still very common up to three years after the event (Kossi et al. 2021). In a 2016 meta-analysis, the total prevalence rate for stroke in Africa was estimated to be 3.5 cases per 1000 people, with an annual increasing trend of 12.0% (Ezejimofor et al. 2016). Relative to high-income countries, stroke affects more young people in middle- and low-income countries, and the affliction of younger people with stroke places an economic burden on such countries (Katan & Luft 2018). Furthermore, the highest incidence, prevalence and case fatality rates of stroke in the world have been reported in Africa (Adoukonou et al. 2021).

The World Health Organization’s international classification of functioning, disability and health (ICF) recognises functional disability as multidimensional and involving nonlinear concepts of the interaction between a person’s health condition and contextual factors. Commonly, stroke survivors experience activity limitations and social participation difficulties, even following a relatively good recovery from impairments (Van der Zee et al. 2013). The reported prevalence of social participation restrictions among stroke patients ranges from 12% to 64% (Adoukonou et al. 2018; Kossi et al. 2019). The improvement of ICF components depends on several factors, including the national economic situation, access to rehabilitation and cultural beliefs (Lekander et al. 2017; Mbada et al. 2019; Mendis 2013). In low-income countries, like Burundi, understanding the challenges faced by people who have suffered a stroke would be useful in identifying the long-term issues that care-based services for stroke will need to address (Ch’ng, French & McLean 2008).

Walking is often severely affected after stroke and the high prevalence of balance impairments remains a problem (Kossi et al. 2021; Pinto et al. 2014). People tend to reduce their walking speed to prevent the risk of falling, and a slow walking speed may limit one’s ability to perform the activities of daily living (ADLs) and limit one’s participation in family and community life (Faria-Fortini et al. 2019; Wang et al. 2020). In a sample of 147 patients, Perry and colleagues (1995) defined functional walking status according to three community walking speed categories: household ambulation (speed < 0.4 m/s, indicating severe walking limitations), limited-community ambulation (speed in the range of 0.4 m/s – 0.8 m/s, indicating moderate walking limitations) and full-community ambulation (speed > 0.8 m/s, reflecting slight or mild walking limitations). Although this classification comes from a high-income context, it can provide information on the functional status of walking in low-income countries. Factors shown to influence ADL abilities and social participation include age, gender, use of medication, time after stroke, number of physiotherapy sessions attended, walking speed, balance and motor function (Faria-Fortini et al. 2018; Kossi et al. 2019). However, consistently, walking speed has been shown to be the best indicator of walking performance (Faria-Fortini et al. 2019). Thus, physical outcomes, such as walking speed, should not be assessed in isolation when considering community ambulation post-stroke (Durcan, Flavin & Horgan 2016). To date, the relationship between walking speed and stroke outcomes has been studied in developing and high-income countries; but to the best of our knowledge, no studies have been conducted in Burundi (Faria-Fortini et al. 2019; Wang et al. 2020). Therefore, our study aimed (1) to describe balance impairments, ADL levels and social participation levels of chronic stroke patients in Burundi and (2) to investigate their relationship with walking speed.

A cross-sectional study was conducted at the National Center of Reference in Physical Therapy and Medical Rehabilitation, University Hospital Roi-Khaled, Bujumbura between September 2019 and February 2020. Our study population consisted of people suffering from a stroke in Burundi at the time of this study. Potential participants were identified by consulting admission records in hospitals and rehabilitation centres. Data for all potential participants were screened to ensure that they met our inclusion criteria: a clinical diagnosis (imaging or neurologist’s diagnosis) of a primary or recurring unilateral stroke, chronic stroke stage (time since stroke ≥ 6 months), age ≥ 18 years and ability to walk 10 m with or without an assistive device. Participants with a major cognitive impairment (community screening interview for dementia score ≥ 7) (Hall et al. 2000) and those who had other neurological impairments with permanent degenerative damage (such as Parkinson’s disease or Alzheimer’s disease) were excluded.

Our study described the level of balance impairments, the ability to perform ADLs and social participation and examined their relationships with walking speed, among people at the chronic stage of stroke in Burundi, a low-income country. Firstly, we found that most patients experienced poor balance, indicative of a risk of falling, low walking speed, moderate ADL limitations and social participation restrictions. Additionally, household ambulators (low walking speed) displayed major ADL limitations and moderate restrictions in their social participation, and there was a moderate and significant relationship between walking speed and balance. Activities of daily living limitation scores correlated strongly with walking speed, so that slower walking speeds were associated with more limitations. Walking speed, however, did not correlate with social participation.

Impaired balance affects the ability of stroke patients to walk, whether in developed or low-income settings. The low walking speed and poor balance in our sample may be linked to limited early stroke management and access to rehabilitation in low-income settings, as has been shown in previous studies (Bernhardt et al. 2020; Urimubenshi et al. 2018).

In his systematic review, Urimubenshi et al. (2018) concluded that stroke care in Africa fell below the recommended standards and invited African policy makers and health care professionals to combine their efforts to improve access to organised stroke care. Recently, there has been a call to action aimed at developing and delivering cost-effective and equitably accessible rehabilitation services in low- and middle-income countries (Bernhardt et al. 2020). Studies examining the frequency of balance impairments following a stroke (Chang, Gajasinghe & Arambepola 2015; Kossi et al. 2021; Vincent-Onabajo, Musa & Joseph 2018) have reported prevalence rates ranging from 16.7% (Vincent-Onabajo et al. 2018) to as high as 83% (Chang et al. 2015; Kossi et al. 2021). Our finding, that a faster walking speed is related to good balance, is consistent with prior studies (Obembe, Olaogun & Adedoyin 2014; Ojagbemi & Owolabi 2013). Balance disorders are characterised by a short support time and an asymmetrical left versus right step length, which leads to a slow walking speed (Khan & Chevidikunnan 2021). Two major goals of rehabilitation are to enable or improve walking ability and to promote independence in ADLs, both of which are highly dependent on standing balance (Kossi et al. 2019).

Overall, participants in our study showed limitations that affected them performing about half of the assessed ADLs, confirming the persistence of substantial ACTIVLIMs in the chronic phase of stroke. These results are similar to those of a 2019 study carried out in a comparable context of low-income countries (Kossi et al. 2019). In that study, Kossi et al. observed a mean activity level of 55% on the ACTIVLIM-Stroke scale in a sample of 64 stroke survivors, 6 months post-stroke. In 2020, Mwaka-Rutare et al. (2020) reported relatively higher ACTIVLIM-Stroke scores in stroke survivors with a longer post-stroke recovery period (average of 38 months). Their sample’s mean activity level was estimated to be 69%, whereas the overall level observed in our study was 51%. Given the similar socioeconomic contexts of the three studies, the differences in ADL ability levels might be attributed to the baseline clinical characteristics of the samples, such as the disability level. Consistent with this possibility, the mean walking speed of Mwaka-Rutare et al.’s sample (0.9 m/s) was faster than that of both our sample (0.68 m/s) and Kossi et al.’s sample (0.73 m/s).

According to the ICF, walking is a critical component in enabling the performance of ADLs and social participation for stroke patients (Geyh et al. 2004). A limited ability to walk following a stroke restricts patients’ independence and mobility at home and in the community (Schmid et al. 2007). In our study, walking speed was very different between individuals who had different ADL levels. A faster walking speed was associated with better ADL performance scores. Compared with household and limited-community ambulators, participants with full-community ambulation status reported higher ADL levels. Similarly, in their study in the Democratic Republic of Congo, in a socio-economic context similar to that of our study, Mwaka-Rutare et al. (2020) found that faster walking speeds were associated with better performance of basic ADLs. Moreover, our findings seem to be consistent with those of a 2007 study by Schmid et al. (2007) in a high-income context, wherein progression from one walking speed-based category to another was associated with ADL gains.

As in previous studies (Mwaka-Rutare et al. 2020; Schmid et al. 2007), walking speed in our did not correlate well with the participants’ social participation levels. This indicates that a faster walking speed is no longer a priority for the patient in the chronic phase of stroke, but a number of other factors, such as impairment, depression and social support, remain determinants of social participation in both the acute and chronic phases (Elloker & Rhoda 2018; Kossi et al. 2019). However, we obtained a lower optimal walking speed cut-off for distinguishing social participation (0.5 m/s) than was applied by Mwaka-Rutare et al. (2020) and Schmid et al. (2007) (>0.8 m/s). These differences in optimal walking speed cut-offs for social participation may be consequent to our sample being younger on average (52.1 ± 11.4 years) than the samples in these two studies (58.4 ± 12.9 years [Mwaka-Rutare et al. 2020] and 71.0 ± 10.64 years [Schmid et al. 2007]). Young patients are likely to develop better active coping strategies that may contribute to limiting the impact of the reduced walking speed on social participation. Secondly, the lack of a significant association between walking speed and social participation scores, in the chronic stage of stroke, suggests that the optimal cut-off walking speed may be sufficient to enable good social participation, which would support the assumption that social support may be the main factor in enabling social participation without a strict dependence on a faster walking speed (Elloker & Rhoda 2018; Mairami et al. 2020). Finally, the optimal walking speed cut-off of 0.5 m/s found here raises questions about the applicability of Perry et al.’s walking speed classification system in socioeconomic contexts that differ from the context where it was developed (the Los Angeles metropolitan area in the USA). In practice, requirements for safe community walking may differ across locations depending on contextual factors, such as environmental adaptations, the availability and affordability of technical assistance apparatuses and social supports. Further studies should investigate walking speed classification cut-offs that are well suited to an African context.

There are several implications from our findings for clinicians involved in rehabilitation after stroke in low-income settings. Our findings suggest that interventions aimed at improving balance and enhancing walking speed may be useful in promoting long-term ADL ability. Although our results did not confirm a clear association between walking speed and social participation, probably because of socioeconomic contexts, there is some evidence indicating that ADL level may be a key determinant of social participation in the subacute and chronic phases of stroke (Kossi et al. 2019).

Our study is the first to report the functional recovery of people post-stroke in Burundi, a low-income country. Functional recovery in our sample was evaluated with socioculturally adapted tools and in accordance with the ICF, considering impairment, activity limitations and participation restrictions. However, our findings need to be interpreted in light of some potential limitations. Firstly, because we excluded patients with major cognitive impairments, severe disability and/or with an unclear diagnosis, our findings may not be generalised to the entire stroke population in Burundi. Secondly, despite the statistical normality of the distributions, the unequal numbers of participants in the three groups could have influenced the statistical analyses. In addition, the use of patient-reported outcome tools to assess activity limitations and participation restrictions could generate a recall bias, but we believe that the use of culturally valid Rasch-based questionnaires with very good psychometric properties limits this risk of bias. Finally, Perry’s walking speed classification may not be well suited to the context of low-income countries, given the relatively low average age of their populations and their poorer access to rehabilitation services, compared with that in most developed countries.

Participants in our study presented major balance impairments with a high risk of falling, slow walking speeds, moderate ADL limitations and moderate restrictions in social participation. It was confirmed that walking speed was associated with balance ability and ADL levels. A faster walking speed was associated with better functional performance, while a moderate speed was sufficient to enable social participation. During the rehabilitation process, particularly in the chronic stage, a special emphasis should be placed on balance and walking speed. Training that targets walking speed and balance can be achieved at a low cost, making it well suited to low-income settings.